xref: /openbmc/linux/net/openvswitch/datapath.h (revision 93f5715e)
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 DATAPATH_H
20 #define DATAPATH_H 1
21 
22 #include <asm/page.h>
23 #include <linux/kernel.h>
24 #include <linux/mutex.h>
25 #include <linux/netdevice.h>
26 #include <linux/skbuff.h>
27 #include <linux/u64_stats_sync.h>
28 #include <net/ip_tunnels.h>
29 
30 #include "conntrack.h"
31 #include "flow.h"
32 #include "flow_table.h"
33 #include "meter.h"
34 #include "vport-internal_dev.h"
35 
36 #define DP_MAX_PORTS           USHRT_MAX
37 #define DP_VPORT_HASH_BUCKETS  1024
38 
39 /**
40  * struct dp_stats_percpu - per-cpu packet processing statistics for a given
41  * datapath.
42  * @n_hit: Number of received packets for which a matching flow was found in
43  * the flow table.
44  * @n_miss: Number of received packets that had no matching flow in the flow
45  * table.  The sum of @n_hit and @n_miss is the number of packets that have
46  * been received by the datapath.
47  * @n_lost: Number of received packets that had no matching flow in the flow
48  * table that could not be sent to userspace (normally due to an overflow in
49  * one of the datapath's queues).
50  * @n_mask_hit: Number of masks looked up for flow match.
51  *   @n_mask_hit / (@n_hit + @n_missed)  will be the average masks looked
52  *   up per packet.
53  */
54 struct dp_stats_percpu {
55 	u64 n_hit;
56 	u64 n_missed;
57 	u64 n_lost;
58 	u64 n_mask_hit;
59 	struct u64_stats_sync syncp;
60 };
61 
62 /**
63  * struct datapath - datapath for flow-based packet switching
64  * @rcu: RCU callback head for deferred destruction.
65  * @list_node: Element in global 'dps' list.
66  * @table: flow table.
67  * @ports: Hash table for ports.  %OVSP_LOCAL port always exists.  Protected by
68  * ovs_mutex and RCU.
69  * @stats_percpu: Per-CPU datapath statistics.
70  * @net: Reference to net namespace.
71  * @max_headroom: the maximum headroom of all vports in this datapath; it will
72  * be used by all the internal vports in this dp.
73  *
74  * Context: See the comment on locking at the top of datapath.c for additional
75  * locking information.
76  */
77 struct datapath {
78 	struct rcu_head rcu;
79 	struct list_head list_node;
80 
81 	/* Flow table. */
82 	struct flow_table table;
83 
84 	/* Switch ports. */
85 	struct hlist_head *ports;
86 
87 	/* Stats. */
88 	struct dp_stats_percpu __percpu *stats_percpu;
89 
90 	/* Network namespace ref. */
91 	possible_net_t net;
92 
93 	u32 user_features;
94 
95 	u32 max_headroom;
96 
97 	/* Switch meters. */
98 	struct hlist_head *meters;
99 };
100 
101 /**
102  * struct ovs_skb_cb - OVS data in skb CB
103  * @input_vport: The original vport packet came in on. This value is cached
104  * when a packet is received by OVS.
105  * @mru: The maximum received fragement size; 0 if the packet is not
106  * fragmented.
107  * @acts_origlen: The netlink size of the flow actions applied to this skb.
108  * @cutlen: The number of bytes from the packet end to be removed.
109  */
110 struct ovs_skb_cb {
111 	struct vport		*input_vport;
112 	u16			mru;
113 	u16			acts_origlen;
114 	u32			cutlen;
115 };
116 #define OVS_CB(skb) ((struct ovs_skb_cb *)(skb)->cb)
117 
118 /**
119  * struct dp_upcall - metadata to include with a packet to send to userspace
120  * @cmd: One of %OVS_PACKET_CMD_*.
121  * @userdata: If nonnull, its variable-length value is passed to userspace as
122  * %OVS_PACKET_ATTR_USERDATA.
123  * @portid: Netlink portid to which packet should be sent.  If @portid is 0
124  * then no packet is sent and the packet is accounted in the datapath's @n_lost
125  * counter.
126  * @egress_tun_info: If nonnull, becomes %OVS_PACKET_ATTR_EGRESS_TUN_KEY.
127  * @mru: If not zero, Maximum received IP fragment size.
128  */
129 struct dp_upcall_info {
130 	struct ip_tunnel_info *egress_tun_info;
131 	const struct nlattr *userdata;
132 	const struct nlattr *actions;
133 	int actions_len;
134 	u32 portid;
135 	u8 cmd;
136 	u16 mru;
137 };
138 
139 /**
140  * struct ovs_net - Per net-namespace data for ovs.
141  * @dps: List of datapaths to enable dumping them all out.
142  * Protected by genl_mutex.
143  */
144 struct ovs_net {
145 	struct list_head dps;
146 	struct work_struct dp_notify_work;
147 #if	IS_ENABLED(CONFIG_NETFILTER_CONNCOUNT)
148 	struct ovs_ct_limit_info *ct_limit_info;
149 #endif
150 
151 	/* Module reference for configuring conntrack. */
152 	bool xt_label;
153 };
154 
155 extern unsigned int ovs_net_id;
156 void ovs_lock(void);
157 void ovs_unlock(void);
158 
159 #ifdef CONFIG_LOCKDEP
160 int lockdep_ovsl_is_held(void);
161 #else
162 #define lockdep_ovsl_is_held()	1
163 #endif
164 
165 #define ASSERT_OVSL()		WARN_ON(!lockdep_ovsl_is_held())
166 #define ovsl_dereference(p)					\
167 	rcu_dereference_protected(p, lockdep_ovsl_is_held())
168 #define rcu_dereference_ovsl(p)					\
169 	rcu_dereference_check(p, lockdep_ovsl_is_held())
170 
171 static inline struct net *ovs_dp_get_net(const struct datapath *dp)
172 {
173 	return read_pnet(&dp->net);
174 }
175 
176 static inline void ovs_dp_set_net(struct datapath *dp, struct net *net)
177 {
178 	write_pnet(&dp->net, net);
179 }
180 
181 struct vport *ovs_lookup_vport(const struct datapath *dp, u16 port_no);
182 
183 static inline struct vport *ovs_vport_rcu(const struct datapath *dp, int port_no)
184 {
185 	WARN_ON_ONCE(!rcu_read_lock_held());
186 	return ovs_lookup_vport(dp, port_no);
187 }
188 
189 static inline struct vport *ovs_vport_ovsl_rcu(const struct datapath *dp, int port_no)
190 {
191 	WARN_ON_ONCE(!rcu_read_lock_held() && !lockdep_ovsl_is_held());
192 	return ovs_lookup_vport(dp, port_no);
193 }
194 
195 static inline struct vport *ovs_vport_ovsl(const struct datapath *dp, int port_no)
196 {
197 	ASSERT_OVSL();
198 	return ovs_lookup_vport(dp, port_no);
199 }
200 
201 /* Must be called with rcu_read_lock. */
202 static inline struct datapath *get_dp_rcu(struct net *net, int dp_ifindex)
203 {
204 	struct net_device *dev = dev_get_by_index_rcu(net, dp_ifindex);
205 
206 	if (dev) {
207 		struct vport *vport = ovs_internal_dev_get_vport(dev);
208 
209 		if (vport)
210 			return vport->dp;
211 	}
212 
213 	return NULL;
214 }
215 
216 /* The caller must hold either ovs_mutex or rcu_read_lock to keep the
217  * returned dp pointer valid.
218  */
219 static inline struct datapath *get_dp(struct net *net, int dp_ifindex)
220 {
221 	struct datapath *dp;
222 
223 	WARN_ON_ONCE(!rcu_read_lock_held() && !lockdep_ovsl_is_held());
224 	rcu_read_lock();
225 	dp = get_dp_rcu(net, dp_ifindex);
226 	rcu_read_unlock();
227 
228 	return dp;
229 }
230 
231 extern struct notifier_block ovs_dp_device_notifier;
232 extern struct genl_family dp_vport_genl_family;
233 
234 void ovs_dp_process_packet(struct sk_buff *skb, struct sw_flow_key *key);
235 void ovs_dp_detach_port(struct vport *);
236 int ovs_dp_upcall(struct datapath *, struct sk_buff *,
237 		  const struct sw_flow_key *, const struct dp_upcall_info *,
238 		  uint32_t cutlen);
239 
240 const char *ovs_dp_name(const struct datapath *dp);
241 struct sk_buff *ovs_vport_cmd_build_info(struct vport *vport, struct net *net,
242 					 u32 portid, u32 seq, u8 cmd);
243 
244 int ovs_execute_actions(struct datapath *dp, struct sk_buff *skb,
245 			const struct sw_flow_actions *, struct sw_flow_key *);
246 
247 void ovs_dp_notify_wq(struct work_struct *work);
248 
249 int action_fifos_init(void);
250 void action_fifos_exit(void);
251 
252 /* 'KEY' must not have any bits set outside of the 'MASK' */
253 #define OVS_MASKED(OLD, KEY, MASK) ((KEY) | ((OLD) & ~(MASK)))
254 #define OVS_SET_MASKED(OLD, KEY, MASK) ((OLD) = OVS_MASKED(OLD, KEY, MASK))
255 
256 #define OVS_NLERR(logging_allowed, fmt, ...)			\
257 do {								\
258 	if (logging_allowed && net_ratelimit())			\
259 		pr_info("netlink: " fmt "\n", ##__VA_ARGS__);	\
260 } while (0)
261 #endif /* datapath.h */
262