David Brazdil | 0f672f6 | 2019-12-10 10:32:29 +0000 | [diff] [blame^] | 1 | /* SPDX-License-Identifier: GPL-2.0-only */ |
Andrew Scull | b4b6d4a | 2019-01-02 15:54:55 +0000 | [diff] [blame] | 2 | /* |
| 3 | * Copyright (c) 2007-2014 Nicira, Inc. |
Andrew Scull | b4b6d4a | 2019-01-02 15:54:55 +0000 | [diff] [blame] | 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 | |
| 26 | /** |
| 27 | * struct dp_stats_percpu - per-cpu packet processing statistics for a given |
| 28 | * datapath. |
| 29 | * @n_hit: Number of received packets for which a matching flow was found in |
| 30 | * the flow table. |
| 31 | * @n_miss: Number of received packets that had no matching flow in the flow |
| 32 | * table. The sum of @n_hit and @n_miss is the number of packets that have |
| 33 | * been received by the datapath. |
| 34 | * @n_lost: Number of received packets that had no matching flow in the flow |
| 35 | * table that could not be sent to userspace (normally due to an overflow in |
| 36 | * one of the datapath's queues). |
| 37 | * @n_mask_hit: Number of masks looked up for flow match. |
| 38 | * @n_mask_hit / (@n_hit + @n_missed) will be the average masks looked |
| 39 | * up per packet. |
| 40 | */ |
| 41 | struct dp_stats_percpu { |
| 42 | u64 n_hit; |
| 43 | u64 n_missed; |
| 44 | u64 n_lost; |
| 45 | u64 n_mask_hit; |
| 46 | struct u64_stats_sync syncp; |
| 47 | }; |
| 48 | |
| 49 | /** |
| 50 | * struct datapath - datapath for flow-based packet switching |
| 51 | * @rcu: RCU callback head for deferred destruction. |
| 52 | * @list_node: Element in global 'dps' list. |
| 53 | * @table: flow table. |
| 54 | * @ports: Hash table for ports. %OVSP_LOCAL port always exists. Protected by |
| 55 | * ovs_mutex and RCU. |
| 56 | * @stats_percpu: Per-CPU datapath statistics. |
| 57 | * @net: Reference to net namespace. |
| 58 | * @max_headroom: the maximum headroom of all vports in this datapath; it will |
| 59 | * be used by all the internal vports in this dp. |
| 60 | * |
| 61 | * Context: See the comment on locking at the top of datapath.c for additional |
| 62 | * locking information. |
| 63 | */ |
| 64 | struct datapath { |
| 65 | struct rcu_head rcu; |
| 66 | struct list_head list_node; |
| 67 | |
| 68 | /* Flow table. */ |
| 69 | struct flow_table table; |
| 70 | |
| 71 | /* Switch ports. */ |
| 72 | struct hlist_head *ports; |
| 73 | |
| 74 | /* Stats. */ |
| 75 | struct dp_stats_percpu __percpu *stats_percpu; |
| 76 | |
| 77 | /* Network namespace ref. */ |
| 78 | possible_net_t net; |
| 79 | |
| 80 | u32 user_features; |
| 81 | |
| 82 | u32 max_headroom; |
| 83 | |
| 84 | /* Switch meters. */ |
| 85 | struct hlist_head *meters; |
| 86 | }; |
| 87 | |
| 88 | /** |
| 89 | * struct ovs_skb_cb - OVS data in skb CB |
| 90 | * @input_vport: The original vport packet came in on. This value is cached |
| 91 | * when a packet is received by OVS. |
| 92 | * @mru: The maximum received fragement size; 0 if the packet is not |
| 93 | * fragmented. |
| 94 | * @acts_origlen: The netlink size of the flow actions applied to this skb. |
| 95 | * @cutlen: The number of bytes from the packet end to be removed. |
| 96 | */ |
| 97 | struct ovs_skb_cb { |
| 98 | struct vport *input_vport; |
| 99 | u16 mru; |
| 100 | u16 acts_origlen; |
| 101 | u32 cutlen; |
| 102 | }; |
| 103 | #define OVS_CB(skb) ((struct ovs_skb_cb *)(skb)->cb) |
| 104 | |
| 105 | /** |
| 106 | * struct dp_upcall - metadata to include with a packet to send to userspace |
| 107 | * @cmd: One of %OVS_PACKET_CMD_*. |
| 108 | * @userdata: If nonnull, its variable-length value is passed to userspace as |
| 109 | * %OVS_PACKET_ATTR_USERDATA. |
| 110 | * @portid: Netlink portid to which packet should be sent. If @portid is 0 |
| 111 | * then no packet is sent and the packet is accounted in the datapath's @n_lost |
| 112 | * counter. |
| 113 | * @egress_tun_info: If nonnull, becomes %OVS_PACKET_ATTR_EGRESS_TUN_KEY. |
| 114 | * @mru: If not zero, Maximum received IP fragment size. |
| 115 | */ |
| 116 | struct dp_upcall_info { |
| 117 | struct ip_tunnel_info *egress_tun_info; |
| 118 | const struct nlattr *userdata; |
| 119 | const struct nlattr *actions; |
| 120 | int actions_len; |
| 121 | u32 portid; |
| 122 | u8 cmd; |
| 123 | u16 mru; |
| 124 | }; |
| 125 | |
| 126 | /** |
| 127 | * struct ovs_net - Per net-namespace data for ovs. |
| 128 | * @dps: List of datapaths to enable dumping them all out. |
| 129 | * Protected by genl_mutex. |
| 130 | */ |
| 131 | struct ovs_net { |
| 132 | struct list_head dps; |
| 133 | struct work_struct dp_notify_work; |
| 134 | #if IS_ENABLED(CONFIG_NETFILTER_CONNCOUNT) |
| 135 | struct ovs_ct_limit_info *ct_limit_info; |
| 136 | #endif |
| 137 | |
| 138 | /* Module reference for configuring conntrack. */ |
| 139 | bool xt_label; |
| 140 | }; |
| 141 | |
| 142 | extern unsigned int ovs_net_id; |
| 143 | void ovs_lock(void); |
| 144 | void ovs_unlock(void); |
| 145 | |
| 146 | #ifdef CONFIG_LOCKDEP |
| 147 | int lockdep_ovsl_is_held(void); |
| 148 | #else |
| 149 | #define lockdep_ovsl_is_held() 1 |
| 150 | #endif |
| 151 | |
| 152 | #define ASSERT_OVSL() WARN_ON(!lockdep_ovsl_is_held()) |
| 153 | #define ovsl_dereference(p) \ |
| 154 | rcu_dereference_protected(p, lockdep_ovsl_is_held()) |
| 155 | #define rcu_dereference_ovsl(p) \ |
| 156 | rcu_dereference_check(p, lockdep_ovsl_is_held()) |
| 157 | |
| 158 | static inline struct net *ovs_dp_get_net(const struct datapath *dp) |
| 159 | { |
| 160 | return read_pnet(&dp->net); |
| 161 | } |
| 162 | |
| 163 | static inline void ovs_dp_set_net(struct datapath *dp, struct net *net) |
| 164 | { |
| 165 | write_pnet(&dp->net, net); |
| 166 | } |
| 167 | |
| 168 | struct vport *ovs_lookup_vport(const struct datapath *dp, u16 port_no); |
| 169 | |
| 170 | static inline struct vport *ovs_vport_rcu(const struct datapath *dp, int port_no) |
| 171 | { |
| 172 | WARN_ON_ONCE(!rcu_read_lock_held()); |
| 173 | return ovs_lookup_vport(dp, port_no); |
| 174 | } |
| 175 | |
| 176 | static inline struct vport *ovs_vport_ovsl_rcu(const struct datapath *dp, int port_no) |
| 177 | { |
| 178 | WARN_ON_ONCE(!rcu_read_lock_held() && !lockdep_ovsl_is_held()); |
| 179 | return ovs_lookup_vport(dp, port_no); |
| 180 | } |
| 181 | |
| 182 | static inline struct vport *ovs_vport_ovsl(const struct datapath *dp, int port_no) |
| 183 | { |
| 184 | ASSERT_OVSL(); |
| 185 | return ovs_lookup_vport(dp, port_no); |
| 186 | } |
| 187 | |
| 188 | /* Must be called with rcu_read_lock. */ |
| 189 | static inline struct datapath *get_dp_rcu(struct net *net, int dp_ifindex) |
| 190 | { |
| 191 | struct net_device *dev = dev_get_by_index_rcu(net, dp_ifindex); |
| 192 | |
| 193 | if (dev) { |
| 194 | struct vport *vport = ovs_internal_dev_get_vport(dev); |
| 195 | |
| 196 | if (vport) |
| 197 | return vport->dp; |
| 198 | } |
| 199 | |
| 200 | return NULL; |
| 201 | } |
| 202 | |
| 203 | /* The caller must hold either ovs_mutex or rcu_read_lock to keep the |
| 204 | * returned dp pointer valid. |
| 205 | */ |
| 206 | static inline struct datapath *get_dp(struct net *net, int dp_ifindex) |
| 207 | { |
| 208 | struct datapath *dp; |
| 209 | |
| 210 | WARN_ON_ONCE(!rcu_read_lock_held() && !lockdep_ovsl_is_held()); |
| 211 | rcu_read_lock(); |
| 212 | dp = get_dp_rcu(net, dp_ifindex); |
| 213 | rcu_read_unlock(); |
| 214 | |
| 215 | return dp; |
| 216 | } |
| 217 | |
| 218 | extern struct notifier_block ovs_dp_device_notifier; |
| 219 | extern struct genl_family dp_vport_genl_family; |
| 220 | |
David Brazdil | 0f672f6 | 2019-12-10 10:32:29 +0000 | [diff] [blame^] | 221 | DECLARE_STATIC_KEY_FALSE(tc_recirc_sharing_support); |
| 222 | |
Andrew Scull | b4b6d4a | 2019-01-02 15:54:55 +0000 | [diff] [blame] | 223 | void ovs_dp_process_packet(struct sk_buff *skb, struct sw_flow_key *key); |
| 224 | void ovs_dp_detach_port(struct vport *); |
| 225 | int ovs_dp_upcall(struct datapath *, struct sk_buff *, |
| 226 | const struct sw_flow_key *, const struct dp_upcall_info *, |
| 227 | uint32_t cutlen); |
| 228 | |
| 229 | const char *ovs_dp_name(const struct datapath *dp); |
| 230 | struct sk_buff *ovs_vport_cmd_build_info(struct vport *vport, struct net *net, |
| 231 | u32 portid, u32 seq, u8 cmd); |
| 232 | |
| 233 | int ovs_execute_actions(struct datapath *dp, struct sk_buff *skb, |
| 234 | const struct sw_flow_actions *, struct sw_flow_key *); |
| 235 | |
| 236 | void ovs_dp_notify_wq(struct work_struct *work); |
| 237 | |
| 238 | int action_fifos_init(void); |
| 239 | void action_fifos_exit(void); |
| 240 | |
| 241 | /* 'KEY' must not have any bits set outside of the 'MASK' */ |
| 242 | #define OVS_MASKED(OLD, KEY, MASK) ((KEY) | ((OLD) & ~(MASK))) |
| 243 | #define OVS_SET_MASKED(OLD, KEY, MASK) ((OLD) = OVS_MASKED(OLD, KEY, MASK)) |
| 244 | |
| 245 | #define OVS_NLERR(logging_allowed, fmt, ...) \ |
| 246 | do { \ |
| 247 | if (logging_allowed && net_ratelimit()) \ |
| 248 | pr_info("netlink: " fmt "\n", ##__VA_ARGS__); \ |
| 249 | } while (0) |
| 250 | #endif /* datapath.h */ |