Loading...
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 */
44struct 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 datapath - datapath for flow-based packet switching
55 * @rcu: RCU callback head for deferred destruction.
56 * @list_node: Element in global 'dps' list.
57 * @table: flow table.
58 * @ports: Hash table for ports. %OVSP_LOCAL port always exists. Protected by
59 * ovs_mutex and RCU.
60 * @stats_percpu: Per-CPU datapath statistics.
61 * @net: Reference to net namespace.
62 * @max_headroom: the maximum headroom of all vports in this datapath; it will
63 * be used by all the internal vports in this dp.
64 *
65 * Context: See the comment on locking at the top of datapath.c for additional
66 * locking information.
67 */
68struct datapath {
69 struct rcu_head rcu;
70 struct list_head list_node;
71
72 /* Flow table. */
73 struct flow_table table;
74
75 /* Switch ports. */
76 struct hlist_head *ports;
77
78 /* Stats. */
79 struct dp_stats_percpu __percpu *stats_percpu;
80
81 /* Network namespace ref. */
82 possible_net_t net;
83
84 u32 user_features;
85
86 u32 max_headroom;
87
88 /* Switch meters. */
89 struct dp_meter_table meter_tbl;
90};
91
92/**
93 * struct ovs_skb_cb - OVS data in skb CB
94 * @input_vport: The original vport packet came in on. This value is cached
95 * when a packet is received by OVS.
96 * @mru: The maximum received fragement size; 0 if the packet is not
97 * fragmented.
98 * @acts_origlen: The netlink size of the flow actions applied to this skb.
99 * @cutlen: The number of bytes from the packet end to be removed.
100 */
101struct ovs_skb_cb {
102 struct vport *input_vport;
103 u16 mru;
104 u16 acts_origlen;
105 u32 cutlen;
106};
107#define OVS_CB(skb) ((struct ovs_skb_cb *)(skb)->cb)
108
109/**
110 * struct dp_upcall - metadata to include with a packet to send to userspace
111 * @cmd: One of %OVS_PACKET_CMD_*.
112 * @userdata: If nonnull, its variable-length value is passed to userspace as
113 * %OVS_PACKET_ATTR_USERDATA.
114 * @portid: Netlink portid to which packet should be sent. If @portid is 0
115 * then no packet is sent and the packet is accounted in the datapath's @n_lost
116 * counter.
117 * @egress_tun_info: If nonnull, becomes %OVS_PACKET_ATTR_EGRESS_TUN_KEY.
118 * @mru: If not zero, Maximum received IP fragment size.
119 */
120struct dp_upcall_info {
121 struct ip_tunnel_info *egress_tun_info;
122 const struct nlattr *userdata;
123 const struct nlattr *actions;
124 int actions_len;
125 u32 portid;
126 u8 cmd;
127 u16 mru;
128};
129
130/**
131 * struct ovs_net - Per net-namespace data for ovs.
132 * @dps: List of datapaths to enable dumping them all out.
133 * Protected by genl_mutex.
134 */
135struct ovs_net {
136 struct list_head dps;
137 struct work_struct dp_notify_work;
138 struct delayed_work masks_rebalance;
139#if IS_ENABLED(CONFIG_NETFILTER_CONNCOUNT)
140 struct ovs_ct_limit_info *ct_limit_info;
141#endif
142
143 /* Module reference for configuring conntrack. */
144 bool xt_label;
145};
146
147/**
148 * enum ovs_pkt_hash_types - hash info to include with a packet
149 * to send to userspace.
150 * @OVS_PACKET_HASH_SW_BIT: indicates hash was computed in software stack.
151 * @OVS_PACKET_HASH_L4_BIT: indicates hash is a canonical 4-tuple hash
152 * over transport ports.
153 */
154enum ovs_pkt_hash_types {
155 OVS_PACKET_HASH_SW_BIT = (1ULL << 32),
156 OVS_PACKET_HASH_L4_BIT = (1ULL << 33),
157};
158
159extern unsigned int ovs_net_id;
160void ovs_lock(void);
161void ovs_unlock(void);
162
163#ifdef CONFIG_LOCKDEP
164int lockdep_ovsl_is_held(void);
165#else
166#define lockdep_ovsl_is_held() 1
167#endif
168
169#define ASSERT_OVSL() WARN_ON(!lockdep_ovsl_is_held())
170#define ovsl_dereference(p) \
171 rcu_dereference_protected(p, lockdep_ovsl_is_held())
172#define rcu_dereference_ovsl(p) \
173 rcu_dereference_check(p, lockdep_ovsl_is_held())
174
175static inline struct net *ovs_dp_get_net(const struct datapath *dp)
176{
177 return read_pnet(&dp->net);
178}
179
180static inline void ovs_dp_set_net(struct datapath *dp, struct net *net)
181{
182 write_pnet(&dp->net, net);
183}
184
185struct vport *ovs_lookup_vport(const struct datapath *dp, u16 port_no);
186
187static inline struct vport *ovs_vport_rcu(const struct datapath *dp, int port_no)
188{
189 WARN_ON_ONCE(!rcu_read_lock_held());
190 return ovs_lookup_vport(dp, port_no);
191}
192
193static inline struct vport *ovs_vport_ovsl_rcu(const struct datapath *dp, int port_no)
194{
195 WARN_ON_ONCE(!rcu_read_lock_held() && !lockdep_ovsl_is_held());
196 return ovs_lookup_vport(dp, port_no);
197}
198
199static inline struct vport *ovs_vport_ovsl(const struct datapath *dp, int port_no)
200{
201 ASSERT_OVSL();
202 return ovs_lookup_vport(dp, port_no);
203}
204
205/* Must be called with rcu_read_lock. */
206static inline struct datapath *get_dp_rcu(struct net *net, int dp_ifindex)
207{
208 struct net_device *dev = dev_get_by_index_rcu(net, dp_ifindex);
209
210 if (dev) {
211 struct vport *vport = ovs_internal_dev_get_vport(dev);
212
213 if (vport)
214 return vport->dp;
215 }
216
217 return NULL;
218}
219
220/* The caller must hold either ovs_mutex or rcu_read_lock to keep the
221 * returned dp pointer valid.
222 */
223static inline struct datapath *get_dp(struct net *net, int dp_ifindex)
224{
225 struct datapath *dp;
226
227 WARN_ON_ONCE(!rcu_read_lock_held() && !lockdep_ovsl_is_held());
228 rcu_read_lock();
229 dp = get_dp_rcu(net, dp_ifindex);
230 rcu_read_unlock();
231
232 return dp;
233}
234
235extern struct notifier_block ovs_dp_device_notifier;
236extern struct genl_family dp_vport_genl_family;
237
238DECLARE_STATIC_KEY_FALSE(tc_recirc_sharing_support);
239
240void ovs_dp_process_packet(struct sk_buff *skb, struct sw_flow_key *key);
241void ovs_dp_detach_port(struct vport *);
242int ovs_dp_upcall(struct datapath *, struct sk_buff *,
243 const struct sw_flow_key *, const struct dp_upcall_info *,
244 uint32_t cutlen);
245
246const char *ovs_dp_name(const struct datapath *dp);
247struct sk_buff *ovs_vport_cmd_build_info(struct vport *vport, struct net *net,
248 u32 portid, u32 seq, u8 cmd);
249
250int ovs_execute_actions(struct datapath *dp, struct sk_buff *skb,
251 const struct sw_flow_actions *, struct sw_flow_key *);
252
253void ovs_dp_notify_wq(struct work_struct *work);
254
255int action_fifos_init(void);
256void action_fifos_exit(void);
257
258/* 'KEY' must not have any bits set outside of the 'MASK' */
259#define OVS_MASKED(OLD, KEY, MASK) ((KEY) | ((OLD) & ~(MASK)))
260#define OVS_SET_MASKED(OLD, KEY, MASK) ((OLD) = OVS_MASKED(OLD, KEY, MASK))
261
262#define OVS_NLERR(logging_allowed, fmt, ...) \
263do { \
264 if (logging_allowed && net_ratelimit()) \
265 pr_info("netlink: " fmt "\n", ##__VA_ARGS__); \
266} while (0)
267#endif /* datapath.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 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 */
54struct 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 */
77struct 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 */
110struct 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 */
129struct 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 */
144struct ovs_net {
145 struct list_head dps;
146 struct work_struct dp_notify_work;
147
148 /* Module reference for configuring conntrack. */
149 bool xt_label;
150};
151
152extern unsigned int ovs_net_id;
153void ovs_lock(void);
154void ovs_unlock(void);
155
156#ifdef CONFIG_LOCKDEP
157int lockdep_ovsl_is_held(void);
158#else
159#define lockdep_ovsl_is_held() 1
160#endif
161
162#define ASSERT_OVSL() WARN_ON(!lockdep_ovsl_is_held())
163#define ovsl_dereference(p) \
164 rcu_dereference_protected(p, lockdep_ovsl_is_held())
165#define rcu_dereference_ovsl(p) \
166 rcu_dereference_check(p, lockdep_ovsl_is_held())
167
168static inline struct net *ovs_dp_get_net(const struct datapath *dp)
169{
170 return read_pnet(&dp->net);
171}
172
173static inline void ovs_dp_set_net(struct datapath *dp, struct net *net)
174{
175 write_pnet(&dp->net, net);
176}
177
178struct vport *ovs_lookup_vport(const struct datapath *dp, u16 port_no);
179
180static inline struct vport *ovs_vport_rcu(const struct datapath *dp, int port_no)
181{
182 WARN_ON_ONCE(!rcu_read_lock_held());
183 return ovs_lookup_vport(dp, port_no);
184}
185
186static inline struct vport *ovs_vport_ovsl_rcu(const struct datapath *dp, int port_no)
187{
188 WARN_ON_ONCE(!rcu_read_lock_held() && !lockdep_ovsl_is_held());
189 return ovs_lookup_vport(dp, port_no);
190}
191
192static inline struct vport *ovs_vport_ovsl(const struct datapath *dp, int port_no)
193{
194 ASSERT_OVSL();
195 return ovs_lookup_vport(dp, port_no);
196}
197
198/* Must be called with rcu_read_lock. */
199static inline struct datapath *get_dp_rcu(struct net *net, int dp_ifindex)
200{
201 struct net_device *dev = dev_get_by_index_rcu(net, dp_ifindex);
202
203 if (dev) {
204 struct vport *vport = ovs_internal_dev_get_vport(dev);
205
206 if (vport)
207 return vport->dp;
208 }
209
210 return NULL;
211}
212
213/* The caller must hold either ovs_mutex or rcu_read_lock to keep the
214 * returned dp pointer valid.
215 */
216static inline struct datapath *get_dp(struct net *net, int dp_ifindex)
217{
218 struct datapath *dp;
219
220 WARN_ON_ONCE(!rcu_read_lock_held() && !lockdep_ovsl_is_held());
221 rcu_read_lock();
222 dp = get_dp_rcu(net, dp_ifindex);
223 rcu_read_unlock();
224
225 return dp;
226}
227
228extern struct notifier_block ovs_dp_device_notifier;
229extern struct genl_family dp_vport_genl_family;
230
231void ovs_dp_process_packet(struct sk_buff *skb, struct sw_flow_key *key);
232void ovs_dp_detach_port(struct vport *);
233int ovs_dp_upcall(struct datapath *, struct sk_buff *,
234 const struct sw_flow_key *, const struct dp_upcall_info *,
235 uint32_t cutlen);
236
237const char *ovs_dp_name(const struct datapath *dp);
238struct sk_buff *ovs_vport_cmd_build_info(struct vport *vport, struct net *net,
239 u32 portid, u32 seq, u8 cmd);
240
241int ovs_execute_actions(struct datapath *dp, struct sk_buff *skb,
242 const struct sw_flow_actions *, struct sw_flow_key *);
243
244void ovs_dp_notify_wq(struct work_struct *work);
245
246int action_fifos_init(void);
247void action_fifos_exit(void);
248
249/* 'KEY' must not have any bits set outside of the 'MASK' */
250#define OVS_MASKED(OLD, KEY, MASK) ((KEY) | ((OLD) & ~(MASK)))
251#define OVS_SET_MASKED(OLD, KEY, MASK) ((OLD) = OVS_MASKED(OLD, KEY, MASK))
252
253#define OVS_NLERR(logging_allowed, fmt, ...) \
254do { \
255 if (logging_allowed && net_ratelimit()) \
256 pr_info("netlink: " fmt "\n", ##__VA_ARGS__); \
257} while (0)
258#endif /* datapath.h */