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 <generated/utsrelease.h>
34 #include <linux/mlx5/fs.h>
35 #include <net/switchdev.h>
36 #include <net/pkt_cls.h>
37 #include <net/act_api.h>
38 #include <net/netevent.h>
40 #include <net/devlink.h>
41 #include <net/ipv6_stubs.h>
44 #include "esw/chains.h"
48 #include "en/tc_tun.h"
50 #include "lib/port_tun.h"
52 #define CREATE_TRACE_POINTS
53 #include "diag/en_rep_tracepoint.h"
55 #define MLX5E_REP_PARAMS_DEF_LOG_SQ_SIZE \
56 max(0x7, MLX5E_PARAMS_MINIMUM_LOG_SQ_SIZE)
57 #define MLX5E_REP_PARAMS_DEF_NUM_CHANNELS 1
59 static const char mlx5e_rep_driver_name[] = "mlx5e_rep";
61 struct mlx5e_rep_indr_block_priv {
62 struct net_device *netdev;
63 struct mlx5e_rep_priv *rpriv;
65 struct list_head list;
68 static void mlx5e_rep_indr_unregister_block(struct mlx5e_rep_priv *rpriv,
69 struct net_device *netdev);
71 static void mlx5e_rep_get_drvinfo(struct net_device *dev,
72 struct ethtool_drvinfo *drvinfo)
74 struct mlx5e_priv *priv = netdev_priv(dev);
75 struct mlx5_core_dev *mdev = priv->mdev;
77 strlcpy(drvinfo->driver, mlx5e_rep_driver_name,
78 sizeof(drvinfo->driver));
79 strlcpy(drvinfo->version, UTS_RELEASE, sizeof(drvinfo->version));
80 snprintf(drvinfo->fw_version, sizeof(drvinfo->fw_version),
82 fw_rev_maj(mdev), fw_rev_min(mdev),
83 fw_rev_sub(mdev), mdev->board_id);
86 static void mlx5e_uplink_rep_get_drvinfo(struct net_device *dev,
87 struct ethtool_drvinfo *drvinfo)
89 struct mlx5e_priv *priv = netdev_priv(dev);
91 mlx5e_rep_get_drvinfo(dev, drvinfo);
92 strlcpy(drvinfo->bus_info, pci_name(priv->mdev->pdev),
93 sizeof(drvinfo->bus_info));
96 static const struct counter_desc sw_rep_stats_desc[] = {
97 { MLX5E_DECLARE_STAT(struct mlx5e_sw_stats, rx_packets) },
98 { MLX5E_DECLARE_STAT(struct mlx5e_sw_stats, rx_bytes) },
99 { MLX5E_DECLARE_STAT(struct mlx5e_sw_stats, tx_packets) },
100 { MLX5E_DECLARE_STAT(struct mlx5e_sw_stats, tx_bytes) },
104 u64 vport_rx_packets;
105 u64 vport_tx_packets;
110 static const struct counter_desc vport_rep_stats_desc[] = {
111 { MLX5E_DECLARE_STAT(struct vport_stats, vport_rx_packets) },
112 { MLX5E_DECLARE_STAT(struct vport_stats, vport_rx_bytes) },
113 { MLX5E_DECLARE_STAT(struct vport_stats, vport_tx_packets) },
114 { MLX5E_DECLARE_STAT(struct vport_stats, vport_tx_bytes) },
117 #define NUM_VPORT_REP_SW_COUNTERS ARRAY_SIZE(sw_rep_stats_desc)
118 #define NUM_VPORT_REP_HW_COUNTERS ARRAY_SIZE(vport_rep_stats_desc)
120 static MLX5E_DECLARE_STATS_GRP_OP_NUM_STATS(sw_rep)
122 return NUM_VPORT_REP_SW_COUNTERS;
125 static MLX5E_DECLARE_STATS_GRP_OP_FILL_STRS(sw_rep)
129 for (i = 0; i < NUM_VPORT_REP_SW_COUNTERS; i++)
130 strcpy(data + (idx++) * ETH_GSTRING_LEN,
131 sw_rep_stats_desc[i].format);
135 static MLX5E_DECLARE_STATS_GRP_OP_FILL_STATS(sw_rep)
139 for (i = 0; i < NUM_VPORT_REP_SW_COUNTERS; i++)
140 data[idx++] = MLX5E_READ_CTR64_CPU(&priv->stats.sw,
141 sw_rep_stats_desc, i);
145 static MLX5E_DECLARE_STATS_GRP_OP_UPDATE_STATS(sw_rep)
147 struct mlx5e_sw_stats *s = &priv->stats.sw;
148 struct rtnl_link_stats64 stats64 = {};
150 memset(s, 0, sizeof(*s));
151 mlx5e_fold_sw_stats64(priv, &stats64);
153 s->rx_packets = stats64.rx_packets;
154 s->rx_bytes = stats64.rx_bytes;
155 s->tx_packets = stats64.tx_packets;
156 s->tx_bytes = stats64.tx_bytes;
157 s->tx_queue_dropped = stats64.tx_dropped;
160 static MLX5E_DECLARE_STATS_GRP_OP_NUM_STATS(vport_rep)
162 return NUM_VPORT_REP_HW_COUNTERS;
165 static MLX5E_DECLARE_STATS_GRP_OP_FILL_STRS(vport_rep)
169 for (i = 0; i < NUM_VPORT_REP_HW_COUNTERS; i++)
170 strcpy(data + (idx++) * ETH_GSTRING_LEN, vport_rep_stats_desc[i].format);
174 static MLX5E_DECLARE_STATS_GRP_OP_FILL_STATS(vport_rep)
178 for (i = 0; i < NUM_VPORT_REP_HW_COUNTERS; i++)
179 data[idx++] = MLX5E_READ_CTR64_CPU(&priv->stats.vf_vport,
180 vport_rep_stats_desc, i);
184 static MLX5E_DECLARE_STATS_GRP_OP_UPDATE_STATS(vport_rep)
186 struct mlx5_eswitch *esw = priv->mdev->priv.eswitch;
187 struct mlx5e_rep_priv *rpriv = priv->ppriv;
188 struct mlx5_eswitch_rep *rep = rpriv->rep;
189 struct rtnl_link_stats64 *vport_stats;
190 struct ifla_vf_stats vf_stats;
193 err = mlx5_eswitch_get_vport_stats(esw, rep->vport, &vf_stats);
195 netdev_warn(priv->netdev, "vport %d error %d reading stats\n",
200 vport_stats = &priv->stats.vf_vport;
201 /* flip tx/rx as we are reporting the counters for the switch vport */
202 vport_stats->rx_packets = vf_stats.tx_packets;
203 vport_stats->rx_bytes = vf_stats.tx_bytes;
204 vport_stats->tx_packets = vf_stats.rx_packets;
205 vport_stats->tx_bytes = vf_stats.rx_bytes;
208 static void mlx5e_rep_get_strings(struct net_device *dev,
209 u32 stringset, uint8_t *data)
211 struct mlx5e_priv *priv = netdev_priv(dev);
215 mlx5e_stats_fill_strings(priv, data);
220 static void mlx5e_rep_get_ethtool_stats(struct net_device *dev,
221 struct ethtool_stats *stats, u64 *data)
223 struct mlx5e_priv *priv = netdev_priv(dev);
225 mlx5e_ethtool_get_ethtool_stats(priv, stats, data);
228 static int mlx5e_rep_get_sset_count(struct net_device *dev, int sset)
230 struct mlx5e_priv *priv = netdev_priv(dev);
234 return mlx5e_stats_total_num(priv);
240 static void mlx5e_rep_get_ringparam(struct net_device *dev,
241 struct ethtool_ringparam *param)
243 struct mlx5e_priv *priv = netdev_priv(dev);
245 mlx5e_ethtool_get_ringparam(priv, param);
248 static int mlx5e_rep_set_ringparam(struct net_device *dev,
249 struct ethtool_ringparam *param)
251 struct mlx5e_priv *priv = netdev_priv(dev);
253 return mlx5e_ethtool_set_ringparam(priv, param);
256 static void mlx5e_rep_get_channels(struct net_device *dev,
257 struct ethtool_channels *ch)
259 struct mlx5e_priv *priv = netdev_priv(dev);
261 mlx5e_ethtool_get_channels(priv, ch);
264 static int mlx5e_rep_set_channels(struct net_device *dev,
265 struct ethtool_channels *ch)
267 struct mlx5e_priv *priv = netdev_priv(dev);
269 return mlx5e_ethtool_set_channels(priv, ch);
272 static int mlx5e_rep_get_coalesce(struct net_device *netdev,
273 struct ethtool_coalesce *coal)
275 struct mlx5e_priv *priv = netdev_priv(netdev);
277 return mlx5e_ethtool_get_coalesce(priv, coal);
280 static int mlx5e_rep_set_coalesce(struct net_device *netdev,
281 struct ethtool_coalesce *coal)
283 struct mlx5e_priv *priv = netdev_priv(netdev);
285 return mlx5e_ethtool_set_coalesce(priv, coal);
288 static u32 mlx5e_rep_get_rxfh_key_size(struct net_device *netdev)
290 struct mlx5e_priv *priv = netdev_priv(netdev);
292 return mlx5e_ethtool_get_rxfh_key_size(priv);
295 static u32 mlx5e_rep_get_rxfh_indir_size(struct net_device *netdev)
297 struct mlx5e_priv *priv = netdev_priv(netdev);
299 return mlx5e_ethtool_get_rxfh_indir_size(priv);
302 static void mlx5e_uplink_rep_get_pauseparam(struct net_device *netdev,
303 struct ethtool_pauseparam *pauseparam)
305 struct mlx5e_priv *priv = netdev_priv(netdev);
307 mlx5e_ethtool_get_pauseparam(priv, pauseparam);
310 static int mlx5e_uplink_rep_set_pauseparam(struct net_device *netdev,
311 struct ethtool_pauseparam *pauseparam)
313 struct mlx5e_priv *priv = netdev_priv(netdev);
315 return mlx5e_ethtool_set_pauseparam(priv, pauseparam);
318 static int mlx5e_uplink_rep_get_link_ksettings(struct net_device *netdev,
319 struct ethtool_link_ksettings *link_ksettings)
321 struct mlx5e_priv *priv = netdev_priv(netdev);
323 return mlx5e_ethtool_get_link_ksettings(priv, link_ksettings);
326 static int mlx5e_uplink_rep_set_link_ksettings(struct net_device *netdev,
327 const struct ethtool_link_ksettings *link_ksettings)
329 struct mlx5e_priv *priv = netdev_priv(netdev);
331 return mlx5e_ethtool_set_link_ksettings(priv, link_ksettings);
334 static const struct ethtool_ops mlx5e_rep_ethtool_ops = {
335 .supported_coalesce_params = ETHTOOL_COALESCE_USECS |
336 ETHTOOL_COALESCE_MAX_FRAMES |
337 ETHTOOL_COALESCE_USE_ADAPTIVE,
338 .get_drvinfo = mlx5e_rep_get_drvinfo,
339 .get_link = ethtool_op_get_link,
340 .get_strings = mlx5e_rep_get_strings,
341 .get_sset_count = mlx5e_rep_get_sset_count,
342 .get_ethtool_stats = mlx5e_rep_get_ethtool_stats,
343 .get_ringparam = mlx5e_rep_get_ringparam,
344 .set_ringparam = mlx5e_rep_set_ringparam,
345 .get_channels = mlx5e_rep_get_channels,
346 .set_channels = mlx5e_rep_set_channels,
347 .get_coalesce = mlx5e_rep_get_coalesce,
348 .set_coalesce = mlx5e_rep_set_coalesce,
349 .get_rxfh_key_size = mlx5e_rep_get_rxfh_key_size,
350 .get_rxfh_indir_size = mlx5e_rep_get_rxfh_indir_size,
353 static const struct ethtool_ops mlx5e_uplink_rep_ethtool_ops = {
354 .supported_coalesce_params = ETHTOOL_COALESCE_USECS |
355 ETHTOOL_COALESCE_MAX_FRAMES |
356 ETHTOOL_COALESCE_USE_ADAPTIVE,
357 .get_drvinfo = mlx5e_uplink_rep_get_drvinfo,
358 .get_link = ethtool_op_get_link,
359 .get_strings = mlx5e_rep_get_strings,
360 .get_sset_count = mlx5e_rep_get_sset_count,
361 .get_ethtool_stats = mlx5e_rep_get_ethtool_stats,
362 .get_ringparam = mlx5e_rep_get_ringparam,
363 .set_ringparam = mlx5e_rep_set_ringparam,
364 .get_channels = mlx5e_rep_get_channels,
365 .set_channels = mlx5e_rep_set_channels,
366 .get_coalesce = mlx5e_rep_get_coalesce,
367 .set_coalesce = mlx5e_rep_set_coalesce,
368 .get_link_ksettings = mlx5e_uplink_rep_get_link_ksettings,
369 .set_link_ksettings = mlx5e_uplink_rep_set_link_ksettings,
370 .get_rxfh_key_size = mlx5e_rep_get_rxfh_key_size,
371 .get_rxfh_indir_size = mlx5e_rep_get_rxfh_indir_size,
372 .get_rxfh = mlx5e_get_rxfh,
373 .set_rxfh = mlx5e_set_rxfh,
374 .get_rxnfc = mlx5e_get_rxnfc,
375 .set_rxnfc = mlx5e_set_rxnfc,
376 .get_pauseparam = mlx5e_uplink_rep_get_pauseparam,
377 .set_pauseparam = mlx5e_uplink_rep_set_pauseparam,
380 static void mlx5e_rep_get_port_parent_id(struct net_device *dev,
381 struct netdev_phys_item_id *ppid)
383 struct mlx5e_priv *priv;
386 priv = netdev_priv(dev);
388 parent_id = mlx5_query_nic_system_image_guid(priv->mdev);
389 ppid->id_len = sizeof(parent_id);
390 memcpy(ppid->id, &parent_id, sizeof(parent_id));
393 static void mlx5e_sqs2vport_stop(struct mlx5_eswitch *esw,
394 struct mlx5_eswitch_rep *rep)
396 struct mlx5e_rep_sq *rep_sq, *tmp;
397 struct mlx5e_rep_priv *rpriv;
399 if (esw->mode != MLX5_ESWITCH_OFFLOADS)
402 rpriv = mlx5e_rep_to_rep_priv(rep);
403 list_for_each_entry_safe(rep_sq, tmp, &rpriv->vport_sqs_list, list) {
404 mlx5_eswitch_del_send_to_vport_rule(rep_sq->send_to_vport_rule);
405 list_del(&rep_sq->list);
410 static int mlx5e_sqs2vport_start(struct mlx5_eswitch *esw,
411 struct mlx5_eswitch_rep *rep,
412 u32 *sqns_array, int sqns_num)
414 struct mlx5_flow_handle *flow_rule;
415 struct mlx5e_rep_priv *rpriv;
416 struct mlx5e_rep_sq *rep_sq;
420 if (esw->mode != MLX5_ESWITCH_OFFLOADS)
423 rpriv = mlx5e_rep_to_rep_priv(rep);
424 for (i = 0; i < sqns_num; i++) {
425 rep_sq = kzalloc(sizeof(*rep_sq), GFP_KERNEL);
431 /* Add re-inject rule to the PF/representor sqs */
432 flow_rule = mlx5_eswitch_add_send_to_vport_rule(esw,
435 if (IS_ERR(flow_rule)) {
436 err = PTR_ERR(flow_rule);
440 rep_sq->send_to_vport_rule = flow_rule;
441 list_add(&rep_sq->list, &rpriv->vport_sqs_list);
446 mlx5e_sqs2vport_stop(esw, rep);
450 int mlx5e_add_sqs_fwd_rules(struct mlx5e_priv *priv)
452 struct mlx5_eswitch *esw = priv->mdev->priv.eswitch;
453 struct mlx5e_rep_priv *rpriv = priv->ppriv;
454 struct mlx5_eswitch_rep *rep = rpriv->rep;
455 struct mlx5e_channel *c;
456 int n, tc, num_sqs = 0;
460 sqs = kcalloc(priv->channels.num * priv->channels.params.num_tc, sizeof(*sqs), GFP_KERNEL);
464 for (n = 0; n < priv->channels.num; n++) {
465 c = priv->channels.c[n];
466 for (tc = 0; tc < c->num_tc; tc++)
467 sqs[num_sqs++] = c->sq[tc].sqn;
470 err = mlx5e_sqs2vport_start(esw, rep, sqs, num_sqs);
475 netdev_warn(priv->netdev, "Failed to add SQs FWD rules %d\n", err);
479 void mlx5e_remove_sqs_fwd_rules(struct mlx5e_priv *priv)
481 struct mlx5_eswitch *esw = priv->mdev->priv.eswitch;
482 struct mlx5e_rep_priv *rpriv = priv->ppriv;
483 struct mlx5_eswitch_rep *rep = rpriv->rep;
485 mlx5e_sqs2vport_stop(esw, rep);
488 static unsigned long mlx5e_rep_ipv6_interval(void)
490 if (IS_ENABLED(CONFIG_IPV6) && ipv6_stub->nd_tbl)
491 return NEIGH_VAR(&ipv6_stub->nd_tbl->parms, DELAY_PROBE_TIME);
496 static void mlx5e_rep_neigh_update_init_interval(struct mlx5e_rep_priv *rpriv)
498 unsigned long ipv4_interval = NEIGH_VAR(&arp_tbl.parms, DELAY_PROBE_TIME);
499 unsigned long ipv6_interval = mlx5e_rep_ipv6_interval();
500 struct net_device *netdev = rpriv->netdev;
501 struct mlx5e_priv *priv = netdev_priv(netdev);
503 rpriv->neigh_update.min_interval = min_t(unsigned long, ipv6_interval, ipv4_interval);
504 mlx5_fc_update_sampling_interval(priv->mdev, rpriv->neigh_update.min_interval);
507 void mlx5e_rep_queue_neigh_stats_work(struct mlx5e_priv *priv)
509 struct mlx5e_rep_priv *rpriv = priv->ppriv;
510 struct mlx5e_neigh_update_table *neigh_update = &rpriv->neigh_update;
512 mlx5_fc_queue_stats_work(priv->mdev,
513 &neigh_update->neigh_stats_work,
514 neigh_update->min_interval);
517 static bool mlx5e_rep_neigh_entry_hold(struct mlx5e_neigh_hash_entry *nhe)
519 return refcount_inc_not_zero(&nhe->refcnt);
522 static void mlx5e_rep_neigh_entry_remove(struct mlx5e_neigh_hash_entry *nhe);
524 static void mlx5e_rep_neigh_entry_release(struct mlx5e_neigh_hash_entry *nhe)
526 if (refcount_dec_and_test(&nhe->refcnt)) {
527 mlx5e_rep_neigh_entry_remove(nhe);
532 static struct mlx5e_neigh_hash_entry *
533 mlx5e_get_next_nhe(struct mlx5e_rep_priv *rpriv,
534 struct mlx5e_neigh_hash_entry *nhe)
536 struct mlx5e_neigh_hash_entry *next = NULL;
541 list_next_or_null_rcu(&rpriv->neigh_update.neigh_list,
543 struct mlx5e_neigh_hash_entry,
545 list_first_or_null_rcu(&rpriv->neigh_update.neigh_list,
546 struct mlx5e_neigh_hash_entry,
549 next = list_next_or_null_rcu(&rpriv->neigh_update.neigh_list,
551 struct mlx5e_neigh_hash_entry,
553 if (mlx5e_rep_neigh_entry_hold(next))
559 mlx5e_rep_neigh_entry_release(nhe);
564 static void mlx5e_rep_neigh_stats_work(struct work_struct *work)
566 struct mlx5e_rep_priv *rpriv = container_of(work, struct mlx5e_rep_priv,
567 neigh_update.neigh_stats_work.work);
568 struct net_device *netdev = rpriv->netdev;
569 struct mlx5e_priv *priv = netdev_priv(netdev);
570 struct mlx5e_neigh_hash_entry *nhe = NULL;
573 if (!list_empty(&rpriv->neigh_update.neigh_list))
574 mlx5e_rep_queue_neigh_stats_work(priv);
576 while ((nhe = mlx5e_get_next_nhe(rpriv, nhe)) != NULL)
577 mlx5e_tc_update_neigh_used_value(nhe);
582 static void mlx5e_rep_update_flows(struct mlx5e_priv *priv,
583 struct mlx5e_encap_entry *e,
584 bool neigh_connected,
585 unsigned char ha[ETH_ALEN])
587 struct ethhdr *eth = (struct ethhdr *)e->encap_header;
588 struct mlx5_eswitch *esw = priv->mdev->priv.eswitch;
589 bool encap_connected;
590 LIST_HEAD(flow_list);
594 /* wait for encap to be fully initialized */
595 wait_for_completion(&e->res_ready);
597 mutex_lock(&esw->offloads.encap_tbl_lock);
598 encap_connected = !!(e->flags & MLX5_ENCAP_ENTRY_VALID);
599 if (e->compl_result < 0 || (encap_connected == neigh_connected &&
600 ether_addr_equal(e->h_dest, ha)))
603 mlx5e_take_all_encap_flows(e, &flow_list);
605 if ((e->flags & MLX5_ENCAP_ENTRY_VALID) &&
606 (!neigh_connected || !ether_addr_equal(e->h_dest, ha)))
607 mlx5e_tc_encap_flows_del(priv, e, &flow_list);
609 if (neigh_connected && !(e->flags & MLX5_ENCAP_ENTRY_VALID)) {
610 ether_addr_copy(e->h_dest, ha);
611 ether_addr_copy(eth->h_dest, ha);
612 /* Update the encap source mac, in case that we delete
613 * the flows when encap source mac changed.
615 ether_addr_copy(eth->h_source, e->route_dev->dev_addr);
617 mlx5e_tc_encap_flows_add(priv, e, &flow_list);
620 mutex_unlock(&esw->offloads.encap_tbl_lock);
621 mlx5e_put_encap_flow_list(priv, &flow_list);
624 static void mlx5e_rep_neigh_update(struct work_struct *work)
626 struct mlx5e_neigh_hash_entry *nhe =
627 container_of(work, struct mlx5e_neigh_hash_entry, neigh_update_work);
628 struct neighbour *n = nhe->n;
629 struct mlx5e_encap_entry *e;
630 unsigned char ha[ETH_ALEN];
631 struct mlx5e_priv *priv;
632 bool neigh_connected;
637 /* If these parameters are changed after we release the lock,
638 * we'll receive another event letting us know about it.
639 * We use this lock to avoid inconsistency between the neigh validity
640 * and it's hw address.
642 read_lock_bh(&n->lock);
643 memcpy(ha, n->ha, ETH_ALEN);
644 nud_state = n->nud_state;
646 read_unlock_bh(&n->lock);
648 neigh_connected = (nud_state & NUD_VALID) && !dead;
650 trace_mlx5e_rep_neigh_update(nhe, ha, neigh_connected);
652 list_for_each_entry(e, &nhe->encap_list, encap_list) {
653 if (!mlx5e_encap_take(e))
656 priv = netdev_priv(e->out_dev);
657 mlx5e_rep_update_flows(priv, e, neigh_connected, ha);
658 mlx5e_encap_put(priv, e);
660 mlx5e_rep_neigh_entry_release(nhe);
665 static struct mlx5e_rep_indr_block_priv *
666 mlx5e_rep_indr_block_priv_lookup(struct mlx5e_rep_priv *rpriv,
667 struct net_device *netdev)
669 struct mlx5e_rep_indr_block_priv *cb_priv;
671 /* All callback list access should be protected by RTNL. */
674 list_for_each_entry(cb_priv,
675 &rpriv->uplink_priv.tc_indr_block_priv_list,
677 if (cb_priv->netdev == netdev)
683 static void mlx5e_rep_indr_clean_block_privs(struct mlx5e_rep_priv *rpriv)
685 struct mlx5e_rep_indr_block_priv *cb_priv, *temp;
686 struct list_head *head = &rpriv->uplink_priv.tc_indr_block_priv_list;
688 list_for_each_entry_safe(cb_priv, temp, head, list) {
689 mlx5e_rep_indr_unregister_block(rpriv, cb_priv->netdev);
695 mlx5e_rep_indr_offload(struct net_device *netdev,
696 struct flow_cls_offload *flower,
697 struct mlx5e_rep_indr_block_priv *indr_priv,
700 struct mlx5e_priv *priv = netdev_priv(indr_priv->rpriv->netdev);
703 switch (flower->command) {
704 case FLOW_CLS_REPLACE:
705 err = mlx5e_configure_flower(netdev, priv, flower, flags);
707 case FLOW_CLS_DESTROY:
708 err = mlx5e_delete_flower(netdev, priv, flower, flags);
711 err = mlx5e_stats_flower(netdev, priv, flower, flags);
720 static int mlx5e_rep_indr_setup_tc_cb(enum tc_setup_type type,
721 void *type_data, void *indr_priv)
723 unsigned long flags = MLX5_TC_FLAG(EGRESS) | MLX5_TC_FLAG(ESW_OFFLOAD);
724 struct mlx5e_rep_indr_block_priv *priv = indr_priv;
727 case TC_SETUP_CLSFLOWER:
728 return mlx5e_rep_indr_offload(priv->netdev, type_data, priv,
735 static int mlx5e_rep_indr_setup_ft_cb(enum tc_setup_type type,
736 void *type_data, void *indr_priv)
738 struct mlx5e_rep_indr_block_priv *priv = indr_priv;
739 struct flow_cls_offload *f = type_data;
740 struct flow_cls_offload tmp;
741 struct mlx5e_priv *mpriv;
742 struct mlx5_eswitch *esw;
746 mpriv = netdev_priv(priv->rpriv->netdev);
747 esw = mpriv->mdev->priv.eswitch;
749 flags = MLX5_TC_FLAG(EGRESS) |
750 MLX5_TC_FLAG(ESW_OFFLOAD) |
751 MLX5_TC_FLAG(FT_OFFLOAD);
754 case TC_SETUP_CLSFLOWER:
755 memcpy(&tmp, f, sizeof(*f));
757 /* Re-use tc offload path by moving the ft flow to the
760 * FT offload can use prio range [0, INT_MAX], so we normalize
761 * it to range [1, mlx5_esw_chains_get_prio_range(esw)]
762 * as with tc, where prio 0 isn't supported.
764 * We only support chain 0 of FT offload.
766 if (!mlx5_esw_chains_prios_supported(esw) ||
767 tmp.common.prio >= mlx5_esw_chains_get_prio_range(esw) ||
768 tmp.common.chain_index)
771 tmp.common.chain_index = mlx5_esw_chains_get_ft_chain(esw);
773 err = mlx5e_rep_indr_offload(priv->netdev, &tmp, priv, flags);
774 memcpy(&f->stats, &tmp.stats, sizeof(f->stats));
781 static void mlx5e_rep_indr_block_unbind(void *cb_priv)
783 struct mlx5e_rep_indr_block_priv *indr_priv = cb_priv;
785 list_del(&indr_priv->list);
789 static LIST_HEAD(mlx5e_block_cb_list);
792 mlx5e_rep_indr_setup_block(struct net_device *netdev,
793 struct mlx5e_rep_priv *rpriv,
794 struct flow_block_offload *f,
795 flow_setup_cb_t *setup_cb)
797 struct mlx5e_rep_indr_block_priv *indr_priv;
798 struct flow_block_cb *block_cb;
800 if (f->binder_type != FLOW_BLOCK_BINDER_TYPE_CLSACT_INGRESS)
803 f->unlocked_driver_cb = true;
804 f->driver_block_list = &mlx5e_block_cb_list;
806 switch (f->command) {
807 case FLOW_BLOCK_BIND:
808 indr_priv = mlx5e_rep_indr_block_priv_lookup(rpriv, netdev);
812 indr_priv = kmalloc(sizeof(*indr_priv), GFP_KERNEL);
816 indr_priv->netdev = netdev;
817 indr_priv->rpriv = rpriv;
818 list_add(&indr_priv->list,
819 &rpriv->uplink_priv.tc_indr_block_priv_list);
821 block_cb = flow_block_cb_alloc(setup_cb, indr_priv, indr_priv,
822 mlx5e_rep_indr_block_unbind);
823 if (IS_ERR(block_cb)) {
824 list_del(&indr_priv->list);
826 return PTR_ERR(block_cb);
828 flow_block_cb_add(block_cb, f);
829 list_add_tail(&block_cb->driver_list, &mlx5e_block_cb_list);
832 case FLOW_BLOCK_UNBIND:
833 indr_priv = mlx5e_rep_indr_block_priv_lookup(rpriv, netdev);
837 block_cb = flow_block_cb_lookup(f->block, setup_cb, indr_priv);
841 flow_block_cb_remove(block_cb, f);
842 list_del(&block_cb->driver_list);
851 int mlx5e_rep_indr_setup_cb(struct net_device *netdev, void *cb_priv,
852 enum tc_setup_type type, void *type_data)
856 return mlx5e_rep_indr_setup_block(netdev, cb_priv, type_data,
857 mlx5e_rep_indr_setup_tc_cb);
859 return mlx5e_rep_indr_setup_block(netdev, cb_priv, type_data,
860 mlx5e_rep_indr_setup_ft_cb);
866 static int mlx5e_rep_indr_register_block(struct mlx5e_rep_priv *rpriv,
867 struct net_device *netdev)
871 err = __flow_indr_block_cb_register(netdev, rpriv,
872 mlx5e_rep_indr_setup_cb,
875 struct mlx5e_priv *priv = netdev_priv(rpriv->netdev);
877 mlx5_core_err(priv->mdev, "Failed to register remote block notifier for %s err=%d\n",
878 netdev_name(netdev), err);
883 static void mlx5e_rep_indr_unregister_block(struct mlx5e_rep_priv *rpriv,
884 struct net_device *netdev)
886 __flow_indr_block_cb_unregister(netdev, mlx5e_rep_indr_setup_cb,
890 static int mlx5e_nic_rep_netdevice_event(struct notifier_block *nb,
891 unsigned long event, void *ptr)
893 struct mlx5e_rep_priv *rpriv = container_of(nb, struct mlx5e_rep_priv,
894 uplink_priv.netdevice_nb);
895 struct mlx5e_priv *priv = netdev_priv(rpriv->netdev);
896 struct net_device *netdev = netdev_notifier_info_to_dev(ptr);
898 if (!mlx5e_tc_tun_device_to_offload(priv, netdev) &&
899 !(is_vlan_dev(netdev) && vlan_dev_real_dev(netdev) == rpriv->netdev))
903 case NETDEV_REGISTER:
904 mlx5e_rep_indr_register_block(rpriv, netdev);
906 case NETDEV_UNREGISTER:
907 mlx5e_rep_indr_unregister_block(rpriv, netdev);
914 mlx5e_rep_queue_neigh_update_work(struct mlx5e_priv *priv,
915 struct mlx5e_neigh_hash_entry *nhe,
918 /* Take a reference to ensure the neighbour and mlx5 encap
919 * entry won't be destructed until we drop the reference in
924 /* This assignment is valid as long as the the neigh reference
929 if (!queue_work(priv->wq, &nhe->neigh_update_work)) {
930 mlx5e_rep_neigh_entry_release(nhe);
935 static struct mlx5e_neigh_hash_entry *
936 mlx5e_rep_neigh_entry_lookup(struct mlx5e_priv *priv,
937 struct mlx5e_neigh *m_neigh);
939 static int mlx5e_rep_netevent_event(struct notifier_block *nb,
940 unsigned long event, void *ptr)
942 struct mlx5e_rep_priv *rpriv = container_of(nb, struct mlx5e_rep_priv,
943 neigh_update.netevent_nb);
944 struct mlx5e_neigh_update_table *neigh_update = &rpriv->neigh_update;
945 struct net_device *netdev = rpriv->netdev;
946 struct mlx5e_priv *priv = netdev_priv(netdev);
947 struct mlx5e_neigh_hash_entry *nhe = NULL;
948 struct mlx5e_neigh m_neigh = {};
949 struct neigh_parms *p;
954 case NETEVENT_NEIGH_UPDATE:
956 #if IS_ENABLED(CONFIG_IPV6)
957 if (n->tbl != ipv6_stub->nd_tbl && n->tbl != &arp_tbl)
959 if (n->tbl != &arp_tbl)
963 m_neigh.dev = n->dev;
964 m_neigh.family = n->ops->family;
965 memcpy(&m_neigh.dst_ip, n->primary_key, n->tbl->key_len);
968 nhe = mlx5e_rep_neigh_entry_lookup(priv, &m_neigh);
973 mlx5e_rep_queue_neigh_update_work(priv, nhe, n);
976 case NETEVENT_DELAY_PROBE_TIME_UPDATE:
979 /* We check the device is present since we don't care about
980 * changes in the default table, we only care about changes
981 * done per device delay prob time parameter.
983 #if IS_ENABLED(CONFIG_IPV6)
984 if (!p->dev || (p->tbl != ipv6_stub->nd_tbl && p->tbl != &arp_tbl))
986 if (!p->dev || p->tbl != &arp_tbl)
991 list_for_each_entry_rcu(nhe, &neigh_update->neigh_list,
993 if (p->dev == nhe->m_neigh.dev) {
1002 neigh_update->min_interval = min_t(unsigned long,
1003 NEIGH_VAR(p, DELAY_PROBE_TIME),
1004 neigh_update->min_interval);
1005 mlx5_fc_update_sampling_interval(priv->mdev,
1006 neigh_update->min_interval);
1012 static const struct rhashtable_params mlx5e_neigh_ht_params = {
1013 .head_offset = offsetof(struct mlx5e_neigh_hash_entry, rhash_node),
1014 .key_offset = offsetof(struct mlx5e_neigh_hash_entry, m_neigh),
1015 .key_len = sizeof(struct mlx5e_neigh),
1016 .automatic_shrinking = true,
1019 static int mlx5e_rep_neigh_init(struct mlx5e_rep_priv *rpriv)
1021 struct mlx5e_neigh_update_table *neigh_update = &rpriv->neigh_update;
1024 err = rhashtable_init(&neigh_update->neigh_ht, &mlx5e_neigh_ht_params);
1028 INIT_LIST_HEAD(&neigh_update->neigh_list);
1029 mutex_init(&neigh_update->encap_lock);
1030 INIT_DELAYED_WORK(&neigh_update->neigh_stats_work,
1031 mlx5e_rep_neigh_stats_work);
1032 mlx5e_rep_neigh_update_init_interval(rpriv);
1034 rpriv->neigh_update.netevent_nb.notifier_call = mlx5e_rep_netevent_event;
1035 err = register_netevent_notifier(&rpriv->neigh_update.netevent_nb);
1041 rhashtable_destroy(&neigh_update->neigh_ht);
1045 static void mlx5e_rep_neigh_cleanup(struct mlx5e_rep_priv *rpriv)
1047 struct mlx5e_neigh_update_table *neigh_update = &rpriv->neigh_update;
1048 struct mlx5e_priv *priv = netdev_priv(rpriv->netdev);
1050 unregister_netevent_notifier(&neigh_update->netevent_nb);
1052 flush_workqueue(priv->wq); /* flush neigh update works */
1054 cancel_delayed_work_sync(&rpriv->neigh_update.neigh_stats_work);
1056 mutex_destroy(&neigh_update->encap_lock);
1057 rhashtable_destroy(&neigh_update->neigh_ht);
1060 static int mlx5e_rep_neigh_entry_insert(struct mlx5e_priv *priv,
1061 struct mlx5e_neigh_hash_entry *nhe)
1063 struct mlx5e_rep_priv *rpriv = priv->ppriv;
1066 err = rhashtable_insert_fast(&rpriv->neigh_update.neigh_ht,
1068 mlx5e_neigh_ht_params);
1072 list_add_rcu(&nhe->neigh_list, &rpriv->neigh_update.neigh_list);
1077 static void mlx5e_rep_neigh_entry_remove(struct mlx5e_neigh_hash_entry *nhe)
1079 struct mlx5e_rep_priv *rpriv = nhe->priv->ppriv;
1081 mutex_lock(&rpriv->neigh_update.encap_lock);
1083 list_del_rcu(&nhe->neigh_list);
1085 rhashtable_remove_fast(&rpriv->neigh_update.neigh_ht,
1087 mlx5e_neigh_ht_params);
1088 mutex_unlock(&rpriv->neigh_update.encap_lock);
1091 /* This function must only be called under the representor's encap_lock or
1092 * inside rcu read lock section.
1094 static struct mlx5e_neigh_hash_entry *
1095 mlx5e_rep_neigh_entry_lookup(struct mlx5e_priv *priv,
1096 struct mlx5e_neigh *m_neigh)
1098 struct mlx5e_rep_priv *rpriv = priv->ppriv;
1099 struct mlx5e_neigh_update_table *neigh_update = &rpriv->neigh_update;
1100 struct mlx5e_neigh_hash_entry *nhe;
1102 nhe = rhashtable_lookup_fast(&neigh_update->neigh_ht, m_neigh,
1103 mlx5e_neigh_ht_params);
1104 return nhe && mlx5e_rep_neigh_entry_hold(nhe) ? nhe : NULL;
1107 static int mlx5e_rep_neigh_entry_create(struct mlx5e_priv *priv,
1108 struct mlx5e_encap_entry *e,
1109 struct mlx5e_neigh_hash_entry **nhe)
1113 *nhe = kzalloc(sizeof(**nhe), GFP_KERNEL);
1117 (*nhe)->priv = priv;
1118 memcpy(&(*nhe)->m_neigh, &e->m_neigh, sizeof(e->m_neigh));
1119 INIT_WORK(&(*nhe)->neigh_update_work, mlx5e_rep_neigh_update);
1120 spin_lock_init(&(*nhe)->encap_list_lock);
1121 INIT_LIST_HEAD(&(*nhe)->encap_list);
1122 refcount_set(&(*nhe)->refcnt, 1);
1124 err = mlx5e_rep_neigh_entry_insert(priv, *nhe);
1134 int mlx5e_rep_encap_entry_attach(struct mlx5e_priv *priv,
1135 struct mlx5e_encap_entry *e)
1137 struct mlx5e_rep_priv *rpriv = priv->ppriv;
1138 struct mlx5_rep_uplink_priv *uplink_priv = &rpriv->uplink_priv;
1139 struct mlx5_tun_entropy *tun_entropy = &uplink_priv->tun_entropy;
1140 struct mlx5e_neigh_hash_entry *nhe;
1143 err = mlx5_tun_entropy_refcount_inc(tun_entropy, e->reformat_type);
1147 mutex_lock(&rpriv->neigh_update.encap_lock);
1148 nhe = mlx5e_rep_neigh_entry_lookup(priv, &e->m_neigh);
1150 err = mlx5e_rep_neigh_entry_create(priv, e, &nhe);
1152 mutex_unlock(&rpriv->neigh_update.encap_lock);
1153 mlx5_tun_entropy_refcount_dec(tun_entropy,
1160 spin_lock(&nhe->encap_list_lock);
1161 list_add_rcu(&e->encap_list, &nhe->encap_list);
1162 spin_unlock(&nhe->encap_list_lock);
1164 mutex_unlock(&rpriv->neigh_update.encap_lock);
1169 void mlx5e_rep_encap_entry_detach(struct mlx5e_priv *priv,
1170 struct mlx5e_encap_entry *e)
1172 struct mlx5e_rep_priv *rpriv = priv->ppriv;
1173 struct mlx5_rep_uplink_priv *uplink_priv = &rpriv->uplink_priv;
1174 struct mlx5_tun_entropy *tun_entropy = &uplink_priv->tun_entropy;
1179 spin_lock(&e->nhe->encap_list_lock);
1180 list_del_rcu(&e->encap_list);
1181 spin_unlock(&e->nhe->encap_list_lock);
1183 mlx5e_rep_neigh_entry_release(e->nhe);
1185 mlx5_tun_entropy_refcount_dec(tun_entropy, e->reformat_type);
1188 static int mlx5e_rep_open(struct net_device *dev)
1190 struct mlx5e_priv *priv = netdev_priv(dev);
1191 struct mlx5e_rep_priv *rpriv = priv->ppriv;
1192 struct mlx5_eswitch_rep *rep = rpriv->rep;
1195 mutex_lock(&priv->state_lock);
1196 err = mlx5e_open_locked(dev);
1200 if (!mlx5_modify_vport_admin_state(priv->mdev,
1201 MLX5_VPORT_STATE_OP_MOD_ESW_VPORT,
1203 MLX5_VPORT_ADMIN_STATE_UP))
1204 netif_carrier_on(dev);
1207 mutex_unlock(&priv->state_lock);
1211 static int mlx5e_rep_close(struct net_device *dev)
1213 struct mlx5e_priv *priv = netdev_priv(dev);
1214 struct mlx5e_rep_priv *rpriv = priv->ppriv;
1215 struct mlx5_eswitch_rep *rep = rpriv->rep;
1218 mutex_lock(&priv->state_lock);
1219 mlx5_modify_vport_admin_state(priv->mdev,
1220 MLX5_VPORT_STATE_OP_MOD_ESW_VPORT,
1222 MLX5_VPORT_ADMIN_STATE_DOWN);
1223 ret = mlx5e_close_locked(dev);
1224 mutex_unlock(&priv->state_lock);
1229 mlx5e_rep_setup_tc_cls_flower(struct mlx5e_priv *priv,
1230 struct flow_cls_offload *cls_flower, int flags)
1232 switch (cls_flower->command) {
1233 case FLOW_CLS_REPLACE:
1234 return mlx5e_configure_flower(priv->netdev, priv, cls_flower,
1236 case FLOW_CLS_DESTROY:
1237 return mlx5e_delete_flower(priv->netdev, priv, cls_flower,
1239 case FLOW_CLS_STATS:
1240 return mlx5e_stats_flower(priv->netdev, priv, cls_flower,
1248 int mlx5e_rep_setup_tc_cls_matchall(struct mlx5e_priv *priv,
1249 struct tc_cls_matchall_offload *ma)
1251 switch (ma->command) {
1252 case TC_CLSMATCHALL_REPLACE:
1253 return mlx5e_tc_configure_matchall(priv, ma);
1254 case TC_CLSMATCHALL_DESTROY:
1255 return mlx5e_tc_delete_matchall(priv, ma);
1256 case TC_CLSMATCHALL_STATS:
1257 mlx5e_tc_stats_matchall(priv, ma);
1264 static int mlx5e_rep_setup_tc_cb(enum tc_setup_type type, void *type_data,
1267 unsigned long flags = MLX5_TC_FLAG(INGRESS) | MLX5_TC_FLAG(ESW_OFFLOAD);
1268 struct mlx5e_priv *priv = cb_priv;
1271 case TC_SETUP_CLSFLOWER:
1272 return mlx5e_rep_setup_tc_cls_flower(priv, type_data, flags);
1273 case TC_SETUP_CLSMATCHALL:
1274 return mlx5e_rep_setup_tc_cls_matchall(priv, type_data);
1280 static int mlx5e_rep_setup_ft_cb(enum tc_setup_type type, void *type_data,
1283 struct flow_cls_offload tmp, *f = type_data;
1284 struct mlx5e_priv *priv = cb_priv;
1285 struct mlx5_eswitch *esw;
1286 unsigned long flags;
1289 flags = MLX5_TC_FLAG(INGRESS) |
1290 MLX5_TC_FLAG(ESW_OFFLOAD) |
1291 MLX5_TC_FLAG(FT_OFFLOAD);
1292 esw = priv->mdev->priv.eswitch;
1295 case TC_SETUP_CLSFLOWER:
1296 memcpy(&tmp, f, sizeof(*f));
1298 if (!mlx5_esw_chains_prios_supported(esw))
1301 /* Re-use tc offload path by moving the ft flow to the
1302 * reserved ft chain.
1304 * FT offload can use prio range [0, INT_MAX], so we normalize
1305 * it to range [1, mlx5_esw_chains_get_prio_range(esw)]
1306 * as with tc, where prio 0 isn't supported.
1308 * We only support chain 0 of FT offload.
1310 if (tmp.common.prio >= mlx5_esw_chains_get_prio_range(esw))
1312 if (tmp.common.chain_index != 0)
1315 tmp.common.chain_index = mlx5_esw_chains_get_ft_chain(esw);
1317 err = mlx5e_rep_setup_tc_cls_flower(priv, &tmp, flags);
1318 memcpy(&f->stats, &tmp.stats, sizeof(f->stats));
1325 static LIST_HEAD(mlx5e_rep_block_tc_cb_list);
1326 static LIST_HEAD(mlx5e_rep_block_ft_cb_list);
1327 static int mlx5e_rep_setup_tc(struct net_device *dev, enum tc_setup_type type,
1330 struct mlx5e_priv *priv = netdev_priv(dev);
1331 struct flow_block_offload *f = type_data;
1333 f->unlocked_driver_cb = true;
1336 case TC_SETUP_BLOCK:
1337 return flow_block_cb_setup_simple(type_data,
1338 &mlx5e_rep_block_tc_cb_list,
1339 mlx5e_rep_setup_tc_cb,
1342 return flow_block_cb_setup_simple(type_data,
1343 &mlx5e_rep_block_ft_cb_list,
1344 mlx5e_rep_setup_ft_cb,
1351 bool mlx5e_is_uplink_rep(struct mlx5e_priv *priv)
1353 struct mlx5e_rep_priv *rpriv = priv->ppriv;
1354 struct mlx5_eswitch_rep *rep;
1356 if (!MLX5_ESWITCH_MANAGER(priv->mdev))
1359 if (!rpriv) /* non vport rep mlx5e instances don't use this field */
1363 return (rep->vport == MLX5_VPORT_UPLINK);
1366 static bool mlx5e_rep_has_offload_stats(const struct net_device *dev, int attr_id)
1369 case IFLA_OFFLOAD_XSTATS_CPU_HIT:
1377 mlx5e_get_sw_stats64(const struct net_device *dev,
1378 struct rtnl_link_stats64 *stats)
1380 struct mlx5e_priv *priv = netdev_priv(dev);
1382 mlx5e_fold_sw_stats64(priv, stats);
1386 static int mlx5e_rep_get_offload_stats(int attr_id, const struct net_device *dev,
1390 case IFLA_OFFLOAD_XSTATS_CPU_HIT:
1391 return mlx5e_get_sw_stats64(dev, sp);
1398 mlx5e_rep_get_stats(struct net_device *dev, struct rtnl_link_stats64 *stats)
1400 struct mlx5e_priv *priv = netdev_priv(dev);
1402 /* update HW stats in background for next time */
1403 mlx5e_queue_update_stats(priv);
1404 memcpy(stats, &priv->stats.vf_vport, sizeof(*stats));
1407 static int mlx5e_rep_change_mtu(struct net_device *netdev, int new_mtu)
1409 return mlx5e_change_mtu(netdev, new_mtu, NULL);
1412 static int mlx5e_uplink_rep_change_mtu(struct net_device *netdev, int new_mtu)
1414 return mlx5e_change_mtu(netdev, new_mtu, mlx5e_set_dev_port_mtu_ctx);
1417 static int mlx5e_uplink_rep_set_mac(struct net_device *netdev, void *addr)
1419 struct sockaddr *saddr = addr;
1421 if (!is_valid_ether_addr(saddr->sa_data))
1422 return -EADDRNOTAVAIL;
1424 ether_addr_copy(netdev->dev_addr, saddr->sa_data);
1428 static int mlx5e_uplink_rep_set_vf_vlan(struct net_device *dev, int vf, u16 vlan, u8 qos,
1431 netdev_warn_once(dev, "legacy vf vlan setting isn't supported in switchdev mode\n");
1436 /* allow setting 0-vid for compatibility with libvirt */
1440 static struct devlink_port *mlx5e_rep_get_devlink_port(struct net_device *dev)
1442 struct mlx5e_priv *priv = netdev_priv(dev);
1443 struct mlx5e_rep_priv *rpriv = priv->ppriv;
1445 return &rpriv->dl_port;
1448 static const struct net_device_ops mlx5e_netdev_ops_rep = {
1449 .ndo_open = mlx5e_rep_open,
1450 .ndo_stop = mlx5e_rep_close,
1451 .ndo_start_xmit = mlx5e_xmit,
1452 .ndo_setup_tc = mlx5e_rep_setup_tc,
1453 .ndo_get_devlink_port = mlx5e_rep_get_devlink_port,
1454 .ndo_get_stats64 = mlx5e_rep_get_stats,
1455 .ndo_has_offload_stats = mlx5e_rep_has_offload_stats,
1456 .ndo_get_offload_stats = mlx5e_rep_get_offload_stats,
1457 .ndo_change_mtu = mlx5e_rep_change_mtu,
1460 static const struct net_device_ops mlx5e_netdev_ops_uplink_rep = {
1461 .ndo_open = mlx5e_open,
1462 .ndo_stop = mlx5e_close,
1463 .ndo_start_xmit = mlx5e_xmit,
1464 .ndo_set_mac_address = mlx5e_uplink_rep_set_mac,
1465 .ndo_setup_tc = mlx5e_rep_setup_tc,
1466 .ndo_get_devlink_port = mlx5e_rep_get_devlink_port,
1467 .ndo_get_stats64 = mlx5e_get_stats,
1468 .ndo_has_offload_stats = mlx5e_rep_has_offload_stats,
1469 .ndo_get_offload_stats = mlx5e_rep_get_offload_stats,
1470 .ndo_change_mtu = mlx5e_uplink_rep_change_mtu,
1471 .ndo_udp_tunnel_add = mlx5e_add_vxlan_port,
1472 .ndo_udp_tunnel_del = mlx5e_del_vxlan_port,
1473 .ndo_features_check = mlx5e_features_check,
1474 .ndo_set_vf_mac = mlx5e_set_vf_mac,
1475 .ndo_set_vf_rate = mlx5e_set_vf_rate,
1476 .ndo_get_vf_config = mlx5e_get_vf_config,
1477 .ndo_get_vf_stats = mlx5e_get_vf_stats,
1478 .ndo_set_vf_vlan = mlx5e_uplink_rep_set_vf_vlan,
1479 .ndo_set_features = mlx5e_set_features,
1482 bool mlx5e_eswitch_uplink_rep(struct net_device *netdev)
1484 return netdev->netdev_ops == &mlx5e_netdev_ops_uplink_rep;
1487 bool mlx5e_eswitch_rep(struct net_device *netdev)
1489 if (netdev->netdev_ops == &mlx5e_netdev_ops_rep ||
1490 netdev->netdev_ops == &mlx5e_netdev_ops_uplink_rep)
1496 static void mlx5e_build_rep_params(struct net_device *netdev)
1498 struct mlx5e_priv *priv = netdev_priv(netdev);
1499 struct mlx5e_rep_priv *rpriv = priv->ppriv;
1500 struct mlx5_eswitch_rep *rep = rpriv->rep;
1501 struct mlx5_core_dev *mdev = priv->mdev;
1502 struct mlx5e_params *params;
1504 u8 cq_period_mode = MLX5_CAP_GEN(mdev, cq_period_start_from_cqe) ?
1505 MLX5_CQ_PERIOD_MODE_START_FROM_CQE :
1506 MLX5_CQ_PERIOD_MODE_START_FROM_EQE;
1508 params = &priv->channels.params;
1509 params->hard_mtu = MLX5E_ETH_HARD_MTU;
1510 params->sw_mtu = netdev->mtu;
1513 if (rep->vport == MLX5_VPORT_UPLINK)
1514 params->log_sq_size = MLX5E_PARAMS_DEFAULT_LOG_SQ_SIZE;
1516 params->log_sq_size = MLX5E_REP_PARAMS_DEF_LOG_SQ_SIZE;
1519 mlx5e_build_rq_params(mdev, params);
1521 /* CQ moderation params */
1522 params->rx_dim_enabled = MLX5_CAP_GEN(mdev, cq_moderation);
1523 mlx5e_set_rx_cq_mode_params(params, cq_period_mode);
1526 params->tunneled_offload_en = false;
1528 mlx5_query_min_inline(mdev, ¶ms->tx_min_inline_mode);
1531 mlx5e_build_rss_params(&priv->rss_params, params->num_channels);
1534 static void mlx5e_build_rep_netdev(struct net_device *netdev)
1536 struct mlx5e_priv *priv = netdev_priv(netdev);
1537 struct mlx5e_rep_priv *rpriv = priv->ppriv;
1538 struct mlx5_eswitch_rep *rep = rpriv->rep;
1539 struct mlx5_core_dev *mdev = priv->mdev;
1541 if (rep->vport == MLX5_VPORT_UPLINK) {
1542 SET_NETDEV_DEV(netdev, mdev->device);
1543 netdev->netdev_ops = &mlx5e_netdev_ops_uplink_rep;
1544 /* we want a persistent mac for the uplink rep */
1545 mlx5_query_mac_address(mdev, netdev->dev_addr);
1546 netdev->ethtool_ops = &mlx5e_uplink_rep_ethtool_ops;
1547 #ifdef CONFIG_MLX5_CORE_EN_DCB
1548 if (MLX5_CAP_GEN(mdev, qos))
1549 netdev->dcbnl_ops = &mlx5e_dcbnl_ops;
1552 netdev->netdev_ops = &mlx5e_netdev_ops_rep;
1553 eth_hw_addr_random(netdev);
1554 netdev->ethtool_ops = &mlx5e_rep_ethtool_ops;
1557 netdev->watchdog_timeo = 15 * HZ;
1559 netdev->features |= NETIF_F_NETNS_LOCAL;
1561 netdev->hw_features |= NETIF_F_HW_TC;
1562 netdev->hw_features |= NETIF_F_SG;
1563 netdev->hw_features |= NETIF_F_IP_CSUM;
1564 netdev->hw_features |= NETIF_F_IPV6_CSUM;
1565 netdev->hw_features |= NETIF_F_GRO;
1566 netdev->hw_features |= NETIF_F_TSO;
1567 netdev->hw_features |= NETIF_F_TSO6;
1568 netdev->hw_features |= NETIF_F_RXCSUM;
1570 if (rep->vport == MLX5_VPORT_UPLINK)
1571 netdev->hw_features |= NETIF_F_HW_VLAN_CTAG_RX;
1573 netdev->features |= NETIF_F_VLAN_CHALLENGED;
1575 netdev->features |= netdev->hw_features;
1578 static int mlx5e_init_rep(struct mlx5_core_dev *mdev,
1579 struct net_device *netdev,
1580 const struct mlx5e_profile *profile,
1583 struct mlx5e_priv *priv = netdev_priv(netdev);
1586 err = mlx5e_netdev_init(netdev, priv, mdev, profile, ppriv);
1590 priv->channels.params.num_channels = MLX5E_REP_PARAMS_DEF_NUM_CHANNELS;
1592 mlx5e_build_rep_params(netdev);
1593 mlx5e_build_rep_netdev(netdev);
1595 mlx5e_timestamp_init(priv);
1600 static void mlx5e_cleanup_rep(struct mlx5e_priv *priv)
1602 mlx5e_netdev_cleanup(priv->netdev, priv);
1605 static int mlx5e_create_rep_ttc_table(struct mlx5e_priv *priv)
1607 struct mlx5e_rep_priv *rpriv = priv->ppriv;
1608 struct mlx5_eswitch_rep *rep = rpriv->rep;
1609 struct ttc_params ttc_params = {};
1612 priv->fs.ns = mlx5_get_flow_namespace(priv->mdev,
1613 MLX5_FLOW_NAMESPACE_KERNEL);
1615 /* The inner_ttc in the ttc params is intentionally not set */
1616 ttc_params.any_tt_tirn = priv->direct_tir[0].tirn;
1617 mlx5e_set_ttc_ft_params(&ttc_params);
1619 if (rep->vport != MLX5_VPORT_UPLINK)
1620 /* To give uplik rep TTC a lower level for chaining from root ft */
1621 ttc_params.ft_attr.level = MLX5E_TTC_FT_LEVEL + 1;
1623 for (tt = 0; tt < MLX5E_NUM_INDIR_TIRS; tt++)
1624 ttc_params.indir_tirn[tt] = priv->indir_tir[tt].tirn;
1626 err = mlx5e_create_ttc_table(priv, &ttc_params, &priv->fs.ttc);
1628 netdev_err(priv->netdev, "Failed to create rep ttc table, err=%d\n", err);
1634 static int mlx5e_create_rep_root_ft(struct mlx5e_priv *priv)
1636 struct mlx5e_rep_priv *rpriv = priv->ppriv;
1637 struct mlx5_eswitch_rep *rep = rpriv->rep;
1638 struct mlx5_flow_table_attr ft_attr = {};
1639 struct mlx5_flow_namespace *ns;
1642 if (rep->vport != MLX5_VPORT_UPLINK) {
1643 /* non uplik reps will skip any bypass tables and go directly to
1646 rpriv->root_ft = priv->fs.ttc.ft.t;
1650 /* uplink root ft will be used to auto chain, to ethtool or ttc tables */
1651 ns = mlx5_get_flow_namespace(priv->mdev, MLX5_FLOW_NAMESPACE_OFFLOADS);
1653 netdev_err(priv->netdev, "Failed to get reps offloads namespace\n");
1657 ft_attr.max_fte = 0; /* Empty table, miss rule will always point to next table */
1661 rpriv->root_ft = mlx5_create_flow_table(ns, &ft_attr);
1662 if (IS_ERR(rpriv->root_ft)) {
1663 err = PTR_ERR(rpriv->root_ft);
1664 rpriv->root_ft = NULL;
1670 static void mlx5e_destroy_rep_root_ft(struct mlx5e_priv *priv)
1672 struct mlx5e_rep_priv *rpriv = priv->ppriv;
1673 struct mlx5_eswitch_rep *rep = rpriv->rep;
1675 if (rep->vport != MLX5_VPORT_UPLINK)
1677 mlx5_destroy_flow_table(rpriv->root_ft);
1680 static int mlx5e_create_rep_vport_rx_rule(struct mlx5e_priv *priv)
1682 struct mlx5_eswitch *esw = priv->mdev->priv.eswitch;
1683 struct mlx5e_rep_priv *rpriv = priv->ppriv;
1684 struct mlx5_eswitch_rep *rep = rpriv->rep;
1685 struct mlx5_flow_handle *flow_rule;
1686 struct mlx5_flow_destination dest;
1688 dest.type = MLX5_FLOW_DESTINATION_TYPE_FLOW_TABLE;
1689 dest.ft = rpriv->root_ft;
1691 flow_rule = mlx5_eswitch_create_vport_rx_rule(esw, rep->vport, &dest);
1692 if (IS_ERR(flow_rule))
1693 return PTR_ERR(flow_rule);
1694 rpriv->vport_rx_rule = flow_rule;
1698 static int mlx5e_init_rep_rx(struct mlx5e_priv *priv)
1700 struct mlx5_core_dev *mdev = priv->mdev;
1703 mlx5e_init_l2_addr(priv);
1705 err = mlx5e_open_drop_rq(priv, &priv->drop_rq);
1707 mlx5_core_err(mdev, "open drop rq failed, %d\n", err);
1711 err = mlx5e_create_indirect_rqt(priv);
1713 goto err_close_drop_rq;
1715 err = mlx5e_create_direct_rqts(priv, priv->direct_tir);
1717 goto err_destroy_indirect_rqts;
1719 err = mlx5e_create_indirect_tirs(priv, false);
1721 goto err_destroy_direct_rqts;
1723 err = mlx5e_create_direct_tirs(priv, priv->direct_tir);
1725 goto err_destroy_indirect_tirs;
1727 err = mlx5e_create_rep_ttc_table(priv);
1729 goto err_destroy_direct_tirs;
1731 err = mlx5e_create_rep_root_ft(priv);
1733 goto err_destroy_ttc_table;
1735 err = mlx5e_create_rep_vport_rx_rule(priv);
1737 goto err_destroy_root_ft;
1739 mlx5e_ethtool_init_steering(priv);
1743 err_destroy_root_ft:
1744 mlx5e_destroy_rep_root_ft(priv);
1745 err_destroy_ttc_table:
1746 mlx5e_destroy_ttc_table(priv, &priv->fs.ttc);
1747 err_destroy_direct_tirs:
1748 mlx5e_destroy_direct_tirs(priv, priv->direct_tir);
1749 err_destroy_indirect_tirs:
1750 mlx5e_destroy_indirect_tirs(priv, false);
1751 err_destroy_direct_rqts:
1752 mlx5e_destroy_direct_rqts(priv, priv->direct_tir);
1753 err_destroy_indirect_rqts:
1754 mlx5e_destroy_rqt(priv, &priv->indir_rqt);
1756 mlx5e_close_drop_rq(&priv->drop_rq);
1760 static void mlx5e_cleanup_rep_rx(struct mlx5e_priv *priv)
1762 struct mlx5e_rep_priv *rpriv = priv->ppriv;
1764 mlx5_del_flow_rules(rpriv->vport_rx_rule);
1765 mlx5e_destroy_rep_root_ft(priv);
1766 mlx5e_destroy_ttc_table(priv, &priv->fs.ttc);
1767 mlx5e_destroy_direct_tirs(priv, priv->direct_tir);
1768 mlx5e_destroy_indirect_tirs(priv, false);
1769 mlx5e_destroy_direct_rqts(priv, priv->direct_tir);
1770 mlx5e_destroy_rqt(priv, &priv->indir_rqt);
1771 mlx5e_close_drop_rq(&priv->drop_rq);
1774 static int mlx5e_init_ul_rep_rx(struct mlx5e_priv *priv)
1776 int err = mlx5e_init_rep_rx(priv);
1781 mlx5e_create_q_counters(priv);
1785 static void mlx5e_cleanup_ul_rep_rx(struct mlx5e_priv *priv)
1787 mlx5e_destroy_q_counters(priv);
1788 mlx5e_cleanup_rep_rx(priv);
1791 static int mlx5e_init_uplink_rep_tx(struct mlx5e_rep_priv *rpriv)
1793 struct mlx5_rep_uplink_priv *uplink_priv;
1794 struct net_device *netdev;
1795 struct mlx5e_priv *priv;
1798 netdev = rpriv->netdev;
1799 priv = netdev_priv(netdev);
1800 uplink_priv = &rpriv->uplink_priv;
1802 mutex_init(&uplink_priv->unready_flows_lock);
1803 INIT_LIST_HEAD(&uplink_priv->unready_flows);
1805 /* init shared tc flow table */
1806 err = mlx5e_tc_esw_init(&uplink_priv->tc_ht);
1810 mlx5_init_port_tun_entropy(&uplink_priv->tun_entropy, priv->mdev);
1812 /* init indirect block notifications */
1813 INIT_LIST_HEAD(&uplink_priv->tc_indr_block_priv_list);
1814 uplink_priv->netdevice_nb.notifier_call = mlx5e_nic_rep_netdevice_event;
1815 err = register_netdevice_notifier_dev_net(rpriv->netdev,
1816 &uplink_priv->netdevice_nb,
1817 &uplink_priv->netdevice_nn);
1819 mlx5_core_err(priv->mdev, "Failed to register netdev notifier\n");
1820 goto tc_esw_cleanup;
1826 mlx5e_tc_esw_cleanup(&uplink_priv->tc_ht);
1830 static int mlx5e_init_rep_tx(struct mlx5e_priv *priv)
1832 struct mlx5e_rep_priv *rpriv = priv->ppriv;
1835 err = mlx5e_create_tises(priv);
1837 mlx5_core_warn(priv->mdev, "create tises failed, %d\n", err);
1841 if (rpriv->rep->vport == MLX5_VPORT_UPLINK) {
1842 err = mlx5e_init_uplink_rep_tx(rpriv);
1850 mlx5e_destroy_tises(priv);
1854 static void mlx5e_cleanup_uplink_rep_tx(struct mlx5e_rep_priv *rpriv)
1856 struct mlx5_rep_uplink_priv *uplink_priv = &rpriv->uplink_priv;
1858 /* clean indirect TC block notifications */
1859 unregister_netdevice_notifier_dev_net(rpriv->netdev,
1860 &uplink_priv->netdevice_nb,
1861 &uplink_priv->netdevice_nn);
1862 mlx5e_rep_indr_clean_block_privs(rpriv);
1864 /* delete shared tc flow table */
1865 mlx5e_tc_esw_cleanup(&rpriv->uplink_priv.tc_ht);
1866 mutex_destroy(&rpriv->uplink_priv.unready_flows_lock);
1869 static void mlx5e_cleanup_rep_tx(struct mlx5e_priv *priv)
1871 struct mlx5e_rep_priv *rpriv = priv->ppriv;
1873 mlx5e_destroy_tises(priv);
1875 if (rpriv->rep->vport == MLX5_VPORT_UPLINK)
1876 mlx5e_cleanup_uplink_rep_tx(rpriv);
1879 static void mlx5e_rep_enable(struct mlx5e_priv *priv)
1881 mlx5e_set_netdev_mtu_boundaries(priv);
1884 static int mlx5e_update_rep_rx(struct mlx5e_priv *priv)
1889 static int uplink_rep_async_event(struct notifier_block *nb, unsigned long event, void *data)
1891 struct mlx5e_priv *priv = container_of(nb, struct mlx5e_priv, events_nb);
1893 if (event == MLX5_EVENT_TYPE_PORT_CHANGE) {
1894 struct mlx5_eqe *eqe = data;
1896 switch (eqe->sub_type) {
1897 case MLX5_PORT_CHANGE_SUBTYPE_DOWN:
1898 case MLX5_PORT_CHANGE_SUBTYPE_ACTIVE:
1899 queue_work(priv->wq, &priv->update_carrier_work);
1908 if (event == MLX5_DEV_EVENT_PORT_AFFINITY) {
1909 struct mlx5e_rep_priv *rpriv = priv->ppriv;
1911 queue_work(priv->wq, &rpriv->uplink_priv.reoffload_flows_work);
1919 static void mlx5e_uplink_rep_enable(struct mlx5e_priv *priv)
1921 struct net_device *netdev = priv->netdev;
1922 struct mlx5_core_dev *mdev = priv->mdev;
1923 struct mlx5e_rep_priv *rpriv = priv->ppriv;
1926 netdev->min_mtu = ETH_MIN_MTU;
1927 mlx5_query_port_max_mtu(priv->mdev, &max_mtu, 1);
1928 netdev->max_mtu = MLX5E_HW2SW_MTU(&priv->channels.params, max_mtu);
1929 mlx5e_set_dev_port_mtu(priv);
1931 INIT_WORK(&rpriv->uplink_priv.reoffload_flows_work,
1932 mlx5e_tc_reoffload_flows_work);
1934 mlx5_lag_add(mdev, netdev);
1935 priv->events_nb.notifier_call = uplink_rep_async_event;
1936 mlx5_notifier_register(mdev, &priv->events_nb);
1937 #ifdef CONFIG_MLX5_CORE_EN_DCB
1938 mlx5e_dcbnl_initialize(priv);
1939 mlx5e_dcbnl_init_app(priv);
1943 static void mlx5e_uplink_rep_disable(struct mlx5e_priv *priv)
1945 struct mlx5_core_dev *mdev = priv->mdev;
1946 struct mlx5e_rep_priv *rpriv = priv->ppriv;
1948 #ifdef CONFIG_MLX5_CORE_EN_DCB
1949 mlx5e_dcbnl_delete_app(priv);
1951 mlx5_notifier_unregister(mdev, &priv->events_nb);
1952 cancel_work_sync(&rpriv->uplink_priv.reoffload_flows_work);
1953 mlx5_lag_remove(mdev);
1956 static MLX5E_DEFINE_STATS_GRP(sw_rep, 0);
1957 static MLX5E_DEFINE_STATS_GRP(vport_rep, MLX5E_NDO_UPDATE_STATS);
1959 /* The stats groups order is opposite to the update_stats() order calls */
1960 static mlx5e_stats_grp_t mlx5e_rep_stats_grps[] = {
1961 &MLX5E_STATS_GRP(sw_rep),
1962 &MLX5E_STATS_GRP(vport_rep),
1965 static unsigned int mlx5e_rep_stats_grps_num(struct mlx5e_priv *priv)
1967 return ARRAY_SIZE(mlx5e_rep_stats_grps);
1970 /* The stats groups order is opposite to the update_stats() order calls */
1971 static mlx5e_stats_grp_t mlx5e_ul_rep_stats_grps[] = {
1972 &MLX5E_STATS_GRP(sw),
1973 &MLX5E_STATS_GRP(qcnt),
1974 &MLX5E_STATS_GRP(vnic_env),
1975 &MLX5E_STATS_GRP(vport),
1976 &MLX5E_STATS_GRP(802_3),
1977 &MLX5E_STATS_GRP(2863),
1978 &MLX5E_STATS_GRP(2819),
1979 &MLX5E_STATS_GRP(phy),
1980 &MLX5E_STATS_GRP(eth_ext),
1981 &MLX5E_STATS_GRP(pcie),
1982 &MLX5E_STATS_GRP(per_prio),
1983 &MLX5E_STATS_GRP(pme),
1984 &MLX5E_STATS_GRP(channels),
1985 &MLX5E_STATS_GRP(per_port_buff_congest),
1988 static unsigned int mlx5e_ul_rep_stats_grps_num(struct mlx5e_priv *priv)
1990 return ARRAY_SIZE(mlx5e_ul_rep_stats_grps);
1993 static const struct mlx5e_profile mlx5e_rep_profile = {
1994 .init = mlx5e_init_rep,
1995 .cleanup = mlx5e_cleanup_rep,
1996 .init_rx = mlx5e_init_rep_rx,
1997 .cleanup_rx = mlx5e_cleanup_rep_rx,
1998 .init_tx = mlx5e_init_rep_tx,
1999 .cleanup_tx = mlx5e_cleanup_rep_tx,
2000 .enable = mlx5e_rep_enable,
2001 .update_rx = mlx5e_update_rep_rx,
2002 .update_stats = mlx5e_update_ndo_stats,
2003 .rx_handlers.handle_rx_cqe = mlx5e_handle_rx_cqe_rep,
2004 .rx_handlers.handle_rx_cqe_mpwqe = mlx5e_handle_rx_cqe_mpwrq_rep,
2006 .rq_groups = MLX5E_NUM_RQ_GROUPS(REGULAR),
2007 .stats_grps = mlx5e_rep_stats_grps,
2008 .stats_grps_num = mlx5e_rep_stats_grps_num,
2011 static const struct mlx5e_profile mlx5e_uplink_rep_profile = {
2012 .init = mlx5e_init_rep,
2013 .cleanup = mlx5e_cleanup_rep,
2014 .init_rx = mlx5e_init_ul_rep_rx,
2015 .cleanup_rx = mlx5e_cleanup_ul_rep_rx,
2016 .init_tx = mlx5e_init_rep_tx,
2017 .cleanup_tx = mlx5e_cleanup_rep_tx,
2018 .enable = mlx5e_uplink_rep_enable,
2019 .disable = mlx5e_uplink_rep_disable,
2020 .update_rx = mlx5e_update_rep_rx,
2021 .update_stats = mlx5e_update_ndo_stats,
2022 .update_carrier = mlx5e_update_carrier,
2023 .rx_handlers.handle_rx_cqe = mlx5e_handle_rx_cqe_rep,
2024 .rx_handlers.handle_rx_cqe_mpwqe = mlx5e_handle_rx_cqe_mpwrq_rep,
2025 .max_tc = MLX5E_MAX_NUM_TC,
2026 .rq_groups = MLX5E_NUM_RQ_GROUPS(REGULAR),
2027 .stats_grps = mlx5e_ul_rep_stats_grps,
2028 .stats_grps_num = mlx5e_ul_rep_stats_grps_num,
2032 is_devlink_port_supported(const struct mlx5_core_dev *dev,
2033 const struct mlx5e_rep_priv *rpriv)
2035 return rpriv->rep->vport == MLX5_VPORT_UPLINK ||
2036 rpriv->rep->vport == MLX5_VPORT_PF ||
2037 mlx5_eswitch_is_vf_vport(dev->priv.eswitch, rpriv->rep->vport);
2041 vport_to_devlink_port_index(const struct mlx5_core_dev *dev, u16 vport_num)
2043 return (MLX5_CAP_GEN(dev, vhca_id) << 16) | vport_num;
2046 static int register_devlink_port(struct mlx5_core_dev *dev,
2047 struct mlx5e_rep_priv *rpriv)
2049 struct devlink *devlink = priv_to_devlink(dev);
2050 struct mlx5_eswitch_rep *rep = rpriv->rep;
2051 struct netdev_phys_item_id ppid = {};
2052 unsigned int dl_port_index = 0;
2054 if (!is_devlink_port_supported(dev, rpriv))
2057 mlx5e_rep_get_port_parent_id(rpriv->netdev, &ppid);
2059 if (rep->vport == MLX5_VPORT_UPLINK) {
2060 devlink_port_attrs_set(&rpriv->dl_port,
2061 DEVLINK_PORT_FLAVOUR_PHYSICAL,
2062 PCI_FUNC(dev->pdev->devfn), false, 0,
2063 &ppid.id[0], ppid.id_len);
2064 dl_port_index = vport_to_devlink_port_index(dev, rep->vport);
2065 } else if (rep->vport == MLX5_VPORT_PF) {
2066 devlink_port_attrs_pci_pf_set(&rpriv->dl_port,
2067 &ppid.id[0], ppid.id_len,
2069 dl_port_index = rep->vport;
2070 } else if (mlx5_eswitch_is_vf_vport(dev->priv.eswitch,
2071 rpriv->rep->vport)) {
2072 devlink_port_attrs_pci_vf_set(&rpriv->dl_port,
2073 &ppid.id[0], ppid.id_len,
2076 dl_port_index = vport_to_devlink_port_index(dev, rep->vport);
2079 return devlink_port_register(devlink, &rpriv->dl_port, dl_port_index);
2082 static void unregister_devlink_port(struct mlx5_core_dev *dev,
2083 struct mlx5e_rep_priv *rpriv)
2085 if (is_devlink_port_supported(dev, rpriv))
2086 devlink_port_unregister(&rpriv->dl_port);
2089 /* e-Switch vport representors */
2091 mlx5e_vport_rep_load(struct mlx5_core_dev *dev, struct mlx5_eswitch_rep *rep)
2093 const struct mlx5e_profile *profile;
2094 struct mlx5e_rep_priv *rpriv;
2095 struct net_device *netdev;
2098 rpriv = kzalloc(sizeof(*rpriv), GFP_KERNEL);
2102 /* rpriv->rep to be looked up when profile->init() is called */
2105 nch = mlx5e_get_max_num_channels(dev);
2106 profile = (rep->vport == MLX5_VPORT_UPLINK) ?
2107 &mlx5e_uplink_rep_profile : &mlx5e_rep_profile;
2108 netdev = mlx5e_create_netdev(dev, profile, nch, rpriv);
2111 "Failed to create representor netdev for vport %d\n",
2117 dev_net_set(netdev, mlx5_core_net(dev));
2118 rpriv->netdev = netdev;
2119 rep->rep_data[REP_ETH].priv = rpriv;
2120 INIT_LIST_HEAD(&rpriv->vport_sqs_list);
2122 if (rep->vport == MLX5_VPORT_UPLINK) {
2123 err = mlx5e_create_mdev_resources(dev);
2125 goto err_destroy_netdev;
2128 err = mlx5e_attach_netdev(netdev_priv(netdev));
2131 "Failed to attach representor netdev for vport %d\n",
2133 goto err_destroy_mdev_resources;
2136 err = mlx5e_rep_neigh_init(rpriv);
2139 "Failed to initialized neighbours handling for vport %d\n",
2141 goto err_detach_netdev;
2144 err = register_devlink_port(dev, rpriv);
2146 netdev_warn(netdev, "Failed to register devlink port %d\n",
2148 goto err_neigh_cleanup;
2151 err = register_netdev(netdev);
2154 "Failed to register representor netdev for vport %d\n",
2156 goto err_devlink_cleanup;
2159 if (is_devlink_port_supported(dev, rpriv))
2160 devlink_port_type_eth_set(&rpriv->dl_port, netdev);
2163 err_devlink_cleanup:
2164 unregister_devlink_port(dev, rpriv);
2167 mlx5e_rep_neigh_cleanup(rpriv);
2170 mlx5e_detach_netdev(netdev_priv(netdev));
2172 err_destroy_mdev_resources:
2173 if (rep->vport == MLX5_VPORT_UPLINK)
2174 mlx5e_destroy_mdev_resources(dev);
2177 mlx5e_destroy_netdev(netdev_priv(netdev));
2183 mlx5e_vport_rep_unload(struct mlx5_eswitch_rep *rep)
2185 struct mlx5e_rep_priv *rpriv = mlx5e_rep_to_rep_priv(rep);
2186 struct net_device *netdev = rpriv->netdev;
2187 struct mlx5e_priv *priv = netdev_priv(netdev);
2188 struct mlx5_core_dev *dev = priv->mdev;
2189 void *ppriv = priv->ppriv;
2191 if (is_devlink_port_supported(dev, rpriv))
2192 devlink_port_type_clear(&rpriv->dl_port);
2193 unregister_netdev(netdev);
2194 unregister_devlink_port(dev, rpriv);
2195 mlx5e_rep_neigh_cleanup(rpriv);
2196 mlx5e_detach_netdev(priv);
2197 if (rep->vport == MLX5_VPORT_UPLINK)
2198 mlx5e_destroy_mdev_resources(priv->mdev);
2199 mlx5e_destroy_netdev(priv);
2200 kfree(ppriv); /* mlx5e_rep_priv */
2203 static void *mlx5e_vport_rep_get_proto_dev(struct mlx5_eswitch_rep *rep)
2205 struct mlx5e_rep_priv *rpriv;
2207 rpriv = mlx5e_rep_to_rep_priv(rep);
2209 return rpriv->netdev;
2212 static const struct mlx5_eswitch_rep_ops rep_ops = {
2213 .load = mlx5e_vport_rep_load,
2214 .unload = mlx5e_vport_rep_unload,
2215 .get_proto_dev = mlx5e_vport_rep_get_proto_dev
2218 void mlx5e_rep_register_vport_reps(struct mlx5_core_dev *mdev)
2220 struct mlx5_eswitch *esw = mdev->priv.eswitch;
2222 mlx5_eswitch_register_vport_reps(esw, &rep_ops, REP_ETH);
2225 void mlx5e_rep_unregister_vport_reps(struct mlx5_core_dev *mdev)
2227 struct mlx5_eswitch *esw = mdev->priv.eswitch;
2229 mlx5_eswitch_unregister_vport_reps(esw, REP_ETH);