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 * @do_ttl_dec: used to indicate IP TTL / Hop Limit should be decremented 52 * @deliver: used to indicate a deliver action should take place 53 * @vlan_tci: tci fields for vlan push actions 54 * @vlan_proto: ethernet types for vlan push actions 55 * @count: counter mapping 56 * @encap_md: encap entry in tc_encap_ht table 57 * @encap_user: linked list of encap users (encap_md->users) 58 * @user: owning action-set-list. Only populated if @encap_md is; used by efx_tc_update_encap() fallback handling 59 * @count_user: linked list of counter users (counter->users) 60 * @dest_mport: destination mport 61 * @src_mac: source mac entry in tc_mac_ht table 62 * @dst_mac: destination mac entry in tc_mac_ht table 63 * @fw_id: index of this entry in firmware actions table 64 * @list: linked list of tc actions 65 * 66 */ 67 struct efx_tc_action_set { 68 u16 vlan_push:2; 69 u16 vlan_pop:2; 70 u16 decap:1; 71 u16 do_ttl_dec:1; 72 u16 deliver:1; 73 __be16 vlan_tci[2]; 74 __be16 vlan_proto[2]; 75 struct efx_tc_counter_index *count; 76 struct efx_tc_encap_action *encap_md; 77 struct list_head encap_user; 78 struct efx_tc_action_set_list *user; 79 struct list_head count_user; 80 u32 dest_mport; 81 struct efx_tc_mac_pedit_action *src_mac; 82 struct efx_tc_mac_pedit_action *dst_mac; 83 u32 fw_id; 84 struct list_head list; 85 }; 86 87 struct efx_tc_match_fields { 88 /* L1 */ 89 u32 ingress_port; 90 u8 recirc_id; /* mapped from (u32) TC chain_index to smaller space */ 91 /* L2 (inner when encap) */ 92 __be16 eth_proto; 93 __be16 vlan_tci[2], vlan_proto[2]; 94 u8 eth_saddr[ETH_ALEN], eth_daddr[ETH_ALEN]; 95 /* L3 (when IP) */ 96 u8 ip_proto, ip_tos, ip_ttl; 97 __be32 src_ip, dst_ip; 98 #ifdef CONFIG_IPV6 99 struct in6_addr src_ip6, dst_ip6; 100 #endif 101 bool ip_frag, ip_firstfrag; 102 /* L4 */ 103 __be16 l4_sport, l4_dport; /* Ports (UDP, TCP) */ 104 __be16 tcp_flags; 105 bool tcp_syn_fin_rst; /* true if ANY of SYN/FIN/RST are set */ 106 /* Encap. The following are *outer* fields. Note that there are no 107 * outer eth (L2) fields; this is because TC doesn't have them. 108 */ 109 __be32 enc_src_ip, enc_dst_ip; 110 struct in6_addr enc_src_ip6, enc_dst_ip6; 111 u8 enc_ip_tos, enc_ip_ttl; 112 __be16 enc_sport, enc_dport; 113 __be32 enc_keyid; /* e.g. VNI, VSID */ 114 /* Conntrack. */ 115 u16 ct_state_trk:1, ct_state_est:1; 116 u32 ct_mark; 117 u16 ct_zone; 118 }; 119 120 static inline bool efx_tc_match_is_encap(const struct efx_tc_match_fields *mask) 121 { 122 return mask->enc_src_ip || mask->enc_dst_ip || 123 !ipv6_addr_any(&mask->enc_src_ip6) || 124 !ipv6_addr_any(&mask->enc_dst_ip6) || mask->enc_ip_tos || 125 mask->enc_ip_ttl || mask->enc_sport || mask->enc_dport; 126 } 127 128 /** 129 * enum efx_tc_em_pseudo_type - &struct efx_tc_encap_match pseudo type 130 * 131 * These are used to classify "pseudo" encap matches, which don't refer 132 * to an entry in hardware but rather indicate that a section of the 133 * match space is in use by another Outer Rule. 134 * 135 * @EFX_TC_EM_DIRECT: real HW entry in Outer Rule table; not a pseudo. 136 * Hardware index in &struct efx_tc_encap_match.fw_id is valid. 137 * @EFX_TC_EM_PSEUDO_MASK: registered by an encap match which includes a 138 * match on an optional field (currently ip_tos and/or udp_sport), 139 * to prevent an overlapping encap match _without_ optional fields. 140 * The pseudo encap match may be referenced again by an encap match 141 * with different values for these fields, but all masks must match the 142 * first (stored in our child_* fields). 143 * @EFX_TC_EM_PSEUDO_OR: registered by an fLHS rule that fits in the OR 144 * table. The &struct efx_tc_lhs_rule already holds the HW OR entry. 145 * Only one reference to this encap match may exist. 146 */ 147 enum efx_tc_em_pseudo_type { 148 EFX_TC_EM_DIRECT, 149 EFX_TC_EM_PSEUDO_MASK, 150 EFX_TC_EM_PSEUDO_OR, 151 }; 152 153 struct efx_tc_encap_match { 154 __be32 src_ip, dst_ip; 155 struct in6_addr src_ip6, dst_ip6; 156 __be16 udp_dport; 157 __be16 udp_sport, udp_sport_mask; 158 u8 ip_tos, ip_tos_mask; 159 struct rhash_head linkage; 160 enum efx_encap_type tun_type; 161 u8 child_ip_tos_mask; 162 __be16 child_udp_sport_mask; 163 refcount_t ref; 164 enum efx_tc_em_pseudo_type type; 165 u32 fw_id; /* index of this entry in firmware encap match table */ 166 struct efx_tc_encap_match *pseudo; /* Referenced pseudo EM if needed */ 167 }; 168 169 struct efx_tc_recirc_id { 170 u32 chain_index; 171 struct net_device *net_dev; 172 struct rhash_head linkage; 173 refcount_t ref; 174 u8 fw_id; /* index allocated for use in the MAE */ 175 }; 176 177 struct efx_tc_match { 178 struct efx_tc_match_fields value; 179 struct efx_tc_match_fields mask; 180 struct efx_tc_encap_match *encap; 181 struct efx_tc_recirc_id *rid; 182 }; 183 184 struct efx_tc_action_set_list { 185 struct list_head list; 186 u32 fw_id; 187 }; 188 189 struct efx_tc_lhs_action { 190 enum efx_encap_type tun_type; 191 struct efx_tc_recirc_id *rid; 192 struct efx_tc_ct_zone *zone; 193 struct efx_tc_counter_index *count; 194 }; 195 196 struct efx_tc_flow_rule { 197 unsigned long cookie; 198 struct rhash_head linkage; 199 struct efx_tc_match match; 200 struct efx_tc_action_set_list acts; 201 struct efx_tc_action_set_list *fallback; /* what to use when unready? */ 202 u32 fw_id; 203 }; 204 205 struct efx_tc_lhs_rule { 206 unsigned long cookie; 207 struct efx_tc_match match; 208 struct efx_tc_lhs_action lhs_act; 209 struct rhash_head linkage; 210 u32 fw_id; 211 bool is_ar; /* Action Rule (for OR-AR-CT-AR sequence) */ 212 }; 213 214 enum efx_tc_rule_prios { 215 EFX_TC_PRIO_TC, /* Rule inserted by TC */ 216 EFX_TC_PRIO_DFLT, /* Default switch rule; one of efx_tc_default_rules */ 217 EFX_TC_PRIO__NUM 218 }; 219 220 struct efx_tc_table_field_fmt { 221 u16 field_id; 222 u16 lbn; 223 u16 width; 224 u8 masking; 225 u8 scheme; 226 }; 227 228 struct efx_tc_table_desc { 229 u16 type; 230 u16 key_width; 231 u16 resp_width; 232 u16 n_keys; 233 u16 n_resps; 234 u16 n_prios; 235 u8 flags; 236 u8 scheme; 237 struct efx_tc_table_field_fmt *keys; 238 struct efx_tc_table_field_fmt *resps; 239 }; 240 241 struct efx_tc_table_ct { /* TABLE_ID_CONNTRACK_TABLE */ 242 struct efx_tc_table_desc desc; 243 bool hooked; 244 struct { /* indices of named fields within @desc.keys */ 245 u8 eth_proto_idx; 246 u8 ip_proto_idx; 247 u8 src_ip_idx; /* either v4 or v6 */ 248 u8 dst_ip_idx; 249 u8 l4_sport_idx; 250 u8 l4_dport_idx; 251 u8 zone_idx; /* for TABLE_FIELD_ID_DOMAIN */ 252 } keys; 253 struct { /* indices of named fields within @desc.resps */ 254 u8 dnat_idx; 255 u8 nat_ip_idx; 256 u8 l4_natport_idx; 257 u8 mark_idx; 258 u8 counter_id_idx; 259 } resps; 260 }; 261 262 /** 263 * struct efx_tc_state - control plane data for TC offload 264 * 265 * @caps: MAE capabilities reported by MCDI 266 * @block_list: List of &struct efx_tc_block_binding 267 * @mutex: Used to serialise operations on TC hashtables 268 * @counter_ht: Hashtable of TC counters (FW IDs and counter values) 269 * @counter_id_ht: Hashtable mapping TC counter cookies to counters 270 * @encap_ht: Hashtable of TC encap actions 271 * @mac_ht: Hashtable of MAC address entries (for pedits) 272 * @encap_match_ht: Hashtable of TC encap matches 273 * @match_action_ht: Hashtable of TC match-action rules 274 * @lhs_rule_ht: Hashtable of TC left-hand (act ct & goto chain) rules 275 * @ct_zone_ht: Hashtable of TC conntrack flowtable bindings 276 * @ct_ht: Hashtable of TC conntrack flow entries 277 * @neigh_ht: Hashtable of neighbour watches (&struct efx_neigh_binder) 278 * @recirc_ht: Hashtable of recirculation ID mappings (&struct efx_tc_recirc_id) 279 * @recirc_ida: Recirculation ID allocator 280 * @meta_ct: MAE table layout for conntrack table 281 * @reps_mport_id: MAE port allocated for representor RX 282 * @reps_filter_uc: VNIC filter for representor unicast RX (promisc) 283 * @reps_filter_mc: VNIC filter for representor multicast RX (allmulti) 284 * @reps_mport_vport_id: vport_id for representor RX filters 285 * @flush_counters: counters have been stopped, waiting for drain 286 * @flush_gen: final generation count per type array as reported by 287 * MC_CMD_MAE_COUNTERS_STREAM_STOP 288 * @seen_gen: most recent generation count per type as seen by efx_tc_rx() 289 * @flush_wq: wait queue used by efx_mae_stop_counters() to wait for 290 * MAE counters RXQ to finish draining 291 * @dflt: Match-action rules for default switching; at priority 292 * %EFX_TC_PRIO_DFLT. Named by *ingress* port 293 * @dflt.pf: rule for traffic ingressing from PF (egresses to wire) 294 * @dflt.wire: rule for traffic ingressing from wire (egresses to PF) 295 * @facts: Fallback action-set-lists for unready rules. Named by *egress* port 296 * @facts.pf: action-set-list for unready rules on PF netdev, hence applying to 297 * traffic from wire, and egressing to PF 298 * @facts.reps: action-set-list for unready rules on representors, hence 299 * applying to traffic from representees, and egressing to the reps mport 300 * @up: have TC datastructures been set up? 301 */ 302 struct efx_tc_state { 303 struct mae_caps *caps; 304 struct list_head block_list; 305 struct mutex mutex; 306 struct rhashtable counter_ht; 307 struct rhashtable counter_id_ht; 308 struct rhashtable encap_ht; 309 struct rhashtable mac_ht; 310 struct rhashtable encap_match_ht; 311 struct rhashtable match_action_ht; 312 struct rhashtable lhs_rule_ht; 313 struct rhashtable ct_zone_ht; 314 struct rhashtable ct_ht; 315 struct rhashtable neigh_ht; 316 struct rhashtable recirc_ht; 317 struct ida recirc_ida; 318 struct efx_tc_table_ct meta_ct; 319 u32 reps_mport_id, reps_mport_vport_id; 320 s32 reps_filter_uc, reps_filter_mc; 321 bool flush_counters; 322 u32 flush_gen[EFX_TC_COUNTER_TYPE_MAX]; 323 u32 seen_gen[EFX_TC_COUNTER_TYPE_MAX]; 324 wait_queue_head_t flush_wq; 325 struct { 326 struct efx_tc_flow_rule pf; 327 struct efx_tc_flow_rule wire; 328 } dflt; 329 struct { 330 struct efx_tc_action_set_list pf; 331 struct efx_tc_action_set_list reps; 332 } facts; 333 bool up; 334 }; 335 336 struct efx_rep; 337 338 enum efx_encap_type efx_tc_indr_netdev_type(struct net_device *net_dev); 339 struct efx_rep *efx_tc_flower_lookup_efv(struct efx_nic *efx, 340 struct net_device *dev); 341 s64 efx_tc_flower_external_mport(struct efx_nic *efx, struct efx_rep *efv); 342 int efx_tc_configure_default_rule_rep(struct efx_rep *efv); 343 void efx_tc_deconfigure_default_rule(struct efx_nic *efx, 344 struct efx_tc_flow_rule *rule); 345 int efx_tc_flower(struct efx_nic *efx, struct net_device *net_dev, 346 struct flow_cls_offload *tc, struct efx_rep *efv); 347 348 int efx_tc_insert_rep_filters(struct efx_nic *efx); 349 void efx_tc_remove_rep_filters(struct efx_nic *efx); 350 351 int efx_init_tc(struct efx_nic *efx); 352 void efx_fini_tc(struct efx_nic *efx); 353 354 int efx_init_struct_tc(struct efx_nic *efx); 355 void efx_fini_struct_tc(struct efx_nic *efx); 356 357 #endif /* EFX_TC_H */ 358