1 /* SPDX-License-Identifier: GPL-2.0-only */ 2 /**************************************************************************** 3 * Driver for Solarflare network controllers and boards 4 * Copyright 2019 Solarflare Communications Inc. 5 * Copyright 2020-2022 Xilinx Inc. 6 * 7 * This program is free software; you can redistribute it and/or modify it 8 * under the terms of the GNU General Public License version 2 as published 9 * by the Free Software Foundation, incorporated herein by reference. 10 */ 11 12 #ifndef EFX_TC_H 13 #define EFX_TC_H 14 #include <net/flow_offload.h> 15 #include <linux/rhashtable.h> 16 #include "net_driver.h" 17 #include "tc_counters.h" 18 19 #define IS_ALL_ONES(v) (!(typeof (v))~(v)) 20 21 /** 22 * struct efx_tc_mac_pedit_action - mac pedit action fields 23 * 24 * @h_addr: mac address field of ethernet header 25 * @linkage: rhashtable reference 26 * @ref: reference count 27 * @fw_id: index of this entry in firmware MAC address table 28 * 29 * MAC address edits are indirected through a table in the hardware 30 */ 31 struct efx_tc_mac_pedit_action { 32 u8 h_addr[ETH_ALEN]; 33 struct rhash_head linkage; 34 refcount_t ref; 35 u32 fw_id; /* index of this entry in firmware MAC address table */ 36 }; 37 38 static inline bool efx_ipv6_addr_all_ones(struct in6_addr *addr) 39 { 40 return !memchr_inv(addr, 0xff, sizeof(*addr)); 41 } 42 43 struct efx_tc_encap_action; /* see tc_encap_actions.h */ 44 45 /** 46 * struct efx_tc_action_set - collection of tc action fields 47 * 48 * @vlan_push: the number of vlan headers to push 49 * @vlan_pop: the number of vlan headers to pop 50 * @decap: used to indicate a tunnel header decapsulation should take place 51 * @deliver: used to indicate a deliver action should take place 52 * @vlan_tci: tci fields for vlan push actions 53 * @vlan_proto: ethernet types for vlan push actions 54 * @count: counter mapping 55 * @encap_md: encap entry in tc_encap_ht table 56 * @encap_user: linked list of encap users (encap_md->users) 57 * @user: owning action-set-list. Only populated if @encap_md is; used by efx_tc_update_encap() fallback handling 58 * @count_user: linked list of counter users (counter->users) 59 * @dest_mport: destination mport 60 * @src_mac: source mac entry in tc_mac_ht table 61 * @dst_mac: destination mac entry in tc_mac_ht table 62 * @fw_id: index of this entry in firmware actions table 63 * @list: linked list of tc actions 64 * 65 */ 66 struct efx_tc_action_set { 67 u16 vlan_push:2; 68 u16 vlan_pop:2; 69 u16 decap:1; 70 u16 deliver:1; 71 __be16 vlan_tci[2]; 72 __be16 vlan_proto[2]; 73 struct efx_tc_counter_index *count; 74 struct efx_tc_encap_action *encap_md; 75 struct list_head encap_user; 76 struct efx_tc_action_set_list *user; 77 struct list_head count_user; 78 u32 dest_mport; 79 struct efx_tc_mac_pedit_action *src_mac; 80 struct efx_tc_mac_pedit_action *dst_mac; 81 u32 fw_id; 82 struct list_head list; 83 }; 84 85 struct efx_tc_match_fields { 86 /* L1 */ 87 u32 ingress_port; 88 u8 recirc_id; /* mapped from (u32) TC chain_index to smaller space */ 89 /* L2 (inner when encap) */ 90 __be16 eth_proto; 91 __be16 vlan_tci[2], vlan_proto[2]; 92 u8 eth_saddr[ETH_ALEN], eth_daddr[ETH_ALEN]; 93 /* L3 (when IP) */ 94 u8 ip_proto, ip_tos, ip_ttl; 95 __be32 src_ip, dst_ip; 96 #ifdef CONFIG_IPV6 97 struct in6_addr src_ip6, dst_ip6; 98 #endif 99 bool ip_frag, ip_firstfrag; 100 /* L4 */ 101 __be16 l4_sport, l4_dport; /* Ports (UDP, TCP) */ 102 __be16 tcp_flags; 103 bool tcp_syn_fin_rst; /* true if ANY of SYN/FIN/RST are set */ 104 /* Encap. The following are *outer* fields. Note that there are no 105 * outer eth (L2) fields; this is because TC doesn't have them. 106 */ 107 __be32 enc_src_ip, enc_dst_ip; 108 struct in6_addr enc_src_ip6, enc_dst_ip6; 109 u8 enc_ip_tos, enc_ip_ttl; 110 __be16 enc_sport, enc_dport; 111 __be32 enc_keyid; /* e.g. VNI, VSID */ 112 /* Conntrack. */ 113 u16 ct_state_trk:1, ct_state_est:1; 114 u32 ct_mark; 115 u16 ct_zone; 116 }; 117 118 static inline bool efx_tc_match_is_encap(const struct efx_tc_match_fields *mask) 119 { 120 return mask->enc_src_ip || mask->enc_dst_ip || 121 !ipv6_addr_any(&mask->enc_src_ip6) || 122 !ipv6_addr_any(&mask->enc_dst_ip6) || mask->enc_ip_tos || 123 mask->enc_ip_ttl || mask->enc_sport || mask->enc_dport; 124 } 125 126 /** 127 * enum efx_tc_em_pseudo_type - &struct efx_tc_encap_match pseudo type 128 * 129 * These are used to classify "pseudo" encap matches, which don't refer 130 * to an entry in hardware but rather indicate that a section of the 131 * match space is in use by another Outer Rule. 132 * 133 * @EFX_TC_EM_DIRECT: real HW entry in Outer Rule table; not a pseudo. 134 * Hardware index in &struct efx_tc_encap_match.fw_id is valid. 135 * @EFX_TC_EM_PSEUDO_MASK: registered by an encap match which includes a 136 * match on an optional field (currently ip_tos and/or udp_sport), 137 * to prevent an overlapping encap match _without_ optional fields. 138 * The pseudo encap match may be referenced again by an encap match 139 * with different values for these fields, but all masks must match the 140 * first (stored in our child_* fields). 141 */ 142 enum efx_tc_em_pseudo_type { 143 EFX_TC_EM_DIRECT, 144 EFX_TC_EM_PSEUDO_MASK, 145 }; 146 147 struct efx_tc_encap_match { 148 __be32 src_ip, dst_ip; 149 struct in6_addr src_ip6, dst_ip6; 150 __be16 udp_dport; 151 __be16 udp_sport, udp_sport_mask; 152 u8 ip_tos, ip_tos_mask; 153 struct rhash_head linkage; 154 enum efx_encap_type tun_type; 155 u8 child_ip_tos_mask; 156 __be16 child_udp_sport_mask; 157 refcount_t ref; 158 enum efx_tc_em_pseudo_type type; 159 u32 fw_id; /* index of this entry in firmware encap match table */ 160 struct efx_tc_encap_match *pseudo; /* Referenced pseudo EM if needed */ 161 }; 162 163 struct efx_tc_recirc_id { 164 u32 chain_index; 165 struct net_device *net_dev; 166 struct rhash_head linkage; 167 refcount_t ref; 168 u8 fw_id; /* index allocated for use in the MAE */ 169 }; 170 171 struct efx_tc_match { 172 struct efx_tc_match_fields value; 173 struct efx_tc_match_fields mask; 174 struct efx_tc_encap_match *encap; 175 struct efx_tc_recirc_id *rid; 176 }; 177 178 struct efx_tc_action_set_list { 179 struct list_head list; 180 u32 fw_id; 181 }; 182 183 struct efx_tc_lhs_action { 184 struct efx_tc_recirc_id *rid; 185 struct efx_tc_ct_zone *zone; 186 struct efx_tc_counter_index *count; 187 }; 188 189 struct efx_tc_flow_rule { 190 unsigned long cookie; 191 struct rhash_head linkage; 192 struct efx_tc_match match; 193 struct efx_tc_action_set_list acts; 194 struct efx_tc_action_set_list *fallback; /* what to use when unready? */ 195 u32 fw_id; 196 }; 197 198 struct efx_tc_lhs_rule { 199 unsigned long cookie; 200 struct efx_tc_match match; 201 struct efx_tc_lhs_action lhs_act; 202 struct rhash_head linkage; 203 u32 fw_id; 204 }; 205 206 enum efx_tc_rule_prios { 207 EFX_TC_PRIO_TC, /* Rule inserted by TC */ 208 EFX_TC_PRIO_DFLT, /* Default switch rule; one of efx_tc_default_rules */ 209 EFX_TC_PRIO__NUM 210 }; 211 212 struct efx_tc_table_field_fmt { 213 u16 field_id; 214 u16 lbn; 215 u16 width; 216 u8 masking; 217 u8 scheme; 218 }; 219 220 struct efx_tc_table_desc { 221 u16 type; 222 u16 key_width; 223 u16 resp_width; 224 u16 n_keys; 225 u16 n_resps; 226 u16 n_prios; 227 u8 flags; 228 u8 scheme; 229 struct efx_tc_table_field_fmt *keys; 230 struct efx_tc_table_field_fmt *resps; 231 }; 232 233 struct efx_tc_table_ct { /* TABLE_ID_CONNTRACK_TABLE */ 234 struct efx_tc_table_desc desc; 235 bool hooked; 236 struct { /* indices of named fields within @desc.keys */ 237 u8 eth_proto_idx; 238 u8 ip_proto_idx; 239 u8 src_ip_idx; /* either v4 or v6 */ 240 u8 dst_ip_idx; 241 u8 l4_sport_idx; 242 u8 l4_dport_idx; 243 u8 zone_idx; /* for TABLE_FIELD_ID_DOMAIN */ 244 } keys; 245 struct { /* indices of named fields within @desc.resps */ 246 u8 dnat_idx; 247 u8 nat_ip_idx; 248 u8 l4_natport_idx; 249 u8 mark_idx; 250 u8 counter_id_idx; 251 } resps; 252 }; 253 254 /** 255 * struct efx_tc_state - control plane data for TC offload 256 * 257 * @caps: MAE capabilities reported by MCDI 258 * @block_list: List of &struct efx_tc_block_binding 259 * @mutex: Used to serialise operations on TC hashtables 260 * @counter_ht: Hashtable of TC counters (FW IDs and counter values) 261 * @counter_id_ht: Hashtable mapping TC counter cookies to counters 262 * @encap_ht: Hashtable of TC encap actions 263 * @mac_ht: Hashtable of MAC address entries (for pedits) 264 * @encap_match_ht: Hashtable of TC encap matches 265 * @match_action_ht: Hashtable of TC match-action rules 266 * @lhs_rule_ht: Hashtable of TC left-hand (act ct & goto chain) rules 267 * @ct_zone_ht: Hashtable of TC conntrack flowtable bindings 268 * @ct_ht: Hashtable of TC conntrack flow entries 269 * @neigh_ht: Hashtable of neighbour watches (&struct efx_neigh_binder) 270 * @recirc_ht: Hashtable of recirculation ID mappings (&struct efx_tc_recirc_id) 271 * @recirc_ida: Recirculation ID allocator 272 * @meta_ct: MAE table layout for conntrack table 273 * @reps_mport_id: MAE port allocated for representor RX 274 * @reps_filter_uc: VNIC filter for representor unicast RX (promisc) 275 * @reps_filter_mc: VNIC filter for representor multicast RX (allmulti) 276 * @reps_mport_vport_id: vport_id for representor RX filters 277 * @flush_counters: counters have been stopped, waiting for drain 278 * @flush_gen: final generation count per type array as reported by 279 * MC_CMD_MAE_COUNTERS_STREAM_STOP 280 * @seen_gen: most recent generation count per type as seen by efx_tc_rx() 281 * @flush_wq: wait queue used by efx_mae_stop_counters() to wait for 282 * MAE counters RXQ to finish draining 283 * @dflt: Match-action rules for default switching; at priority 284 * %EFX_TC_PRIO_DFLT. Named by *ingress* port 285 * @dflt.pf: rule for traffic ingressing from PF (egresses to wire) 286 * @dflt.wire: rule for traffic ingressing from wire (egresses to PF) 287 * @facts: Fallback action-set-lists for unready rules. Named by *egress* port 288 * @facts.pf: action-set-list for unready rules on PF netdev, hence applying to 289 * traffic from wire, and egressing to PF 290 * @facts.reps: action-set-list for unready rules on representors, hence 291 * applying to traffic from representees, and egressing to the reps mport 292 * @up: have TC datastructures been set up? 293 */ 294 struct efx_tc_state { 295 struct mae_caps *caps; 296 struct list_head block_list; 297 struct mutex mutex; 298 struct rhashtable counter_ht; 299 struct rhashtable counter_id_ht; 300 struct rhashtable encap_ht; 301 struct rhashtable mac_ht; 302 struct rhashtable encap_match_ht; 303 struct rhashtable match_action_ht; 304 struct rhashtable lhs_rule_ht; 305 struct rhashtable ct_zone_ht; 306 struct rhashtable ct_ht; 307 struct rhashtable neigh_ht; 308 struct rhashtable recirc_ht; 309 struct ida recirc_ida; 310 struct efx_tc_table_ct meta_ct; 311 u32 reps_mport_id, reps_mport_vport_id; 312 s32 reps_filter_uc, reps_filter_mc; 313 bool flush_counters; 314 u32 flush_gen[EFX_TC_COUNTER_TYPE_MAX]; 315 u32 seen_gen[EFX_TC_COUNTER_TYPE_MAX]; 316 wait_queue_head_t flush_wq; 317 struct { 318 struct efx_tc_flow_rule pf; 319 struct efx_tc_flow_rule wire; 320 } dflt; 321 struct { 322 struct efx_tc_action_set_list pf; 323 struct efx_tc_action_set_list reps; 324 } facts; 325 bool up; 326 }; 327 328 struct efx_rep; 329 330 enum efx_encap_type efx_tc_indr_netdev_type(struct net_device *net_dev); 331 struct efx_rep *efx_tc_flower_lookup_efv(struct efx_nic *efx, 332 struct net_device *dev); 333 s64 efx_tc_flower_external_mport(struct efx_nic *efx, struct efx_rep *efv); 334 int efx_tc_configure_default_rule_rep(struct efx_rep *efv); 335 void efx_tc_deconfigure_default_rule(struct efx_nic *efx, 336 struct efx_tc_flow_rule *rule); 337 int efx_tc_flower(struct efx_nic *efx, struct net_device *net_dev, 338 struct flow_cls_offload *tc, struct efx_rep *efv); 339 340 int efx_tc_insert_rep_filters(struct efx_nic *efx); 341 void efx_tc_remove_rep_filters(struct efx_nic *efx); 342 343 int efx_init_tc(struct efx_nic *efx); 344 void efx_fini_tc(struct efx_nic *efx); 345 346 int efx_init_struct_tc(struct efx_nic *efx); 347 void efx_fini_struct_tc(struct efx_nic *efx); 348 349 #endif /* EFX_TC_H */ 350