1 /* 2 * Copyright (c) 2015, Mellanox Technologies. All rights reserved. 3 * 4 * This software is available to you under a choice of one of two 5 * licenses. You may choose to be licensed under the terms of the GNU 6 * General Public License (GPL) Version 2, available from the file 7 * COPYING in the main directory of this source tree, or the 8 * OpenIB.org BSD license below: 9 * 10 * Redistribution and use in source and binary forms, with or 11 * without modification, are permitted provided that the following 12 * conditions are met: 13 * 14 * - Redistributions of source code must retain the above 15 * copyright notice, this list of conditions and the following 16 * disclaimer. 17 * 18 * - Redistributions in binary form must reproduce the above 19 * copyright notice, this list of conditions and the following 20 * disclaimer in the documentation and/or other materials 21 * provided with the distribution. 22 * 23 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, 24 * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF 25 * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND 26 * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS 27 * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN 28 * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN 29 * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE 30 * SOFTWARE. 31 */ 32 33 #ifndef _MLX5_FS_CORE_ 34 #define _MLX5_FS_CORE_ 35 36 #include <linux/refcount.h> 37 #include <linux/mlx5/fs.h> 38 #include <linux/rhashtable.h> 39 #include <linux/llist.h> 40 #include <steering/fs_dr.h> 41 42 struct mlx5_modify_hdr { 43 enum mlx5_flow_namespace_type ns_type; 44 union { 45 struct mlx5_fs_dr_action action; 46 u32 id; 47 }; 48 }; 49 50 struct mlx5_pkt_reformat { 51 enum mlx5_flow_namespace_type ns_type; 52 int reformat_type; /* from mlx5_ifc */ 53 union { 54 struct mlx5_fs_dr_action action; 55 u32 id; 56 }; 57 }; 58 59 /* FS_TYPE_PRIO_CHAINS is a PRIO that will have namespaces only, 60 * and those are in parallel to one another when going over them to connect 61 * a new flow table. Meaning the last flow table in a TYPE_PRIO prio in one 62 * parallel namespace will not automatically connect to the first flow table 63 * found in any prio in any next namespace, but skip the entire containing 64 * TYPE_PRIO_CHAINS prio. 65 * 66 * This is used to implement tc chains, each chain of prios is a different 67 * namespace inside a containing TYPE_PRIO_CHAINS prio. 68 */ 69 70 enum fs_node_type { 71 FS_TYPE_NAMESPACE, 72 FS_TYPE_PRIO, 73 FS_TYPE_PRIO_CHAINS, 74 FS_TYPE_FLOW_TABLE, 75 FS_TYPE_FLOW_GROUP, 76 FS_TYPE_FLOW_ENTRY, 77 FS_TYPE_FLOW_DEST 78 }; 79 80 enum fs_flow_table_type { 81 FS_FT_NIC_RX = 0x0, 82 FS_FT_NIC_TX = 0x1, 83 FS_FT_ESW_EGRESS_ACL = 0x2, 84 FS_FT_ESW_INGRESS_ACL = 0x3, 85 FS_FT_FDB = 0X4, 86 FS_FT_SNIFFER_RX = 0X5, 87 FS_FT_SNIFFER_TX = 0X6, 88 FS_FT_RDMA_RX = 0X7, 89 FS_FT_MAX_TYPE = FS_FT_RDMA_RX, 90 }; 91 92 enum fs_flow_table_op_mod { 93 FS_FT_OP_MOD_NORMAL, 94 FS_FT_OP_MOD_LAG_DEMUX, 95 }; 96 97 enum fs_fte_status { 98 FS_FTE_STATUS_EXISTING = 1UL << 0, 99 }; 100 101 enum mlx5_flow_steering_mode { 102 MLX5_FLOW_STEERING_MODE_DMFS, 103 MLX5_FLOW_STEERING_MODE_SMFS 104 }; 105 106 struct mlx5_flow_steering { 107 struct mlx5_core_dev *dev; 108 enum mlx5_flow_steering_mode mode; 109 struct kmem_cache *fgs_cache; 110 struct kmem_cache *ftes_cache; 111 struct mlx5_flow_root_namespace *root_ns; 112 struct mlx5_flow_root_namespace *fdb_root_ns; 113 struct mlx5_flow_namespace **fdb_sub_ns; 114 struct mlx5_flow_root_namespace **esw_egress_root_ns; 115 struct mlx5_flow_root_namespace **esw_ingress_root_ns; 116 struct mlx5_flow_root_namespace *sniffer_tx_root_ns; 117 struct mlx5_flow_root_namespace *sniffer_rx_root_ns; 118 struct mlx5_flow_root_namespace *rdma_rx_root_ns; 119 struct mlx5_flow_root_namespace *egress_root_ns; 120 }; 121 122 struct fs_node { 123 struct list_head list; 124 struct list_head children; 125 enum fs_node_type type; 126 struct fs_node *parent; 127 struct fs_node *root; 128 /* lock the node for writing and traversing */ 129 struct rw_semaphore lock; 130 refcount_t refcount; 131 bool active; 132 void (*del_hw_func)(struct fs_node *); 133 void (*del_sw_func)(struct fs_node *); 134 atomic_t version; 135 }; 136 137 struct mlx5_flow_rule { 138 struct fs_node node; 139 struct mlx5_flow_destination dest_attr; 140 /* next_ft should be accessed under chain_lock and only of 141 * destination type is FWD_NEXT_fT. 142 */ 143 struct list_head next_ft; 144 u32 sw_action; 145 }; 146 147 struct mlx5_flow_handle { 148 int num_rules; 149 struct mlx5_flow_rule *rule[]; 150 }; 151 152 /* Type of children is mlx5_flow_group */ 153 struct mlx5_flow_table { 154 struct fs_node node; 155 struct mlx5_fs_dr_table fs_dr_table; 156 u32 id; 157 u16 vport; 158 unsigned int max_fte; 159 unsigned int level; 160 enum fs_flow_table_type type; 161 enum fs_flow_table_op_mod op_mod; 162 struct { 163 bool active; 164 unsigned int required_groups; 165 unsigned int group_size; 166 unsigned int num_groups; 167 } autogroup; 168 /* Protect fwd_rules */ 169 struct mutex lock; 170 /* FWD rules that point on this flow table */ 171 struct list_head fwd_rules; 172 u32 flags; 173 struct rhltable fgs_hash; 174 enum mlx5_flow_table_miss_action def_miss_action; 175 }; 176 177 struct mlx5_ft_underlay_qp { 178 struct list_head list; 179 u32 qpn; 180 }; 181 182 #define MLX5_FTE_MATCH_PARAM_RESERVED reserved_at_a00 183 /* Calculate the fte_match_param length and without the reserved length. 184 * Make sure the reserved field is the last. 185 */ 186 #define MLX5_ST_SZ_DW_MATCH_PARAM \ 187 ((MLX5_BYTE_OFF(fte_match_param, MLX5_FTE_MATCH_PARAM_RESERVED) / sizeof(u32)) + \ 188 BUILD_BUG_ON_ZERO(MLX5_ST_SZ_BYTES(fte_match_param) != \ 189 MLX5_FLD_SZ_BYTES(fte_match_param, \ 190 MLX5_FTE_MATCH_PARAM_RESERVED) +\ 191 MLX5_BYTE_OFF(fte_match_param, \ 192 MLX5_FTE_MATCH_PARAM_RESERVED))) 193 194 /* Type of children is mlx5_flow_rule */ 195 struct fs_fte { 196 struct fs_node node; 197 struct mlx5_fs_dr_rule fs_dr_rule; 198 u32 val[MLX5_ST_SZ_DW_MATCH_PARAM]; 199 u32 dests_size; 200 u32 index; 201 struct mlx5_flow_context flow_context; 202 struct mlx5_flow_act action; 203 enum fs_fte_status status; 204 struct mlx5_fc *counter; 205 struct rhash_head hash; 206 int modify_mask; 207 }; 208 209 /* Type of children is mlx5_flow_table/namespace */ 210 struct fs_prio { 211 struct fs_node node; 212 unsigned int num_levels; 213 unsigned int start_level; 214 unsigned int prio; 215 unsigned int num_ft; 216 }; 217 218 /* Type of children is fs_prio */ 219 struct mlx5_flow_namespace { 220 /* parent == NULL => root ns */ 221 struct fs_node node; 222 enum mlx5_flow_table_miss_action def_miss_action; 223 }; 224 225 struct mlx5_flow_group_mask { 226 u8 match_criteria_enable; 227 u32 match_criteria[MLX5_ST_SZ_DW_MATCH_PARAM]; 228 }; 229 230 /* Type of children is fs_fte */ 231 struct mlx5_flow_group { 232 struct fs_node node; 233 struct mlx5_fs_dr_matcher fs_dr_matcher; 234 struct mlx5_flow_group_mask mask; 235 u32 start_index; 236 u32 max_ftes; 237 struct ida fte_allocator; 238 u32 id; 239 struct rhashtable ftes_hash; 240 struct rhlist_head hash; 241 }; 242 243 struct mlx5_flow_root_namespace { 244 struct mlx5_flow_namespace ns; 245 enum mlx5_flow_steering_mode mode; 246 struct mlx5_fs_dr_domain fs_dr_domain; 247 enum fs_flow_table_type table_type; 248 struct mlx5_core_dev *dev; 249 struct mlx5_flow_table *root_ft; 250 /* Should be held when chaining flow tables */ 251 struct mutex chain_lock; 252 struct list_head underlay_qpns; 253 const struct mlx5_flow_cmds *cmds; 254 }; 255 256 int mlx5_init_fc_stats(struct mlx5_core_dev *dev); 257 void mlx5_cleanup_fc_stats(struct mlx5_core_dev *dev); 258 void mlx5_fc_queue_stats_work(struct mlx5_core_dev *dev, 259 struct delayed_work *dwork, 260 unsigned long delay); 261 void mlx5_fc_update_sampling_interval(struct mlx5_core_dev *dev, 262 unsigned long interval); 263 264 const struct mlx5_flow_cmds *mlx5_fs_cmd_get_fw_cmds(void); 265 266 int mlx5_flow_namespace_set_peer(struct mlx5_flow_root_namespace *ns, 267 struct mlx5_flow_root_namespace *peer_ns); 268 269 int mlx5_flow_namespace_set_mode(struct mlx5_flow_namespace *ns, 270 enum mlx5_flow_steering_mode mode); 271 272 int mlx5_init_fs(struct mlx5_core_dev *dev); 273 void mlx5_cleanup_fs(struct mlx5_core_dev *dev); 274 275 #define fs_get_obj(v, _node) {v = container_of((_node), typeof(*v), node); } 276 277 #define fs_list_for_each_entry(pos, root) \ 278 list_for_each_entry(pos, root, node.list) 279 280 #define fs_list_for_each_entry_safe(pos, tmp, root) \ 281 list_for_each_entry_safe(pos, tmp, root, node.list) 282 283 #define fs_for_each_ns_or_ft_reverse(pos, prio) \ 284 list_for_each_entry_reverse(pos, &(prio)->node.children, list) 285 286 #define fs_for_each_ns_or_ft(pos, prio) \ 287 list_for_each_entry(pos, (&(prio)->node.children), list) 288 289 #define fs_for_each_prio(pos, ns) \ 290 fs_list_for_each_entry(pos, &(ns)->node.children) 291 292 #define fs_for_each_ns(pos, prio) \ 293 fs_list_for_each_entry(pos, &(prio)->node.children) 294 295 #define fs_for_each_ft(pos, prio) \ 296 fs_list_for_each_entry(pos, &(prio)->node.children) 297 298 #define fs_for_each_ft_safe(pos, tmp, prio) \ 299 fs_list_for_each_entry_safe(pos, tmp, &(prio)->node.children) 300 301 #define fs_for_each_fg(pos, ft) \ 302 fs_list_for_each_entry(pos, &(ft)->node.children) 303 304 #define fs_for_each_fte(pos, fg) \ 305 fs_list_for_each_entry(pos, &(fg)->node.children) 306 307 #define fs_for_each_dst(pos, fte) \ 308 fs_list_for_each_entry(pos, &(fte)->node.children) 309 310 #define MLX5_CAP_FLOWTABLE_TYPE(mdev, cap, type) ( \ 311 (type == FS_FT_NIC_RX) ? MLX5_CAP_FLOWTABLE_NIC_RX(mdev, cap) : \ 312 (type == FS_FT_ESW_EGRESS_ACL) ? MLX5_CAP_ESW_EGRESS_ACL(mdev, cap) : \ 313 (type == FS_FT_ESW_INGRESS_ACL) ? MLX5_CAP_ESW_INGRESS_ACL(mdev, cap) : \ 314 (type == FS_FT_FDB) ? MLX5_CAP_ESW_FLOWTABLE_FDB(mdev, cap) : \ 315 (type == FS_FT_SNIFFER_RX) ? MLX5_CAP_FLOWTABLE_SNIFFER_RX(mdev, cap) : \ 316 (type == FS_FT_SNIFFER_TX) ? MLX5_CAP_FLOWTABLE_SNIFFER_TX(mdev, cap) : \ 317 (type == FS_FT_RDMA_RX) ? MLX5_CAP_FLOWTABLE_RDMA_RX(mdev, cap) : \ 318 (BUILD_BUG_ON_ZERO(FS_FT_RDMA_RX != FS_FT_MAX_TYPE))\ 319 ) 320 321 #endif 322