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 #include <net/mpls.h>
17
18 #include "conntrack.h"
19 #include "flow.h"
20 #include "flow_table.h"
21 #include "meter.h"
22 #include "vport-internal_dev.h"
23
24 #define DP_MAX_PORTS USHRT_MAX
25 #define DP_VPORT_HASH_BUCKETS 1024
26 #define DP_MASKS_REBALANCE_INTERVAL 4000
27
28 /**
29 * struct dp_stats_percpu - per-cpu packet processing statistics for a given
30 * datapath.
31 * @n_hit: Number of received packets for which a matching flow was found in
32 * the flow table.
33 * @n_missed: Number of received packets that had no matching flow in the flow
34 * table. The sum of @n_hit and @n_missed is the number of packets that have
35 * been received by the datapath.
36 * @n_lost: Number of received packets that had no matching flow in the flow
37 * table that could not be sent to userspace (normally due to an overflow in
38 * one of the datapath's queues).
39 * @n_mask_hit: Number of masks looked up for flow match.
40 * @n_mask_hit / (@n_hit + @n_missed) will be the average masks looked
41 * up per packet.
42 * @n_cache_hit: The number of received packets that had their mask found using
43 * the mask cache.
44 * @syncp: Synchronization point for 64bit counters.
45 */
46 struct dp_stats_percpu {
47 u64 n_hit;
48 u64 n_missed;
49 u64 n_lost;
50 u64 n_mask_hit;
51 u64 n_cache_hit;
52 struct u64_stats_sync syncp;
53 };
54
55 /**
56 * struct dp_nlsk_pids - array of netlink portids of for a datapath.
57 * This is used when OVS_DP_F_DISPATCH_UPCALL_PER_CPU
58 * is enabled and must be protected by rcu.
59 * @rcu: RCU callback head for deferred destruction.
60 * @n_pids: Size of @pids array.
61 * @pids: Array storing the Netlink socket PIDs indexed by CPU ID for packets
62 * that miss the flow table.
63 */
64 struct dp_nlsk_pids {
65 struct rcu_head rcu;
66 u32 n_pids;
67 u32 pids[];
68 };
69
70 /**
71 * struct datapath - datapath for flow-based packet switching
72 * @rcu: RCU callback head for deferred destruction.
73 * @list_node: Element in global 'dps' list.
74 * @table: flow table.
75 * @ports: Hash table for ports. %OVSP_LOCAL port always exists. Protected by
76 * ovs_mutex and RCU.
77 * @stats_percpu: Per-CPU datapath statistics.
78 * @net: Reference to net namespace.
79 * @user_features: Bitmap of enabled %OVS_DP_F_* features.
80 * @max_headroom: The maximum headroom of all vports in this datapath; it will
81 * be used by all the internal vports in this dp.
82 * @meter_tbl: Meter table.
83 * @upcall_portids: RCU protected 'struct dp_nlsk_pids'.
84 *
85 * Context: See the comment on locking at the top of datapath.c for additional
86 * locking information.
87 */
88 struct datapath {
89 struct rcu_head rcu;
90 struct list_head list_node;
91
92 /* Flow table. */
93 struct flow_table table;
94
95 /* Switch ports. */
96 struct hlist_head *ports;
97
98 /* Stats. */
99 struct dp_stats_percpu __percpu *stats_percpu;
100
101 /* Network namespace ref. */
102 possible_net_t net;
103
104 u32 user_features;
105
106 u32 max_headroom;
107
108 /* Switch meters. */
109 struct dp_meter_table meter_tbl;
110
111 struct dp_nlsk_pids __rcu *upcall_portids;
112 };
113
114 /**
115 * struct ovs_skb_cb - OVS data in skb CB
116 * @input_vport: The original vport packet came in on. This value is cached
117 * when a packet is received by OVS.
118 * @mru: The maximum received fragement size; 0 if the packet is not
119 * fragmented.
120 * @acts_origlen: The netlink size of the flow actions applied to this skb.
121 * @cutlen: The number of bytes in the packet to preserve on output.
122 * @probability: The sampling probability that was applied to this skb; 0 means
123 * no sampling has occurred; U32_MAX means 100% probability.
124 * @upcall_pid: Netlink socket PID to use for sending this packet to userspace;
125 * 0 means "not set" and default per-CPU or per-vport dispatch should be used.
126 */
127 struct ovs_skb_cb {
128 struct vport *input_vport;
129 u16 mru;
130 u16 acts_origlen;
131 u32 cutlen;
132 u32 probability;
133 u32 upcall_pid;
134 };
135 #define OVS_CB(skb) ((struct ovs_skb_cb *)(skb)->cb)
136
137 /**
138 * struct dp_upcall_info - metadata to include with a packet sent to userspace
139 * @cmd: One of %OVS_PACKET_CMD_*.
140 * @userdata: If nonnull, its variable-length value is passed to userspace as
141 * %OVS_PACKET_ATTR_USERDATA.
142 * @actions: If nonnull, its variable-length value is passed to userspace as
143 * %OVS_PACKET_ATTR_ACTIONS.
144 * @actions_len: The length of the @actions.
145 * @portid: Netlink portid to which packet should be sent. If @portid is 0
146 * then no packet is sent and the packet is accounted in the datapath's @n_lost
147 * counter.
148 * @egress_tun_info: If nonnull, becomes %OVS_PACKET_ATTR_EGRESS_TUN_KEY.
149 * @mru: If not zero, Maximum received IP fragment size.
150 */
151 struct dp_upcall_info {
152 struct ip_tunnel_info *egress_tun_info;
153 const struct nlattr *userdata;
154 const struct nlattr *actions;
155 int actions_len;
156 u32 portid;
157 u8 cmd;
158 u16 mru;
159 };
160
161 /**
162 * struct ovs_net - Per net-namespace data for ovs.
163 * @dps: List of datapaths to enable dumping them all out.
164 * Protected by genl_mutex.
165 * @dp_notify_work: A work notifier to handle port unregistering.
166 * @masks_rebalance: A work to periodically optimize flow table caches.
167 * @ct_limit_info: Hash table of conntrack zone connection limits. Protected
168 * by RCU; updates and teardown are serialized by ovs_mutex. May be NULL during
169 * netns teardown.
170 * @ct_limit_exit_data: CT limit state detached at .pre_exit, freed at .exit.
171 * @xt_label: Whether connlables are configured for the network or not.
172 */
173 struct ovs_net {
174 struct list_head dps;
175 struct work_struct dp_notify_work;
176 struct delayed_work masks_rebalance;
177 #if IS_ENABLED(CONFIG_NETFILTER_CONNCOUNT)
178 struct ovs_ct_limit_info __rcu *ct_limit_info;
179 struct ovs_ct_limit_info *ct_limit_exit_data;
180 #endif
181 bool xt_label;
182 };
183
184 #define MAX_L2_LEN (VLAN_ETH_HLEN + 3 * MPLS_HLEN)
185 struct ovs_frag_data {
186 unsigned long dst;
187 struct vport *vport;
188 struct ovs_skb_cb cb;
189 __be16 inner_protocol;
190 u16 network_offset; /* valid only for MPLS */
191 u16 vlan_tci;
192 __be16 vlan_proto;
193 unsigned int l2_len;
194 u8 mac_proto;
195 u8 l2_data[MAX_L2_LEN];
196 };
197
198 struct deferred_action {
199 struct sk_buff *skb;
200 const struct nlattr *actions;
201 int actions_len;
202
203 /* Store pkt_key clone when creating deferred action. */
204 struct sw_flow_key pkt_key;
205 };
206
207 #define DEFERRED_ACTION_FIFO_SIZE 10
208 #define OVS_RECURSION_LIMIT 5
209 #define OVS_DEFERRED_ACTION_THRESHOLD (OVS_RECURSION_LIMIT - 2)
210
211 struct action_fifo {
212 int head;
213 int tail;
214 /* Deferred action fifo queue storage. */
215 struct deferred_action fifo[DEFERRED_ACTION_FIFO_SIZE];
216 };
217
218 struct action_flow_keys {
219 struct sw_flow_key key[OVS_DEFERRED_ACTION_THRESHOLD];
220 };
221
222 struct ovs_pcpu_storage {
223 struct action_fifo action_fifos;
224 struct action_flow_keys flow_keys;
225 struct ovs_frag_data frag_data;
226 int exec_level;
227 struct task_struct *owner;
228 local_lock_t bh_lock;
229 };
230
231 extern struct ovs_pcpu_storage __percpu *ovs_pcpu_storage;
232
233 /**
234 * enum ovs_pkt_hash_types - hash info to include with a packet
235 * to send to userspace.
236 * @OVS_PACKET_HASH_SW_BIT: indicates hash was computed in software stack.
237 * @OVS_PACKET_HASH_L4_BIT: indicates hash is a canonical 4-tuple hash
238 * over transport ports.
239 */
240 enum ovs_pkt_hash_types {
241 OVS_PACKET_HASH_SW_BIT = (1ULL << 32),
242 OVS_PACKET_HASH_L4_BIT = (1ULL << 33),
243 };
244
245 extern unsigned int ovs_net_id;
246 void ovs_lock(void);
247 void ovs_unlock(void);
248
249 #ifdef CONFIG_LOCKDEP
250 int lockdep_ovsl_is_held(void);
251 #else
252 #define lockdep_ovsl_is_held() 1
253 #endif
254
255 #define ASSERT_OVSL() WARN_ON(!lockdep_ovsl_is_held())
256 #define ovsl_dereference(p) \
257 rcu_dereference_protected(p, lockdep_ovsl_is_held())
258 #define rcu_dereference_ovsl(p) \
259 rcu_dereference_check(p, lockdep_ovsl_is_held())
260
ovs_dp_get_net(const struct datapath * dp)261 static inline struct net *ovs_dp_get_net(const struct datapath *dp)
262 {
263 return read_pnet(&dp->net);
264 }
265
ovs_dp_set_net(struct datapath * dp,struct net * net)266 static inline void ovs_dp_set_net(struct datapath *dp, struct net *net)
267 {
268 write_pnet(&dp->net, net);
269 }
270
271 struct vport *ovs_lookup_vport(const struct datapath *dp, u16 port_no);
272
ovs_vport_rcu(const struct datapath * dp,int port_no)273 static inline struct vport *ovs_vport_rcu(const struct datapath *dp, int port_no)
274 {
275 WARN_ON_ONCE(!rcu_read_lock_held());
276 return ovs_lookup_vport(dp, port_no);
277 }
278
ovs_vport_ovsl_rcu(const struct datapath * dp,int port_no)279 static inline struct vport *ovs_vport_ovsl_rcu(const struct datapath *dp, int port_no)
280 {
281 WARN_ON_ONCE(!rcu_read_lock_held() && !lockdep_ovsl_is_held());
282 return ovs_lookup_vport(dp, port_no);
283 }
284
ovs_vport_ovsl(const struct datapath * dp,int port_no)285 static inline struct vport *ovs_vport_ovsl(const struct datapath *dp, int port_no)
286 {
287 ASSERT_OVSL();
288 return ovs_lookup_vport(dp, port_no);
289 }
290
291 /* Must be called with rcu_read_lock. */
get_dp_rcu(struct net * net,int dp_ifindex)292 static inline struct datapath *get_dp_rcu(struct net *net, int dp_ifindex)
293 {
294 struct net_device *dev = dev_get_by_index_rcu(net, dp_ifindex);
295
296 if (dev) {
297 struct vport *vport = ovs_internal_dev_get_vport(dev);
298
299 if (vport)
300 return vport->dp;
301 }
302
303 return NULL;
304 }
305
306 /* The caller must hold either ovs_mutex or rcu_read_lock to keep the
307 * returned dp pointer valid.
308 */
get_dp(struct net * net,int dp_ifindex)309 static inline struct datapath *get_dp(struct net *net, int dp_ifindex)
310 {
311 struct datapath *dp;
312
313 WARN_ON_ONCE(!rcu_read_lock_held() && !lockdep_ovsl_is_held());
314 rcu_read_lock();
315 dp = get_dp_rcu(net, dp_ifindex);
316 rcu_read_unlock();
317
318 return dp;
319 }
320
321 extern struct notifier_block ovs_dp_device_notifier;
322 extern struct genl_family dp_vport_genl_family;
323
324 void ovs_dp_process_packet(struct sk_buff *skb, struct sw_flow_key *key);
325 void ovs_dp_detach_port(struct vport *);
326 int ovs_dp_upcall(struct datapath *, struct sk_buff *,
327 const struct sw_flow_key *, const struct dp_upcall_info *,
328 uint32_t cutlen);
329
330 u32 ovs_dp_get_upcall_portid(const struct datapath *dp, uint32_t cpu_id);
331
332 const char *ovs_dp_name(const struct datapath *dp);
333 struct sk_buff *ovs_vport_cmd_build_info(struct vport *vport, struct net *net,
334 u32 portid, u32 seq, u8 cmd);
335
336 int ovs_execute_actions(struct datapath *dp, struct sk_buff *skb,
337 const struct sw_flow_actions *, struct sw_flow_key *);
338
339 void ovs_dp_notify_wq(struct work_struct *work);
340
341 /* 'KEY' must not have any bits set outside of the 'MASK' */
342 #define OVS_MASKED(OLD, KEY, MASK) ((KEY) | ((OLD) & ~(MASK)))
343 #define OVS_SET_MASKED(OLD, KEY, MASK) ((OLD) = OVS_MASKED(OLD, KEY, MASK))
344
345 #define OVS_NLERR(logging_allowed, fmt, ...) \
346 do { \
347 if (logging_allowed && net_ratelimit()) \
348 pr_info("netlink: " fmt "\n", ##__VA_ARGS__); \
349 } while (0)
350 #endif /* datapath.h */
351