2 * Copyright (c) 2016, Mellanox Technologies. All rights reserved.
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:
10 * Redistribution and use in source and binary forms, with or
11 * without modification, are permitted provided that the following
14 * - Redistributions of source code must retain the above
15 * copyright notice, this list of conditions and the following
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.
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
33 #include <linux/mlx5/fs.h>
34 #include <net/switchdev.h>
35 #include <net/pkt_cls.h>
36 #include <net/act_api.h>
37 #include <net/devlink.h>
38 #include <net/ipv6_stubs.h>
43 #include "en/params.h"
46 #include "en/rep/tc.h"
47 #include "en/rep/neigh.h"
48 #include "en/rep/bridge.h"
49 #include "en/devlink.h"
52 #define CREATE_TRACE_POINTS
53 #include "diag/en_rep_tracepoint.h"
54 #include "en_accel/ipsec.h"
56 #define MLX5E_REP_PARAMS_DEF_LOG_SQ_SIZE \
57 max(0x7, MLX5E_PARAMS_MINIMUM_LOG_SQ_SIZE)
58 #define MLX5E_REP_PARAMS_DEF_NUM_CHANNELS 1
60 static const char mlx5e_rep_driver_name[] = "mlx5e_rep";
62 static void mlx5e_rep_get_drvinfo(struct net_device *dev,
63 struct ethtool_drvinfo *drvinfo)
65 struct mlx5e_priv *priv = netdev_priv(dev);
66 struct mlx5_core_dev *mdev = priv->mdev;
68 strlcpy(drvinfo->driver, mlx5e_rep_driver_name,
69 sizeof(drvinfo->driver));
70 snprintf(drvinfo->fw_version, sizeof(drvinfo->fw_version),
72 fw_rev_maj(mdev), fw_rev_min(mdev),
73 fw_rev_sub(mdev), mdev->board_id);
76 static const struct counter_desc sw_rep_stats_desc[] = {
77 { MLX5E_DECLARE_STAT(struct mlx5e_sw_stats, rx_packets) },
78 { MLX5E_DECLARE_STAT(struct mlx5e_sw_stats, rx_bytes) },
79 { MLX5E_DECLARE_STAT(struct mlx5e_sw_stats, tx_packets) },
80 { MLX5E_DECLARE_STAT(struct mlx5e_sw_stats, tx_bytes) },
90 static const struct counter_desc vport_rep_stats_desc[] = {
91 { MLX5E_DECLARE_STAT(struct vport_stats, vport_rx_packets) },
92 { MLX5E_DECLARE_STAT(struct vport_stats, vport_rx_bytes) },
93 { MLX5E_DECLARE_STAT(struct vport_stats, vport_tx_packets) },
94 { MLX5E_DECLARE_STAT(struct vport_stats, vport_tx_bytes) },
97 #define NUM_VPORT_REP_SW_COUNTERS ARRAY_SIZE(sw_rep_stats_desc)
98 #define NUM_VPORT_REP_HW_COUNTERS ARRAY_SIZE(vport_rep_stats_desc)
100 static MLX5E_DECLARE_STATS_GRP_OP_NUM_STATS(sw_rep)
102 return NUM_VPORT_REP_SW_COUNTERS;
105 static MLX5E_DECLARE_STATS_GRP_OP_FILL_STRS(sw_rep)
109 for (i = 0; i < NUM_VPORT_REP_SW_COUNTERS; i++)
110 strcpy(data + (idx++) * ETH_GSTRING_LEN,
111 sw_rep_stats_desc[i].format);
115 static MLX5E_DECLARE_STATS_GRP_OP_FILL_STATS(sw_rep)
119 for (i = 0; i < NUM_VPORT_REP_SW_COUNTERS; i++)
120 data[idx++] = MLX5E_READ_CTR64_CPU(&priv->stats.sw,
121 sw_rep_stats_desc, i);
125 static MLX5E_DECLARE_STATS_GRP_OP_UPDATE_STATS(sw_rep)
127 struct mlx5e_sw_stats *s = &priv->stats.sw;
128 struct rtnl_link_stats64 stats64 = {};
130 memset(s, 0, sizeof(*s));
131 mlx5e_fold_sw_stats64(priv, &stats64);
133 s->rx_packets = stats64.rx_packets;
134 s->rx_bytes = stats64.rx_bytes;
135 s->tx_packets = stats64.tx_packets;
136 s->tx_bytes = stats64.tx_bytes;
137 s->tx_queue_dropped = stats64.tx_dropped;
140 static MLX5E_DECLARE_STATS_GRP_OP_NUM_STATS(vport_rep)
142 return NUM_VPORT_REP_HW_COUNTERS;
145 static MLX5E_DECLARE_STATS_GRP_OP_FILL_STRS(vport_rep)
149 for (i = 0; i < NUM_VPORT_REP_HW_COUNTERS; i++)
150 strcpy(data + (idx++) * ETH_GSTRING_LEN, vport_rep_stats_desc[i].format);
154 static MLX5E_DECLARE_STATS_GRP_OP_FILL_STATS(vport_rep)
158 for (i = 0; i < NUM_VPORT_REP_HW_COUNTERS; i++)
159 data[idx++] = MLX5E_READ_CTR64_CPU(&priv->stats.vf_vport,
160 vport_rep_stats_desc, i);
164 static MLX5E_DECLARE_STATS_GRP_OP_UPDATE_STATS(vport_rep)
166 struct mlx5_eswitch *esw = priv->mdev->priv.eswitch;
167 struct mlx5e_rep_priv *rpriv = priv->ppriv;
168 struct mlx5_eswitch_rep *rep = rpriv->rep;
169 struct rtnl_link_stats64 *vport_stats;
170 struct ifla_vf_stats vf_stats;
173 err = mlx5_eswitch_get_vport_stats(esw, rep->vport, &vf_stats);
175 netdev_warn(priv->netdev, "vport %d error %d reading stats\n",
180 vport_stats = &priv->stats.vf_vport;
181 /* flip tx/rx as we are reporting the counters for the switch vport */
182 vport_stats->rx_packets = vf_stats.tx_packets;
183 vport_stats->rx_bytes = vf_stats.tx_bytes;
184 vport_stats->tx_packets = vf_stats.rx_packets;
185 vport_stats->tx_bytes = vf_stats.rx_bytes;
188 static void mlx5e_rep_get_strings(struct net_device *dev,
189 u32 stringset, uint8_t *data)
191 struct mlx5e_priv *priv = netdev_priv(dev);
195 mlx5e_stats_fill_strings(priv, data);
200 static void mlx5e_rep_get_ethtool_stats(struct net_device *dev,
201 struct ethtool_stats *stats, u64 *data)
203 struct mlx5e_priv *priv = netdev_priv(dev);
205 mlx5e_ethtool_get_ethtool_stats(priv, stats, data);
208 static int mlx5e_rep_get_sset_count(struct net_device *dev, int sset)
210 struct mlx5e_priv *priv = netdev_priv(dev);
214 return mlx5e_stats_total_num(priv);
220 static void mlx5e_rep_get_ringparam(struct net_device *dev,
221 struct ethtool_ringparam *param)
223 struct mlx5e_priv *priv = netdev_priv(dev);
225 mlx5e_ethtool_get_ringparam(priv, param);
228 static int mlx5e_rep_set_ringparam(struct net_device *dev,
229 struct ethtool_ringparam *param)
231 struct mlx5e_priv *priv = netdev_priv(dev);
233 return mlx5e_ethtool_set_ringparam(priv, param);
236 static void mlx5e_rep_get_channels(struct net_device *dev,
237 struct ethtool_channels *ch)
239 struct mlx5e_priv *priv = netdev_priv(dev);
241 mlx5e_ethtool_get_channels(priv, ch);
244 static int mlx5e_rep_set_channels(struct net_device *dev,
245 struct ethtool_channels *ch)
247 struct mlx5e_priv *priv = netdev_priv(dev);
249 return mlx5e_ethtool_set_channels(priv, ch);
252 static int mlx5e_rep_get_coalesce(struct net_device *netdev,
253 struct ethtool_coalesce *coal)
255 struct mlx5e_priv *priv = netdev_priv(netdev);
257 return mlx5e_ethtool_get_coalesce(priv, coal);
260 static int mlx5e_rep_set_coalesce(struct net_device *netdev,
261 struct ethtool_coalesce *coal)
263 struct mlx5e_priv *priv = netdev_priv(netdev);
265 return mlx5e_ethtool_set_coalesce(priv, coal);
268 static u32 mlx5e_rep_get_rxfh_key_size(struct net_device *netdev)
270 struct mlx5e_priv *priv = netdev_priv(netdev);
272 return mlx5e_ethtool_get_rxfh_key_size(priv);
275 static u32 mlx5e_rep_get_rxfh_indir_size(struct net_device *netdev)
277 struct mlx5e_priv *priv = netdev_priv(netdev);
279 return mlx5e_ethtool_get_rxfh_indir_size(priv);
282 static const struct ethtool_ops mlx5e_rep_ethtool_ops = {
283 .supported_coalesce_params = ETHTOOL_COALESCE_USECS |
284 ETHTOOL_COALESCE_MAX_FRAMES |
285 ETHTOOL_COALESCE_USE_ADAPTIVE,
286 .get_drvinfo = mlx5e_rep_get_drvinfo,
287 .get_link = ethtool_op_get_link,
288 .get_strings = mlx5e_rep_get_strings,
289 .get_sset_count = mlx5e_rep_get_sset_count,
290 .get_ethtool_stats = mlx5e_rep_get_ethtool_stats,
291 .get_ringparam = mlx5e_rep_get_ringparam,
292 .set_ringparam = mlx5e_rep_set_ringparam,
293 .get_channels = mlx5e_rep_get_channels,
294 .set_channels = mlx5e_rep_set_channels,
295 .get_coalesce = mlx5e_rep_get_coalesce,
296 .set_coalesce = mlx5e_rep_set_coalesce,
297 .get_rxfh_key_size = mlx5e_rep_get_rxfh_key_size,
298 .get_rxfh_indir_size = mlx5e_rep_get_rxfh_indir_size,
301 static void mlx5e_sqs2vport_stop(struct mlx5_eswitch *esw,
302 struct mlx5_eswitch_rep *rep)
304 struct mlx5e_rep_sq *rep_sq, *tmp;
305 struct mlx5e_rep_priv *rpriv;
307 if (esw->mode != MLX5_ESWITCH_OFFLOADS)
310 rpriv = mlx5e_rep_to_rep_priv(rep);
311 list_for_each_entry_safe(rep_sq, tmp, &rpriv->vport_sqs_list, list) {
312 mlx5_eswitch_del_send_to_vport_rule(rep_sq->send_to_vport_rule);
313 list_del(&rep_sq->list);
318 static int mlx5e_sqs2vport_start(struct mlx5_eswitch *esw,
319 struct mlx5_eswitch_rep *rep,
320 u32 *sqns_array, int sqns_num)
322 struct mlx5_flow_handle *flow_rule;
323 struct mlx5e_rep_priv *rpriv;
324 struct mlx5e_rep_sq *rep_sq;
328 if (esw->mode != MLX5_ESWITCH_OFFLOADS)
331 rpriv = mlx5e_rep_to_rep_priv(rep);
332 for (i = 0; i < sqns_num; i++) {
333 rep_sq = kzalloc(sizeof(*rep_sq), GFP_KERNEL);
339 /* Add re-inject rule to the PF/representor sqs */
340 flow_rule = mlx5_eswitch_add_send_to_vport_rule(esw, rep,
342 if (IS_ERR(flow_rule)) {
343 err = PTR_ERR(flow_rule);
347 rep_sq->send_to_vport_rule = flow_rule;
348 list_add(&rep_sq->list, &rpriv->vport_sqs_list);
353 mlx5e_sqs2vport_stop(esw, rep);
357 int mlx5e_add_sqs_fwd_rules(struct mlx5e_priv *priv)
359 struct mlx5_eswitch *esw = priv->mdev->priv.eswitch;
360 struct mlx5e_rep_priv *rpriv = priv->ppriv;
361 struct mlx5_eswitch_rep *rep = rpriv->rep;
362 struct mlx5e_channel *c;
363 int n, tc, num_sqs = 0;
367 sqs = kcalloc(priv->channels.num * priv->channels.params.num_tc, sizeof(*sqs), GFP_KERNEL);
371 for (n = 0; n < priv->channels.num; n++) {
372 c = priv->channels.c[n];
373 for (tc = 0; tc < c->num_tc; tc++)
374 sqs[num_sqs++] = c->sq[tc].sqn;
377 err = mlx5e_sqs2vport_start(esw, rep, sqs, num_sqs);
382 netdev_warn(priv->netdev, "Failed to add SQs FWD rules %d\n", err);
386 void mlx5e_remove_sqs_fwd_rules(struct mlx5e_priv *priv)
388 struct mlx5_eswitch *esw = priv->mdev->priv.eswitch;
389 struct mlx5e_rep_priv *rpriv = priv->ppriv;
390 struct mlx5_eswitch_rep *rep = rpriv->rep;
392 mlx5e_sqs2vport_stop(esw, rep);
395 static int mlx5e_rep_open(struct net_device *dev)
397 struct mlx5e_priv *priv = netdev_priv(dev);
398 struct mlx5e_rep_priv *rpriv = priv->ppriv;
399 struct mlx5_eswitch_rep *rep = rpriv->rep;
402 mutex_lock(&priv->state_lock);
403 err = mlx5e_open_locked(dev);
407 if (!mlx5_modify_vport_admin_state(priv->mdev,
408 MLX5_VPORT_STATE_OP_MOD_ESW_VPORT,
410 MLX5_VPORT_ADMIN_STATE_UP))
411 netif_carrier_on(dev);
414 mutex_unlock(&priv->state_lock);
418 static int mlx5e_rep_close(struct net_device *dev)
420 struct mlx5e_priv *priv = netdev_priv(dev);
421 struct mlx5e_rep_priv *rpriv = priv->ppriv;
422 struct mlx5_eswitch_rep *rep = rpriv->rep;
425 mutex_lock(&priv->state_lock);
426 mlx5_modify_vport_admin_state(priv->mdev,
427 MLX5_VPORT_STATE_OP_MOD_ESW_VPORT,
429 MLX5_VPORT_ADMIN_STATE_DOWN);
430 ret = mlx5e_close_locked(dev);
431 mutex_unlock(&priv->state_lock);
435 bool mlx5e_is_uplink_rep(struct mlx5e_priv *priv)
437 struct mlx5e_rep_priv *rpriv = priv->ppriv;
438 struct mlx5_eswitch_rep *rep;
440 if (!MLX5_ESWITCH_MANAGER(priv->mdev))
443 if (!rpriv) /* non vport rep mlx5e instances don't use this field */
447 return (rep->vport == MLX5_VPORT_UPLINK);
450 bool mlx5e_rep_has_offload_stats(const struct net_device *dev, int attr_id)
453 case IFLA_OFFLOAD_XSTATS_CPU_HIT:
461 mlx5e_get_sw_stats64(const struct net_device *dev,
462 struct rtnl_link_stats64 *stats)
464 struct mlx5e_priv *priv = netdev_priv(dev);
466 mlx5e_fold_sw_stats64(priv, stats);
470 int mlx5e_rep_get_offload_stats(int attr_id, const struct net_device *dev,
474 case IFLA_OFFLOAD_XSTATS_CPU_HIT:
475 return mlx5e_get_sw_stats64(dev, sp);
482 mlx5e_rep_get_stats(struct net_device *dev, struct rtnl_link_stats64 *stats)
484 struct mlx5e_priv *priv = netdev_priv(dev);
486 /* update HW stats in background for next time */
487 mlx5e_queue_update_stats(priv);
488 memcpy(stats, &priv->stats.vf_vport, sizeof(*stats));
491 static int mlx5e_rep_change_mtu(struct net_device *netdev, int new_mtu)
493 return mlx5e_change_mtu(netdev, new_mtu, NULL);
496 static struct devlink_port *mlx5e_rep_get_devlink_port(struct net_device *netdev)
498 struct mlx5e_priv *priv = netdev_priv(netdev);
499 struct mlx5e_rep_priv *rpriv = priv->ppriv;
500 struct mlx5_core_dev *dev = priv->mdev;
502 return mlx5_esw_offloads_devlink_port(dev->priv.eswitch, rpriv->rep->vport);
505 static int mlx5e_rep_change_carrier(struct net_device *dev, bool new_carrier)
507 struct mlx5e_priv *priv = netdev_priv(dev);
508 struct mlx5e_rep_priv *rpriv = priv->ppriv;
509 struct mlx5_eswitch_rep *rep = rpriv->rep;
513 err = mlx5_modify_vport_admin_state(priv->mdev, MLX5_VPORT_STATE_OP_MOD_ESW_VPORT,
514 rep->vport, 1, MLX5_VPORT_ADMIN_STATE_UP);
517 netif_carrier_on(dev);
519 err = mlx5_modify_vport_admin_state(priv->mdev, MLX5_VPORT_STATE_OP_MOD_ESW_VPORT,
520 rep->vport, 1, MLX5_VPORT_ADMIN_STATE_DOWN);
523 netif_carrier_off(dev);
528 static const struct net_device_ops mlx5e_netdev_ops_rep = {
529 .ndo_open = mlx5e_rep_open,
530 .ndo_stop = mlx5e_rep_close,
531 .ndo_start_xmit = mlx5e_xmit,
532 .ndo_setup_tc = mlx5e_rep_setup_tc,
533 .ndo_get_devlink_port = mlx5e_rep_get_devlink_port,
534 .ndo_get_stats64 = mlx5e_rep_get_stats,
535 .ndo_has_offload_stats = mlx5e_rep_has_offload_stats,
536 .ndo_get_offload_stats = mlx5e_rep_get_offload_stats,
537 .ndo_change_mtu = mlx5e_rep_change_mtu,
538 .ndo_change_carrier = mlx5e_rep_change_carrier,
541 bool mlx5e_eswitch_uplink_rep(const struct net_device *netdev)
543 return netdev->netdev_ops == &mlx5e_netdev_ops &&
544 mlx5e_is_uplink_rep(netdev_priv(netdev));
547 bool mlx5e_eswitch_vf_rep(const struct net_device *netdev)
549 return netdev->netdev_ops == &mlx5e_netdev_ops_rep;
552 static void mlx5e_build_rep_params(struct net_device *netdev)
554 struct mlx5e_priv *priv = netdev_priv(netdev);
555 struct mlx5e_rep_priv *rpriv = priv->ppriv;
556 struct mlx5_eswitch_rep *rep = rpriv->rep;
557 struct mlx5_core_dev *mdev = priv->mdev;
558 struct mlx5e_params *params;
560 u8 cq_period_mode = MLX5_CAP_GEN(mdev, cq_period_start_from_cqe) ?
561 MLX5_CQ_PERIOD_MODE_START_FROM_CQE :
562 MLX5_CQ_PERIOD_MODE_START_FROM_EQE;
564 priv->max_nch = mlx5e_calc_max_nch(priv, priv->profile);
565 params = &priv->channels.params;
567 params->num_channels = MLX5E_REP_PARAMS_DEF_NUM_CHANNELS;
568 params->hard_mtu = MLX5E_ETH_HARD_MTU;
569 params->sw_mtu = netdev->mtu;
572 if (rep->vport == MLX5_VPORT_UPLINK)
573 params->log_sq_size = MLX5E_PARAMS_DEFAULT_LOG_SQ_SIZE;
575 params->log_sq_size = MLX5E_REP_PARAMS_DEF_LOG_SQ_SIZE;
578 mlx5e_build_rq_params(mdev, params);
580 /* CQ moderation params */
581 params->rx_dim_enabled = MLX5_CAP_GEN(mdev, cq_moderation);
582 mlx5e_set_rx_cq_mode_params(params, cq_period_mode);
585 params->tunneled_offload_en = false;
587 mlx5_query_min_inline(mdev, ¶ms->tx_min_inline_mode);
590 static void mlx5e_build_rep_netdev(struct net_device *netdev,
591 struct mlx5_core_dev *mdev)
593 SET_NETDEV_DEV(netdev, mdev->device);
594 netdev->netdev_ops = &mlx5e_netdev_ops_rep;
595 eth_hw_addr_random(netdev);
596 netdev->ethtool_ops = &mlx5e_rep_ethtool_ops;
598 netdev->watchdog_timeo = 15 * HZ;
600 #if IS_ENABLED(CONFIG_MLX5_CLS_ACT)
601 netdev->hw_features |= NETIF_F_HW_TC;
603 netdev->hw_features |= NETIF_F_SG;
604 netdev->hw_features |= NETIF_F_IP_CSUM;
605 netdev->hw_features |= NETIF_F_IPV6_CSUM;
606 netdev->hw_features |= NETIF_F_GRO;
607 netdev->hw_features |= NETIF_F_TSO;
608 netdev->hw_features |= NETIF_F_TSO6;
609 netdev->hw_features |= NETIF_F_RXCSUM;
611 netdev->features |= netdev->hw_features;
612 netdev->features |= NETIF_F_VLAN_CHALLENGED;
613 netdev->features |= NETIF_F_NETNS_LOCAL;
616 static int mlx5e_init_rep(struct mlx5_core_dev *mdev,
617 struct net_device *netdev)
619 struct mlx5e_priv *priv = netdev_priv(netdev);
621 mlx5e_build_rep_params(netdev);
622 mlx5e_timestamp_init(priv);
627 static int mlx5e_init_ul_rep(struct mlx5_core_dev *mdev,
628 struct net_device *netdev)
630 struct mlx5e_priv *priv = netdev_priv(netdev);
633 err = mlx5e_ipsec_init(priv);
635 mlx5_core_err(mdev, "Uplink rep IPsec initialization failed, %d\n", err);
637 mlx5e_vxlan_set_netdev_info(priv);
638 return mlx5e_init_rep(mdev, netdev);
641 static void mlx5e_cleanup_rep(struct mlx5e_priv *priv)
643 mlx5e_ipsec_cleanup(priv);
646 static int mlx5e_create_rep_ttc_table(struct mlx5e_priv *priv)
648 struct mlx5e_rep_priv *rpriv = priv->ppriv;
649 struct mlx5_eswitch_rep *rep = rpriv->rep;
650 struct ttc_params ttc_params = {};
653 priv->fs.ns = mlx5_get_flow_namespace(priv->mdev,
654 MLX5_FLOW_NAMESPACE_KERNEL);
656 /* The inner_ttc in the ttc params is intentionally not set */
657 ttc_params.any_tt_tirn = priv->direct_tir[0].tirn;
658 mlx5e_set_ttc_ft_params(&ttc_params);
660 if (rep->vport != MLX5_VPORT_UPLINK)
661 /* To give uplik rep TTC a lower level for chaining from root ft */
662 ttc_params.ft_attr.level = MLX5E_TTC_FT_LEVEL + 1;
664 for (tt = 0; tt < MLX5E_NUM_INDIR_TIRS; tt++)
665 ttc_params.indir_tirn[tt] = priv->indir_tir[tt].tirn;
667 err = mlx5e_create_ttc_table(priv, &ttc_params, &priv->fs.ttc);
669 netdev_err(priv->netdev, "Failed to create rep ttc table, err=%d\n", err);
675 static int mlx5e_create_rep_root_ft(struct mlx5e_priv *priv)
677 struct mlx5e_rep_priv *rpriv = priv->ppriv;
678 struct mlx5_eswitch_rep *rep = rpriv->rep;
679 struct mlx5_flow_table_attr ft_attr = {};
680 struct mlx5_flow_namespace *ns;
683 if (rep->vport != MLX5_VPORT_UPLINK) {
684 /* non uplik reps will skip any bypass tables and go directly to
687 rpriv->root_ft = priv->fs.ttc.ft.t;
691 /* uplink root ft will be used to auto chain, to ethtool or ttc tables */
692 ns = mlx5_get_flow_namespace(priv->mdev, MLX5_FLOW_NAMESPACE_OFFLOADS);
694 netdev_err(priv->netdev, "Failed to get reps offloads namespace\n");
698 ft_attr.max_fte = 0; /* Empty table, miss rule will always point to next table */
702 rpriv->root_ft = mlx5_create_flow_table(ns, &ft_attr);
703 if (IS_ERR(rpriv->root_ft)) {
704 err = PTR_ERR(rpriv->root_ft);
705 rpriv->root_ft = NULL;
711 static void mlx5e_destroy_rep_root_ft(struct mlx5e_priv *priv)
713 struct mlx5e_rep_priv *rpriv = priv->ppriv;
714 struct mlx5_eswitch_rep *rep = rpriv->rep;
716 if (rep->vport != MLX5_VPORT_UPLINK)
718 mlx5_destroy_flow_table(rpriv->root_ft);
721 static int mlx5e_create_rep_vport_rx_rule(struct mlx5e_priv *priv)
723 struct mlx5_eswitch *esw = priv->mdev->priv.eswitch;
724 struct mlx5e_rep_priv *rpriv = priv->ppriv;
725 struct mlx5_eswitch_rep *rep = rpriv->rep;
726 struct mlx5_flow_handle *flow_rule;
727 struct mlx5_flow_destination dest;
729 dest.type = MLX5_FLOW_DESTINATION_TYPE_FLOW_TABLE;
730 dest.ft = rpriv->root_ft;
732 flow_rule = mlx5_eswitch_create_vport_rx_rule(esw, rep->vport, &dest);
733 if (IS_ERR(flow_rule))
734 return PTR_ERR(flow_rule);
735 rpriv->vport_rx_rule = flow_rule;
739 static void rep_vport_rx_rule_destroy(struct mlx5e_priv *priv)
741 struct mlx5e_rep_priv *rpriv = priv->ppriv;
743 if (!rpriv->vport_rx_rule)
746 mlx5_del_flow_rules(rpriv->vport_rx_rule);
747 rpriv->vport_rx_rule = NULL;
750 int mlx5e_rep_bond_update(struct mlx5e_priv *priv, bool cleanup)
752 rep_vport_rx_rule_destroy(priv);
754 return cleanup ? 0 : mlx5e_create_rep_vport_rx_rule(priv);
757 static int mlx5e_init_rep_rx(struct mlx5e_priv *priv)
759 struct mlx5_core_dev *mdev = priv->mdev;
760 u16 max_nch = priv->max_nch;
763 mlx5e_build_rss_params(&priv->rss_params, priv->channels.params.num_channels);
765 mlx5e_init_l2_addr(priv);
767 err = mlx5e_open_drop_rq(priv, &priv->drop_rq);
769 mlx5_core_err(mdev, "open drop rq failed, %d\n", err);
773 err = mlx5e_create_indirect_rqt(priv);
775 goto err_close_drop_rq;
777 err = mlx5e_create_direct_rqts(priv, priv->direct_tir, max_nch);
779 goto err_destroy_indirect_rqts;
781 err = mlx5e_create_indirect_tirs(priv, false);
783 goto err_destroy_direct_rqts;
785 err = mlx5e_create_direct_tirs(priv, priv->direct_tir, max_nch);
787 goto err_destroy_indirect_tirs;
789 err = mlx5e_create_rep_ttc_table(priv);
791 goto err_destroy_direct_tirs;
793 err = mlx5e_create_rep_root_ft(priv);
795 goto err_destroy_ttc_table;
797 err = mlx5e_create_rep_vport_rx_rule(priv);
799 goto err_destroy_root_ft;
801 mlx5e_ethtool_init_steering(priv);
806 mlx5e_destroy_rep_root_ft(priv);
807 err_destroy_ttc_table:
808 mlx5e_destroy_ttc_table(priv, &priv->fs.ttc);
809 err_destroy_direct_tirs:
810 mlx5e_destroy_direct_tirs(priv, priv->direct_tir, max_nch);
811 err_destroy_indirect_tirs:
812 mlx5e_destroy_indirect_tirs(priv);
813 err_destroy_direct_rqts:
814 mlx5e_destroy_direct_rqts(priv, priv->direct_tir, max_nch);
815 err_destroy_indirect_rqts:
816 priv->indir_rqt_enabled = false;
817 mlx5e_rqt_destroy(&priv->indir_rqt);
819 mlx5e_close_drop_rq(&priv->drop_rq);
823 static void mlx5e_cleanup_rep_rx(struct mlx5e_priv *priv)
825 u16 max_nch = priv->max_nch;
827 mlx5e_ethtool_cleanup_steering(priv);
828 rep_vport_rx_rule_destroy(priv);
829 mlx5e_destroy_rep_root_ft(priv);
830 mlx5e_destroy_ttc_table(priv, &priv->fs.ttc);
831 mlx5e_destroy_direct_tirs(priv, priv->direct_tir, max_nch);
832 mlx5e_destroy_indirect_tirs(priv);
833 mlx5e_destroy_direct_rqts(priv, priv->direct_tir, max_nch);
834 priv->indir_rqt_enabled = false;
835 mlx5e_rqt_destroy(&priv->indir_rqt);
836 mlx5e_close_drop_rq(&priv->drop_rq);
839 static int mlx5e_init_ul_rep_rx(struct mlx5e_priv *priv)
841 mlx5e_create_q_counters(priv);
842 return mlx5e_init_rep_rx(priv);
845 static void mlx5e_cleanup_ul_rep_rx(struct mlx5e_priv *priv)
847 mlx5e_cleanup_rep_rx(priv);
848 mlx5e_destroy_q_counters(priv);
851 static int mlx5e_init_uplink_rep_tx(struct mlx5e_rep_priv *rpriv)
853 struct mlx5_rep_uplink_priv *uplink_priv;
854 struct net_device *netdev;
855 struct mlx5e_priv *priv;
858 netdev = rpriv->netdev;
859 priv = netdev_priv(netdev);
860 uplink_priv = &rpriv->uplink_priv;
862 err = mlx5e_rep_tc_init(rpriv);
866 mlx5_init_port_tun_entropy(&uplink_priv->tun_entropy, priv->mdev);
868 mlx5e_rep_bond_init(rpriv);
869 err = mlx5e_rep_tc_netdevice_event_register(rpriv);
871 mlx5_core_err(priv->mdev, "Failed to register netdev notifier, err: %d\n",
879 mlx5e_rep_bond_cleanup(rpriv);
880 mlx5e_rep_tc_cleanup(rpriv);
884 static int mlx5e_init_rep_tx(struct mlx5e_priv *priv)
886 struct mlx5e_rep_priv *rpriv = priv->ppriv;
889 err = mlx5e_create_tises(priv);
891 mlx5_core_warn(priv->mdev, "create tises failed, %d\n", err);
895 if (rpriv->rep->vport == MLX5_VPORT_UPLINK) {
896 err = mlx5e_init_uplink_rep_tx(rpriv);
904 mlx5e_destroy_tises(priv);
908 static void mlx5e_cleanup_uplink_rep_tx(struct mlx5e_rep_priv *rpriv)
910 mlx5e_rep_tc_netdevice_event_unregister(rpriv);
911 mlx5e_rep_bond_cleanup(rpriv);
912 mlx5e_rep_tc_cleanup(rpriv);
915 static void mlx5e_cleanup_rep_tx(struct mlx5e_priv *priv)
917 struct mlx5e_rep_priv *rpriv = priv->ppriv;
919 mlx5e_destroy_tises(priv);
921 if (rpriv->rep->vport == MLX5_VPORT_UPLINK)
922 mlx5e_cleanup_uplink_rep_tx(rpriv);
925 static void mlx5e_rep_enable(struct mlx5e_priv *priv)
927 struct mlx5e_rep_priv *rpriv = priv->ppriv;
929 mlx5e_set_netdev_mtu_boundaries(priv);
930 mlx5e_rep_neigh_init(rpriv);
933 static void mlx5e_rep_disable(struct mlx5e_priv *priv)
935 struct mlx5e_rep_priv *rpriv = priv->ppriv;
937 mlx5e_rep_neigh_cleanup(rpriv);
940 static int mlx5e_update_rep_rx(struct mlx5e_priv *priv)
945 static int uplink_rep_async_event(struct notifier_block *nb, unsigned long event, void *data)
947 struct mlx5e_priv *priv = container_of(nb, struct mlx5e_priv, events_nb);
949 if (event == MLX5_EVENT_TYPE_PORT_CHANGE) {
950 struct mlx5_eqe *eqe = data;
952 switch (eqe->sub_type) {
953 case MLX5_PORT_CHANGE_SUBTYPE_DOWN:
954 case MLX5_PORT_CHANGE_SUBTYPE_ACTIVE:
955 queue_work(priv->wq, &priv->update_carrier_work);
964 if (event == MLX5_DEV_EVENT_PORT_AFFINITY)
965 return mlx5e_rep_tc_event_port_affinity(priv);
970 static void mlx5e_uplink_rep_enable(struct mlx5e_priv *priv)
972 struct mlx5e_rep_priv *rpriv = priv->ppriv;
973 struct net_device *netdev = priv->netdev;
974 struct mlx5_core_dev *mdev = priv->mdev;
977 netdev->min_mtu = ETH_MIN_MTU;
978 mlx5_query_port_max_mtu(priv->mdev, &max_mtu, 1);
979 netdev->max_mtu = MLX5E_HW2SW_MTU(&priv->channels.params, max_mtu);
980 mlx5e_set_dev_port_mtu(priv);
982 mlx5e_rep_tc_enable(priv);
984 if (MLX5_CAP_GEN(mdev, uplink_follow))
985 mlx5_modify_vport_admin_state(mdev, MLX5_VPORT_STATE_OP_MOD_UPLINK,
986 0, 0, MLX5_VPORT_ADMIN_STATE_AUTO);
987 mlx5_lag_add_netdev(mdev, netdev);
988 priv->events_nb.notifier_call = uplink_rep_async_event;
989 mlx5_notifier_register(mdev, &priv->events_nb);
990 mlx5e_dcbnl_initialize(priv);
991 mlx5e_dcbnl_init_app(priv);
992 mlx5e_rep_neigh_init(rpriv);
993 mlx5e_rep_bridge_init(priv);
995 netdev->wanted_features |= NETIF_F_HW_TC;
998 if (netif_running(netdev))
1000 netif_device_attach(netdev);
1004 static void mlx5e_uplink_rep_disable(struct mlx5e_priv *priv)
1006 struct mlx5e_rep_priv *rpriv = priv->ppriv;
1007 struct mlx5_core_dev *mdev = priv->mdev;
1010 if (netif_running(priv->netdev))
1011 mlx5e_close(priv->netdev);
1012 netif_device_detach(priv->netdev);
1015 mlx5e_rep_bridge_cleanup(priv);
1016 mlx5e_rep_neigh_cleanup(rpriv);
1017 mlx5e_dcbnl_delete_app(priv);
1018 mlx5_notifier_unregister(mdev, &priv->events_nb);
1019 mlx5e_rep_tc_disable(priv);
1020 mlx5_lag_remove_netdev(mdev, priv->netdev);
1023 static MLX5E_DEFINE_STATS_GRP(sw_rep, 0);
1024 static MLX5E_DEFINE_STATS_GRP(vport_rep, MLX5E_NDO_UPDATE_STATS);
1026 /* The stats groups order is opposite to the update_stats() order calls */
1027 static mlx5e_stats_grp_t mlx5e_rep_stats_grps[] = {
1028 &MLX5E_STATS_GRP(sw_rep),
1029 &MLX5E_STATS_GRP(vport_rep),
1032 static unsigned int mlx5e_rep_stats_grps_num(struct mlx5e_priv *priv)
1034 return ARRAY_SIZE(mlx5e_rep_stats_grps);
1037 /* The stats groups order is opposite to the update_stats() order calls */
1038 static mlx5e_stats_grp_t mlx5e_ul_rep_stats_grps[] = {
1039 &MLX5E_STATS_GRP(sw),
1040 &MLX5E_STATS_GRP(qcnt),
1041 &MLX5E_STATS_GRP(vnic_env),
1042 &MLX5E_STATS_GRP(vport),
1043 &MLX5E_STATS_GRP(802_3),
1044 &MLX5E_STATS_GRP(2863),
1045 &MLX5E_STATS_GRP(2819),
1046 &MLX5E_STATS_GRP(phy),
1047 &MLX5E_STATS_GRP(eth_ext),
1048 &MLX5E_STATS_GRP(pcie),
1049 &MLX5E_STATS_GRP(per_prio),
1050 &MLX5E_STATS_GRP(pme),
1051 &MLX5E_STATS_GRP(channels),
1052 &MLX5E_STATS_GRP(per_port_buff_congest),
1055 static unsigned int mlx5e_ul_rep_stats_grps_num(struct mlx5e_priv *priv)
1057 return ARRAY_SIZE(mlx5e_ul_rep_stats_grps);
1060 static const struct mlx5e_profile mlx5e_rep_profile = {
1061 .init = mlx5e_init_rep,
1062 .cleanup = mlx5e_cleanup_rep,
1063 .init_rx = mlx5e_init_rep_rx,
1064 .cleanup_rx = mlx5e_cleanup_rep_rx,
1065 .init_tx = mlx5e_init_rep_tx,
1066 .cleanup_tx = mlx5e_cleanup_rep_tx,
1067 .enable = mlx5e_rep_enable,
1068 .disable = mlx5e_rep_disable,
1069 .update_rx = mlx5e_update_rep_rx,
1070 .update_stats = mlx5e_stats_update_ndo_stats,
1071 .rx_handlers = &mlx5e_rx_handlers_rep,
1073 .rq_groups = MLX5E_NUM_RQ_GROUPS(REGULAR),
1074 .stats_grps = mlx5e_rep_stats_grps,
1075 .stats_grps_num = mlx5e_rep_stats_grps_num,
1076 .rx_ptp_support = false,
1079 static const struct mlx5e_profile mlx5e_uplink_rep_profile = {
1080 .init = mlx5e_init_ul_rep,
1081 .cleanup = mlx5e_cleanup_rep,
1082 .init_rx = mlx5e_init_ul_rep_rx,
1083 .cleanup_rx = mlx5e_cleanup_ul_rep_rx,
1084 .init_tx = mlx5e_init_rep_tx,
1085 .cleanup_tx = mlx5e_cleanup_rep_tx,
1086 .enable = mlx5e_uplink_rep_enable,
1087 .disable = mlx5e_uplink_rep_disable,
1088 .update_rx = mlx5e_update_rep_rx,
1089 .update_stats = mlx5e_stats_update_ndo_stats,
1090 .update_carrier = mlx5e_update_carrier,
1091 .rx_handlers = &mlx5e_rx_handlers_rep,
1092 .max_tc = MLX5E_MAX_NUM_TC,
1093 /* XSK is needed so we can replace profile with NIC netdev */
1094 .rq_groups = MLX5E_NUM_RQ_GROUPS(XSK),
1095 .stats_grps = mlx5e_ul_rep_stats_grps,
1096 .stats_grps_num = mlx5e_ul_rep_stats_grps_num,
1097 .rx_ptp_support = false,
1100 /* e-Switch vport representors */
1102 mlx5e_vport_uplink_rep_load(struct mlx5_core_dev *dev, struct mlx5_eswitch_rep *rep)
1104 struct mlx5e_priv *priv = netdev_priv(mlx5_uplink_netdev_get(dev));
1105 struct mlx5e_rep_priv *rpriv = mlx5e_rep_to_rep_priv(rep);
1106 struct devlink_port *dl_port;
1109 rpriv->netdev = priv->netdev;
1111 err = mlx5e_netdev_change_profile(priv, &mlx5e_uplink_rep_profile,
1116 dl_port = mlx5_esw_offloads_devlink_port(dev->priv.eswitch, rpriv->rep->vport);
1118 devlink_port_type_eth_set(dl_port, rpriv->netdev);
1124 mlx5e_vport_uplink_rep_unload(struct mlx5e_rep_priv *rpriv)
1126 struct net_device *netdev = rpriv->netdev;
1127 struct devlink_port *dl_port;
1128 struct mlx5_core_dev *dev;
1129 struct mlx5e_priv *priv;
1131 priv = netdev_priv(netdev);
1134 dl_port = mlx5_esw_offloads_devlink_port(dev->priv.eswitch, rpriv->rep->vport);
1136 devlink_port_type_clear(dl_port);
1137 mlx5e_netdev_attach_nic_profile(priv);
1141 mlx5e_vport_vf_rep_load(struct mlx5_core_dev *dev, struct mlx5_eswitch_rep *rep)
1143 struct mlx5e_rep_priv *rpriv = mlx5e_rep_to_rep_priv(rep);
1144 const struct mlx5e_profile *profile;
1145 struct devlink_port *dl_port;
1146 struct net_device *netdev;
1147 struct mlx5e_priv *priv;
1148 unsigned int txqs, rxqs;
1151 profile = &mlx5e_rep_profile;
1152 nch = mlx5e_get_max_num_channels(dev);
1153 txqs = nch * profile->max_tc;
1154 rxqs = nch * profile->rq_groups;
1155 netdev = mlx5e_create_netdev(dev, txqs, rxqs);
1158 "Failed to create representor netdev for vport %d\n",
1163 mlx5e_build_rep_netdev(netdev, dev);
1164 rpriv->netdev = netdev;
1166 priv = netdev_priv(netdev);
1167 priv->profile = profile;
1168 priv->ppriv = rpriv;
1169 err = profile->init(dev, netdev);
1171 netdev_warn(netdev, "rep profile init failed, %d\n", err);
1172 goto err_destroy_netdev;
1175 err = mlx5e_attach_netdev(netdev_priv(netdev));
1178 "Failed to attach representor netdev for vport %d\n",
1180 goto err_cleanup_profile;
1183 err = register_netdev(netdev);
1186 "Failed to register representor netdev for vport %d\n",
1188 goto err_detach_netdev;
1191 dl_port = mlx5_esw_offloads_devlink_port(dev->priv.eswitch, rpriv->rep->vport);
1193 devlink_port_type_eth_set(dl_port, netdev);
1197 mlx5e_detach_netdev(netdev_priv(netdev));
1199 err_cleanup_profile:
1200 priv->profile->cleanup(priv);
1203 mlx5e_destroy_netdev(netdev_priv(netdev));
1208 mlx5e_vport_rep_load(struct mlx5_core_dev *dev, struct mlx5_eswitch_rep *rep)
1210 struct mlx5e_rep_priv *rpriv;
1213 rpriv = kzalloc(sizeof(*rpriv), GFP_KERNEL);
1217 /* rpriv->rep to be looked up when profile->init() is called */
1219 rep->rep_data[REP_ETH].priv = rpriv;
1220 INIT_LIST_HEAD(&rpriv->vport_sqs_list);
1222 if (rep->vport == MLX5_VPORT_UPLINK)
1223 err = mlx5e_vport_uplink_rep_load(dev, rep);
1225 err = mlx5e_vport_vf_rep_load(dev, rep);
1234 mlx5e_vport_rep_unload(struct mlx5_eswitch_rep *rep)
1236 struct mlx5e_rep_priv *rpriv = mlx5e_rep_to_rep_priv(rep);
1237 struct net_device *netdev = rpriv->netdev;
1238 struct mlx5e_priv *priv = netdev_priv(netdev);
1239 struct mlx5_core_dev *dev = priv->mdev;
1240 struct devlink_port *dl_port;
1241 void *ppriv = priv->ppriv;
1243 if (rep->vport == MLX5_VPORT_UPLINK) {
1244 mlx5e_vport_uplink_rep_unload(rpriv);
1248 dl_port = mlx5_esw_offloads_devlink_port(dev->priv.eswitch, rpriv->rep->vport);
1250 devlink_port_type_clear(dl_port);
1251 unregister_netdev(netdev);
1252 mlx5e_detach_netdev(priv);
1253 priv->profile->cleanup(priv);
1254 mlx5e_destroy_netdev(priv);
1256 kfree(ppriv); /* mlx5e_rep_priv */
1259 static void *mlx5e_vport_rep_get_proto_dev(struct mlx5_eswitch_rep *rep)
1261 struct mlx5e_rep_priv *rpriv;
1263 rpriv = mlx5e_rep_to_rep_priv(rep);
1265 return rpriv->netdev;
1268 static const struct mlx5_eswitch_rep_ops rep_ops = {
1269 .load = mlx5e_vport_rep_load,
1270 .unload = mlx5e_vport_rep_unload,
1271 .get_proto_dev = mlx5e_vport_rep_get_proto_dev
1274 static int mlx5e_rep_probe(struct auxiliary_device *adev,
1275 const struct auxiliary_device_id *id)
1277 struct mlx5_adev *edev = container_of(adev, struct mlx5_adev, adev);
1278 struct mlx5_core_dev *mdev = edev->mdev;
1279 struct mlx5_eswitch *esw;
1281 esw = mdev->priv.eswitch;
1282 mlx5_eswitch_register_vport_reps(esw, &rep_ops, REP_ETH);
1286 static void mlx5e_rep_remove(struct auxiliary_device *adev)
1288 struct mlx5_adev *vdev = container_of(adev, struct mlx5_adev, adev);
1289 struct mlx5_core_dev *mdev = vdev->mdev;
1290 struct mlx5_eswitch *esw;
1292 esw = mdev->priv.eswitch;
1293 mlx5_eswitch_unregister_vport_reps(esw, REP_ETH);
1296 static const struct auxiliary_device_id mlx5e_rep_id_table[] = {
1297 { .name = MLX5_ADEV_NAME ".eth-rep", },
1301 MODULE_DEVICE_TABLE(auxiliary, mlx5e_rep_id_table);
1303 static struct auxiliary_driver mlx5e_rep_driver = {
1305 .probe = mlx5e_rep_probe,
1306 .remove = mlx5e_rep_remove,
1307 .id_table = mlx5e_rep_id_table,
1310 int mlx5e_rep_init(void)
1312 return auxiliary_driver_register(&mlx5e_rep_driver);
1315 void mlx5e_rep_cleanup(void)
1317 auxiliary_driver_unregister(&mlx5e_rep_driver);