drivers/net/ethernet/mellanox/mlx5/core/eswitch.c

/* [<][>][^][v][top][bottom][index][help] */
This source file includes following definitions.
mlx5_eswitch_get_vport
arm_vport_context_events_cmd
modify_esw_vport_context_cmd
mlx5_eswitch_modify_esw_vport_context
query_esw_vport_context_cmd
mlx5_eswitch_query_esw_vport_context
modify_esw_vport_cvlan
__esw_fdb_set_vport_rule
esw_fdb_set_vport_rule
esw_fdb_set_vport_allmulti_rule
esw_fdb_set_vport_promisc_rule
esw_create_legacy_vepa_table
esw_create_legacy_fdb_table
esw_destroy_legacy_vepa_table
esw_destroy_legacy_fdb_table
esw_create_legacy_table
esw_legacy_enable
esw_destroy_legacy_table
esw_legacy_disable
esw_add_uc_addr
esw_del_uc_addr
update_allmulti_vports
esw_add_mc_addr
esw_del_mc_addr
esw_apply_vport_addr_list
esw_update_vport_addr_list
esw_update_vport_mc_promisc
esw_apply_vport_rx_mode
esw_update_vport_rx_mode
esw_vport_change_handle_locked
esw_vport_change_handler
esw_vport_enable_egress_acl
esw_vport_cleanup_egress_rules
esw_vport_disable_egress_acl
esw_vport_enable_ingress_acl
esw_vport_cleanup_ingress_rules
esw_vport_disable_ingress_acl
esw_vport_ingress_config
esw_vport_egress_config
element_type_supported
esw_create_tsar
esw_destroy_tsar
esw_vport_enable_qos
esw_vport_disable_qos
esw_vport_qos_config
mlx5_esw_modify_vport_rate
node_guid_gen_from_mac
esw_apply_vport_conf
esw_vport_create_drop_counters
esw_vport_destroy_drop_counters
esw_enable_vport
esw_disable_vport
eswitch_vport_event
mlx5_esw_query_functions
mlx5_eswitch_event_handlers_register
mlx5_eswitch_event_handlers_unregister
mlx5_eswitch_enable_pf_vf_vports
mlx5_eswitch_disable_pf_vf_vports
mlx5_eswitch_enable
mlx5_eswitch_disable
mlx5_eswitch_init
mlx5_eswitch_cleanup
mlx5_eswitch_set_vport_mac
mlx5_eswitch_set_vport_state
mlx5_eswitch_get_vport_config
__mlx5_eswitch_set_vport_vlan
mlx5_eswitch_set_vport_vlan
mlx5_eswitch_set_vport_spoofchk
esw_cleanup_vepa_rules
_mlx5_eswitch_set_vepa_locked
mlx5_eswitch_set_vepa
mlx5_eswitch_get_vepa
mlx5_eswitch_set_vport_trust
calculate_vports_min_rate_divider
normalize_vports_min_rate
mlx5_eswitch_set_vport_rate
mlx5_eswitch_query_vport_drop_stats
mlx5_eswitch_get_vport_stats
mlx5_eswitch_mode
mlx5_eswitch_get_encap_mode
mlx5_esw_lag_prereq
mlx5_esw_multipath_prereq
mlx5_eswitch_update_num_of_vfs
   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 #include <linux/etherdevice.h>
  34 #include <linux/mlx5/driver.h>
  35 #include <linux/mlx5/mlx5_ifc.h>
  36 #include <linux/mlx5/vport.h>
  37 #include <linux/mlx5/fs.h>
  38 #include "mlx5_core.h"
  39 #include "lib/eq.h"
  40 #include "eswitch.h"
  41 #include "fs_core.h"
  42 #include "ecpf.h"
  43 
  44 enum {
  45         MLX5_ACTION_NONE = 0,
  46         MLX5_ACTION_ADD  = 1,
  47         MLX5_ACTION_DEL  = 2,
  48 };
  49 
  50 /* Vport UC/MC hash node */
  51 struct vport_addr {
  52         struct l2addr_node     node;
  53         u8                     action;
  54         u16                    vport;
  55         struct mlx5_flow_handle *flow_rule;
  56         bool mpfs; /* UC MAC was added to MPFs */
  57         /* A flag indicating that mac was added due to mc promiscuous vport */
  58         bool mc_promisc;
  59 };
  60 
  61 static void esw_destroy_legacy_fdb_table(struct mlx5_eswitch *esw);
  62 static void esw_cleanup_vepa_rules(struct mlx5_eswitch *esw);
  63 
  64 struct mlx5_vport *__must_check
  65 mlx5_eswitch_get_vport(struct mlx5_eswitch *esw, u16 vport_num)
  66 {
  67         u16 idx;
  68 
  69         if (!esw || !MLX5_CAP_GEN(esw->dev, vport_group_manager))
  70                 return ERR_PTR(-EPERM);
  71 
  72         idx = mlx5_eswitch_vport_num_to_index(esw, vport_num);
  73 
  74         if (idx > esw->total_vports - 1) {
  75                 esw_debug(esw->dev, "vport out of range: num(0x%x), idx(0x%x)\n",
  76                           vport_num, idx);
  77                 return ERR_PTR(-EINVAL);
  78         }
  79 
  80         return &esw->vports[idx];
  81 }
  82 
  83 static int arm_vport_context_events_cmd(struct mlx5_core_dev *dev, u16 vport,
  84                                         u32 events_mask)
  85 {
  86         int in[MLX5_ST_SZ_DW(modify_nic_vport_context_in)]   = {0};
  87         int out[MLX5_ST_SZ_DW(modify_nic_vport_context_out)] = {0};
  88         void *nic_vport_ctx;
  89 
  90         MLX5_SET(modify_nic_vport_context_in, in,
  91                  opcode, MLX5_CMD_OP_MODIFY_NIC_VPORT_CONTEXT);
  92         MLX5_SET(modify_nic_vport_context_in, in, field_select.change_event, 1);
  93         MLX5_SET(modify_nic_vport_context_in, in, vport_number, vport);
  94         MLX5_SET(modify_nic_vport_context_in, in, other_vport, 1);
  95         nic_vport_ctx = MLX5_ADDR_OF(modify_nic_vport_context_in,
  96                                      in, nic_vport_context);
  97 
  98         MLX5_SET(nic_vport_context, nic_vport_ctx, arm_change_event, 1);
  99 
 100         if (events_mask & MLX5_VPORT_UC_ADDR_CHANGE)
 101                 MLX5_SET(nic_vport_context, nic_vport_ctx,
 102                          event_on_uc_address_change, 1);
 103         if (events_mask & MLX5_VPORT_MC_ADDR_CHANGE)
 104                 MLX5_SET(nic_vport_context, nic_vport_ctx,
 105                          event_on_mc_address_change, 1);
 106         if (events_mask & MLX5_VPORT_PROMISC_CHANGE)
 107                 MLX5_SET(nic_vport_context, nic_vport_ctx,
 108                          event_on_promisc_change, 1);
 109 
 110         return mlx5_cmd_exec(dev, in, sizeof(in), out, sizeof(out));
 111 }
 112 
 113 /* E-Switch vport context HW commands */
 114 static int modify_esw_vport_context_cmd(struct mlx5_core_dev *dev, u16 vport,
 115                                         void *in, int inlen)
 116 {
 117         u32 out[MLX5_ST_SZ_DW(modify_esw_vport_context_out)] = {0};
 118 
 119         MLX5_SET(modify_esw_vport_context_in, in, opcode,
 120                  MLX5_CMD_OP_MODIFY_ESW_VPORT_CONTEXT);
 121         MLX5_SET(modify_esw_vport_context_in, in, vport_number, vport);
 122         MLX5_SET(modify_esw_vport_context_in, in, other_vport, 1);
 123         return mlx5_cmd_exec(dev, in, inlen, out, sizeof(out));
 124 }
 125 
 126 int mlx5_eswitch_modify_esw_vport_context(struct mlx5_eswitch *esw, u16 vport,
 127                                           void *in, int inlen)
 128 {
 129         return modify_esw_vport_context_cmd(esw->dev, vport, in, inlen);
 130 }
 131 
 132 static int query_esw_vport_context_cmd(struct mlx5_core_dev *dev, u16 vport,
 133                                        void *out, int outlen)
 134 {
 135         u32 in[MLX5_ST_SZ_DW(query_esw_vport_context_in)] = {};
 136 
 137         MLX5_SET(query_esw_vport_context_in, in, opcode,
 138                  MLX5_CMD_OP_QUERY_ESW_VPORT_CONTEXT);
 139         MLX5_SET(modify_esw_vport_context_in, in, vport_number, vport);
 140         MLX5_SET(modify_esw_vport_context_in, in, other_vport, 1);
 141         return mlx5_cmd_exec(dev, in, sizeof(in), out, outlen);
 142 }
 143 
 144 int mlx5_eswitch_query_esw_vport_context(struct mlx5_eswitch *esw, u16 vport,
 145                                          void *out, int outlen)
 146 {
 147         return query_esw_vport_context_cmd(esw->dev, vport, out, outlen);
 148 }
 149 
 150 static int modify_esw_vport_cvlan(struct mlx5_core_dev *dev, u16 vport,
 151                                   u16 vlan, u8 qos, u8 set_flags)
 152 {
 153         u32 in[MLX5_ST_SZ_DW(modify_esw_vport_context_in)] = {0};
 154 
 155         if (!MLX5_CAP_ESW(dev, vport_cvlan_strip) ||
 156             !MLX5_CAP_ESW(dev, vport_cvlan_insert_if_not_exist))
 157                 return -EOPNOTSUPP;
 158 
 159         esw_debug(dev, "Set Vport[%d] VLAN %d qos %d set=%x\n",
 160                   vport, vlan, qos, set_flags);
 161 
 162         if (set_flags & SET_VLAN_STRIP)
 163                 MLX5_SET(modify_esw_vport_context_in, in,
 164                          esw_vport_context.vport_cvlan_strip, 1);
 165 
 166         if (set_flags & SET_VLAN_INSERT) {
 167                 /* insert only if no vlan in packet */
 168                 MLX5_SET(modify_esw_vport_context_in, in,
 169                          esw_vport_context.vport_cvlan_insert, 1);
 170 
 171                 MLX5_SET(modify_esw_vport_context_in, in,
 172                          esw_vport_context.cvlan_pcp, qos);
 173                 MLX5_SET(modify_esw_vport_context_in, in,
 174                          esw_vport_context.cvlan_id, vlan);
 175         }
 176 
 177         MLX5_SET(modify_esw_vport_context_in, in,
 178                  field_select.vport_cvlan_strip, 1);
 179         MLX5_SET(modify_esw_vport_context_in, in,
 180                  field_select.vport_cvlan_insert, 1);
 181 
 182         return modify_esw_vport_context_cmd(dev, vport, in, sizeof(in));
 183 }
 184 
 185 /* E-Switch FDB */
 186 static struct mlx5_flow_handle *
 187 __esw_fdb_set_vport_rule(struct mlx5_eswitch *esw, u16 vport, bool rx_rule,
 188                          u8 mac_c[ETH_ALEN], u8 mac_v[ETH_ALEN])
 189 {
 190         int match_header = (is_zero_ether_addr(mac_c) ? 0 :
 191                             MLX5_MATCH_OUTER_HEADERS);
 192         struct mlx5_flow_handle *flow_rule = NULL;
 193         struct mlx5_flow_act flow_act = {0};
 194         struct mlx5_flow_destination dest = {};
 195         struct mlx5_flow_spec *spec;
 196         void *mv_misc = NULL;
 197         void *mc_misc = NULL;
 198         u8 *dmac_v = NULL;
 199         u8 *dmac_c = NULL;
 200 
 201         if (rx_rule)
 202                 match_header |= MLX5_MATCH_MISC_PARAMETERS;
 203 
 204         spec = kvzalloc(sizeof(*spec), GFP_KERNEL);
 205         if (!spec)
 206                 return NULL;
 207 
 208         dmac_v = MLX5_ADDR_OF(fte_match_param, spec->match_value,
 209                               outer_headers.dmac_47_16);
 210         dmac_c = MLX5_ADDR_OF(fte_match_param, spec->match_criteria,
 211                               outer_headers.dmac_47_16);
 212 
 213         if (match_header & MLX5_MATCH_OUTER_HEADERS) {
 214                 ether_addr_copy(dmac_v, mac_v);
 215                 ether_addr_copy(dmac_c, mac_c);
 216         }
 217 
 218         if (match_header & MLX5_MATCH_MISC_PARAMETERS) {
 219                 mv_misc  = MLX5_ADDR_OF(fte_match_param, spec->match_value,
 220                                         misc_parameters);
 221                 mc_misc  = MLX5_ADDR_OF(fte_match_param, spec->match_criteria,
 222                                         misc_parameters);
 223                 MLX5_SET(fte_match_set_misc, mv_misc, source_port, MLX5_VPORT_UPLINK);
 224                 MLX5_SET_TO_ONES(fte_match_set_misc, mc_misc, source_port);
 225         }
 226 
 227         dest.type = MLX5_FLOW_DESTINATION_TYPE_VPORT;
 228         dest.vport.num = vport;
 229 
 230         esw_debug(esw->dev,
 231                   "\tFDB add rule dmac_v(%pM) dmac_c(%pM) -> vport(%d)\n",
 232                   dmac_v, dmac_c, vport);
 233         spec->match_criteria_enable = match_header;
 234         flow_act.action =  MLX5_FLOW_CONTEXT_ACTION_FWD_DEST;
 235         flow_rule =
 236                 mlx5_add_flow_rules(esw->fdb_table.legacy.fdb, spec,
 237                                     &flow_act, &dest, 1);
 238         if (IS_ERR(flow_rule)) {
 239                 esw_warn(esw->dev,
 240                          "FDB: Failed to add flow rule: dmac_v(%pM) dmac_c(%pM) -> vport(%d), err(%ld)\n",
 241                          dmac_v, dmac_c, vport, PTR_ERR(flow_rule));
 242                 flow_rule = NULL;
 243         }
 244 
 245         kvfree(spec);
 246         return flow_rule;
 247 }
 248 
 249 static struct mlx5_flow_handle *
 250 esw_fdb_set_vport_rule(struct mlx5_eswitch *esw, u8 mac[ETH_ALEN], u16 vport)
 251 {
 252         u8 mac_c[ETH_ALEN];
 253 
 254         eth_broadcast_addr(mac_c);
 255         return __esw_fdb_set_vport_rule(esw, vport, false, mac_c, mac);
 256 }
 257 
 258 static struct mlx5_flow_handle *
 259 esw_fdb_set_vport_allmulti_rule(struct mlx5_eswitch *esw, u16 vport)
 260 {
 261         u8 mac_c[ETH_ALEN];
 262         u8 mac_v[ETH_ALEN];
 263 
 264         eth_zero_addr(mac_c);
 265         eth_zero_addr(mac_v);
 266         mac_c[0] = 0x01;
 267         mac_v[0] = 0x01;
 268         return __esw_fdb_set_vport_rule(esw, vport, false, mac_c, mac_v);
 269 }
 270 
 271 static struct mlx5_flow_handle *
 272 esw_fdb_set_vport_promisc_rule(struct mlx5_eswitch *esw, u16 vport)
 273 {
 274         u8 mac_c[ETH_ALEN];
 275         u8 mac_v[ETH_ALEN];
 276 
 277         eth_zero_addr(mac_c);
 278         eth_zero_addr(mac_v);
 279         return __esw_fdb_set_vport_rule(esw, vport, true, mac_c, mac_v);
 280 }
 281 
 282 enum {
 283         LEGACY_VEPA_PRIO = 0,
 284         LEGACY_FDB_PRIO,
 285 };
 286 
 287 static int esw_create_legacy_vepa_table(struct mlx5_eswitch *esw)
 288 {
 289         struct mlx5_core_dev *dev = esw->dev;
 290         struct mlx5_flow_namespace *root_ns;
 291         struct mlx5_flow_table *fdb;
 292         int err;
 293 
 294         root_ns = mlx5_get_fdb_sub_ns(dev, 0);
 295         if (!root_ns) {
 296                 esw_warn(dev, "Failed to get FDB flow namespace\n");
 297                 return -EOPNOTSUPP;
 298         }
 299 
 300         /* num FTE 2, num FG 2 */
 301         fdb = mlx5_create_auto_grouped_flow_table(root_ns, LEGACY_VEPA_PRIO,
 302                                                   2, 2, 0, 0);
 303         if (IS_ERR(fdb)) {
 304                 err = PTR_ERR(fdb);
 305                 esw_warn(dev, "Failed to create VEPA FDB err %d\n", err);
 306                 return err;
 307         }
 308         esw->fdb_table.legacy.vepa_fdb = fdb;
 309 
 310         return 0;
 311 }
 312 
 313 static int esw_create_legacy_fdb_table(struct mlx5_eswitch *esw)
 314 {
 315         int inlen = MLX5_ST_SZ_BYTES(create_flow_group_in);
 316         struct mlx5_flow_table_attr ft_attr = {};
 317         struct mlx5_core_dev *dev = esw->dev;
 318         struct mlx5_flow_namespace *root_ns;
 319         struct mlx5_flow_table *fdb;
 320         struct mlx5_flow_group *g;
 321         void *match_criteria;
 322         int table_size;
 323         u32 *flow_group_in;
 324         u8 *dmac;
 325         int err = 0;
 326 
 327         esw_debug(dev, "Create FDB log_max_size(%d)\n",
 328                   MLX5_CAP_ESW_FLOWTABLE_FDB(dev, log_max_ft_size));
 329 
 330         root_ns = mlx5_get_fdb_sub_ns(dev, 0);
 331         if (!root_ns) {
 332                 esw_warn(dev, "Failed to get FDB flow namespace\n");
 333                 return -EOPNOTSUPP;
 334         }
 335 
 336         flow_group_in = kvzalloc(inlen, GFP_KERNEL);
 337         if (!flow_group_in)
 338                 return -ENOMEM;
 339 
 340         table_size = BIT(MLX5_CAP_ESW_FLOWTABLE_FDB(dev, log_max_ft_size));
 341         ft_attr.max_fte = table_size;
 342         ft_attr.prio = LEGACY_FDB_PRIO;
 343         fdb = mlx5_create_flow_table(root_ns, &ft_attr);
 344         if (IS_ERR(fdb)) {
 345                 err = PTR_ERR(fdb);
 346                 esw_warn(dev, "Failed to create FDB Table err %d\n", err);
 347                 goto out;
 348         }
 349         esw->fdb_table.legacy.fdb = fdb;
 350 
 351         /* Addresses group : Full match unicast/multicast addresses */
 352         MLX5_SET(create_flow_group_in, flow_group_in, match_criteria_enable,
 353                  MLX5_MATCH_OUTER_HEADERS);
 354         match_criteria = MLX5_ADDR_OF(create_flow_group_in, flow_group_in, match_criteria);
 355         dmac = MLX5_ADDR_OF(fte_match_param, match_criteria, outer_headers.dmac_47_16);
 356         MLX5_SET(create_flow_group_in, flow_group_in, start_flow_index, 0);
 357         /* Preserve 2 entries for allmulti and promisc rules*/
 358         MLX5_SET(create_flow_group_in, flow_group_in, end_flow_index, table_size - 3);
 359         eth_broadcast_addr(dmac);
 360         g = mlx5_create_flow_group(fdb, flow_group_in);
 361         if (IS_ERR(g)) {
 362                 err = PTR_ERR(g);
 363                 esw_warn(dev, "Failed to create flow group err(%d)\n", err);
 364                 goto out;
 365         }
 366         esw->fdb_table.legacy.addr_grp = g;
 367 
 368         /* Allmulti group : One rule that forwards any mcast traffic */
 369         MLX5_SET(create_flow_group_in, flow_group_in, match_criteria_enable,
 370                  MLX5_MATCH_OUTER_HEADERS);
 371         MLX5_SET(create_flow_group_in, flow_group_in, start_flow_index, table_size - 2);
 372         MLX5_SET(create_flow_group_in, flow_group_in, end_flow_index, table_size - 2);
 373         eth_zero_addr(dmac);
 374         dmac[0] = 0x01;
 375         g = mlx5_create_flow_group(fdb, flow_group_in);
 376         if (IS_ERR(g)) {
 377                 err = PTR_ERR(g);
 378                 esw_warn(dev, "Failed to create allmulti flow group err(%d)\n", err);
 379                 goto out;
 380         }
 381         esw->fdb_table.legacy.allmulti_grp = g;
 382 
 383         /* Promiscuous group :
 384          * One rule that forward all unmatched traffic from previous groups
 385          */
 386         eth_zero_addr(dmac);
 387         MLX5_SET(create_flow_group_in, flow_group_in, match_criteria_enable,
 388                  MLX5_MATCH_MISC_PARAMETERS);
 389         MLX5_SET_TO_ONES(fte_match_param, match_criteria, misc_parameters.source_port);
 390         MLX5_SET(create_flow_group_in, flow_group_in, start_flow_index, table_size - 1);
 391         MLX5_SET(create_flow_group_in, flow_group_in, end_flow_index, table_size - 1);
 392         g = mlx5_create_flow_group(fdb, flow_group_in);
 393         if (IS_ERR(g)) {
 394                 err = PTR_ERR(g);
 395                 esw_warn(dev, "Failed to create promisc flow group err(%d)\n", err);
 396                 goto out;
 397         }
 398         esw->fdb_table.legacy.promisc_grp = g;
 399 
 400 out:
 401         if (err)
 402                 esw_destroy_legacy_fdb_table(esw);
 403 
 404         kvfree(flow_group_in);
 405         return err;
 406 }
 407 
 408 static void esw_destroy_legacy_vepa_table(struct mlx5_eswitch *esw)
 409 {
 410         esw_debug(esw->dev, "Destroy VEPA Table\n");
 411         if (!esw->fdb_table.legacy.vepa_fdb)
 412                 return;
 413 
 414         mlx5_destroy_flow_table(esw->fdb_table.legacy.vepa_fdb);
 415         esw->fdb_table.legacy.vepa_fdb = NULL;
 416 }
 417 
 418 static void esw_destroy_legacy_fdb_table(struct mlx5_eswitch *esw)
 419 {
 420         esw_debug(esw->dev, "Destroy FDB Table\n");
 421         if (!esw->fdb_table.legacy.fdb)
 422                 return;
 423 
 424         if (esw->fdb_table.legacy.promisc_grp)
 425                 mlx5_destroy_flow_group(esw->fdb_table.legacy.promisc_grp);
 426         if (esw->fdb_table.legacy.allmulti_grp)
 427                 mlx5_destroy_flow_group(esw->fdb_table.legacy.allmulti_grp);
 428         if (esw->fdb_table.legacy.addr_grp)
 429                 mlx5_destroy_flow_group(esw->fdb_table.legacy.addr_grp);
 430         mlx5_destroy_flow_table(esw->fdb_table.legacy.fdb);
 431 
 432         esw->fdb_table.legacy.fdb = NULL;
 433         esw->fdb_table.legacy.addr_grp = NULL;
 434         esw->fdb_table.legacy.allmulti_grp = NULL;
 435         esw->fdb_table.legacy.promisc_grp = NULL;
 436 }
 437 
 438 static int esw_create_legacy_table(struct mlx5_eswitch *esw)
 439 {
 440         int err;
 441 
 442         memset(&esw->fdb_table.legacy, 0, sizeof(struct legacy_fdb));
 443 
 444         err = esw_create_legacy_vepa_table(esw);
 445         if (err)
 446                 return err;
 447 
 448         err = esw_create_legacy_fdb_table(esw);
 449         if (err)
 450                 esw_destroy_legacy_vepa_table(esw);
 451 
 452         return err;
 453 }
 454 
 455 #define MLX5_LEGACY_SRIOV_VPORT_EVENTS (MLX5_VPORT_UC_ADDR_CHANGE | \
 456                                         MLX5_VPORT_MC_ADDR_CHANGE | \
 457                                         MLX5_VPORT_PROMISC_CHANGE)
 458 
 459 static int esw_legacy_enable(struct mlx5_eswitch *esw)
 460 {
 461         int ret;
 462 
 463         ret = esw_create_legacy_table(esw);
 464         if (ret)
 465                 return ret;
 466 
 467         mlx5_eswitch_enable_pf_vf_vports(esw, MLX5_LEGACY_SRIOV_VPORT_EVENTS);
 468         return 0;
 469 }
 470 
 471 static void esw_destroy_legacy_table(struct mlx5_eswitch *esw)
 472 {
 473         esw_cleanup_vepa_rules(esw);
 474         esw_destroy_legacy_fdb_table(esw);
 475         esw_destroy_legacy_vepa_table(esw);
 476 }
 477 
 478 static void esw_legacy_disable(struct mlx5_eswitch *esw)
 479 {
 480         struct esw_mc_addr *mc_promisc;
 481 
 482         mlx5_eswitch_disable_pf_vf_vports(esw);
 483 
 484         mc_promisc = &esw->mc_promisc;
 485         if (mc_promisc->uplink_rule)
 486                 mlx5_del_flow_rules(mc_promisc->uplink_rule);
 487 
 488         esw_destroy_legacy_table(esw);
 489 }
 490 
 491 /* E-Switch vport UC/MC lists management */
 492 typedef int (*vport_addr_action)(struct mlx5_eswitch *esw,
 493                                  struct vport_addr *vaddr);
 494 
 495 static int esw_add_uc_addr(struct mlx5_eswitch *esw, struct vport_addr *vaddr)
 496 {
 497         u8 *mac = vaddr->node.addr;
 498         u16 vport = vaddr->vport;
 499         int err;
 500 
 501         /* Skip mlx5_mpfs_add_mac for eswitch_managers,
 502          * it is already done by its netdev in mlx5e_execute_l2_action
 503          */
 504         if (esw->manager_vport == vport)
 505                 goto fdb_add;
 506 
 507         err = mlx5_mpfs_add_mac(esw->dev, mac);
 508         if (err) {
 509                 esw_warn(esw->dev,
 510                          "Failed to add L2 table mac(%pM) for vport(0x%x), err(%d)\n",
 511                          mac, vport, err);
 512                 return err;
 513         }
 514         vaddr->mpfs = true;
 515 
 516 fdb_add:
 517         /* SRIOV is enabled: Forward UC MAC to vport */
 518         if (esw->fdb_table.legacy.fdb && esw->mode == MLX5_ESWITCH_LEGACY)
 519                 vaddr->flow_rule = esw_fdb_set_vport_rule(esw, mac, vport);
 520 
 521         esw_debug(esw->dev, "\tADDED UC MAC: vport[%d] %pM fr(%p)\n",
 522                   vport, mac, vaddr->flow_rule);
 523 
 524         return 0;
 525 }
 526 
 527 static int esw_del_uc_addr(struct mlx5_eswitch *esw, struct vport_addr *vaddr)
 528 {
 529         u8 *mac = vaddr->node.addr;
 530         u16 vport = vaddr->vport;
 531         int err = 0;
 532 
 533         /* Skip mlx5_mpfs_del_mac for eswitch managerss,
 534          * it is already done by its netdev in mlx5e_execute_l2_action
 535          */
 536         if (!vaddr->mpfs || esw->manager_vport == vport)
 537                 goto fdb_del;
 538 
 539         err = mlx5_mpfs_del_mac(esw->dev, mac);
 540         if (err)
 541                 esw_warn(esw->dev,
 542                          "Failed to del L2 table mac(%pM) for vport(%d), err(%d)\n",
 543                          mac, vport, err);
 544         vaddr->mpfs = false;
 545 
 546 fdb_del:
 547         if (vaddr->flow_rule)
 548                 mlx5_del_flow_rules(vaddr->flow_rule);
 549         vaddr->flow_rule = NULL;
 550 
 551         return 0;
 552 }
 553 
 554 static void update_allmulti_vports(struct mlx5_eswitch *esw,
 555                                    struct vport_addr *vaddr,
 556                                    struct esw_mc_addr *esw_mc)
 557 {
 558         u8 *mac = vaddr->node.addr;
 559         struct mlx5_vport *vport;
 560         u16 i, vport_num;
 561 
 562         mlx5_esw_for_all_vports(esw, i, vport) {
 563                 struct hlist_head *vport_hash = vport->mc_list;
 564                 struct vport_addr *iter_vaddr =
 565                                         l2addr_hash_find(vport_hash,
 566                                                          mac,
 567                                                          struct vport_addr);
 568                 vport_num = vport->vport;
 569                 if (IS_ERR_OR_NULL(vport->allmulti_rule) ||
 570                     vaddr->vport == vport_num)
 571                         continue;
 572                 switch (vaddr->action) {
 573                 case MLX5_ACTION_ADD:
 574                         if (iter_vaddr)
 575                                 continue;
 576                         iter_vaddr = l2addr_hash_add(vport_hash, mac,
 577                                                      struct vport_addr,
 578                                                      GFP_KERNEL);
 579                         if (!iter_vaddr) {
 580                                 esw_warn(esw->dev,
 581                                          "ALL-MULTI: Failed to add MAC(%pM) to vport[%d] DB\n",
 582                                          mac, vport_num);
 583                                 continue;
 584                         }
 585                         iter_vaddr->vport = vport_num;
 586                         iter_vaddr->flow_rule =
 587                                         esw_fdb_set_vport_rule(esw,
 588                                                                mac,
 589                                                                vport_num);
 590                         iter_vaddr->mc_promisc = true;
 591                         break;
 592                 case MLX5_ACTION_DEL:
 593                         if (!iter_vaddr)
 594                                 continue;
 595                         mlx5_del_flow_rules(iter_vaddr->flow_rule);
 596                         l2addr_hash_del(iter_vaddr);
 597                         break;
 598                 }
 599         }
 600 }
 601 
 602 static int esw_add_mc_addr(struct mlx5_eswitch *esw, struct vport_addr *vaddr)
 603 {
 604         struct hlist_head *hash = esw->mc_table;
 605         struct esw_mc_addr *esw_mc;
 606         u8 *mac = vaddr->node.addr;
 607         u16 vport = vaddr->vport;
 608 
 609         if (!esw->fdb_table.legacy.fdb)
 610                 return 0;
 611 
 612         esw_mc = l2addr_hash_find(hash, mac, struct esw_mc_addr);
 613         if (esw_mc)
 614                 goto add;
 615 
 616         esw_mc = l2addr_hash_add(hash, mac, struct esw_mc_addr, GFP_KERNEL);
 617         if (!esw_mc)
 618                 return -ENOMEM;
 619 
 620         esw_mc->uplink_rule = /* Forward MC MAC to Uplink */
 621                 esw_fdb_set_vport_rule(esw, mac, MLX5_VPORT_UPLINK);
 622 
 623         /* Add this multicast mac to all the mc promiscuous vports */
 624         update_allmulti_vports(esw, vaddr, esw_mc);
 625 
 626 add:
 627         /* If the multicast mac is added as a result of mc promiscuous vport,
 628          * don't increment the multicast ref count
 629          */
 630         if (!vaddr->mc_promisc)
 631                 esw_mc->refcnt++;
 632 
 633         /* Forward MC MAC to vport */
 634         vaddr->flow_rule = esw_fdb_set_vport_rule(esw, mac, vport);
 635         esw_debug(esw->dev,
 636                   "\tADDED MC MAC: vport[%d] %pM fr(%p) refcnt(%d) uplinkfr(%p)\n",
 637                   vport, mac, vaddr->flow_rule,
 638                   esw_mc->refcnt, esw_mc->uplink_rule);
 639         return 0;
 640 }
 641 
 642 static int esw_del_mc_addr(struct mlx5_eswitch *esw, struct vport_addr *vaddr)
 643 {
 644         struct hlist_head *hash = esw->mc_table;
 645         struct esw_mc_addr *esw_mc;
 646         u8 *mac = vaddr->node.addr;
 647         u16 vport = vaddr->vport;
 648 
 649         if (!esw->fdb_table.legacy.fdb)
 650                 return 0;
 651 
 652         esw_mc = l2addr_hash_find(hash, mac, struct esw_mc_addr);
 653         if (!esw_mc) {
 654                 esw_warn(esw->dev,
 655                          "Failed to find eswitch MC addr for MAC(%pM) vport(%d)",
 656                          mac, vport);
 657                 return -EINVAL;
 658         }
 659         esw_debug(esw->dev,
 660                   "\tDELETE MC MAC: vport[%d] %pM fr(%p) refcnt(%d) uplinkfr(%p)\n",
 661                   vport, mac, vaddr->flow_rule, esw_mc->refcnt,
 662                   esw_mc->uplink_rule);
 663 
 664         if (vaddr->flow_rule)
 665                 mlx5_del_flow_rules(vaddr->flow_rule);
 666         vaddr->flow_rule = NULL;
 667 
 668         /* If the multicast mac is added as a result of mc promiscuous vport,
 669          * don't decrement the multicast ref count.
 670          */
 671         if (vaddr->mc_promisc || (--esw_mc->refcnt > 0))
 672                 return 0;
 673 
 674         /* Remove this multicast mac from all the mc promiscuous vports */
 675         update_allmulti_vports(esw, vaddr, esw_mc);
 676 
 677         if (esw_mc->uplink_rule)
 678                 mlx5_del_flow_rules(esw_mc->uplink_rule);
 679 
 680         l2addr_hash_del(esw_mc);
 681         return 0;
 682 }
 683 
 684 /* Apply vport UC/MC list to HW l2 table and FDB table */
 685 static void esw_apply_vport_addr_list(struct mlx5_eswitch *esw,
 686                                       struct mlx5_vport *vport, int list_type)
 687 {
 688         bool is_uc = list_type == MLX5_NVPRT_LIST_TYPE_UC;
 689         vport_addr_action vport_addr_add;
 690         vport_addr_action vport_addr_del;
 691         struct vport_addr *addr;
 692         struct l2addr_node *node;
 693         struct hlist_head *hash;
 694         struct hlist_node *tmp;
 695         int hi;
 696 
 697         vport_addr_add = is_uc ? esw_add_uc_addr :
 698                                  esw_add_mc_addr;
 699         vport_addr_del = is_uc ? esw_del_uc_addr :
 700                                  esw_del_mc_addr;
 701 
 702         hash = is_uc ? vport->uc_list : vport->mc_list;
 703         for_each_l2hash_node(node, tmp, hash, hi) {
 704                 addr = container_of(node, struct vport_addr, node);
 705                 switch (addr->action) {
 706                 case MLX5_ACTION_ADD:
 707                         vport_addr_add(esw, addr);
 708                         addr->action = MLX5_ACTION_NONE;
 709                         break;
 710                 case MLX5_ACTION_DEL:
 711                         vport_addr_del(esw, addr);
 712                         l2addr_hash_del(addr);
 713                         break;
 714                 }
 715         }
 716 }
 717 
 718 /* Sync vport UC/MC list from vport context */
 719 static void esw_update_vport_addr_list(struct mlx5_eswitch *esw,
 720                                        struct mlx5_vport *vport, int list_type)
 721 {
 722         bool is_uc = list_type == MLX5_NVPRT_LIST_TYPE_UC;
 723         u8 (*mac_list)[ETH_ALEN];
 724         struct l2addr_node *node;
 725         struct vport_addr *addr;
 726         struct hlist_head *hash;
 727         struct hlist_node *tmp;
 728         int size;
 729         int err;
 730         int hi;
 731         int i;
 732 
 733         size = is_uc ? MLX5_MAX_UC_PER_VPORT(esw->dev) :
 734                        MLX5_MAX_MC_PER_VPORT(esw->dev);
 735 
 736         mac_list = kcalloc(size, ETH_ALEN, GFP_KERNEL);
 737         if (!mac_list)
 738                 return;
 739 
 740         hash = is_uc ? vport->uc_list : vport->mc_list;
 741 
 742         for_each_l2hash_node(node, tmp, hash, hi) {
 743                 addr = container_of(node, struct vport_addr, node);
 744                 addr->action = MLX5_ACTION_DEL;
 745         }
 746 
 747         if (!vport->enabled)
 748                 goto out;
 749 
 750         err = mlx5_query_nic_vport_mac_list(esw->dev, vport->vport, list_type,
 751                                             mac_list, &size);
 752         if (err)
 753                 goto out;
 754         esw_debug(esw->dev, "vport[%d] context update %s list size (%d)\n",
 755                   vport->vport, is_uc ? "UC" : "MC", size);
 756 
 757         for (i = 0; i < size; i++) {
 758                 if (is_uc && !is_valid_ether_addr(mac_list[i]))
 759                         continue;
 760 
 761                 if (!is_uc && !is_multicast_ether_addr(mac_list[i]))
 762                         continue;
 763 
 764                 addr = l2addr_hash_find(hash, mac_list[i], struct vport_addr);
 765                 if (addr) {
 766                         addr->action = MLX5_ACTION_NONE;
 767                         /* If this mac was previously added because of allmulti
 768                          * promiscuous rx mode, its now converted to be original
 769                          * vport mac.
 770                          */
 771                         if (addr->mc_promisc) {
 772                                 struct esw_mc_addr *esw_mc =
 773                                         l2addr_hash_find(esw->mc_table,
 774                                                          mac_list[i],
 775                                                          struct esw_mc_addr);
 776                                 if (!esw_mc) {
 777                                         esw_warn(esw->dev,
 778                                                  "Failed to MAC(%pM) in mcast DB\n",
 779                                                  mac_list[i]);
 780                                         continue;
 781                                 }
 782                                 esw_mc->refcnt++;
 783                                 addr->mc_promisc = false;
 784                         }
 785                         continue;
 786                 }
 787 
 788                 addr = l2addr_hash_add(hash, mac_list[i], struct vport_addr,
 789                                        GFP_KERNEL);
 790                 if (!addr) {
 791                         esw_warn(esw->dev,
 792                                  "Failed to add MAC(%pM) to vport[%d] DB\n",
 793                                  mac_list[i], vport->vport);
 794                         continue;
 795                 }
 796                 addr->vport = vport->vport;
 797                 addr->action = MLX5_ACTION_ADD;
 798         }
 799 out:
 800         kfree(mac_list);
 801 }
 802 
 803 /* Sync vport UC/MC list from vport context
 804  * Must be called after esw_update_vport_addr_list
 805  */
 806 static void esw_update_vport_mc_promisc(struct mlx5_eswitch *esw,
 807                                         struct mlx5_vport *vport)
 808 {
 809         struct l2addr_node *node;
 810         struct vport_addr *addr;
 811         struct hlist_head *hash;
 812         struct hlist_node *tmp;
 813         int hi;
 814 
 815         hash = vport->mc_list;
 816 
 817         for_each_l2hash_node(node, tmp, esw->mc_table, hi) {
 818                 u8 *mac = node->addr;
 819 
 820                 addr = l2addr_hash_find(hash, mac, struct vport_addr);
 821                 if (addr) {
 822                         if (addr->action == MLX5_ACTION_DEL)
 823                                 addr->action = MLX5_ACTION_NONE;
 824                         continue;
 825                 }
 826                 addr = l2addr_hash_add(hash, mac, struct vport_addr,
 827                                        GFP_KERNEL);
 828                 if (!addr) {
 829                         esw_warn(esw->dev,
 830                                  "Failed to add allmulti MAC(%pM) to vport[%d] DB\n",
 831                                  mac, vport->vport);
 832                         continue;
 833                 }
 834                 addr->vport = vport->vport;
 835                 addr->action = MLX5_ACTION_ADD;
 836                 addr->mc_promisc = true;
 837         }
 838 }
 839 
 840 /* Apply vport rx mode to HW FDB table */
 841 static void esw_apply_vport_rx_mode(struct mlx5_eswitch *esw,
 842                                     struct mlx5_vport *vport,
 843                                     bool promisc, bool mc_promisc)
 844 {
 845         struct esw_mc_addr *allmulti_addr = &esw->mc_promisc;
 846 
 847         if (IS_ERR_OR_NULL(vport->allmulti_rule) != mc_promisc)
 848                 goto promisc;
 849 
 850         if (mc_promisc) {
 851                 vport->allmulti_rule =
 852                         esw_fdb_set_vport_allmulti_rule(esw, vport->vport);
 853                 if (!allmulti_addr->uplink_rule)
 854                         allmulti_addr->uplink_rule =
 855                                 esw_fdb_set_vport_allmulti_rule(esw,
 856                                                                 MLX5_VPORT_UPLINK);
 857                 allmulti_addr->refcnt++;
 858         } else if (vport->allmulti_rule) {
 859                 mlx5_del_flow_rules(vport->allmulti_rule);
 860                 vport->allmulti_rule = NULL;
 861 
 862                 if (--allmulti_addr->refcnt > 0)
 863                         goto promisc;
 864 
 865                 if (allmulti_addr->uplink_rule)
 866                         mlx5_del_flow_rules(allmulti_addr->uplink_rule);
 867                 allmulti_addr->uplink_rule = NULL;
 868         }
 869 
 870 promisc:
 871         if (IS_ERR_OR_NULL(vport->promisc_rule) != promisc)
 872                 return;
 873 
 874         if (promisc) {
 875                 vport->promisc_rule =
 876                         esw_fdb_set_vport_promisc_rule(esw, vport->vport);
 877         } else if (vport->promisc_rule) {
 878                 mlx5_del_flow_rules(vport->promisc_rule);
 879                 vport->promisc_rule = NULL;
 880         }
 881 }
 882 
 883 /* Sync vport rx mode from vport context */
 884 static void esw_update_vport_rx_mode(struct mlx5_eswitch *esw,
 885                                      struct mlx5_vport *vport)
 886 {
 887         int promisc_all = 0;
 888         int promisc_uc = 0;
 889         int promisc_mc = 0;
 890         int err;
 891 
 892         err = mlx5_query_nic_vport_promisc(esw->dev,
 893                                            vport->vport,
 894                                            &promisc_uc,
 895                                            &promisc_mc,
 896                                            &promisc_all);
 897         if (err)
 898                 return;
 899         esw_debug(esw->dev, "vport[%d] context update rx mode promisc_all=%d, all_multi=%d\n",
 900                   vport->vport, promisc_all, promisc_mc);
 901 
 902         if (!vport->info.trusted || !vport->enabled) {
 903                 promisc_uc = 0;
 904                 promisc_mc = 0;
 905                 promisc_all = 0;
 906         }
 907 
 908         esw_apply_vport_rx_mode(esw, vport, promisc_all,
 909                                 (promisc_all || promisc_mc));
 910 }
 911 
 912 static void esw_vport_change_handle_locked(struct mlx5_vport *vport)
 913 {
 914         struct mlx5_core_dev *dev = vport->dev;
 915         struct mlx5_eswitch *esw = dev->priv.eswitch;
 916         u8 mac[ETH_ALEN];
 917 
 918         mlx5_query_nic_vport_mac_address(dev, vport->vport, true, mac);
 919         esw_debug(dev, "vport[%d] Context Changed: perm mac: %pM\n",
 920                   vport->vport, mac);
 921 
 922         if (vport->enabled_events & MLX5_VPORT_UC_ADDR_CHANGE) {
 923                 esw_update_vport_addr_list(esw, vport, MLX5_NVPRT_LIST_TYPE_UC);
 924                 esw_apply_vport_addr_list(esw, vport, MLX5_NVPRT_LIST_TYPE_UC);
 925         }
 926 
 927         if (vport->enabled_events & MLX5_VPORT_MC_ADDR_CHANGE)
 928                 esw_update_vport_addr_list(esw, vport, MLX5_NVPRT_LIST_TYPE_MC);
 929 
 930         if (vport->enabled_events & MLX5_VPORT_PROMISC_CHANGE) {
 931                 esw_update_vport_rx_mode(esw, vport);
 932                 if (!IS_ERR_OR_NULL(vport->allmulti_rule))
 933                         esw_update_vport_mc_promisc(esw, vport);
 934         }
 935 
 936         if (vport->enabled_events & (MLX5_VPORT_PROMISC_CHANGE | MLX5_VPORT_MC_ADDR_CHANGE))
 937                 esw_apply_vport_addr_list(esw, vport, MLX5_NVPRT_LIST_TYPE_MC);
 938 
 939         esw_debug(esw->dev, "vport[%d] Context Changed: Done\n", vport->vport);
 940         if (vport->enabled)
 941                 arm_vport_context_events_cmd(dev, vport->vport,
 942                                              vport->enabled_events);
 943 }
 944 
 945 static void esw_vport_change_handler(struct work_struct *work)
 946 {
 947         struct mlx5_vport *vport =
 948                 container_of(work, struct mlx5_vport, vport_change_handler);
 949         struct mlx5_eswitch *esw = vport->dev->priv.eswitch;
 950 
 951         mutex_lock(&esw->state_lock);
 952         esw_vport_change_handle_locked(vport);
 953         mutex_unlock(&esw->state_lock);
 954 }
 955 
 956 int esw_vport_enable_egress_acl(struct mlx5_eswitch *esw,
 957                                 struct mlx5_vport *vport)
 958 {
 959         int inlen = MLX5_ST_SZ_BYTES(create_flow_group_in);
 960         struct mlx5_flow_group *vlan_grp = NULL;
 961         struct mlx5_flow_group *drop_grp = NULL;
 962         struct mlx5_core_dev *dev = esw->dev;
 963         struct mlx5_flow_namespace *root_ns;
 964         struct mlx5_flow_table *acl;
 965         void *match_criteria;
 966         u32 *flow_group_in;
 967         /* The egress acl table contains 2 rules:
 968          * 1)Allow traffic with vlan_tag=vst_vlan_id
 969          * 2)Drop all other traffic.
 970          */
 971         int table_size = 2;
 972         int err = 0;
 973 
 974         if (!MLX5_CAP_ESW_EGRESS_ACL(dev, ft_support))
 975                 return -EOPNOTSUPP;
 976 
 977         if (!IS_ERR_OR_NULL(vport->egress.acl))
 978                 return 0;
 979 
 980         esw_debug(dev, "Create vport[%d] egress ACL log_max_size(%d)\n",
 981                   vport->vport, MLX5_CAP_ESW_EGRESS_ACL(dev, log_max_ft_size));
 982 
 983         root_ns = mlx5_get_flow_vport_acl_namespace(dev, MLX5_FLOW_NAMESPACE_ESW_EGRESS,
 984                         mlx5_eswitch_vport_num_to_index(esw, vport->vport));
 985         if (!root_ns) {
 986                 esw_warn(dev, "Failed to get E-Switch egress flow namespace for vport (%d)\n", vport->vport);
 987                 return -EOPNOTSUPP;
 988         }
 989 
 990         flow_group_in = kvzalloc(inlen, GFP_KERNEL);
 991         if (!flow_group_in)
 992                 return -ENOMEM;
 993 
 994         acl = mlx5_create_vport_flow_table(root_ns, 0, table_size, 0, vport->vport);
 995         if (IS_ERR(acl)) {
 996                 err = PTR_ERR(acl);
 997                 esw_warn(dev, "Failed to create E-Switch vport[%d] egress flow Table, err(%d)\n",
 998                          vport->vport, err);
 999                 goto out;
1000         }
1001 
1002         MLX5_SET(create_flow_group_in, flow_group_in, match_criteria_enable, MLX5_MATCH_OUTER_HEADERS);
1003         match_criteria = MLX5_ADDR_OF(create_flow_group_in, flow_group_in, match_criteria);
1004         MLX5_SET_TO_ONES(fte_match_param, match_criteria, outer_headers.cvlan_tag);
1005         MLX5_SET_TO_ONES(fte_match_param, match_criteria, outer_headers.first_vid);
1006         MLX5_SET(create_flow_group_in, flow_group_in, start_flow_index, 0);
1007         MLX5_SET(create_flow_group_in, flow_group_in, end_flow_index, 0);
1008 
1009         vlan_grp = mlx5_create_flow_group(acl, flow_group_in);
1010         if (IS_ERR(vlan_grp)) {
1011                 err = PTR_ERR(vlan_grp);
1012                 esw_warn(dev, "Failed to create E-Switch vport[%d] egress allowed vlans flow group, err(%d)\n",
1013                          vport->vport, err);
1014                 goto out;
1015         }
1016 
1017         memset(flow_group_in, 0, inlen);
1018         MLX5_SET(create_flow_group_in, flow_group_in, start_flow_index, 1);
1019         MLX5_SET(create_flow_group_in, flow_group_in, end_flow_index, 1);
1020         drop_grp = mlx5_create_flow_group(acl, flow_group_in);
1021         if (IS_ERR(drop_grp)) {
1022                 err = PTR_ERR(drop_grp);
1023                 esw_warn(dev, "Failed to create E-Switch vport[%d] egress drop flow group, err(%d)\n",
1024                          vport->vport, err);
1025                 goto out;
1026         }
1027 
1028         vport->egress.acl = acl;
1029         vport->egress.drop_grp = drop_grp;
1030         vport->egress.allowed_vlans_grp = vlan_grp;
1031 out:
1032         kvfree(flow_group_in);
1033         if (err && !IS_ERR_OR_NULL(vlan_grp))
1034                 mlx5_destroy_flow_group(vlan_grp);
1035         if (err && !IS_ERR_OR_NULL(acl))
1036                 mlx5_destroy_flow_table(acl);
1037         return err;
1038 }
1039 
1040 void esw_vport_cleanup_egress_rules(struct mlx5_eswitch *esw,
1041                                     struct mlx5_vport *vport)
1042 {
1043         if (!IS_ERR_OR_NULL(vport->egress.allowed_vlan))
1044                 mlx5_del_flow_rules(vport->egress.allowed_vlan);
1045 
1046         if (!IS_ERR_OR_NULL(vport->egress.drop_rule))
1047                 mlx5_del_flow_rules(vport->egress.drop_rule);
1048 
1049         vport->egress.allowed_vlan = NULL;
1050         vport->egress.drop_rule = NULL;
1051 }
1052 
1053 void esw_vport_disable_egress_acl(struct mlx5_eswitch *esw,
1054                                   struct mlx5_vport *vport)
1055 {
1056         if (IS_ERR_OR_NULL(vport->egress.acl))
1057                 return;
1058 
1059         esw_debug(esw->dev, "Destroy vport[%d] E-Switch egress ACL\n", vport->vport);
1060 
1061         esw_vport_cleanup_egress_rules(esw, vport);
1062         mlx5_destroy_flow_group(vport->egress.allowed_vlans_grp);
1063         mlx5_destroy_flow_group(vport->egress.drop_grp);
1064         mlx5_destroy_flow_table(vport->egress.acl);
1065         vport->egress.allowed_vlans_grp = NULL;
1066         vport->egress.drop_grp = NULL;
1067         vport->egress.acl = NULL;
1068 }
1069 
1070 int esw_vport_enable_ingress_acl(struct mlx5_eswitch *esw,
1071                                  struct mlx5_vport *vport)
1072 {
1073         int inlen = MLX5_ST_SZ_BYTES(create_flow_group_in);
1074         struct mlx5_core_dev *dev = esw->dev;
1075         struct mlx5_flow_namespace *root_ns;
1076         struct mlx5_flow_table *acl;
1077         struct mlx5_flow_group *g;
1078         void *match_criteria;
1079         u32 *flow_group_in;
1080         /* The ingress acl table contains 4 groups
1081          * (2 active rules at the same time -
1082          *      1 allow rule from one of the first 3 groups.
1083          *      1 drop rule from the last group):
1084          * 1)Allow untagged traffic with smac=original mac.
1085          * 2)Allow untagged traffic.
1086          * 3)Allow traffic with smac=original mac.
1087          * 4)Drop all other traffic.
1088          */
1089         int table_size = 4;
1090         int err = 0;
1091 
1092         if (!MLX5_CAP_ESW_INGRESS_ACL(dev, ft_support))
1093                 return -EOPNOTSUPP;
1094 
1095         if (!IS_ERR_OR_NULL(vport->ingress.acl))
1096                 return 0;
1097 
1098         esw_debug(dev, "Create vport[%d] ingress ACL log_max_size(%d)\n",
1099                   vport->vport, MLX5_CAP_ESW_INGRESS_ACL(dev, log_max_ft_size));
1100 
1101         root_ns = mlx5_get_flow_vport_acl_namespace(dev, MLX5_FLOW_NAMESPACE_ESW_INGRESS,
1102                         mlx5_eswitch_vport_num_to_index(esw, vport->vport));
1103         if (!root_ns) {
1104                 esw_warn(dev, "Failed to get E-Switch ingress flow namespace for vport (%d)\n", vport->vport);
1105                 return -EOPNOTSUPP;
1106         }
1107 
1108         flow_group_in = kvzalloc(inlen, GFP_KERNEL);
1109         if (!flow_group_in)
1110                 return -ENOMEM;
1111 
1112         acl = mlx5_create_vport_flow_table(root_ns, 0, table_size, 0, vport->vport);
1113         if (IS_ERR(acl)) {
1114                 err = PTR_ERR(acl);
1115                 esw_warn(dev, "Failed to create E-Switch vport[%d] ingress flow Table, err(%d)\n",
1116                          vport->vport, err);
1117                 goto out;
1118         }
1119         vport->ingress.acl = acl;
1120 
1121         match_criteria = MLX5_ADDR_OF(create_flow_group_in, flow_group_in, match_criteria);
1122 
1123         MLX5_SET(create_flow_group_in, flow_group_in, match_criteria_enable, MLX5_MATCH_OUTER_HEADERS);
1124         MLX5_SET_TO_ONES(fte_match_param, match_criteria, outer_headers.cvlan_tag);
1125         MLX5_SET_TO_ONES(fte_match_param, match_criteria, outer_headers.smac_47_16);
1126         MLX5_SET_TO_ONES(fte_match_param, match_criteria, outer_headers.smac_15_0);
1127         MLX5_SET(create_flow_group_in, flow_group_in, start_flow_index, 0);
1128         MLX5_SET(create_flow_group_in, flow_group_in, end_flow_index, 0);
1129 
1130         g = mlx5_create_flow_group(acl, flow_group_in);
1131         if (IS_ERR(g)) {
1132                 err = PTR_ERR(g);
1133                 esw_warn(dev, "Failed to create E-Switch vport[%d] ingress untagged spoofchk flow group, err(%d)\n",
1134                          vport->vport, err);
1135                 goto out;
1136         }
1137         vport->ingress.allow_untagged_spoofchk_grp = g;
1138 
1139         memset(flow_group_in, 0, inlen);
1140         MLX5_SET(create_flow_group_in, flow_group_in, match_criteria_enable, MLX5_MATCH_OUTER_HEADERS);
1141         MLX5_SET_TO_ONES(fte_match_param, match_criteria, outer_headers.cvlan_tag);
1142         MLX5_SET(create_flow_group_in, flow_group_in, start_flow_index, 1);
1143         MLX5_SET(create_flow_group_in, flow_group_in, end_flow_index, 1);
1144 
1145         g = mlx5_create_flow_group(acl, flow_group_in);
1146         if (IS_ERR(g)) {
1147                 err = PTR_ERR(g);
1148                 esw_warn(dev, "Failed to create E-Switch vport[%d] ingress untagged flow group, err(%d)\n",
1149                          vport->vport, err);
1150                 goto out;
1151         }
1152         vport->ingress.allow_untagged_only_grp = g;
1153 
1154         memset(flow_group_in, 0, inlen);
1155         MLX5_SET(create_flow_group_in, flow_group_in, match_criteria_enable, MLX5_MATCH_OUTER_HEADERS);
1156         MLX5_SET_TO_ONES(fte_match_param, match_criteria, outer_headers.smac_47_16);
1157         MLX5_SET_TO_ONES(fte_match_param, match_criteria, outer_headers.smac_15_0);
1158         MLX5_SET(create_flow_group_in, flow_group_in, start_flow_index, 2);
1159         MLX5_SET(create_flow_group_in, flow_group_in, end_flow_index, 2);
1160 
1161         g = mlx5_create_flow_group(acl, flow_group_in);
1162         if (IS_ERR(g)) {
1163                 err = PTR_ERR(g);
1164                 esw_warn(dev, "Failed to create E-Switch vport[%d] ingress spoofchk flow group, err(%d)\n",
1165                          vport->vport, err);
1166                 goto out;
1167         }
1168         vport->ingress.allow_spoofchk_only_grp = g;
1169 
1170         memset(flow_group_in, 0, inlen);
1171         MLX5_SET(create_flow_group_in, flow_group_in, start_flow_index, 3);
1172         MLX5_SET(create_flow_group_in, flow_group_in, end_flow_index, 3);
1173 
1174         g = mlx5_create_flow_group(acl, flow_group_in);
1175         if (IS_ERR(g)) {
1176                 err = PTR_ERR(g);
1177                 esw_warn(dev, "Failed to create E-Switch vport[%d] ingress drop flow group, err(%d)\n",
1178                          vport->vport, err);
1179                 goto out;
1180         }
1181         vport->ingress.drop_grp = g;
1182 
1183 out:
1184         if (err) {
1185                 if (!IS_ERR_OR_NULL(vport->ingress.allow_spoofchk_only_grp))
1186                         mlx5_destroy_flow_group(
1187                                         vport->ingress.allow_spoofchk_only_grp);
1188                 if (!IS_ERR_OR_NULL(vport->ingress.allow_untagged_only_grp))
1189                         mlx5_destroy_flow_group(
1190                                         vport->ingress.allow_untagged_only_grp);
1191                 if (!IS_ERR_OR_NULL(vport->ingress.allow_untagged_spoofchk_grp))
1192                         mlx5_destroy_flow_group(
1193                                 vport->ingress.allow_untagged_spoofchk_grp);
1194                 if (!IS_ERR_OR_NULL(vport->ingress.acl))
1195                         mlx5_destroy_flow_table(vport->ingress.acl);
1196         }
1197 
1198         kvfree(flow_group_in);
1199         return err;
1200 }
1201 
1202 void esw_vport_cleanup_ingress_rules(struct mlx5_eswitch *esw,
1203                                      struct mlx5_vport *vport)
1204 {
1205         if (!IS_ERR_OR_NULL(vport->ingress.drop_rule))
1206                 mlx5_del_flow_rules(vport->ingress.drop_rule);
1207 
1208         if (!IS_ERR_OR_NULL(vport->ingress.allow_rule))
1209                 mlx5_del_flow_rules(vport->ingress.allow_rule);
1210 
1211         vport->ingress.drop_rule = NULL;
1212         vport->ingress.allow_rule = NULL;
1213 
1214         esw_vport_del_ingress_acl_modify_metadata(esw, vport);
1215 }
1216 
1217 void esw_vport_disable_ingress_acl(struct mlx5_eswitch *esw,
1218                                    struct mlx5_vport *vport)
1219 {
1220         if (IS_ERR_OR_NULL(vport->ingress.acl))
1221                 return;
1222 
1223         esw_debug(esw->dev, "Destroy vport[%d] E-Switch ingress ACL\n", vport->vport);
1224 
1225         esw_vport_cleanup_ingress_rules(esw, vport);
1226         mlx5_destroy_flow_group(vport->ingress.allow_spoofchk_only_grp);
1227         mlx5_destroy_flow_group(vport->ingress.allow_untagged_only_grp);
1228         mlx5_destroy_flow_group(vport->ingress.allow_untagged_spoofchk_grp);
1229         mlx5_destroy_flow_group(vport->ingress.drop_grp);
1230         mlx5_destroy_flow_table(vport->ingress.acl);
1231         vport->ingress.acl = NULL;
1232         vport->ingress.drop_grp = NULL;
1233         vport->ingress.allow_spoofchk_only_grp = NULL;
1234         vport->ingress.allow_untagged_only_grp = NULL;
1235         vport->ingress.allow_untagged_spoofchk_grp = NULL;
1236 }
1237 
1238 static int esw_vport_ingress_config(struct mlx5_eswitch *esw,
1239                                     struct mlx5_vport *vport)
1240 {
1241         struct mlx5_fc *counter = vport->ingress.drop_counter;
1242         struct mlx5_flow_destination drop_ctr_dst = {0};
1243         struct mlx5_flow_destination *dst = NULL;
1244         struct mlx5_flow_act flow_act = {0};
1245         struct mlx5_flow_spec *spec;
1246         int dest_num = 0;
1247         int err = 0;
1248         u8 *smac_v;
1249 
1250         esw_vport_cleanup_ingress_rules(esw, vport);
1251 
1252         if (!vport->info.vlan && !vport->info.qos && !vport->info.spoofchk) {
1253                 esw_vport_disable_ingress_acl(esw, vport);
1254                 return 0;
1255         }
1256 
1257         err = esw_vport_enable_ingress_acl(esw, vport);
1258         if (err) {
1259                 mlx5_core_warn(esw->dev,
1260                                "failed to enable ingress acl (%d) on vport[%d]\n",
1261                                err, vport->vport);
1262                 return err;
1263         }
1264 
1265         esw_debug(esw->dev,
1266                   "vport[%d] configure ingress rules, vlan(%d) qos(%d)\n",
1267                   vport->vport, vport->info.vlan, vport->info.qos);
1268 
1269         spec = kvzalloc(sizeof(*spec), GFP_KERNEL);
1270         if (!spec) {
1271                 err = -ENOMEM;
1272                 goto out;
1273         }
1274 
1275         if (vport->info.vlan || vport->info.qos)
1276                 MLX5_SET_TO_ONES(fte_match_param, spec->match_criteria, outer_headers.cvlan_tag);
1277 
1278         if (vport->info.spoofchk) {
1279                 MLX5_SET_TO_ONES(fte_match_param, spec->match_criteria, outer_headers.smac_47_16);
1280                 MLX5_SET_TO_ONES(fte_match_param, spec->match_criteria, outer_headers.smac_15_0);
1281                 smac_v = MLX5_ADDR_OF(fte_match_param,
1282                                       spec->match_value,
1283                                       outer_headers.smac_47_16);
1284                 ether_addr_copy(smac_v, vport->info.mac);
1285         }
1286 
1287         spec->match_criteria_enable = MLX5_MATCH_OUTER_HEADERS;
1288         flow_act.action = MLX5_FLOW_CONTEXT_ACTION_ALLOW;
1289         vport->ingress.allow_rule =
1290                 mlx5_add_flow_rules(vport->ingress.acl, spec,
1291                                     &flow_act, NULL, 0);
1292         if (IS_ERR(vport->ingress.allow_rule)) {
1293                 err = PTR_ERR(vport->ingress.allow_rule);
1294                 esw_warn(esw->dev,
1295                          "vport[%d] configure ingress allow rule, err(%d)\n",
1296                          vport->vport, err);
1297                 vport->ingress.allow_rule = NULL;
1298                 goto out;
1299         }
1300 
1301         memset(spec, 0, sizeof(*spec));
1302         flow_act.action = MLX5_FLOW_CONTEXT_ACTION_DROP;
1303 
1304         /* Attach drop flow counter */
1305         if (counter) {
1306                 flow_act.action |= MLX5_FLOW_CONTEXT_ACTION_COUNT;
1307                 drop_ctr_dst.type = MLX5_FLOW_DESTINATION_TYPE_COUNTER;
1308                 drop_ctr_dst.counter_id = mlx5_fc_id(counter);
1309                 dst = &drop_ctr_dst;
1310                 dest_num++;
1311         }
1312         vport->ingress.drop_rule =
1313                 mlx5_add_flow_rules(vport->ingress.acl, spec,
1314                                     &flow_act, dst, dest_num);
1315         if (IS_ERR(vport->ingress.drop_rule)) {
1316                 err = PTR_ERR(vport->ingress.drop_rule);
1317                 esw_warn(esw->dev,
1318                          "vport[%d] configure ingress drop rule, err(%d)\n",
1319                          vport->vport, err);
1320                 vport->ingress.drop_rule = NULL;
1321                 goto out;
1322         }
1323 
1324 out:
1325         if (err)
1326                 esw_vport_cleanup_ingress_rules(esw, vport);
1327         kvfree(spec);
1328         return err;
1329 }
1330 
1331 static int esw_vport_egress_config(struct mlx5_eswitch *esw,
1332                                    struct mlx5_vport *vport)
1333 {
1334         struct mlx5_fc *counter = vport->egress.drop_counter;
1335         struct mlx5_flow_destination drop_ctr_dst = {0};
1336         struct mlx5_flow_destination *dst = NULL;
1337         struct mlx5_flow_act flow_act = {0};
1338         struct mlx5_flow_spec *spec;
1339         int dest_num = 0;
1340         int err = 0;
1341 
1342         esw_vport_cleanup_egress_rules(esw, vport);
1343 
1344         if (!vport->info.vlan && !vport->info.qos) {
1345                 esw_vport_disable_egress_acl(esw, vport);
1346                 return 0;
1347         }
1348 
1349         err = esw_vport_enable_egress_acl(esw, vport);
1350         if (err) {
1351                 mlx5_core_warn(esw->dev,
1352                                "failed to enable egress acl (%d) on vport[%d]\n",
1353                                err, vport->vport);
1354                 return err;
1355         }
1356 
1357         esw_debug(esw->dev,
1358                   "vport[%d] configure egress rules, vlan(%d) qos(%d)\n",
1359                   vport->vport, vport->info.vlan, vport->info.qos);
1360 
1361         spec = kvzalloc(sizeof(*spec), GFP_KERNEL);
1362         if (!spec) {
1363                 err = -ENOMEM;
1364                 goto out;
1365         }
1366 
1367         /* Allowed vlan rule */
1368         MLX5_SET_TO_ONES(fte_match_param, spec->match_criteria, outer_headers.cvlan_tag);
1369         MLX5_SET_TO_ONES(fte_match_param, spec->match_value, outer_headers.cvlan_tag);
1370         MLX5_SET_TO_ONES(fte_match_param, spec->match_criteria, outer_headers.first_vid);
1371         MLX5_SET(fte_match_param, spec->match_value, outer_headers.first_vid, vport->info.vlan);
1372 
1373         spec->match_criteria_enable = MLX5_MATCH_OUTER_HEADERS;
1374         flow_act.action = MLX5_FLOW_CONTEXT_ACTION_ALLOW;
1375         vport->egress.allowed_vlan =
1376                 mlx5_add_flow_rules(vport->egress.acl, spec,
1377                                     &flow_act, NULL, 0);
1378         if (IS_ERR(vport->egress.allowed_vlan)) {
1379                 err = PTR_ERR(vport->egress.allowed_vlan);
1380                 esw_warn(esw->dev,
1381                          "vport[%d] configure egress allowed vlan rule failed, err(%d)\n",
1382                          vport->vport, err);
1383                 vport->egress.allowed_vlan = NULL;
1384                 goto out;
1385         }
1386 
1387         /* Drop others rule (star rule) */
1388         memset(spec, 0, sizeof(*spec));
1389         flow_act.action = MLX5_FLOW_CONTEXT_ACTION_DROP;
1390 
1391         /* Attach egress drop flow counter */
1392         if (counter) {
1393                 flow_act.action |= MLX5_FLOW_CONTEXT_ACTION_COUNT;
1394                 drop_ctr_dst.type = MLX5_FLOW_DESTINATION_TYPE_COUNTER;
1395                 drop_ctr_dst.counter_id = mlx5_fc_id(counter);
1396                 dst = &drop_ctr_dst;
1397                 dest_num++;
1398         }
1399         vport->egress.drop_rule =
1400                 mlx5_add_flow_rules(vport->egress.acl, spec,
1401                                     &flow_act, dst, dest_num);
1402         if (IS_ERR(vport->egress.drop_rule)) {
1403                 err = PTR_ERR(vport->egress.drop_rule);
1404                 esw_warn(esw->dev,
1405                          "vport[%d] configure egress drop rule failed, err(%d)\n",
1406                          vport->vport, err);
1407                 vport->egress.drop_rule = NULL;
1408         }
1409 out:
1410         kvfree(spec);
1411         return err;
1412 }
1413 
1414 static bool element_type_supported(struct mlx5_eswitch *esw, int type)
1415 {
1416         const struct mlx5_core_dev *dev = esw->dev;
1417 
1418         switch (type) {
1419         case SCHEDULING_CONTEXT_ELEMENT_TYPE_TSAR:
1420                 return MLX5_CAP_QOS(dev, esw_element_type) &
1421                        ELEMENT_TYPE_CAP_MASK_TASR;
1422         case SCHEDULING_CONTEXT_ELEMENT_TYPE_VPORT:
1423                 return MLX5_CAP_QOS(dev, esw_element_type) &
1424                        ELEMENT_TYPE_CAP_MASK_VPORT;
1425         case SCHEDULING_CONTEXT_ELEMENT_TYPE_VPORT_TC:
1426                 return MLX5_CAP_QOS(dev, esw_element_type) &
1427                        ELEMENT_TYPE_CAP_MASK_VPORT_TC;
1428         case SCHEDULING_CONTEXT_ELEMENT_TYPE_PARA_VPORT_TC:
1429                 return MLX5_CAP_QOS(dev, esw_element_type) &
1430                        ELEMENT_TYPE_CAP_MASK_PARA_VPORT_TC;
1431         }
1432         return false;
1433 }
1434 
1435 /* Vport QoS management */
1436 static void esw_create_tsar(struct mlx5_eswitch *esw)
1437 {
1438         u32 tsar_ctx[MLX5_ST_SZ_DW(scheduling_context)] = {0};
1439         struct mlx5_core_dev *dev = esw->dev;
1440         __be32 *attr;
1441         int err;
1442 
1443         if (!MLX5_CAP_GEN(dev, qos) || !MLX5_CAP_QOS(dev, esw_scheduling))
1444                 return;
1445 
1446         if (!element_type_supported(esw, SCHEDULING_CONTEXT_ELEMENT_TYPE_TSAR))
1447                 return;
1448 
1449         if (esw->qos.enabled)
1450                 return;
1451 
1452         MLX5_SET(scheduling_context, tsar_ctx, element_type,
1453                  SCHEDULING_CONTEXT_ELEMENT_TYPE_TSAR);
1454 
1455         attr = MLX5_ADDR_OF(scheduling_context, tsar_ctx, element_attributes);
1456         *attr = cpu_to_be32(TSAR_ELEMENT_TSAR_TYPE_DWRR << 16);
1457 
1458         err = mlx5_create_scheduling_element_cmd(dev,
1459                                                  SCHEDULING_HIERARCHY_E_SWITCH,
1460                                                  tsar_ctx,
1461                                                  &esw->qos.root_tsar_id);
1462         if (err) {
1463                 esw_warn(esw->dev, "E-Switch create TSAR failed (%d)\n", err);
1464                 return;
1465         }
1466 
1467         esw->qos.enabled = true;
1468 }
1469 
1470 static void esw_destroy_tsar(struct mlx5_eswitch *esw)
1471 {
1472         int err;
1473 
1474         if (!esw->qos.enabled)
1475                 return;
1476 
1477         err = mlx5_destroy_scheduling_element_cmd(esw->dev,
1478                                                   SCHEDULING_HIERARCHY_E_SWITCH,
1479                                                   esw->qos.root_tsar_id);
1480         if (err)
1481                 esw_warn(esw->dev, "E-Switch destroy TSAR failed (%d)\n", err);
1482 
1483         esw->qos.enabled = false;
1484 }
1485 
1486 static int esw_vport_enable_qos(struct mlx5_eswitch *esw,
1487                                 struct mlx5_vport *vport,
1488                                 u32 initial_max_rate, u32 initial_bw_share)
1489 {
1490         u32 sched_ctx[MLX5_ST_SZ_DW(scheduling_context)] = {0};
1491         struct mlx5_core_dev *dev = esw->dev;
1492         void *vport_elem;
1493         int err = 0;
1494 
1495         if (!esw->qos.enabled || !MLX5_CAP_GEN(dev, qos) ||
1496             !MLX5_CAP_QOS(dev, esw_scheduling))
1497                 return 0;
1498 
1499         if (vport->qos.enabled)
1500                 return -EEXIST;
1501 
1502         MLX5_SET(scheduling_context, sched_ctx, element_type,
1503                  SCHEDULING_CONTEXT_ELEMENT_TYPE_VPORT);
1504         vport_elem = MLX5_ADDR_OF(scheduling_context, sched_ctx,
1505                                   element_attributes);
1506         MLX5_SET(vport_element, vport_elem, vport_number, vport->vport);
1507         MLX5_SET(scheduling_context, sched_ctx, parent_element_id,
1508                  esw->qos.root_tsar_id);
1509         MLX5_SET(scheduling_context, sched_ctx, max_average_bw,
1510                  initial_max_rate);
1511         MLX5_SET(scheduling_context, sched_ctx, bw_share, initial_bw_share);
1512 
1513         err = mlx5_create_scheduling_element_cmd(dev,
1514                                                  SCHEDULING_HIERARCHY_E_SWITCH,
1515                                                  sched_ctx,
1516                                                  &vport->qos.esw_tsar_ix);
1517         if (err) {
1518                 esw_warn(esw->dev, "E-Switch create TSAR vport element failed (vport=%d,err=%d)\n",
1519                          vport->vport, err);
1520                 return err;
1521         }
1522 
1523         vport->qos.enabled = true;
1524         return 0;
1525 }
1526 
1527 static void esw_vport_disable_qos(struct mlx5_eswitch *esw,
1528                                   struct mlx5_vport *vport)
1529 {
1530         int err;
1531 
1532         if (!vport->qos.enabled)
1533                 return;
1534 
1535         err = mlx5_destroy_scheduling_element_cmd(esw->dev,
1536                                                   SCHEDULING_HIERARCHY_E_SWITCH,
1537                                                   vport->qos.esw_tsar_ix);
1538         if (err)
1539                 esw_warn(esw->dev, "E-Switch destroy TSAR vport element failed (vport=%d,err=%d)\n",
1540                          vport->vport, err);
1541 
1542         vport->qos.enabled = false;
1543 }
1544 
1545 static int esw_vport_qos_config(struct mlx5_eswitch *esw,
1546                                 struct mlx5_vport *vport,
1547                                 u32 max_rate, u32 bw_share)
1548 {
1549         u32 sched_ctx[MLX5_ST_SZ_DW(scheduling_context)] = {0};
1550         struct mlx5_core_dev *dev = esw->dev;
1551         void *vport_elem;
1552         u32 bitmask = 0;
1553         int err = 0;
1554 
1555         if (!MLX5_CAP_GEN(dev, qos) || !MLX5_CAP_QOS(dev, esw_scheduling))
1556                 return -EOPNOTSUPP;
1557 
1558         if (!vport->qos.enabled)
1559                 return -EIO;
1560 
1561         MLX5_SET(scheduling_context, sched_ctx, element_type,
1562                  SCHEDULING_CONTEXT_ELEMENT_TYPE_VPORT);
1563         vport_elem = MLX5_ADDR_OF(scheduling_context, sched_ctx,
1564                                   element_attributes);
1565         MLX5_SET(vport_element, vport_elem, vport_number, vport->vport);
1566         MLX5_SET(scheduling_context, sched_ctx, parent_element_id,
1567                  esw->qos.root_tsar_id);
1568         MLX5_SET(scheduling_context, sched_ctx, max_average_bw,
1569                  max_rate);
1570         MLX5_SET(scheduling_context, sched_ctx, bw_share, bw_share);
1571         bitmask |= MODIFY_SCHEDULING_ELEMENT_IN_MODIFY_BITMASK_MAX_AVERAGE_BW;
1572         bitmask |= MODIFY_SCHEDULING_ELEMENT_IN_MODIFY_BITMASK_BW_SHARE;
1573 
1574         err = mlx5_modify_scheduling_element_cmd(dev,
1575                                                  SCHEDULING_HIERARCHY_E_SWITCH,
1576                                                  sched_ctx,
1577                                                  vport->qos.esw_tsar_ix,
1578                                                  bitmask);
1579         if (err) {
1580                 esw_warn(esw->dev, "E-Switch modify TSAR vport element failed (vport=%d,err=%d)\n",
1581                          vport->vport, err);
1582                 return err;
1583         }
1584 
1585         return 0;
1586 }
1587 
1588 int mlx5_esw_modify_vport_rate(struct mlx5_eswitch *esw, u16 vport_num,
1589                                u32 rate_mbps)
1590 {
1591         u32 ctx[MLX5_ST_SZ_DW(scheduling_context)] = {};
1592         struct mlx5_vport *vport;
1593 
1594         vport = mlx5_eswitch_get_vport(esw, vport_num);
1595         MLX5_SET(scheduling_context, ctx, max_average_bw, rate_mbps);
1596 
1597         return mlx5_modify_scheduling_element_cmd(esw->dev,
1598                                                   SCHEDULING_HIERARCHY_E_SWITCH,
1599                                                   ctx,
1600                                                   vport->qos.esw_tsar_ix,
1601                                                   MODIFY_SCHEDULING_ELEMENT_IN_MODIFY_BITMASK_MAX_AVERAGE_BW);
1602 }
1603 
1604 static void node_guid_gen_from_mac(u64 *node_guid, u8 mac[ETH_ALEN])
1605 {
1606         ((u8 *)node_guid)[7] = mac[0];
1607         ((u8 *)node_guid)[6] = mac[1];
1608         ((u8 *)node_guid)[5] = mac[2];
1609         ((u8 *)node_guid)[4] = 0xff;
1610         ((u8 *)node_guid)[3] = 0xfe;
1611         ((u8 *)node_guid)[2] = mac[3];
1612         ((u8 *)node_guid)[1] = mac[4];
1613         ((u8 *)node_guid)[0] = mac[5];
1614 }
1615 
1616 static void esw_apply_vport_conf(struct mlx5_eswitch *esw,
1617                                  struct mlx5_vport *vport)
1618 {
1619         u16 vport_num = vport->vport;
1620         int flags;
1621 
1622         if (esw->manager_vport == vport_num)
1623                 return;
1624 
1625         mlx5_modify_vport_admin_state(esw->dev,
1626                                       MLX5_VPORT_STATE_OP_MOD_ESW_VPORT,
1627                                       vport_num, 1,
1628                                       vport->info.link_state);
1629 
1630         /* Host PF has its own mac/guid. */
1631         if (vport_num) {
1632                 mlx5_modify_nic_vport_mac_address(esw->dev, vport_num,
1633                                                   vport->info.mac);
1634                 mlx5_modify_nic_vport_node_guid(esw->dev, vport_num,
1635                                                 vport->info.node_guid);
1636         }
1637 
1638         flags = (vport->info.vlan || vport->info.qos) ?
1639                 SET_VLAN_STRIP | SET_VLAN_INSERT : 0;
1640         modify_esw_vport_cvlan(esw->dev, vport_num, vport->info.vlan, vport->info.qos,
1641                                flags);
1642 
1643         /* Only legacy mode needs ACLs */
1644         if (esw->mode == MLX5_ESWITCH_LEGACY) {
1645                 esw_vport_ingress_config(esw, vport);
1646                 esw_vport_egress_config(esw, vport);
1647         }
1648 }
1649 
1650 static void esw_vport_create_drop_counters(struct mlx5_vport *vport)
1651 {
1652         struct mlx5_core_dev *dev = vport->dev;
1653 
1654         if (MLX5_CAP_ESW_INGRESS_ACL(dev, flow_counter)) {
1655                 vport->ingress.drop_counter = mlx5_fc_create(dev, false);
1656                 if (IS_ERR(vport->ingress.drop_counter)) {
1657                         esw_warn(dev,
1658                                  "vport[%d] configure ingress drop rule counter failed\n",
1659                                  vport->vport);
1660                         vport->ingress.drop_counter = NULL;
1661                 }
1662         }
1663 
1664         if (MLX5_CAP_ESW_EGRESS_ACL(dev, flow_counter)) {
1665                 vport->egress.drop_counter = mlx5_fc_create(dev, false);
1666                 if (IS_ERR(vport->egress.drop_counter)) {
1667                         esw_warn(dev,
1668                                  "vport[%d] configure egress drop rule counter failed\n",
1669                                  vport->vport);
1670                         vport->egress.drop_counter = NULL;
1671                 }
1672         }
1673 }
1674 
1675 static void esw_vport_destroy_drop_counters(struct mlx5_vport *vport)
1676 {
1677         struct mlx5_core_dev *dev = vport->dev;
1678 
1679         if (vport->ingress.drop_counter)
1680                 mlx5_fc_destroy(dev, vport->ingress.drop_counter);
1681         if (vport->egress.drop_counter)
1682                 mlx5_fc_destroy(dev, vport->egress.drop_counter);
1683 }
1684 
1685 static void esw_enable_vport(struct mlx5_eswitch *esw, struct mlx5_vport *vport,
1686                              enum mlx5_eswitch_vport_event enabled_events)
1687 {
1688         u16 vport_num = vport->vport;
1689 
1690         mutex_lock(&esw->state_lock);
1691         WARN_ON(vport->enabled);
1692 
1693         esw_debug(esw->dev, "Enabling VPORT(%d)\n", vport_num);
1694 
1695         /* Create steering drop counters for ingress and egress ACLs */
1696         if (vport_num && esw->mode == MLX5_ESWITCH_LEGACY)
1697                 esw_vport_create_drop_counters(vport);
1698 
1699         /* Restore old vport configuration */
1700         esw_apply_vport_conf(esw, vport);
1701 
1702         /* Attach vport to the eswitch rate limiter */
1703         if (esw_vport_enable_qos(esw, vport, vport->info.max_rate,
1704                                  vport->qos.bw_share))
1705                 esw_warn(esw->dev, "Failed to attach vport %d to eswitch rate limiter", vport_num);
1706 
1707         /* Sync with current vport context */
1708         vport->enabled_events = enabled_events;
1709         vport->enabled = true;
1710 
1711         /* Esw manager is trusted by default. Host PF (vport 0) is trusted as well
1712          * in smartNIC as it's a vport group manager.
1713          */
1714         if (esw->manager_vport == vport_num ||
1715             (!vport_num && mlx5_core_is_ecpf(esw->dev)))
1716                 vport->info.trusted = true;
1717 
1718         esw_vport_change_handle_locked(vport);
1719 
1720         esw->enabled_vports++;
1721         esw_debug(esw->dev, "Enabled VPORT(%d)\n", vport_num);
1722         mutex_unlock(&esw->state_lock);
1723 }
1724 
1725 static void esw_disable_vport(struct mlx5_eswitch *esw,
1726                               struct mlx5_vport *vport)
1727 {
1728         u16 vport_num = vport->vport;
1729 
1730         if (!vport->enabled)
1731                 return;
1732 
1733         esw_debug(esw->dev, "Disabling vport(%d)\n", vport_num);
1734         /* Mark this vport as disabled to discard new events */
1735         vport->enabled = false;
1736 
1737         /* Wait for current already scheduled events to complete */
1738         flush_workqueue(esw->work_queue);
1739         /* Disable events from this vport */
1740         arm_vport_context_events_cmd(esw->dev, vport->vport, 0);
1741         mutex_lock(&esw->state_lock);
1742         /* We don't assume VFs will cleanup after themselves.
1743          * Calling vport change handler while vport is disabled will cleanup
1744          * the vport resources.
1745          */
1746         esw_vport_change_handle_locked(vport);
1747         vport->enabled_events = 0;
1748         esw_vport_disable_qos(esw, vport);
1749         if (esw->manager_vport != vport_num &&
1750             esw->mode == MLX5_ESWITCH_LEGACY) {
1751                 mlx5_modify_vport_admin_state(esw->dev,
1752                                               MLX5_VPORT_STATE_OP_MOD_ESW_VPORT,
1753                                               vport_num, 1,
1754                                               MLX5_VPORT_ADMIN_STATE_DOWN);
1755                 esw_vport_disable_egress_acl(esw, vport);
1756                 esw_vport_disable_ingress_acl(esw, vport);
1757                 esw_vport_destroy_drop_counters(vport);
1758         }
1759         esw->enabled_vports--;
1760         mutex_unlock(&esw->state_lock);
1761 }
1762 
1763 static int eswitch_vport_event(struct notifier_block *nb,
1764                                unsigned long type, void *data)
1765 {
1766         struct mlx5_eswitch *esw = mlx5_nb_cof(nb, struct mlx5_eswitch, nb);
1767         struct mlx5_eqe *eqe = data;
1768         struct mlx5_vport *vport;
1769         u16 vport_num;
1770 
1771         vport_num = be16_to_cpu(eqe->data.vport_change.vport_num);
1772         vport = mlx5_eswitch_get_vport(esw, vport_num);
1773         if (IS_ERR(vport))
1774                 return NOTIFY_OK;
1775 
1776         if (vport->enabled)
1777                 queue_work(esw->work_queue, &vport->vport_change_handler);
1778 
1779         return NOTIFY_OK;
1780 }
1781 
1782 /**
1783  * mlx5_esw_query_functions - Returns raw output about functions state
1784  * @dev:        Pointer to device to query
1785  *
1786  * mlx5_esw_query_functions() allocates and returns functions changed
1787  * raw output memory pointer from device on success. Otherwise returns ERR_PTR.
1788  * Caller must free the memory using kvfree() when valid pointer is returned.
1789  */
1790 const u32 *mlx5_esw_query_functions(struct mlx5_core_dev *dev)
1791 {
1792         int outlen = MLX5_ST_SZ_BYTES(query_esw_functions_out);
1793         u32 in[MLX5_ST_SZ_DW(query_esw_functions_in)] = {};
1794         u32 *out;
1795         int err;
1796 
1797         out = kvzalloc(outlen, GFP_KERNEL);
1798         if (!out)
1799                 return ERR_PTR(-ENOMEM);
1800 
1801         MLX5_SET(query_esw_functions_in, in, opcode,
1802                  MLX5_CMD_OP_QUERY_ESW_FUNCTIONS);
1803 
1804         err = mlx5_cmd_exec(dev, in, sizeof(in), out, outlen);
1805         if (!err)
1806                 return out;
1807 
1808         kvfree(out);
1809         return ERR_PTR(err);
1810 }
1811 
1812 static void mlx5_eswitch_event_handlers_register(struct mlx5_eswitch *esw)
1813 {
1814         MLX5_NB_INIT(&esw->nb, eswitch_vport_event, NIC_VPORT_CHANGE);
1815         mlx5_eq_notifier_register(esw->dev, &esw->nb);
1816 
1817         if (esw->mode == MLX5_ESWITCH_OFFLOADS && mlx5_eswitch_is_funcs_handler(esw->dev)) {
1818                 MLX5_NB_INIT(&esw->esw_funcs.nb, mlx5_esw_funcs_changed_handler,
1819                              ESW_FUNCTIONS_CHANGED);
1820                 mlx5_eq_notifier_register(esw->dev, &esw->esw_funcs.nb);
1821         }
1822 }
1823 
1824 static void mlx5_eswitch_event_handlers_unregister(struct mlx5_eswitch *esw)
1825 {
1826         if (esw->mode == MLX5_ESWITCH_OFFLOADS && mlx5_eswitch_is_funcs_handler(esw->dev))
1827                 mlx5_eq_notifier_unregister(esw->dev, &esw->esw_funcs.nb);
1828 
1829         mlx5_eq_notifier_unregister(esw->dev, &esw->nb);
1830 
1831         flush_workqueue(esw->work_queue);
1832 }
1833 
1834 /* Public E-Switch API */
1835 #define ESW_ALLOWED(esw) ((esw) && MLX5_ESWITCH_MANAGER((esw)->dev))
1836 
1837 /* mlx5_eswitch_enable_pf_vf_vports() enables vports of PF, ECPF and VFs
1838  * whichever are present on the eswitch.
1839  */
1840 void
1841 mlx5_eswitch_enable_pf_vf_vports(struct mlx5_eswitch *esw,
1842                                  enum mlx5_eswitch_vport_event enabled_events)
1843 {
1844         struct mlx5_vport *vport;
1845         int i;
1846 
1847         /* Enable PF vport */
1848         vport = mlx5_eswitch_get_vport(esw, MLX5_VPORT_PF);
1849         esw_enable_vport(esw, vport, enabled_events);
1850 
1851         /* Enable ECPF vports */
1852         if (mlx5_ecpf_vport_exists(esw->dev)) {
1853                 vport = mlx5_eswitch_get_vport(esw, MLX5_VPORT_ECPF);
1854                 esw_enable_vport(esw, vport, enabled_events);
1855         }
1856 
1857         /* Enable VF vports */
1858         mlx5_esw_for_each_vf_vport(esw, i, vport, esw->esw_funcs.num_vfs)
1859                 esw_enable_vport(esw, vport, enabled_events);
1860 }
1861 
1862 /* mlx5_eswitch_disable_pf_vf_vports() disables vports of PF, ECPF and VFs
1863  * whichever are previously enabled on the eswitch.
1864  */
1865 void mlx5_eswitch_disable_pf_vf_vports(struct mlx5_eswitch *esw)
1866 {
1867         struct mlx5_vport *vport;
1868         int i;
1869 
1870         mlx5_esw_for_all_vports_reverse(esw, i, vport)
1871                 esw_disable_vport(esw, vport);
1872 }
1873 
1874 int mlx5_eswitch_enable(struct mlx5_eswitch *esw, int mode)
1875 {
1876         int err;
1877 
1878         if (!ESW_ALLOWED(esw) ||
1879             !MLX5_CAP_ESW_FLOWTABLE_FDB(esw->dev, ft_support)) {
1880                 esw_warn(esw->dev, "FDB is not supported, aborting ...\n");
1881                 return -EOPNOTSUPP;
1882         }
1883 
1884         if (!MLX5_CAP_ESW_INGRESS_ACL(esw->dev, ft_support))
1885                 esw_warn(esw->dev, "ingress ACL is not supported by FW\n");
1886 
1887         if (!MLX5_CAP_ESW_EGRESS_ACL(esw->dev, ft_support))
1888                 esw_warn(esw->dev, "engress ACL is not supported by FW\n");
1889 
1890         esw_create_tsar(esw);
1891 
1892         esw->mode = mode;
1893 
1894         mlx5_lag_update(esw->dev);
1895 
1896         if (mode == MLX5_ESWITCH_LEGACY) {
1897                 err = esw_legacy_enable(esw);
1898         } else {
1899                 mlx5_reload_interface(esw->dev, MLX5_INTERFACE_PROTOCOL_ETH);
1900                 mlx5_reload_interface(esw->dev, MLX5_INTERFACE_PROTOCOL_IB);
1901                 err = esw_offloads_enable(esw);
1902         }
1903 
1904         if (err)
1905                 goto abort;
1906 
1907         mlx5_eswitch_event_handlers_register(esw);
1908 
1909         esw_info(esw->dev, "Enable: mode(%s), nvfs(%d), active vports(%d)\n",
1910                  mode == MLX5_ESWITCH_LEGACY ? "LEGACY" : "OFFLOADS",
1911                  esw->esw_funcs.num_vfs, esw->enabled_vports);
1912 
1913         return 0;
1914 
1915 abort:
1916         esw->mode = MLX5_ESWITCH_NONE;
1917 
1918         if (mode == MLX5_ESWITCH_OFFLOADS) {
1919                 mlx5_reload_interface(esw->dev, MLX5_INTERFACE_PROTOCOL_IB);
1920                 mlx5_reload_interface(esw->dev, MLX5_INTERFACE_PROTOCOL_ETH);
1921         }
1922 
1923         return err;
1924 }
1925 
1926 void mlx5_eswitch_disable(struct mlx5_eswitch *esw)
1927 {
1928         int old_mode;
1929 
1930         if (!ESW_ALLOWED(esw) || esw->mode == MLX5_ESWITCH_NONE)
1931                 return;
1932 
1933         esw_info(esw->dev, "Disable: mode(%s), nvfs(%d), active vports(%d)\n",
1934                  esw->mode == MLX5_ESWITCH_LEGACY ? "LEGACY" : "OFFLOADS",
1935                  esw->esw_funcs.num_vfs, esw->enabled_vports);
1936 
1937         mlx5_eswitch_event_handlers_unregister(esw);
1938 
1939         if (esw->mode == MLX5_ESWITCH_LEGACY)
1940                 esw_legacy_disable(esw);
1941         else if (esw->mode == MLX5_ESWITCH_OFFLOADS)
1942                 esw_offloads_disable(esw);
1943 
1944         esw_destroy_tsar(esw);
1945 
1946         old_mode = esw->mode;
1947         esw->mode = MLX5_ESWITCH_NONE;
1948 
1949         mlx5_lag_update(esw->dev);
1950 
1951         if (old_mode == MLX5_ESWITCH_OFFLOADS) {
1952                 mlx5_reload_interface(esw->dev, MLX5_INTERFACE_PROTOCOL_IB);
1953                 mlx5_reload_interface(esw->dev, MLX5_INTERFACE_PROTOCOL_ETH);
1954         }
1955 }
1956 
1957 int mlx5_eswitch_init(struct mlx5_core_dev *dev)
1958 {
1959         struct mlx5_eswitch *esw;
1960         struct mlx5_vport *vport;
1961         int total_vports;
1962         int err, i;
1963 
1964         if (!MLX5_VPORT_MANAGER(dev))
1965                 return 0;
1966 
1967         total_vports = mlx5_eswitch_get_total_vports(dev);
1968 
1969         esw_info(dev,
1970                  "Total vports %d, per vport: max uc(%d) max mc(%d)\n",
1971                  total_vports,
1972                  MLX5_MAX_UC_PER_VPORT(dev),
1973                  MLX5_MAX_MC_PER_VPORT(dev));
1974 
1975         esw = kzalloc(sizeof(*esw), GFP_KERNEL);
1976         if (!esw)
1977                 return -ENOMEM;
1978 
1979         esw->dev = dev;
1980         esw->manager_vport = mlx5_eswitch_manager_vport(dev);
1981         esw->first_host_vport = mlx5_eswitch_first_host_vport_num(dev);
1982 
1983         esw->work_queue = create_singlethread_workqueue("mlx5_esw_wq");
1984         if (!esw->work_queue) {
1985                 err = -ENOMEM;
1986                 goto abort;
1987         }
1988 
1989         esw->vports = kcalloc(total_vports, sizeof(struct mlx5_vport),
1990                               GFP_KERNEL);
1991         if (!esw->vports) {
1992                 err = -ENOMEM;
1993                 goto abort;
1994         }
1995 
1996         esw->total_vports = total_vports;
1997 
1998         err = esw_offloads_init_reps(esw);
1999         if (err)
2000                 goto abort;
2001 
2002         mutex_init(&esw->offloads.encap_tbl_lock);
2003         hash_init(esw->offloads.encap_tbl);
2004         mutex_init(&esw->offloads.mod_hdr.lock);
2005         hash_init(esw->offloads.mod_hdr.hlist);
2006         atomic64_set(&esw->offloads.num_flows, 0);
2007         mutex_init(&esw->state_lock);
2008 
2009         mlx5_esw_for_all_vports(esw, i, vport) {
2010                 vport->vport = mlx5_eswitch_index_to_vport_num(esw, i);
2011                 vport->info.link_state = MLX5_VPORT_ADMIN_STATE_AUTO;
2012                 vport->dev = dev;
2013                 INIT_WORK(&vport->vport_change_handler,
2014                           esw_vport_change_handler);
2015         }
2016 
2017         esw->enabled_vports = 0;
2018         esw->mode = MLX5_ESWITCH_NONE;
2019         esw->offloads.inline_mode = MLX5_INLINE_MODE_NONE;
2020 
2021         dev->priv.eswitch = esw;
2022         return 0;
2023 abort:
2024         if (esw->work_queue)
2025                 destroy_workqueue(esw->work_queue);
2026         esw_offloads_cleanup_reps(esw);
2027         kfree(esw->vports);
2028         kfree(esw);
2029         return err;
2030 }
2031 
2032 void mlx5_eswitch_cleanup(struct mlx5_eswitch *esw)
2033 {
2034         if (!esw || !MLX5_VPORT_MANAGER(esw->dev))
2035                 return;
2036 
2037         esw_info(esw->dev, "cleanup\n");
2038 
2039         esw->dev->priv.eswitch = NULL;
2040         destroy_workqueue(esw->work_queue);
2041         esw_offloads_cleanup_reps(esw);
2042         mutex_destroy(&esw->offloads.mod_hdr.lock);
2043         mutex_destroy(&esw->offloads.encap_tbl_lock);
2044         kfree(esw->vports);
2045         kfree(esw);
2046 }
2047 
2048 /* Vport Administration */
2049 int mlx5_eswitch_set_vport_mac(struct mlx5_eswitch *esw,
2050                                u16 vport, u8 mac[ETH_ALEN])
2051 {
2052         struct mlx5_vport *evport = mlx5_eswitch_get_vport(esw, vport);
2053         u64 node_guid;
2054         int err = 0;
2055 
2056         if (IS_ERR(evport))
2057                 return PTR_ERR(evport);
2058         if (is_multicast_ether_addr(mac))
2059                 return -EINVAL;
2060 
2061         mutex_lock(&esw->state_lock);
2062 
2063         if (evport->info.spoofchk && !is_valid_ether_addr(mac))
2064                 mlx5_core_warn(esw->dev,
2065                                "Set invalid MAC while spoofchk is on, vport(%d)\n",
2066                                vport);
2067 
2068         err = mlx5_modify_nic_vport_mac_address(esw->dev, vport, mac);
2069         if (err) {
2070                 mlx5_core_warn(esw->dev,
2071                                "Failed to mlx5_modify_nic_vport_mac vport(%d) err=(%d)\n",
2072                                vport, err);
2073                 goto unlock;
2074         }
2075 
2076         node_guid_gen_from_mac(&node_guid, mac);
2077         err = mlx5_modify_nic_vport_node_guid(esw->dev, vport, node_guid);
2078         if (err)
2079                 mlx5_core_warn(esw->dev,
2080                                "Failed to set vport %d node guid, err = %d. RDMA_CM will not function properly for this VF.\n",
2081                                vport, err);
2082 
2083         ether_addr_copy(evport->info.mac, mac);
2084         evport->info.node_guid = node_guid;
2085         if (evport->enabled && esw->mode == MLX5_ESWITCH_LEGACY)
2086                 err = esw_vport_ingress_config(esw, evport);
2087 
2088 unlock:
2089         mutex_unlock(&esw->state_lock);
2090         return err;
2091 }
2092 
2093 int mlx5_eswitch_set_vport_state(struct mlx5_eswitch *esw,
2094                                  u16 vport, int link_state)
2095 {
2096         struct mlx5_vport *evport = mlx5_eswitch_get_vport(esw, vport);
2097         int err = 0;
2098 
2099         if (!ESW_ALLOWED(esw))
2100                 return -EPERM;
2101         if (IS_ERR(evport))
2102                 return PTR_ERR(evport);
2103 
2104         mutex_lock(&esw->state_lock);
2105 
2106         err = mlx5_modify_vport_admin_state(esw->dev,
2107                                             MLX5_VPORT_STATE_OP_MOD_ESW_VPORT,
2108                                             vport, 1, link_state);
2109         if (err) {
2110                 mlx5_core_warn(esw->dev,
2111                                "Failed to set vport %d link state, err = %d",
2112                                vport, err);
2113                 goto unlock;
2114         }
2115 
2116         evport->info.link_state = link_state;
2117 
2118 unlock:
2119         mutex_unlock(&esw->state_lock);
2120         return err;
2121 }
2122 
2123 int mlx5_eswitch_get_vport_config(struct mlx5_eswitch *esw,
2124                                   u16 vport, struct ifla_vf_info *ivi)
2125 {
2126         struct mlx5_vport *evport = mlx5_eswitch_get_vport(esw, vport);
2127 
2128         if (IS_ERR(evport))
2129                 return PTR_ERR(evport);
2130 
2131         memset(ivi, 0, sizeof(*ivi));
2132         ivi->vf = vport - 1;
2133 
2134         mutex_lock(&esw->state_lock);
2135         ether_addr_copy(ivi->mac, evport->info.mac);
2136         ivi->linkstate = evport->info.link_state;
2137         ivi->vlan = evport->info.vlan;
2138         ivi->qos = evport->info.qos;
2139         ivi->spoofchk = evport->info.spoofchk;
2140         ivi->trusted = evport->info.trusted;
2141         ivi->min_tx_rate = evport->info.min_rate;
2142         ivi->max_tx_rate = evport->info.max_rate;
2143         mutex_unlock(&esw->state_lock);
2144 
2145         return 0;
2146 }
2147 
2148 int __mlx5_eswitch_set_vport_vlan(struct mlx5_eswitch *esw,
2149                                   u16 vport, u16 vlan, u8 qos, u8 set_flags)
2150 {
2151         struct mlx5_vport *evport = mlx5_eswitch_get_vport(esw, vport);
2152         int err = 0;
2153 
2154         if (!ESW_ALLOWED(esw))
2155                 return -EPERM;
2156         if (IS_ERR(evport))
2157                 return PTR_ERR(evport);
2158         if (vlan > 4095 || qos > 7)
2159                 return -EINVAL;
2160 
2161         err = modify_esw_vport_cvlan(esw->dev, vport, vlan, qos, set_flags);
2162         if (err)
2163                 return err;
2164 
2165         evport->info.vlan = vlan;
2166         evport->info.qos = qos;
2167         if (evport->enabled && esw->mode == MLX5_ESWITCH_LEGACY) {
2168                 err = esw_vport_ingress_config(esw, evport);
2169                 if (err)
2170                         return err;
2171                 err = esw_vport_egress_config(esw, evport);
2172         }
2173 
2174         return err;
2175 }
2176 
2177 int mlx5_eswitch_set_vport_vlan(struct mlx5_eswitch *esw,
2178                                 u16 vport, u16 vlan, u8 qos)
2179 {
2180         u8 set_flags = 0;
2181         int err;
2182 
2183         if (vlan || qos)
2184                 set_flags = SET_VLAN_STRIP | SET_VLAN_INSERT;
2185 
2186         mutex_lock(&esw->state_lock);
2187         err = __mlx5_eswitch_set_vport_vlan(esw, vport, vlan, qos, set_flags);
2188         mutex_unlock(&esw->state_lock);
2189 
2190         return err;
2191 }
2192 
2193 int mlx5_eswitch_set_vport_spoofchk(struct mlx5_eswitch *esw,
2194                                     u16 vport, bool spoofchk)
2195 {
2196         struct mlx5_vport *evport = mlx5_eswitch_get_vport(esw, vport);
2197         bool pschk;
2198         int err = 0;
2199 
2200         if (!ESW_ALLOWED(esw))
2201                 return -EPERM;
2202         if (IS_ERR(evport))
2203                 return PTR_ERR(evport);
2204 
2205         mutex_lock(&esw->state_lock);
2206         pschk = evport->info.spoofchk;
2207         evport->info.spoofchk = spoofchk;
2208         if (pschk && !is_valid_ether_addr(evport->info.mac))
2209                 mlx5_core_warn(esw->dev,
2210                                "Spoofchk in set while MAC is invalid, vport(%d)\n",
2211                                evport->vport);
2212         if (evport->enabled && esw->mode == MLX5_ESWITCH_LEGACY)
2213                 err = esw_vport_ingress_config(esw, evport);
2214         if (err)
2215                 evport->info.spoofchk = pschk;
2216         mutex_unlock(&esw->state_lock);
2217 
2218         return err;
2219 }
2220 
2221 static void esw_cleanup_vepa_rules(struct mlx5_eswitch *esw)
2222 {
2223         if (esw->fdb_table.legacy.vepa_uplink_rule)
2224                 mlx5_del_flow_rules(esw->fdb_table.legacy.vepa_uplink_rule);
2225 
2226         if (esw->fdb_table.legacy.vepa_star_rule)
2227                 mlx5_del_flow_rules(esw->fdb_table.legacy.vepa_star_rule);
2228 
2229         esw->fdb_table.legacy.vepa_uplink_rule = NULL;
2230         esw->fdb_table.legacy.vepa_star_rule = NULL;
2231 }
2232 
2233 static int _mlx5_eswitch_set_vepa_locked(struct mlx5_eswitch *esw,
2234                                          u8 setting)
2235 {
2236         struct mlx5_flow_destination dest = {};
2237         struct mlx5_flow_act flow_act = {};
2238         struct mlx5_flow_handle *flow_rule;
2239         struct mlx5_flow_spec *spec;
2240         int err = 0;
2241         void *misc;
2242 
2243         if (!setting) {
2244                 esw_cleanup_vepa_rules(esw);
2245                 return 0;
2246         }
2247 
2248         if (esw->fdb_table.legacy.vepa_uplink_rule)
2249                 return 0;
2250 
2251         spec = kvzalloc(sizeof(*spec), GFP_KERNEL);
2252         if (!spec)
2253                 return -ENOMEM;
2254 
2255         /* Uplink rule forward uplink traffic to FDB */
2256         misc = MLX5_ADDR_OF(fte_match_param, spec->match_value, misc_parameters);
2257         MLX5_SET(fte_match_set_misc, misc, source_port, MLX5_VPORT_UPLINK);
2258 
2259         misc = MLX5_ADDR_OF(fte_match_param, spec->match_criteria, misc_parameters);
2260         MLX5_SET_TO_ONES(fte_match_set_misc, misc, source_port);
2261 
2262         spec->match_criteria_enable = MLX5_MATCH_MISC_PARAMETERS;
2263         dest.type = MLX5_FLOW_DESTINATION_TYPE_FLOW_TABLE;
2264         dest.ft = esw->fdb_table.legacy.fdb;
2265         flow_act.action = MLX5_FLOW_CONTEXT_ACTION_FWD_DEST;
2266         flow_rule = mlx5_add_flow_rules(esw->fdb_table.legacy.vepa_fdb, spec,
2267                                         &flow_act, &dest, 1);
2268         if (IS_ERR(flow_rule)) {
2269                 err = PTR_ERR(flow_rule);
2270                 goto out;
2271         } else {
2272                 esw->fdb_table.legacy.vepa_uplink_rule = flow_rule;
2273         }
2274 
2275         /* Star rule to forward all traffic to uplink vport */
2276         memset(spec, 0, sizeof(*spec));
2277         memset(&dest, 0, sizeof(dest));
2278         dest.type = MLX5_FLOW_DESTINATION_TYPE_VPORT;
2279         dest.vport.num = MLX5_VPORT_UPLINK;
2280         flow_act.action = MLX5_FLOW_CONTEXT_ACTION_FWD_DEST;
2281         flow_rule = mlx5_add_flow_rules(esw->fdb_table.legacy.vepa_fdb, spec,
2282                                         &flow_act, &dest, 1);
2283         if (IS_ERR(flow_rule)) {
2284                 err = PTR_ERR(flow_rule);
2285                 goto out;
2286         } else {
2287                 esw->fdb_table.legacy.vepa_star_rule = flow_rule;
2288         }
2289 
2290 out:
2291         kvfree(spec);
2292         if (err)
2293                 esw_cleanup_vepa_rules(esw);
2294         return err;
2295 }
2296 
2297 int mlx5_eswitch_set_vepa(struct mlx5_eswitch *esw, u8 setting)
2298 {
2299         int err = 0;
2300 
2301         if (!esw)
2302                 return -EOPNOTSUPP;
2303 
2304         if (!ESW_ALLOWED(esw))
2305                 return -EPERM;
2306 
2307         mutex_lock(&esw->state_lock);
2308         if (esw->mode != MLX5_ESWITCH_LEGACY) {
2309                 err = -EOPNOTSUPP;
2310                 goto out;
2311         }
2312 
2313         err = _mlx5_eswitch_set_vepa_locked(esw, setting);
2314 
2315 out:
2316         mutex_unlock(&esw->state_lock);
2317         return err;
2318 }
2319 
2320 int mlx5_eswitch_get_vepa(struct mlx5_eswitch *esw, u8 *setting)
2321 {
2322         if (!esw)
2323                 return -EOPNOTSUPP;
2324 
2325         if (!ESW_ALLOWED(esw))
2326                 return -EPERM;
2327 
2328         if (esw->mode != MLX5_ESWITCH_LEGACY)
2329                 return -EOPNOTSUPP;
2330 
2331         *setting = esw->fdb_table.legacy.vepa_uplink_rule ? 1 : 0;
2332         return 0;
2333 }
2334 
2335 int mlx5_eswitch_set_vport_trust(struct mlx5_eswitch *esw,
2336                                  u16 vport, bool setting)
2337 {
2338         struct mlx5_vport *evport = mlx5_eswitch_get_vport(esw, vport);
2339 
2340         if (!ESW_ALLOWED(esw))
2341                 return -EPERM;
2342         if (IS_ERR(evport))
2343                 return PTR_ERR(evport);
2344 
2345         mutex_lock(&esw->state_lock);
2346         evport->info.trusted = setting;
2347         if (evport->enabled)
2348                 esw_vport_change_handle_locked(evport);
2349         mutex_unlock(&esw->state_lock);
2350 
2351         return 0;
2352 }
2353 
2354 static u32 calculate_vports_min_rate_divider(struct mlx5_eswitch *esw)
2355 {
2356         u32 fw_max_bw_share = MLX5_CAP_QOS(esw->dev, max_tsar_bw_share);
2357         struct mlx5_vport *evport;
2358         u32 max_guarantee = 0;
2359         int i;
2360 
2361         mlx5_esw_for_all_vports(esw, i, evport) {
2362                 if (!evport->enabled || evport->info.min_rate < max_guarantee)
2363                         continue;
2364                 max_guarantee = evport->info.min_rate;
2365         }
2366 
2367         return max_t(u32, max_guarantee / fw_max_bw_share, 1);
2368 }
2369 
2370 static int normalize_vports_min_rate(struct mlx5_eswitch *esw, u32 divider)
2371 {
2372         u32 fw_max_bw_share = MLX5_CAP_QOS(esw->dev, max_tsar_bw_share);
2373         struct mlx5_vport *evport;
2374         u32 vport_max_rate;
2375         u32 vport_min_rate;
2376         u32 bw_share;
2377         int err;
2378         int i;
2379 
2380         mlx5_esw_for_all_vports(esw, i, evport) {
2381                 if (!evport->enabled)
2382                         continue;
2383                 vport_min_rate = evport->info.min_rate;
2384                 vport_max_rate = evport->info.max_rate;
2385                 bw_share = MLX5_MIN_BW_SHARE;
2386 
2387                 if (vport_min_rate)
2388                         bw_share = MLX5_RATE_TO_BW_SHARE(vport_min_rate,
2389                                                          divider,
2390                                                          fw_max_bw_share);
2391 
2392                 if (bw_share == evport->qos.bw_share)
2393                         continue;
2394 
2395                 err = esw_vport_qos_config(esw, evport, vport_max_rate,
2396                                            bw_share);
2397                 if (!err)
2398                         evport->qos.bw_share = bw_share;
2399                 else
2400                         return err;
2401         }
2402 
2403         return 0;
2404 }
2405 
2406 int mlx5_eswitch_set_vport_rate(struct mlx5_eswitch *esw, u16 vport,
2407                                 u32 max_rate, u32 min_rate)
2408 {
2409         struct mlx5_vport *evport = mlx5_eswitch_get_vport(esw, vport);
2410         u32 fw_max_bw_share;
2411         u32 previous_min_rate;
2412         u32 divider;
2413         bool min_rate_supported;
2414         bool max_rate_supported;
2415         int err = 0;
2416 
2417         if (!ESW_ALLOWED(esw))
2418                 return -EPERM;
2419         if (IS_ERR(evport))
2420                 return PTR_ERR(evport);
2421 
2422         fw_max_bw_share = MLX5_CAP_QOS(esw->dev, max_tsar_bw_share);
2423         min_rate_supported = MLX5_CAP_QOS(esw->dev, esw_bw_share) &&
2424                                 fw_max_bw_share >= MLX5_MIN_BW_SHARE;
2425         max_rate_supported = MLX5_CAP_QOS(esw->dev, esw_rate_limit);
2426 
2427         if ((min_rate && !min_rate_supported) || (max_rate && !max_rate_supported))
2428                 return -EOPNOTSUPP;
2429 
2430         mutex_lock(&esw->state_lock);
2431 
2432         if (min_rate == evport->info.min_rate)
2433                 goto set_max_rate;
2434 
2435         previous_min_rate = evport->info.min_rate;
2436         evport->info.min_rate = min_rate;
2437         divider = calculate_vports_min_rate_divider(esw);
2438         err = normalize_vports_min_rate(esw, divider);
2439         if (err) {
2440                 evport->info.min_rate = previous_min_rate;
2441                 goto unlock;
2442         }
2443 
2444 set_max_rate:
2445         if (max_rate == evport->info.max_rate)
2446                 goto unlock;
2447 
2448         err = esw_vport_qos_config(esw, evport, max_rate, evport->qos.bw_share);
2449         if (!err)
2450                 evport->info.max_rate = max_rate;
2451 
2452 unlock:
2453         mutex_unlock(&esw->state_lock);
2454         return err;
2455 }
2456 
2457 static int mlx5_eswitch_query_vport_drop_stats(struct mlx5_core_dev *dev,
2458                                                struct mlx5_vport *vport,
2459                                                struct mlx5_vport_drop_stats *stats)
2460 {
2461         struct mlx5_eswitch *esw = dev->priv.eswitch;
2462         u64 rx_discard_vport_down, tx_discard_vport_down;
2463         u64 bytes = 0;
2464         int err = 0;
2465 
2466         if (!vport->enabled || esw->mode != MLX5_ESWITCH_LEGACY)
2467                 return 0;
2468 
2469         if (vport->egress.drop_counter)
2470                 mlx5_fc_query(dev, vport->egress.drop_counter,
2471                               &stats->rx_dropped, &bytes);
2472 
2473         if (vport->ingress.drop_counter)
2474                 mlx5_fc_query(dev, vport->ingress.drop_counter,
2475                               &stats->tx_dropped, &bytes);
2476 
2477         if (!MLX5_CAP_GEN(dev, receive_discard_vport_down) &&
2478             !MLX5_CAP_GEN(dev, transmit_discard_vport_down))
2479                 return 0;
2480 
2481         err = mlx5_query_vport_down_stats(dev, vport->vport, 1,
2482                                           &rx_discard_vport_down,
2483                                           &tx_discard_vport_down);
2484         if (err)
2485                 return err;
2486 
2487         if (MLX5_CAP_GEN(dev, receive_discard_vport_down))
2488                 stats->rx_dropped += rx_discard_vport_down;
2489         if (MLX5_CAP_GEN(dev, transmit_discard_vport_down))
2490                 stats->tx_dropped += tx_discard_vport_down;
2491 
2492         return 0;
2493 }
2494 
2495 int mlx5_eswitch_get_vport_stats(struct mlx5_eswitch *esw,
2496                                  u16 vport_num,
2497                                  struct ifla_vf_stats *vf_stats)
2498 {
2499         struct mlx5_vport *vport = mlx5_eswitch_get_vport(esw, vport_num);
2500         int outlen = MLX5_ST_SZ_BYTES(query_vport_counter_out);
2501         u32 in[MLX5_ST_SZ_DW(query_vport_counter_in)] = {0};
2502         struct mlx5_vport_drop_stats stats = {0};
2503         int err = 0;
2504         u32 *out;
2505 
2506         if (IS_ERR(vport))
2507                 return PTR_ERR(vport);
2508 
2509         out = kvzalloc(outlen, GFP_KERNEL);
2510         if (!out)
2511                 return -ENOMEM;
2512 
2513         MLX5_SET(query_vport_counter_in, in, opcode,
2514                  MLX5_CMD_OP_QUERY_VPORT_COUNTER);
2515         MLX5_SET(query_vport_counter_in, in, op_mod, 0);
2516         MLX5_SET(query_vport_counter_in, in, vport_number, vport->vport);
2517         MLX5_SET(query_vport_counter_in, in, other_vport, 1);
2518 
2519         err = mlx5_cmd_exec(esw->dev, in, sizeof(in), out, outlen);
2520         if (err)
2521                 goto free_out;
2522 
2523         #define MLX5_GET_CTR(p, x) \
2524                 MLX5_GET64(query_vport_counter_out, p, x)
2525 
2526         memset(vf_stats, 0, sizeof(*vf_stats));
2527         vf_stats->rx_packets =
2528                 MLX5_GET_CTR(out, received_eth_unicast.packets) +
2529                 MLX5_GET_CTR(out, received_ib_unicast.packets) +
2530                 MLX5_GET_CTR(out, received_eth_multicast.packets) +
2531                 MLX5_GET_CTR(out, received_ib_multicast.packets) +
2532                 MLX5_GET_CTR(out, received_eth_broadcast.packets);
2533 
2534         vf_stats->rx_bytes =
2535                 MLX5_GET_CTR(out, received_eth_unicast.octets) +
2536                 MLX5_GET_CTR(out, received_ib_unicast.octets) +
2537                 MLX5_GET_CTR(out, received_eth_multicast.octets) +
2538                 MLX5_GET_CTR(out, received_ib_multicast.octets) +
2539                 MLX5_GET_CTR(out, received_eth_broadcast.octets);
2540 
2541         vf_stats->tx_packets =
2542                 MLX5_GET_CTR(out, transmitted_eth_unicast.packets) +
2543                 MLX5_GET_CTR(out, transmitted_ib_unicast.packets) +
2544                 MLX5_GET_CTR(out, transmitted_eth_multicast.packets) +
2545                 MLX5_GET_CTR(out, transmitted_ib_multicast.packets) +
2546                 MLX5_GET_CTR(out, transmitted_eth_broadcast.packets);
2547 
2548         vf_stats->tx_bytes =
2549                 MLX5_GET_CTR(out, transmitted_eth_unicast.octets) +
2550                 MLX5_GET_CTR(out, transmitted_ib_unicast.octets) +
2551                 MLX5_GET_CTR(out, transmitted_eth_multicast.octets) +
2552                 MLX5_GET_CTR(out, transmitted_ib_multicast.octets) +
2553                 MLX5_GET_CTR(out, transmitted_eth_broadcast.octets);
2554 
2555         vf_stats->multicast =
2556                 MLX5_GET_CTR(out, received_eth_multicast.packets) +
2557                 MLX5_GET_CTR(out, received_ib_multicast.packets);
2558 
2559         vf_stats->broadcast =
2560                 MLX5_GET_CTR(out, received_eth_broadcast.packets);
2561 
2562         err = mlx5_eswitch_query_vport_drop_stats(esw->dev, vport, &stats);
2563         if (err)
2564                 goto free_out;
2565         vf_stats->rx_dropped = stats.rx_dropped;
2566         vf_stats->tx_dropped = stats.tx_dropped;
2567 
2568 free_out:
2569         kvfree(out);
2570         return err;
2571 }
2572 
2573 u8 mlx5_eswitch_mode(struct mlx5_eswitch *esw)
2574 {
2575         return ESW_ALLOWED(esw) ? esw->mode : MLX5_ESWITCH_NONE;
2576 }
2577 EXPORT_SYMBOL_GPL(mlx5_eswitch_mode);
2578 
2579 enum devlink_eswitch_encap_mode
2580 mlx5_eswitch_get_encap_mode(const struct mlx5_core_dev *dev)
2581 {
2582         struct mlx5_eswitch *esw;
2583 
2584         esw = dev->priv.eswitch;
2585         return ESW_ALLOWED(esw) ? esw->offloads.encap :
2586                 DEVLINK_ESWITCH_ENCAP_MODE_NONE;
2587 }
2588 EXPORT_SYMBOL(mlx5_eswitch_get_encap_mode);
2589 
2590 bool mlx5_esw_lag_prereq(struct mlx5_core_dev *dev0, struct mlx5_core_dev *dev1)
2591 {
2592         if ((dev0->priv.eswitch->mode == MLX5_ESWITCH_NONE &&
2593              dev1->priv.eswitch->mode == MLX5_ESWITCH_NONE) ||
2594             (dev0->priv.eswitch->mode == MLX5_ESWITCH_OFFLOADS &&
2595              dev1->priv.eswitch->mode == MLX5_ESWITCH_OFFLOADS))
2596                 return true;
2597 
2598         return false;
2599 }
2600 
2601 bool mlx5_esw_multipath_prereq(struct mlx5_core_dev *dev0,
2602                                struct mlx5_core_dev *dev1)
2603 {
2604         return (dev0->priv.eswitch->mode == MLX5_ESWITCH_OFFLOADS &&
2605                 dev1->priv.eswitch->mode == MLX5_ESWITCH_OFFLOADS);
2606 }
2607 
2608 void mlx5_eswitch_update_num_of_vfs(struct mlx5_eswitch *esw, const int num_vfs)
2609 {
2610         const u32 *out;
2611 
2612         WARN_ON_ONCE(esw->mode != MLX5_ESWITCH_NONE);
2613 
2614         if (!mlx5_core_is_ecpf_esw_manager(esw->dev)) {
2615                 esw->esw_funcs.num_vfs = num_vfs;
2616                 return;
2617         }
2618 
2619         out = mlx5_esw_query_functions(esw->dev);
2620         if (IS_ERR(out))
2621                 return;
2622 
2623         esw->esw_funcs.num_vfs = MLX5_GET(query_esw_functions_out, out,
2624                                           host_params_context.host_num_of_vfs);
2625         kvfree(out);
2626 }
/* [<][>][^][v][top][bottom][index][help] */
root/drivers/net/ethernet/mellanox/mlx5/core/eswitch.c

DEFINITIONS
