Loading...
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
34#define DP_MAX_PORTS USHRT_MAX
35#define DP_VPORT_HASH_BUCKETS 1024
36
37#define SAMPLE_ACTION_DEPTH 3
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
98/**
99 * struct ovs_skb_cb - OVS data in skb CB
100 * @input_vport: The original vport packet came in on. This value is cached
101 * when a packet is received by OVS.
102 * @mru: The maximum received fragement size; 0 if the packet is not
103 * @cutlen: The number of bytes from the packet end to be removed.
104 * fragmented.
105 */
106struct ovs_skb_cb {
107 struct vport *input_vport;
108 u16 mru;
109 u32 cutlen;
110};
111#define OVS_CB(skb) ((struct ovs_skb_cb *)(skb)->cb)
112
113/**
114 * struct dp_upcall - metadata to include with a packet to send to userspace
115 * @cmd: One of %OVS_PACKET_CMD_*.
116 * @userdata: If nonnull, its variable-length value is passed to userspace as
117 * %OVS_PACKET_ATTR_USERDATA.
118 * @portid: Netlink portid to which packet should be sent. If @portid is 0
119 * then no packet is sent and the packet is accounted in the datapath's @n_lost
120 * counter.
121 * @egress_tun_info: If nonnull, becomes %OVS_PACKET_ATTR_EGRESS_TUN_KEY.
122 * @mru: If not zero, Maximum received IP fragment size.
123 */
124struct dp_upcall_info {
125 struct ip_tunnel_info *egress_tun_info;
126 const struct nlattr *userdata;
127 const struct nlattr *actions;
128 int actions_len;
129 u32 portid;
130 u8 cmd;
131 u16 mru;
132};
133
134/**
135 * struct ovs_net - Per net-namespace data for ovs.
136 * @dps: List of datapaths to enable dumping them all out.
137 * Protected by genl_mutex.
138 */
139struct ovs_net {
140 struct list_head dps;
141 struct work_struct dp_notify_work;
142
143 /* Module reference for configuring conntrack. */
144 bool xt_label;
145};
146
147extern unsigned int ovs_net_id;
148void ovs_lock(void);
149void ovs_unlock(void);
150
151#ifdef CONFIG_LOCKDEP
152int lockdep_ovsl_is_held(void);
153#else
154#define lockdep_ovsl_is_held() 1
155#endif
156
157#define ASSERT_OVSL() WARN_ON(!lockdep_ovsl_is_held())
158#define ovsl_dereference(p) \
159 rcu_dereference_protected(p, lockdep_ovsl_is_held())
160#define rcu_dereference_ovsl(p) \
161 rcu_dereference_check(p, lockdep_ovsl_is_held())
162
163static inline struct net *ovs_dp_get_net(const struct datapath *dp)
164{
165 return read_pnet(&dp->net);
166}
167
168static inline void ovs_dp_set_net(struct datapath *dp, struct net *net)
169{
170 write_pnet(&dp->net, net);
171}
172
173struct vport *ovs_lookup_vport(const struct datapath *dp, u16 port_no);
174
175static inline struct vport *ovs_vport_rcu(const struct datapath *dp, int port_no)
176{
177 WARN_ON_ONCE(!rcu_read_lock_held());
178 return ovs_lookup_vport(dp, port_no);
179}
180
181static inline struct vport *ovs_vport_ovsl_rcu(const struct datapath *dp, int port_no)
182{
183 WARN_ON_ONCE(!rcu_read_lock_held() && !lockdep_ovsl_is_held());
184 return ovs_lookup_vport(dp, port_no);
185}
186
187static inline struct vport *ovs_vport_ovsl(const struct datapath *dp, int port_no)
188{
189 ASSERT_OVSL();
190 return ovs_lookup_vport(dp, port_no);
191}
192
193extern struct notifier_block ovs_dp_device_notifier;
194extern struct genl_family dp_vport_genl_family;
195
196void ovs_dp_process_packet(struct sk_buff *skb, struct sw_flow_key *key);
197void ovs_dp_detach_port(struct vport *);
198int ovs_dp_upcall(struct datapath *, struct sk_buff *,
199 const struct sw_flow_key *, const struct dp_upcall_info *,
200 uint32_t cutlen);
201
202const char *ovs_dp_name(const struct datapath *dp);
203struct sk_buff *ovs_vport_cmd_build_info(struct vport *, u32 pid, u32 seq,
204 u8 cmd);
205
206int ovs_execute_actions(struct datapath *dp, struct sk_buff *skb,
207 const struct sw_flow_actions *, struct sw_flow_key *);
208
209void ovs_dp_notify_wq(struct work_struct *work);
210
211int action_fifos_init(void);
212void action_fifos_exit(void);
213
214/* 'KEY' must not have any bits set outside of the 'MASK' */
215#define OVS_MASKED(OLD, KEY, MASK) ((KEY) | ((OLD) & ~(MASK)))
216#define OVS_SET_MASKED(OLD, KEY, MASK) ((OLD) = OVS_MASKED(OLD, KEY, MASK))
217
218#define OVS_NLERR(logging_allowed, fmt, ...) \
219do { \
220 if (logging_allowed && net_ratelimit()) \
221 pr_info("netlink: " fmt "\n", ##__VA_ARGS__); \
222} while (0)
223#endif /* datapath.h */
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 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 */
62struct 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 */
84struct 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 */
121struct 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 */
141struct 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 */
156struct 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
165/**
166 * enum ovs_pkt_hash_types - hash info to include with a packet
167 * to send to userspace.
168 * @OVS_PACKET_HASH_SW_BIT: indicates hash was computed in software stack.
169 * @OVS_PACKET_HASH_L4_BIT: indicates hash is a canonical 4-tuple hash
170 * over transport ports.
171 */
172enum ovs_pkt_hash_types {
173 OVS_PACKET_HASH_SW_BIT = (1ULL << 32),
174 OVS_PACKET_HASH_L4_BIT = (1ULL << 33),
175};
176
177extern unsigned int ovs_net_id;
178void ovs_lock(void);
179void ovs_unlock(void);
180
181#ifdef CONFIG_LOCKDEP
182int lockdep_ovsl_is_held(void);
183#else
184#define lockdep_ovsl_is_held() 1
185#endif
186
187#define ASSERT_OVSL() WARN_ON(!lockdep_ovsl_is_held())
188#define ovsl_dereference(p) \
189 rcu_dereference_protected(p, lockdep_ovsl_is_held())
190#define rcu_dereference_ovsl(p) \
191 rcu_dereference_check(p, lockdep_ovsl_is_held())
192
193static inline struct net *ovs_dp_get_net(const struct datapath *dp)
194{
195 return read_pnet(&dp->net);
196}
197
198static inline void ovs_dp_set_net(struct datapath *dp, struct net *net)
199{
200 write_pnet(&dp->net, net);
201}
202
203struct vport *ovs_lookup_vport(const struct datapath *dp, u16 port_no);
204
205static inline struct vport *ovs_vport_rcu(const struct datapath *dp, int port_no)
206{
207 WARN_ON_ONCE(!rcu_read_lock_held());
208 return ovs_lookup_vport(dp, port_no);
209}
210
211static inline struct vport *ovs_vport_ovsl_rcu(const struct datapath *dp, int port_no)
212{
213 WARN_ON_ONCE(!rcu_read_lock_held() && !lockdep_ovsl_is_held());
214 return ovs_lookup_vport(dp, port_no);
215}
216
217static inline struct vport *ovs_vport_ovsl(const struct datapath *dp, int port_no)
218{
219 ASSERT_OVSL();
220 return ovs_lookup_vport(dp, port_no);
221}
222
223/* Must be called with rcu_read_lock. */
224static inline struct datapath *get_dp_rcu(struct net *net, int dp_ifindex)
225{
226 struct net_device *dev = dev_get_by_index_rcu(net, dp_ifindex);
227
228 if (dev) {
229 struct vport *vport = ovs_internal_dev_get_vport(dev);
230
231 if (vport)
232 return vport->dp;
233 }
234
235 return NULL;
236}
237
238/* The caller must hold either ovs_mutex or rcu_read_lock to keep the
239 * returned dp pointer valid.
240 */
241static inline struct datapath *get_dp(struct net *net, int dp_ifindex)
242{
243 struct datapath *dp;
244
245 WARN_ON_ONCE(!rcu_read_lock_held() && !lockdep_ovsl_is_held());
246 rcu_read_lock();
247 dp = get_dp_rcu(net, dp_ifindex);
248 rcu_read_unlock();
249
250 return dp;
251}
252
253extern struct notifier_block ovs_dp_device_notifier;
254extern struct genl_family dp_vport_genl_family;
255
256void ovs_dp_process_packet(struct sk_buff *skb, struct sw_flow_key *key);
257void ovs_dp_detach_port(struct vport *);
258int ovs_dp_upcall(struct datapath *, struct sk_buff *,
259 const struct sw_flow_key *, const struct dp_upcall_info *,
260 uint32_t cutlen);
261
262u32 ovs_dp_get_upcall_portid(const struct datapath *dp, uint32_t cpu_id);
263
264const char *ovs_dp_name(const struct datapath *dp);
265struct sk_buff *ovs_vport_cmd_build_info(struct vport *vport, struct net *net,
266 u32 portid, u32 seq, u8 cmd);
267
268int ovs_execute_actions(struct datapath *dp, struct sk_buff *skb,
269 const struct sw_flow_actions *, struct sw_flow_key *);
270
271void ovs_dp_notify_wq(struct work_struct *work);
272
273int action_fifos_init(void);
274void action_fifos_exit(void);
275
276/* 'KEY' must not have any bits set outside of the 'MASK' */
277#define OVS_MASKED(OLD, KEY, MASK) ((KEY) | ((OLD) & ~(MASK)))
278#define OVS_SET_MASKED(OLD, KEY, MASK) ((OLD) = OVS_MASKED(OLD, KEY, MASK))
279
280#define OVS_NLERR(logging_allowed, fmt, ...) \
281do { \
282 if (logging_allowed && net_ratelimit()) \
283 pr_info("netlink: " fmt "\n", ##__VA_ARGS__); \
284} while (0)
285#endif /* datapath.h */