drivers/net/ethernet/mellanox/mlx4/en_netdev.c

   1 /*
   2  * Copyright (c) 2007 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
  34 #include <linux/bpf.h>
  35 #include <linux/etherdevice.h>
  36 #include <linux/tcp.h>
  37 #include <linux/if_vlan.h>
  38 #include <linux/delay.h>
  39 #include <linux/slab.h>
  40 #include <linux/hash.h>
  41 #include <net/ip.h>
  42 #include <net/busy_poll.h>
  43 #include <net/vxlan.h>
  44 #include <net/devlink.h>
  45
  46 #include <linux/mlx4/driver.h>
  47 #include <linux/mlx4/device.h>
  48 #include <linux/mlx4/cmd.h>
  49 #include <linux/mlx4/cq.h>
  50
  51 #include "mlx4_en.h"
  52 #include "en_port.h"
  53
  54 int mlx4_en_setup_tc(struct net_device *dev, u8 up)
  55 {
  56         struct mlx4_en_priv *priv = netdev_priv(dev);
  57         int i;
  58         unsigned int offset = 0;
  59
  60         if (up && up != MLX4_EN_NUM_UP)
  61                 return -EINVAL;
  62
  63         netdev_set_num_tc(dev, up);
  64
  65         /* Partition Tx queues evenly amongst UP's */
  66         for (i = 0; i < up; i++) {
  67                 netdev_set_tc_queue(dev, i, priv->num_tx_rings_p_up, offset);
  68                 offset += priv->num_tx_rings_p_up;
  69         }
  70
  71 #ifdef CONFIG_MLX4_EN_DCB
  72         if (!mlx4_is_slave(priv->mdev->dev)) {
  73                 if (up) {
  74                         if (priv->dcbx_cap)
  75                                 priv->flags |= MLX4_EN_FLAG_DCB_ENABLED;
  76                 } else {
  77                         priv->flags &= ~MLX4_EN_FLAG_DCB_ENABLED;
  78                         priv->cee_config.pfc_state = false;
  79                 }
  80         }
  81 #endif /* CONFIG_MLX4_EN_DCB */
  82
  83         return 0;
  84 }
  85
  86 static int __mlx4_en_setup_tc(struct net_device *dev, u32 handle, __be16 proto,
  87                               struct tc_to_netdev *tc)
  88 {
  89         if (tc->type != TC_SETUP_MQPRIO)
  90                 return -EINVAL;
  91
  92         return mlx4_en_setup_tc(dev, tc->tc);
  93 }
  94
  95 #ifdef CONFIG_RFS_ACCEL
  96
  97 struct mlx4_en_filter {
  98         struct list_head next;
  99         struct work_struct work;
 100
 101         u8     ip_proto;
 102         __be32 src_ip;
 103         __be32 dst_ip;
 104         __be16 src_port;
 105         __be16 dst_port;
 106
 107         int rxq_index;
 108         struct mlx4_en_priv *priv;
 109         u32 flow_id;                    /* RFS infrastructure id */
 110         int id;                         /* mlx4_en driver id */
 111         u64 reg_id;                     /* Flow steering API id */
 112         u8 activated;                   /* Used to prevent expiry before filter
 113                                          * is attached
 114                                          */
 115         struct hlist_node filter_chain;
 116 };
 117
 118 static void mlx4_en_filter_rfs_expire(struct mlx4_en_priv *priv);
 119
 120 static enum mlx4_net_trans_rule_id mlx4_ip_proto_to_trans_rule_id(u8 ip_proto)
 121 {
 122         switch (ip_proto) {
 123         case IPPROTO_UDP:
 124                 return MLX4_NET_TRANS_RULE_ID_UDP;
 125         case IPPROTO_TCP:
 126                 return MLX4_NET_TRANS_RULE_ID_TCP;
 127         default:
 128                 return MLX4_NET_TRANS_RULE_NUM;
 129         }
 130 };
 131
 132 /* Must not acquire state_lock, as its corresponding work_sync
 133  * is done under it.
 134  */
 135 static void mlx4_en_filter_work(struct work_struct *work)
 136 {
 137         struct mlx4_en_filter *filter = container_of(work,
 138                                                      struct mlx4_en_filter,
 139                                                      work);
 140         struct mlx4_en_priv *priv = filter->priv;
 141         struct mlx4_spec_list spec_tcp_udp = {
 142                 .id = mlx4_ip_proto_to_trans_rule_id(filter->ip_proto),
 143                 {
 144                         .tcp_udp = {
 145                                 .dst_port = filter->dst_port,
 146                                 .dst_port_msk = (__force __be16)-1,
 147                                 .src_port = filter->src_port,
 148                                 .src_port_msk = (__force __be16)-1,
 149                         },
 150                 },
 151         };
 152         struct mlx4_spec_list spec_ip = {
 153                 .id = MLX4_NET_TRANS_RULE_ID_IPV4,
 154                 {
 155                         .ipv4 = {
 156                                 .dst_ip = filter->dst_ip,
 157                                 .dst_ip_msk = (__force __be32)-1,
 158                                 .src_ip = filter->src_ip,
 159                                 .src_ip_msk = (__force __be32)-1,
 160                         },
 161                 },
 162         };
 163         struct mlx4_spec_list spec_eth = {
 164                 .id = MLX4_NET_TRANS_RULE_ID_ETH,
 165         };
 166         struct mlx4_net_trans_rule rule = {
 167                 .list = LIST_HEAD_INIT(rule.list),
 168                 .queue_mode = MLX4_NET_TRANS_Q_LIFO,
 169                 .exclusive = 1,
 170                 .allow_loopback = 1,
 171                 .promisc_mode = MLX4_FS_REGULAR,
 172                 .port = priv->port,
 173                 .priority = MLX4_DOMAIN_RFS,
 174         };
 175         int rc;
 176         __be64 mac_mask = cpu_to_be64(MLX4_MAC_MASK << 16);
 177
 178         if (spec_tcp_udp.id >= MLX4_NET_TRANS_RULE_NUM) {
 179                 en_warn(priv, "RFS: ignoring unsupported ip protocol (%d)\n",
 180                         filter->ip_proto);
 181                 goto ignore;
 182         }
 183         list_add_tail(&spec_eth.list, &rule.list);
 184         list_add_tail(&spec_ip.list, &rule.list);
 185         list_add_tail(&spec_tcp_udp.list, &rule.list);
 186
 187         rule.qpn = priv->rss_map.qps[filter->rxq_index].qpn;
 188         memcpy(spec_eth.eth.dst_mac, priv->dev->dev_addr, ETH_ALEN);
 189         memcpy(spec_eth.eth.dst_mac_msk, &mac_mask, ETH_ALEN);
 190
 191         filter->activated = 0;
 192
 193         if (filter->reg_id) {
 194                 rc = mlx4_flow_detach(priv->mdev->dev, filter->reg_id);
 195                 if (rc && rc != -ENOENT)
 196                         en_err(priv, "Error detaching flow. rc = %d\n", rc);
 197         }
 198
 199         rc = mlx4_flow_attach(priv->mdev->dev, &rule, &filter->reg_id);
 200         if (rc)
 201                 en_err(priv, "Error attaching flow. err = %d\n", rc);
 202
 203 ignore:
 204         mlx4_en_filter_rfs_expire(priv);
 205
 206         filter->activated = 1;
 207 }
 208
 209 static inline struct hlist_head *
 210 filter_hash_bucket(struct mlx4_en_priv *priv, __be32 src_ip, __be32 dst_ip,
 211                    __be16 src_port, __be16 dst_port)
 212 {
 213         unsigned long l;
 214         int bucket_idx;
 215
 216         l = (__force unsigned long)src_port |
 217             ((__force unsigned long)dst_port << 2);
 218         l ^= (__force unsigned long)(src_ip ^ dst_ip);
 219
 220         bucket_idx = hash_long(l, MLX4_EN_FILTER_HASH_SHIFT);
 221
 222         return &priv->filter_hash[bucket_idx];
 223 }
 224
 225 static struct mlx4_en_filter *
 226 mlx4_en_filter_alloc(struct mlx4_en_priv *priv, int rxq_index, __be32 src_ip,
 227                      __be32 dst_ip, u8 ip_proto, __be16 src_port,
 228                      __be16 dst_port, u32 flow_id)
 229 {
 230         struct mlx4_en_filter *filter = NULL;
 231
 232         filter = kzalloc(sizeof(struct mlx4_en_filter), GFP_ATOMIC);
 233         if (!filter)
 234                 return NULL;
 235
 236         filter->priv = priv;
 237         filter->rxq_index = rxq_index;
 238         INIT_WORK(&filter->work, mlx4_en_filter_work);
 239
 240         filter->src_ip = src_ip;
 241         filter->dst_ip = dst_ip;
 242         filter->ip_proto = ip_proto;
 243         filter->src_port = src_port;
 244         filter->dst_port = dst_port;
 245
 246         filter->flow_id = flow_id;
 247
 248         filter->id = priv->last_filter_id++ % RPS_NO_FILTER;
 249
 250         list_add_tail(&filter->next, &priv->filters);
 251         hlist_add_head(&filter->filter_chain,
 252                        filter_hash_bucket(priv, src_ip, dst_ip, src_port,
 253                                           dst_port));
 254
 255         return filter;
 256 }
 257
 258 static void mlx4_en_filter_free(struct mlx4_en_filter *filter)
 259 {
 260         struct mlx4_en_priv *priv = filter->priv;
 261         int rc;
 262
 263         list_del(&filter->next);
 264
 265         rc = mlx4_flow_detach(priv->mdev->dev, filter->reg_id);
 266         if (rc && rc != -ENOENT)
 267                 en_err(priv, "Error detaching flow. rc = %d\n", rc);
 268
 269         kfree(filter);
 270 }
 271
 272 static inline struct mlx4_en_filter *
 273 mlx4_en_filter_find(struct mlx4_en_priv *priv, __be32 src_ip, __be32 dst_ip,
 274                     u8 ip_proto, __be16 src_port, __be16 dst_port)
 275 {
 276         struct mlx4_en_filter *filter;
 277         struct mlx4_en_filter *ret = NULL;
 278
 279         hlist_for_each_entry(filter,
 280                              filter_hash_bucket(priv, src_ip, dst_ip,
 281                                                 src_port, dst_port),
 282                              filter_chain) {
 283                 if (filter->src_ip == src_ip &&
 284                     filter->dst_ip == dst_ip &&
 285                     filter->ip_proto == ip_proto &&
 286                     filter->src_port == src_port &&
 287                     filter->dst_port == dst_port) {
 288                         ret = filter;
 289                         break;
 290                 }
 291         }
 292
 293         return ret;
 294 }
 295
 296 static int
 297 mlx4_en_filter_rfs(struct net_device *net_dev, const struct sk_buff *skb,
 298                    u16 rxq_index, u32 flow_id)
 299 {
 300         struct mlx4_en_priv *priv = netdev_priv(net_dev);
 301         struct mlx4_en_filter *filter;
 302         const struct iphdr *ip;
 303         const __be16 *ports;
 304         u8 ip_proto;
 305         __be32 src_ip;
 306         __be32 dst_ip;
 307         __be16 src_port;
 308         __be16 dst_port;
 309         int nhoff = skb_network_offset(skb);
 310         int ret = 0;
 311
 312         if (skb->protocol != htons(ETH_P_IP))
 313                 return -EPROTONOSUPPORT;
 314
 315         ip = (const struct iphdr *)(skb->data + nhoff);
 316         if (ip_is_fragment(ip))
 317                 return -EPROTONOSUPPORT;
 318
 319         if ((ip->protocol != IPPROTO_TCP) && (ip->protocol != IPPROTO_UDP))
 320                 return -EPROTONOSUPPORT;
 321         ports = (const __be16 *)(skb->data + nhoff + 4 * ip->ihl);
 322
 323         ip_proto = ip->protocol;
 324         src_ip = ip->saddr;
 325         dst_ip = ip->daddr;
 326         src_port = ports[0];
 327         dst_port = ports[1];
 328
 329         spin_lock_bh(&priv->filters_lock);
 330         filter = mlx4_en_filter_find(priv, src_ip, dst_ip, ip_proto,
 331                                      src_port, dst_port);
 332         if (filter) {
 333                 if (filter->rxq_index == rxq_index)
 334                         goto out;
 335
 336                 filter->rxq_index = rxq_index;
 337         } else {
 338                 filter = mlx4_en_filter_alloc(priv, rxq_index,
 339                                               src_ip, dst_ip, ip_proto,
 340                                               src_port, dst_port, flow_id);
 341                 if (!filter) {
 342                         ret = -ENOMEM;
 343                         goto err;
 344                 }
 345         }
 346
 347         queue_work(priv->mdev->workqueue, &filter->work);
 348
 349 out:
 350         ret = filter->id;
 351 err:
 352         spin_unlock_bh(&priv->filters_lock);
 353
 354         return ret;
 355 }
 356
 357 void mlx4_en_cleanup_filters(struct mlx4_en_priv *priv)
 358 {
 359         struct mlx4_en_filter *filter, *tmp;
 360         LIST_HEAD(del_list);
 361
 362         spin_lock_bh(&priv->filters_lock);
 363         list_for_each_entry_safe(filter, tmp, &priv->filters, next) {
 364                 list_move(&filter->next, &del_list);
 365                 hlist_del(&filter->filter_chain);
 366         }
 367         spin_unlock_bh(&priv->filters_lock);
 368
 369         list_for_each_entry_safe(filter, tmp, &del_list, next) {
 370                 cancel_work_sync(&filter->work);
 371                 mlx4_en_filter_free(filter);
 372         }
 373 }
 374
 375 static void mlx4_en_filter_rfs_expire(struct mlx4_en_priv *priv)
 376 {
 377         struct mlx4_en_filter *filter = NULL, *tmp, *last_filter = NULL;
 378         LIST_HEAD(del_list);
 379         int i = 0;
 380
 381         spin_lock_bh(&priv->filters_lock);
 382         list_for_each_entry_safe(filter, tmp, &priv->filters, next) {
 383                 if (i > MLX4_EN_FILTER_EXPIRY_QUOTA)
 384                         break;
 385
 386                 if (filter->activated &&
 387                     !work_pending(&filter->work) &&
 388                     rps_may_expire_flow(priv->dev,
 389                                         filter->rxq_index, filter->flow_id,
 390                                         filter->id)) {
 391                         list_move(&filter->next, &del_list);
 392                         hlist_del(&filter->filter_chain);
 393                 } else
 394                         last_filter = filter;
 395
 396                 i++;
 397         }
 398
 399         if (last_filter && (&last_filter->next != priv->filters.next))
 400                 list_move(&priv->filters, &last_filter->next);
 401
 402         spin_unlock_bh(&priv->filters_lock);
 403
 404         list_for_each_entry_safe(filter, tmp, &del_list, next)
 405                 mlx4_en_filter_free(filter);
 406 }
 407 #endif
 408
 409 static int mlx4_en_vlan_rx_add_vid(struct net_device *dev,
 410                                    __be16 proto, u16 vid)
 411 {
 412         struct mlx4_en_priv *priv = netdev_priv(dev);
 413         struct mlx4_en_dev *mdev = priv->mdev;
 414         int err;
 415         int idx;
 416
 417         en_dbg(HW, priv, "adding VLAN:%d\n", vid);
 418
 419         set_bit(vid, priv->active_vlans);
 420
 421         /* Add VID to port VLAN filter */
 422         mutex_lock(&mdev->state_lock);
 423         if (mdev->device_up && priv->port_up) {
 424                 err = mlx4_SET_VLAN_FLTR(mdev->dev, priv);
 425                 if (err) {
 426                         en_err(priv, "Failed configuring VLAN filter\n");
 427                         goto out;
 428                 }
 429         }
 430         err = mlx4_register_vlan(mdev->dev, priv->port, vid, &idx);
 431         if (err)
 432                 en_dbg(HW, priv, "Failed adding vlan %d\n", vid);
 433
 434 out:
 435         mutex_unlock(&mdev->state_lock);
 436         return err;
 437 }
 438
 439 static int mlx4_en_vlan_rx_kill_vid(struct net_device *dev,
 440                                     __be16 proto, u16 vid)
 441 {
 442         struct mlx4_en_priv *priv = netdev_priv(dev);
 443         struct mlx4_en_dev *mdev = priv->mdev;
 444         int err = 0;
 445
 446         en_dbg(HW, priv, "Killing VID:%d\n", vid);
 447
 448         clear_bit(vid, priv->active_vlans);
 449
 450         /* Remove VID from port VLAN filter */
 451         mutex_lock(&mdev->state_lock);
 452         mlx4_unregister_vlan(mdev->dev, priv->port, vid);
 453
 454         if (mdev->device_up && priv->port_up) {
 455                 err = mlx4_SET_VLAN_FLTR(mdev->dev, priv);
 456                 if (err)
 457                         en_err(priv, "Failed configuring VLAN filter\n");
 458         }
 459         mutex_unlock(&mdev->state_lock);
 460
 461         return err;
 462 }
 463
 464 static void mlx4_en_u64_to_mac(unsigned char dst_mac[ETH_ALEN + 2], u64 src_mac)
 465 {
 466         int i;
 467         for (i = ETH_ALEN - 1; i >= 0; --i) {
 468                 dst_mac[i] = src_mac & 0xff;
 469                 src_mac >>= 8;
 470         }
 471         memset(&dst_mac[ETH_ALEN], 0, 2);
 472 }
 473
 474
 475 static int mlx4_en_tunnel_steer_add(struct mlx4_en_priv *priv, unsigned char *addr,
 476                                     int qpn, u64 *reg_id)
 477 {
 478         int err;
 479
 480         if (priv->mdev->dev->caps.tunnel_offload_mode != MLX4_TUNNEL_OFFLOAD_MODE_VXLAN ||
 481             priv->mdev->dev->caps.dmfs_high_steer_mode == MLX4_STEERING_DMFS_A0_STATIC)
 482                 return 0; /* do nothing */
 483
 484         err = mlx4_tunnel_steer_add(priv->mdev->dev, addr, priv->port, qpn,
 485                                     MLX4_DOMAIN_NIC, reg_id);
 486         if (err) {
 487                 en_err(priv, "failed to add vxlan steering rule, err %d\n", err);
 488                 return err;
 489         }
 490         en_dbg(DRV, priv, "added vxlan steering rule, mac %pM reg_id %llx\n", addr, *reg_id);
 491         return 0;
 492 }
 493
 494
 495 static int mlx4_en_uc_steer_add(struct mlx4_en_priv *priv,
 496                                 unsigned char *mac, int *qpn, u64 *reg_id)
 497 {
 498         struct mlx4_en_dev *mdev = priv->mdev;
 499         struct mlx4_dev *dev = mdev->dev;
 500         int err;
 501
 502         switch (dev->caps.steering_mode) {
 503         case MLX4_STEERING_MODE_B0: {
 504                 struct mlx4_qp qp;
 505                 u8 gid[16] = {0};
 506
 507                 qp.qpn = *qpn;
 508                 memcpy(&gid[10], mac, ETH_ALEN);
 509                 gid[5] = priv->port;
 510
 511                 err = mlx4_unicast_attach(dev, &qp, gid, 0, MLX4_PROT_ETH);
 512                 break;
 513         }
 514         case MLX4_STEERING_MODE_DEVICE_MANAGED: {
 515                 struct mlx4_spec_list spec_eth = { {NULL} };
 516                 __be64 mac_mask = cpu_to_be64(MLX4_MAC_MASK << 16);
 517
 518                 struct mlx4_net_trans_rule rule = {
 519                         .queue_mode = MLX4_NET_TRANS_Q_FIFO,
 520                         .exclusive = 0,
 521                         .allow_loopback = 1,
 522                         .promisc_mode = MLX4_FS_REGULAR,
 523                         .priority = MLX4_DOMAIN_NIC,
 524                 };
 525
 526                 rule.port = priv->port;
 527                 rule.qpn = *qpn;
 528                 INIT_LIST_HEAD(&rule.list);
 529
 530                 spec_eth.id = MLX4_NET_TRANS_RULE_ID_ETH;
 531                 memcpy(spec_eth.eth.dst_mac, mac, ETH_ALEN);
 532                 memcpy(spec_eth.eth.dst_mac_msk, &mac_mask, ETH_ALEN);
 533                 list_add_tail(&spec_eth.list, &rule.list);
 534
 535                 err = mlx4_flow_attach(dev, &rule, reg_id);
 536                 break;
 537         }
 538         default:
 539                 return -EINVAL;
 540         }
 541         if (err)
 542                 en_warn(priv, "Failed Attaching Unicast\n");
 543
 544         return err;
 545 }
 546
 547 static void mlx4_en_uc_steer_release(struct mlx4_en_priv *priv,
 548                                      unsigned char *mac, int qpn, u64 reg_id)
 549 {
 550         struct mlx4_en_dev *mdev = priv->mdev;
 551         struct mlx4_dev *dev = mdev->dev;
 552
 553         switch (dev->caps.steering_mode) {
 554         case MLX4_STEERING_MODE_B0: {
 555                 struct mlx4_qp qp;
 556                 u8 gid[16] = {0};
 557
 558                 qp.qpn = qpn;
 559                 memcpy(&gid[10], mac, ETH_ALEN);
 560                 gid[5] = priv->port;
 561
 562                 mlx4_unicast_detach(dev, &qp, gid, MLX4_PROT_ETH);
 563                 break;
 564         }
 565         case MLX4_STEERING_MODE_DEVICE_MANAGED: {
 566                 mlx4_flow_detach(dev, reg_id);
 567                 break;
 568         }
 569         default:
 570                 en_err(priv, "Invalid steering mode.\n");
 571         }
 572 }
 573
 574 static int mlx4_en_get_qp(struct mlx4_en_priv *priv)
 575 {
 576         struct mlx4_en_dev *mdev = priv->mdev;
 577         struct mlx4_dev *dev = mdev->dev;
 578         int index = 0;
 579         int err = 0;
 580         int *qpn = &priv->base_qpn;
 581         u64 mac = mlx4_mac_to_u64(priv->dev->dev_addr);
 582
 583         en_dbg(DRV, priv, "Registering MAC: %pM for adding\n",
 584                priv->dev->dev_addr);
 585         index = mlx4_register_mac(dev, priv->port, mac);
 586         if (index < 0) {
 587                 err = index;
 588                 en_err(priv, "Failed adding MAC: %pM\n",
 589                        priv->dev->dev_addr);
 590                 return err;
 591         }
 592
 593         if (dev->caps.steering_mode == MLX4_STEERING_MODE_A0) {
 594                 int base_qpn = mlx4_get_base_qpn(dev, priv->port);
 595                 *qpn = base_qpn + index;
 596                 return 0;
 597         }
 598
 599         err = mlx4_qp_reserve_range(dev, 1, 1, qpn, MLX4_RESERVE_A0_QP);
 600         en_dbg(DRV, priv, "Reserved qp %d\n", *qpn);
 601         if (err) {
 602                 en_err(priv, "Failed to reserve qp for mac registration\n");
 603                 mlx4_unregister_mac(dev, priv->port, mac);
 604                 return err;
 605         }
 606
 607         return 0;
 608 }
 609
 610 static void mlx4_en_put_qp(struct mlx4_en_priv *priv)
 611 {
 612         struct mlx4_en_dev *mdev = priv->mdev;
 613         struct mlx4_dev *dev = mdev->dev;
 614         int qpn = priv->base_qpn;
 615
 616         if (dev->caps.steering_mode == MLX4_STEERING_MODE_A0) {
 617                 u64 mac = mlx4_mac_to_u64(priv->dev->dev_addr);
 618                 en_dbg(DRV, priv, "Registering MAC: %pM for deleting\n",
 619                        priv->dev->dev_addr);
 620                 mlx4_unregister_mac(dev, priv->port, mac);
 621         } else {
 622                 en_dbg(DRV, priv, "Releasing qp: port %d, qpn %d\n",
 623                        priv->port, qpn);
 624                 mlx4_qp_release_range(dev, qpn, 1);
 625                 priv->flags &= ~MLX4_EN_FLAG_FORCE_PROMISC;
 626         }
 627 }
 628
 629 static int mlx4_en_replace_mac(struct mlx4_en_priv *priv, int qpn,
 630                                unsigned char *new_mac, unsigned char *prev_mac)
 631 {
 632         struct mlx4_en_dev *mdev = priv->mdev;
 633         struct mlx4_dev *dev = mdev->dev;
 634         int err = 0;
 635         u64 new_mac_u64 = mlx4_mac_to_u64(new_mac);
 636
 637         if (dev->caps.steering_mode != MLX4_STEERING_MODE_A0) {
 638                 struct hlist_head *bucket;
 639                 unsigned int mac_hash;
 640                 struct mlx4_mac_entry *entry;
 641                 struct hlist_node *tmp;
 642                 u64 prev_mac_u64 = mlx4_mac_to_u64(prev_mac);
 643
 644                 bucket = &priv->mac_hash[prev_mac[MLX4_EN_MAC_HASH_IDX]];
 645                 hlist_for_each_entry_safe(entry, tmp, bucket, hlist) {
 646                         if (ether_addr_equal_64bits(entry->mac, prev_mac)) {
 647                                 mlx4_en_uc_steer_release(priv, entry->mac,
 648                                                          qpn, entry->reg_id);
 649                                 mlx4_unregister_mac(dev, priv->port,
 650                                                     prev_mac_u64);
 651                                 hlist_del_rcu(&entry->hlist);
 652                                 synchronize_rcu();
 653                                 memcpy(entry->mac, new_mac, ETH_ALEN);
 654                                 entry->reg_id = 0;
 655                                 mac_hash = new_mac[MLX4_EN_MAC_HASH_IDX];
 656                                 hlist_add_head_rcu(&entry->hlist,
 657                                                    &priv->mac_hash[mac_hash]);
 658                                 mlx4_register_mac(dev, priv->port, new_mac_u64);
 659                                 err = mlx4_en_uc_steer_add(priv, new_mac,
 660                                                            &qpn,
 661                                                            &entry->reg_id);
 662                                 if (err)
 663                                         return err;
 664                                 if (priv->tunnel_reg_id) {
 665                                         mlx4_flow_detach(priv->mdev->dev, priv->tunnel_reg_id);
 666                                         priv->tunnel_reg_id = 0;
 667                                 }
 668                                 err = mlx4_en_tunnel_steer_add(priv, new_mac, qpn,
 669                                                                &priv->tunnel_reg_id);
 670                                 return err;
 671                         }
 672                 }
 673                 return -EINVAL;
 674         }
 675
 676         return __mlx4_replace_mac(dev, priv->port, qpn, new_mac_u64);
 677 }
 678
 679 static int mlx4_en_do_set_mac(struct mlx4_en_priv *priv,
 680                               unsigned char new_mac[ETH_ALEN + 2])
 681 {
 682         int err = 0;
 683
 684         if (priv->port_up) {
 685                 /* Remove old MAC and insert the new one */
 686                 err = mlx4_en_replace_mac(priv, priv->base_qpn,
 687                                           new_mac, priv->current_mac);
 688                 if (err)
 689                         en_err(priv, "Failed changing HW MAC address\n");
 690         } else
 691                 en_dbg(HW, priv, "Port is down while registering mac, exiting...\n");
 692
 693         if (!err)
 694                 memcpy(priv->current_mac, new_mac, sizeof(priv->current_mac));
 695
 696         return err;
 697 }
 698
 699 static int mlx4_en_set_mac(struct net_device *dev, void *addr)
 700 {
 701         struct mlx4_en_priv *priv = netdev_priv(dev);
 702         struct mlx4_en_dev *mdev = priv->mdev;
 703         struct sockaddr *saddr = addr;
 704         unsigned char new_mac[ETH_ALEN + 2];
 705         int err;
 706
 707         if (!is_valid_ether_addr(saddr->sa_data))
 708                 return -EADDRNOTAVAIL;
 709
 710         mutex_lock(&mdev->state_lock);
 711         memcpy(new_mac, saddr->sa_data, ETH_ALEN);
 712         err = mlx4_en_do_set_mac(priv, new_mac);
 713         if (!err)
 714                 memcpy(dev->dev_addr, saddr->sa_data, ETH_ALEN);
 715         mutex_unlock(&mdev->state_lock);
 716
 717         return err;
 718 }
 719
 720 static void mlx4_en_clear_list(struct net_device *dev)
 721 {
 722         struct mlx4_en_priv *priv = netdev_priv(dev);
 723         struct mlx4_en_mc_list *tmp, *mc_to_del;
 724
 725         list_for_each_entry_safe(mc_to_del, tmp, &priv->mc_list, list) {
 726                 list_del(&mc_to_del->list);
 727                 kfree(mc_to_del);
 728         }
 729 }
 730
 731 static void mlx4_en_cache_mclist(struct net_device *dev)
 732 {
 733         struct mlx4_en_priv *priv = netdev_priv(dev);
 734         struct netdev_hw_addr *ha;
 735         struct mlx4_en_mc_list *tmp;
 736
 737         mlx4_en_clear_list(dev);
 738         netdev_for_each_mc_addr(ha, dev) {
 739                 tmp = kzalloc(sizeof(struct mlx4_en_mc_list), GFP_ATOMIC);
 740                 if (!tmp) {
 741                         mlx4_en_clear_list(dev);
 742                         return;
 743                 }
 744                 memcpy(tmp->addr, ha->addr, ETH_ALEN);
 745                 list_add_tail(&tmp->list, &priv->mc_list);
 746         }
 747 }
 748
 749 static void update_mclist_flags(struct mlx4_en_priv *priv,
 750                                 struct list_head *dst,
 751                                 struct list_head *src)
 752 {
 753         struct mlx4_en_mc_list *dst_tmp, *src_tmp, *new_mc;
 754         bool found;
 755
 756         /* Find all the entries that should be removed from dst,
 757          * These are the entries that are not found in src
 758          */
 759         list_for_each_entry(dst_tmp, dst, list) {
 760                 found = false;
 761                 list_for_each_entry(src_tmp, src, list) {
 762                         if (ether_addr_equal(dst_tmp->addr, src_tmp->addr)) {
 763                                 found = true;
 764                                 break;
 765                         }
 766                 }
 767                 if (!found)
 768                         dst_tmp->action = MCLIST_REM;
 769         }
 770
 771         /* Add entries that exist in src but not in dst
 772          * mark them as need to add
 773          */
 774         list_for_each_entry(src_tmp, src, list) {
 775                 found = false;
 776                 list_for_each_entry(dst_tmp, dst, list) {
 777                         if (ether_addr_equal(dst_tmp->addr, src_tmp->addr)) {
 778                                 dst_tmp->action = MCLIST_NONE;
 779                                 found = true;
 780                                 break;
 781                         }
 782                 }
 783                 if (!found) {
 784                         new_mc = kmemdup(src_tmp,
 785                                          sizeof(struct mlx4_en_mc_list),
 786                                          GFP_KERNEL);
 787                         if (!new_mc)
 788                                 return;
 789
 790                         new_mc->action = MCLIST_ADD;
 791                         list_add_tail(&new_mc->list, dst);
 792                 }
 793         }
 794 }
 795
 796 static void mlx4_en_set_rx_mode(struct net_device *dev)
 797 {
 798         struct mlx4_en_priv *priv = netdev_priv(dev);
 799
 800         if (!priv->port_up)
 801                 return;
 802
 803         queue_work(priv->mdev->workqueue, &priv->rx_mode_task);
 804 }
 805
 806 static void mlx4_en_set_promisc_mode(struct mlx4_en_priv *priv,
 807                                      struct mlx4_en_dev *mdev)
 808 {
 809         int err = 0;
 810
 811         if (!(priv->flags & MLX4_EN_FLAG_PROMISC)) {
 812                 if (netif_msg_rx_status(priv))
 813                         en_warn(priv, "Entering promiscuous mode\n");
 814                 priv->flags |= MLX4_EN_FLAG_PROMISC;
 815
 816                 /* Enable promiscouos mode */
 817                 switch (mdev->dev->caps.steering_mode) {
 818                 case MLX4_STEERING_MODE_DEVICE_MANAGED:
 819                         err = mlx4_flow_steer_promisc_add(mdev->dev,
 820                                                           priv->port,
 821                                                           priv->base_qpn,
 822                                                           MLX4_FS_ALL_DEFAULT);
 823                         if (err)
 824                                 en_err(priv, "Failed enabling promiscuous mode\n");
 825                         priv->flags |= MLX4_EN_FLAG_MC_PROMISC;
 826                         break;
 827
 828                 case MLX4_STEERING_MODE_B0:
 829                         err = mlx4_unicast_promisc_add(mdev->dev,
 830                                                        priv->base_qpn,
 831                                                        priv->port);
 832                         if (err)
 833                                 en_err(priv, "Failed enabling unicast promiscuous mode\n");
 834
 835                         /* Add the default qp number as multicast
 836                          * promisc
 837                          */
 838                         if (!(priv->flags & MLX4_EN_FLAG_MC_PROMISC)) {
 839                                 err = mlx4_multicast_promisc_add(mdev->dev,
 840                                                                  priv->base_qpn,
 841                                                                  priv->port);
 842                                 if (err)
 843                                         en_err(priv, "Failed enabling multicast promiscuous mode\n");
 844                                 priv->flags |= MLX4_EN_FLAG_MC_PROMISC;
 845                         }
 846                         break;
 847
 848                 case MLX4_STEERING_MODE_A0:
 849                         err = mlx4_SET_PORT_qpn_calc(mdev->dev,
 850                                                      priv->port,
 851                                                      priv->base_qpn,
 852                                                      1);
 853                         if (err)
 854                                 en_err(priv, "Failed enabling promiscuous mode\n");
 855                         break;
 856                 }
 857
 858                 /* Disable port multicast filter (unconditionally) */
 859                 err = mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, 0,
 860                                           0, MLX4_MCAST_DISABLE);
 861                 if (err)
 862                         en_err(priv, "Failed disabling multicast filter\n");
 863         }
 864 }
 865
 866 static void mlx4_en_clear_promisc_mode(struct mlx4_en_priv *priv,
 867                                        struct mlx4_en_dev *mdev)
 868 {
 869         int err = 0;
 870
 871         if (netif_msg_rx_status(priv))
 872                 en_warn(priv, "Leaving promiscuous mode\n");
 873         priv->flags &= ~MLX4_EN_FLAG_PROMISC;
 874
 875         /* Disable promiscouos mode */
 876         switch (mdev->dev->caps.steering_mode) {
 877         case MLX4_STEERING_MODE_DEVICE_MANAGED:
 878                 err = mlx4_flow_steer_promisc_remove(mdev->dev,
 879                                                      priv->port,
 880                                                      MLX4_FS_ALL_DEFAULT);
 881                 if (err)
 882                         en_err(priv, "Failed disabling promiscuous mode\n");
 883                 priv->flags &= ~MLX4_EN_FLAG_MC_PROMISC;
 884                 break;
 885
 886         case MLX4_STEERING_MODE_B0:
 887                 err = mlx4_unicast_promisc_remove(mdev->dev,
 888                                                   priv->base_qpn,
 889                                                   priv->port);
 890                 if (err)
 891                         en_err(priv, "Failed disabling unicast promiscuous mode\n");
 892                 /* Disable Multicast promisc */
 893                 if (priv->flags & MLX4_EN_FLAG_MC_PROMISC) {
 894                         err = mlx4_multicast_promisc_remove(mdev->dev,
 895                                                             priv->base_qpn,
 896                                                             priv->port);
 897                         if (err)
 898                                 en_err(priv, "Failed disabling multicast promiscuous mode\n");
 899                         priv->flags &= ~MLX4_EN_FLAG_MC_PROMISC;
 900                 }
 901                 break;
 902
 903         case MLX4_STEERING_MODE_A0:
 904                 err = mlx4_SET_PORT_qpn_calc(mdev->dev,
 905                                              priv->port,
 906                                              priv->base_qpn, 0);
 907                 if (err)
 908                         en_err(priv, "Failed disabling promiscuous mode\n");
 909                 break;
 910         }
 911 }
 912
 913 static void mlx4_en_do_multicast(struct mlx4_en_priv *priv,
 914                                  struct net_device *dev,
 915                                  struct mlx4_en_dev *mdev)
 916 {
 917         struct mlx4_en_mc_list *mclist, *tmp;
 918         u64 mcast_addr = 0;
 919         u8 mc_list[16] = {0};
 920         int err = 0;
 921
 922         /* Enable/disable the multicast filter according to IFF_ALLMULTI */
 923         if (dev->flags & IFF_ALLMULTI) {
 924                 err = mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, 0,
 925                                           0, MLX4_MCAST_DISABLE);
 926                 if (err)
 927                         en_err(priv, "Failed disabling multicast filter\n");
 928
 929                 /* Add the default qp number as multicast promisc */
 930                 if (!(priv->flags & MLX4_EN_FLAG_MC_PROMISC)) {
 931                         switch (mdev->dev->caps.steering_mode) {
 932                         case MLX4_STEERING_MODE_DEVICE_MANAGED:
 933                                 err = mlx4_flow_steer_promisc_add(mdev->dev,
 934                                                                   priv->port,
 935                                                                   priv->base_qpn,
 936                                                                   MLX4_FS_MC_DEFAULT);
 937                                 break;
 938
 939                         case MLX4_STEERING_MODE_B0:
 940                                 err = mlx4_multicast_promisc_add(mdev->dev,
 941                                                                  priv->base_qpn,
 942                                                                  priv->port);
 943                                 break;
 944
 945                         case MLX4_STEERING_MODE_A0:
 946                                 break;
 947                         }
 948                         if (err)
 949                                 en_err(priv, "Failed entering multicast promisc mode\n");
 950                         priv->flags |= MLX4_EN_FLAG_MC_PROMISC;
 951                 }
 952         } else {
 953                 /* Disable Multicast promisc */
 954                 if (priv->flags & MLX4_EN_FLAG_MC_PROMISC) {
 955                         switch (mdev->dev->caps.steering_mode) {
 956                         case MLX4_STEERING_MODE_DEVICE_MANAGED:
 957                                 err = mlx4_flow_steer_promisc_remove(mdev->dev,
 958                                                                      priv->port,
 959                                                                      MLX4_FS_MC_DEFAULT);
 960                                 break;
 961
 962                         case MLX4_STEERING_MODE_B0:
 963                                 err = mlx4_multicast_promisc_remove(mdev->dev,
 964                                                                     priv->base_qpn,
 965                                                                     priv->port);
 966                                 break;
 967
 968                         case MLX4_STEERING_MODE_A0:
 969                                 break;
 970                         }
 971                         if (err)
 972                                 en_err(priv, "Failed disabling multicast promiscuous mode\n");
 973                         priv->flags &= ~MLX4_EN_FLAG_MC_PROMISC;
 974                 }
 975
 976                 err = mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, 0,
 977                                           0, MLX4_MCAST_DISABLE);
 978                 if (err)
 979                         en_err(priv, "Failed disabling multicast filter\n");
 980
 981                 /* Flush mcast filter and init it with broadcast address */
 982                 mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, ETH_BCAST,
 983                                     1, MLX4_MCAST_CONFIG);
 984
 985                 /* Update multicast list - we cache all addresses so they won't
 986                  * change while HW is updated holding the command semaphor */
 987                 netif_addr_lock_bh(dev);
 988                 mlx4_en_cache_mclist(dev);
 989                 netif_addr_unlock_bh(dev);
 990                 list_for_each_entry(mclist, &priv->mc_list, list) {
 991                         mcast_addr = mlx4_mac_to_u64(mclist->addr);
 992                         mlx4_SET_MCAST_FLTR(mdev->dev, priv->port,
 993                                             mcast_addr, 0, MLX4_MCAST_CONFIG);
 994                 }
 995                 err = mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, 0,
 996                                           0, MLX4_MCAST_ENABLE);
 997                 if (err)
 998                         en_err(priv, "Failed enabling multicast filter\n");
 999
1000                 update_mclist_flags(priv, &priv->curr_list, &priv->mc_list);
1001                 list_for_each_entry_safe(mclist, tmp, &priv->curr_list, list) {
1002                         if (mclist->action == MCLIST_REM) {
1003                                 /* detach this address and delete from list */
1004                                 memcpy(&mc_list[10], mclist->addr, ETH_ALEN);
1005                                 mc_list[5] = priv->port;
1006                                 err = mlx4_multicast_detach(mdev->dev,
1007                                                             &priv->rss_map.indir_qp,
1008                                                             mc_list,
1009                                                             MLX4_PROT_ETH,
1010                                                             mclist->reg_id);
1011                                 if (err)
1012                                         en_err(priv, "Fail to detach multicast address\n");
1013
1014                                 if (mclist->tunnel_reg_id) {
1015                                         err = mlx4_flow_detach(priv->mdev->dev, mclist->tunnel_reg_id);
1016                                         if (err)
1017                                                 en_err(priv, "Failed to detach multicast address\n");
1018                                 }
1019
1020                                 /* remove from list */
1021                                 list_del(&mclist->list);
1022                                 kfree(mclist);
1023                         } else if (mclist->action == MCLIST_ADD) {
1024                                 /* attach the address */
1025                                 memcpy(&mc_list[10], mclist->addr, ETH_ALEN);
1026                                 /* needed for B0 steering support */
1027                                 mc_list[5] = priv->port;
1028                                 err = mlx4_multicast_attach(mdev->dev,
1029                                                             &priv->rss_map.indir_qp,
1030                                                             mc_list,
1031                                                             priv->port, 0,
1032                                                             MLX4_PROT_ETH,
1033                                                             &mclist->reg_id);
1034                                 if (err)
1035                                         en_err(priv, "Fail to attach multicast address\n");
1036
1037                                 err = mlx4_en_tunnel_steer_add(priv, &mc_list[10], priv->base_qpn,
1038                                                                &mclist->tunnel_reg_id);
1039                                 if (err)
1040                                         en_err(priv, "Failed to attach multicast address\n");
1041                         }
1042                 }
1043         }
1044 }
1045
1046 static void mlx4_en_do_uc_filter(struct mlx4_en_priv *priv,
1047                                  struct net_device *dev,
1048                                  struct mlx4_en_dev *mdev)
1049 {
1050         struct netdev_hw_addr *ha;
1051         struct mlx4_mac_entry *entry;
1052         struct hlist_node *tmp;
1053         bool found;
1054         u64 mac;
1055         int err = 0;
1056         struct hlist_head *bucket;
1057         unsigned int i;
1058         int removed = 0;
1059         u32 prev_flags;
1060
1061         /* Note that we do not need to protect our mac_hash traversal with rcu,
1062          * since all modification code is protected by mdev->state_lock
1063          */
1064
1065         /* find what to remove */
1066         for (i = 0; i < MLX4_EN_MAC_HASH_SIZE; ++i) {
1067                 bucket = &priv->mac_hash[i];
1068                 hlist_for_each_entry_safe(entry, tmp, bucket, hlist) {
1069                         found = false;
1070                         netdev_for_each_uc_addr(ha, dev) {
1071                                 if (ether_addr_equal_64bits(entry->mac,
1072                                                             ha->addr)) {
1073                                         found = true;
1074                                         break;
1075                                 }
1076                         }
1077
1078                         /* MAC address of the port is not in uc list */
1079                         if (ether_addr_equal_64bits(entry->mac,
1080                                                     priv->current_mac))
1081                                 found = true;
1082
1083                         if (!found) {
1084                                 mac = mlx4_mac_to_u64(entry->mac);
1085                                 mlx4_en_uc_steer_release(priv, entry->mac,
1086                                                          priv->base_qpn,
1087                                                          entry->reg_id);
1088                                 mlx4_unregister_mac(mdev->dev, priv->port, mac);
1089
1090                                 hlist_del_rcu(&entry->hlist);
1091                                 kfree_rcu(entry, rcu);
1092                                 en_dbg(DRV, priv, "Removed MAC %pM on port:%d\n",
1093                                        entry->mac, priv->port);
1094                                 ++removed;
1095                         }
1096                 }
1097         }
1098
1099         /* if we didn't remove anything, there is no use in trying to add
1100          * again once we are in a forced promisc mode state
1101          */
1102         if ((priv->flags & MLX4_EN_FLAG_FORCE_PROMISC) && 0 == removed)
1103                 return;
1104
1105         prev_flags = priv->flags;
1106         priv->flags &= ~MLX4_EN_FLAG_FORCE_PROMISC;
1107
1108         /* find what to add */
1109         netdev_for_each_uc_addr(ha, dev) {
1110                 found = false;
1111                 bucket = &priv->mac_hash[ha->addr[MLX4_EN_MAC_HASH_IDX]];
1112                 hlist_for_each_entry(entry, bucket, hlist) {
1113                         if (ether_addr_equal_64bits(entry->mac, ha->addr)) {
1114                                 found = true;
1115                                 break;
1116                         }
1117                 }
1118
1119                 if (!found) {
1120                         entry = kmalloc(sizeof(*entry), GFP_KERNEL);
1121                         if (!entry) {
1122                                 en_err(priv, "Failed adding MAC %pM on port:%d (out of memory)\n",
1123                                        ha->addr, priv->port);
1124                                 priv->flags |= MLX4_EN_FLAG_FORCE_PROMISC;
1125                                 break;
1126                         }
1127                         mac = mlx4_mac_to_u64(ha->addr);
1128                         memcpy(entry->mac, ha->addr, ETH_ALEN);
1129                         err = mlx4_register_mac(mdev->dev, priv->port, mac);
1130                         if (err < 0) {
1131                                 en_err(priv, "Failed registering MAC %pM on port %d: %d\n",
1132                                        ha->addr, priv->port, err);
1133                                 kfree(entry);
1134                                 priv->flags |= MLX4_EN_FLAG_FORCE_PROMISC;
1135                                 break;
1136                         }
1137                         err = mlx4_en_uc_steer_add(priv, ha->addr,
1138                                                    &priv->base_qpn,
1139                                                    &entry->reg_id);
1140                         if (err) {
1141                                 en_err(priv, "Failed adding MAC %pM on port %d: %d\n",
1142                                        ha->addr, priv->port, err);
1143                                 mlx4_unregister_mac(mdev->dev, priv->port, mac);
1144                                 kfree(entry);
1145                                 priv->flags |= MLX4_EN_FLAG_FORCE_PROMISC;
1146                                 break;
1147                         } else {
1148                                 unsigned int mac_hash;
1149                                 en_dbg(DRV, priv, "Added MAC %pM on port:%d\n",
1150                                        ha->addr, priv->port);
1151                                 mac_hash = ha->addr[MLX4_EN_MAC_HASH_IDX];
1152                                 bucket = &priv->mac_hash[mac_hash];
1153                                 hlist_add_head_rcu(&entry->hlist, bucket);
1154                         }
1155                 }
1156         }
1157
1158         if (priv->flags & MLX4_EN_FLAG_FORCE_PROMISC) {
1159                 en_warn(priv, "Forcing promiscuous mode on port:%d\n",
1160                         priv->port);
1161         } else if (prev_flags & MLX4_EN_FLAG_FORCE_PROMISC) {
1162                 en_warn(priv, "Stop forcing promiscuous mode on port:%d\n",
1163                         priv->port);
1164         }
1165 }
1166
1167 static void mlx4_en_do_set_rx_mode(struct work_struct *work)
1168 {
1169         struct mlx4_en_priv *priv = container_of(work, struct mlx4_en_priv,
1170                                                  rx_mode_task);
1171         struct mlx4_en_dev *mdev = priv->mdev;
1172         struct net_device *dev = priv->dev;
1173
1174         mutex_lock(&mdev->state_lock);
1175         if (!mdev->device_up) {
1176                 en_dbg(HW, priv, "Card is not up, ignoring rx mode change.\n");
1177                 goto out;
1178         }
1179         if (!priv->port_up) {
1180                 en_dbg(HW, priv, "Port is down, ignoring rx mode change.\n");
1181                 goto out;
1182         }
1183
1184         if (!netif_carrier_ok(dev)) {
1185                 if (!mlx4_en_QUERY_PORT(mdev, priv->port)) {
1186                         if (priv->port_state.link_state) {
1187                                 priv->last_link_state = MLX4_DEV_EVENT_PORT_UP;
1188                                 netif_carrier_on(dev);
1189                                 en_dbg(LINK, priv, "Link Up\n");
1190                         }
1191                 }
1192         }
1193
1194         if (dev->priv_flags & IFF_UNICAST_FLT)
1195                 mlx4_en_do_uc_filter(priv, dev, mdev);
1196
1197         /* Promsicuous mode: disable all filters */
1198         if ((dev->flags & IFF_PROMISC) ||
1199             (priv->flags & MLX4_EN_FLAG_FORCE_PROMISC)) {
1200                 mlx4_en_set_promisc_mode(priv, mdev);
1201                 goto out;
1202         }
1203
1204         /* Not in promiscuous mode */
1205         if (priv->flags & MLX4_EN_FLAG_PROMISC)
1206                 mlx4_en_clear_promisc_mode(priv, mdev);
1207
1208         mlx4_en_do_multicast(priv, dev, mdev);
1209 out:
1210         mutex_unlock(&mdev->state_lock);
1211 }
1212
1213 #ifdef CONFIG_NET_POLL_CONTROLLER
1214 static void mlx4_en_netpoll(struct net_device *dev)
1215 {
1216         struct mlx4_en_priv *priv = netdev_priv(dev);
1217         struct mlx4_en_cq *cq;
1218         int i;
1219
1220         for (i = 0; i < priv->tx_ring_num; i++) {
1221                 cq = priv->tx_cq[i];
1222                 napi_schedule(&cq->napi);
1223         }
1224 }
1225 #endif
1226
1227 static int mlx4_en_set_rss_steer_rules(struct mlx4_en_priv *priv)
1228 {
1229         u64 reg_id;
1230         int err = 0;
1231         int *qpn = &priv->base_qpn;
1232         struct mlx4_mac_entry *entry;
1233
1234         err = mlx4_en_uc_steer_add(priv, priv->dev->dev_addr, qpn, &reg_id);
1235         if (err)
1236                 return err;
1237
1238         err = mlx4_en_tunnel_steer_add(priv, priv->dev->dev_addr, *qpn,
1239                                        &priv->tunnel_reg_id);
1240         if (err)
1241                 goto tunnel_err;
1242
1243         entry = kmalloc(sizeof(*entry), GFP_KERNEL);
1244         if (!entry) {
1245                 err = -ENOMEM;
1246                 goto alloc_err;
1247         }
1248
1249         memcpy(entry->mac, priv->dev->dev_addr, sizeof(entry->mac));
1250         memcpy(priv->current_mac, entry->mac, sizeof(priv->current_mac));
1251         entry->reg_id = reg_id;
1252         hlist_add_head_rcu(&entry->hlist,
1253                            &priv->mac_hash[entry->mac[MLX4_EN_MAC_HASH_IDX]]);
1254
1255         return 0;
1256
1257 alloc_err:
1258         if (priv->tunnel_reg_id)
1259                 mlx4_flow_detach(priv->mdev->dev, priv->tunnel_reg_id);
1260
1261 tunnel_err:
1262         mlx4_en_uc_steer_release(priv, priv->dev->dev_addr, *qpn, reg_id);
1263         return err;
1264 }
1265
1266 static void mlx4_en_delete_rss_steer_rules(struct mlx4_en_priv *priv)
1267 {
1268         u64 mac;
1269         unsigned int i;
1270         int qpn = priv->base_qpn;
1271         struct hlist_head *bucket;
1272         struct hlist_node *tmp;
1273         struct mlx4_mac_entry *entry;
1274
1275         for (i = 0; i < MLX4_EN_MAC_HASH_SIZE; ++i) {
1276                 bucket = &priv->mac_hash[i];
1277                 hlist_for_each_entry_safe(entry, tmp, bucket, hlist) {
1278                         mac = mlx4_mac_to_u64(entry->mac);
1279                         en_dbg(DRV, priv, "Registering MAC:%pM for deleting\n",
1280                                entry->mac);
1281                         mlx4_en_uc_steer_release(priv, entry->mac,
1282                                                  qpn, entry->reg_id);
1283
1284                         mlx4_unregister_mac(priv->mdev->dev, priv->port, mac);
1285                         hlist_del_rcu(&entry->hlist);
1286                         kfree_rcu(entry, rcu);
1287                 }
1288         }
1289
1290         if (priv->tunnel_reg_id) {
1291                 mlx4_flow_detach(priv->mdev->dev, priv->tunnel_reg_id);
1292                 priv->tunnel_reg_id = 0;
1293         }
1294 }
1295
1296 static void mlx4_en_tx_timeout(struct net_device *dev)
1297 {
1298         struct mlx4_en_priv *priv = netdev_priv(dev);
1299         struct mlx4_en_dev *mdev = priv->mdev;
1300         int i;
1301
1302         if (netif_msg_timer(priv))
1303                 en_warn(priv, "Tx timeout called on port:%d\n", priv->port);
1304
1305         for (i = 0; i < priv->tx_ring_num; i++) {
1306                 if (!netif_tx_queue_stopped(netdev_get_tx_queue(dev, i)))
1307                         continue;
1308                 en_warn(priv, "TX timeout on queue: %d, QP: 0x%x, CQ: 0x%x, Cons: 0x%x, Prod: 0x%x\n",
1309                         i, priv->tx_ring[i]->qpn, priv->tx_ring[i]->cqn,
1310                         priv->tx_ring[i]->cons, priv->tx_ring[i]->prod);
1311         }
1312
1313         priv->port_stats.tx_timeout++;
1314         if (!test_and_set_bit(MLX4_EN_STATE_FLAG_RESTARTING, &priv->state)) {
1315                 en_dbg(DRV, priv, "Scheduling port restart\n");
1316                 queue_work(mdev->workqueue, &priv->restart_task);
1317         }
1318 }
1319
1320
1321 static struct rtnl_link_stats64 *
1322 mlx4_en_get_stats64(struct net_device *dev, struct rtnl_link_stats64 *stats)
1323 {
1324         struct mlx4_en_priv *priv = netdev_priv(dev);
1325
1326         spin_lock_bh(&priv->stats_lock);
1327         netdev_stats_to_stats64(stats, &dev->stats);
1328         spin_unlock_bh(&priv->stats_lock);
1329
1330         return stats;
1331 }
1332
1333 static void mlx4_en_set_default_moderation(struct mlx4_en_priv *priv)
1334 {
1335         struct mlx4_en_cq *cq;
1336         int i;
1337
1338         /* If we haven't received a specific coalescing setting
1339          * (module param), we set the moderation parameters as follows:
1340          * - moder_cnt is set to the number of mtu sized packets to
1341          *   satisfy our coalescing target.
1342          * - moder_time is set to a fixed value.
1343          */
1344         priv->rx_frames = MLX4_EN_RX_COAL_TARGET;
1345         priv->rx_usecs = MLX4_EN_RX_COAL_TIME;
1346         priv->tx_frames = MLX4_EN_TX_COAL_PKTS;
1347         priv->tx_usecs = MLX4_EN_TX_COAL_TIME;
1348         en_dbg(INTR, priv, "Default coalesing params for mtu:%d - rx_frames:%d rx_usecs:%d\n",
1349                priv->dev->mtu, priv->rx_frames, priv->rx_usecs);
1350
1351         /* Setup cq moderation params */
1352         for (i = 0; i < priv->rx_ring_num; i++) {
1353                 cq = priv->rx_cq[i];
1354                 cq->moder_cnt = priv->rx_frames;
1355                 cq->moder_time = priv->rx_usecs;
1356                 priv->last_moder_time[i] = MLX4_EN_AUTO_CONF;
1357                 priv->last_moder_packets[i] = 0;
1358                 priv->last_moder_bytes[i] = 0;
1359         }
1360
1361         for (i = 0; i < priv->tx_ring_num; i++) {
1362                 cq = priv->tx_cq[i];
1363                 cq->moder_cnt = priv->tx_frames;
1364                 cq->moder_time = priv->tx_usecs;
1365         }
1366
1367         /* Reset auto-moderation params */
1368         priv->pkt_rate_low = MLX4_EN_RX_RATE_LOW;
1369         priv->rx_usecs_low = MLX4_EN_RX_COAL_TIME_LOW;
1370         priv->pkt_rate_high = MLX4_EN_RX_RATE_HIGH;
1371         priv->rx_usecs_high = MLX4_EN_RX_COAL_TIME_HIGH;
1372         priv->sample_interval = MLX4_EN_SAMPLE_INTERVAL;
1373         priv->adaptive_rx_coal = 1;
1374         priv->last_moder_jiffies = 0;
1375         priv->last_moder_tx_packets = 0;
1376 }
1377
1378 static void mlx4_en_auto_moderation(struct mlx4_en_priv *priv)
1379 {
1380         unsigned long period = (unsigned long) (jiffies - priv->last_moder_jiffies);
1381         struct mlx4_en_cq *cq;
1382         unsigned long packets;
1383         unsigned long rate;
1384         unsigned long avg_pkt_size;
1385         unsigned long rx_packets;
1386         unsigned long rx_bytes;
1387         unsigned long rx_pkt_diff;
1388         int moder_time;
1389         int ring, err;
1390
1391         if (!priv->adaptive_rx_coal || period < priv->sample_interval * HZ)
1392                 return;
1393
1394         for (ring = 0; ring < priv->rx_ring_num; ring++) {
1395                 spin_lock_bh(&priv->stats_lock);
1396                 rx_packets = priv->rx_ring[ring]->packets;
1397                 rx_bytes = priv->rx_ring[ring]->bytes;
1398                 spin_unlock_bh(&priv->stats_lock);
1399
1400                 rx_pkt_diff = ((unsigned long) (rx_packets -
1401                                 priv->last_moder_packets[ring]));
1402                 packets = rx_pkt_diff;
1403                 rate = packets * HZ / period;
1404                 avg_pkt_size = packets ? ((unsigned long) (rx_bytes -
1405                                 priv->last_moder_bytes[ring])) / packets : 0;
1406
1407                 /* Apply auto-moderation only when packet rate
1408                  * exceeds a rate that it matters */
1409                 if (rate > (MLX4_EN_RX_RATE_THRESH / priv->rx_ring_num) &&
1410                     avg_pkt_size > MLX4_EN_AVG_PKT_SMALL) {
1411                         if (rate < priv->pkt_rate_low)
1412                                 moder_time = priv->rx_usecs_low;
1413                         else if (rate > priv->pkt_rate_high)
1414                                 moder_time = priv->rx_usecs_high;
1415                         else
1416                                 moder_time = (rate - priv->pkt_rate_low) *
1417                                         (priv->rx_usecs_high - priv->rx_usecs_low) /
1418                                         (priv->pkt_rate_high - priv->pkt_rate_low) +
1419                                         priv->rx_usecs_low;
1420                 } else {
1421                         moder_time = priv->rx_usecs_low;
1422                 }
1423
1424                 if (moder_time != priv->last_moder_time[ring]) {
1425                         priv->last_moder_time[ring] = moder_time;
1426                         cq = priv->rx_cq[ring];
1427                         cq->moder_time = moder_time;
1428                         cq->moder_cnt = priv->rx_frames;
1429                         err = mlx4_en_set_cq_moder(priv, cq);
1430                         if (err)
1431                                 en_err(priv, "Failed modifying moderation for cq:%d\n",
1432                                        ring);
1433                 }
1434                 priv->last_moder_packets[ring] = rx_packets;
1435                 priv->last_moder_bytes[ring] = rx_bytes;
1436         }
1437
1438         priv->last_moder_jiffies = jiffies;
1439 }
1440
1441 static void mlx4_en_do_get_stats(struct work_struct *work)
1442 {
1443         struct delayed_work *delay = to_delayed_work(work);
1444         struct mlx4_en_priv *priv = container_of(delay, struct mlx4_en_priv,
1445                                                  stats_task);
1446         struct mlx4_en_dev *mdev = priv->mdev;
1447         int err;
1448
1449         mutex_lock(&mdev->state_lock);
1450         if (mdev->device_up) {
1451                 if (priv->port_up) {
1452                         err = mlx4_en_DUMP_ETH_STATS(mdev, priv->port, 0);
1453                         if (err)
1454                                 en_dbg(HW, priv, "Could not update stats\n");
1455
1456                         mlx4_en_auto_moderation(priv);
1457                 }
1458
1459                 queue_delayed_work(mdev->workqueue, &priv->stats_task, STATS_DELAY);
1460         }
1461         if (mdev->mac_removed[MLX4_MAX_PORTS + 1 - priv->port]) {
1462                 mlx4_en_do_set_mac(priv, priv->current_mac);
1463                 mdev->mac_removed[MLX4_MAX_PORTS + 1 - priv->port] = 0;
1464         }
1465         mutex_unlock(&mdev->state_lock);
1466 }
1467
1468 /* mlx4_en_service_task - Run service task for tasks that needed to be done
1469  * periodically
1470  */
1471 static void mlx4_en_service_task(struct work_struct *work)
1472 {
1473         struct delayed_work *delay = to_delayed_work(work);
1474         struct mlx4_en_priv *priv = container_of(delay, struct mlx4_en_priv,
1475                                                  service_task);
1476         struct mlx4_en_dev *mdev = priv->mdev;
1477
1478         mutex_lock(&mdev->state_lock);
1479         if (mdev->device_up) {
1480                 if (mdev->dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_TS)
1481                         mlx4_en_ptp_overflow_check(mdev);
1482
1483                 mlx4_en_recover_from_oom(priv);
1484                 queue_delayed_work(mdev->workqueue, &priv->service_task,
1485                                    SERVICE_TASK_DELAY);
1486         }
1487         mutex_unlock(&mdev->state_lock);
1488 }
1489
1490 static void mlx4_en_linkstate(struct work_struct *work)
1491 {
1492         struct mlx4_en_priv *priv = container_of(work, struct mlx4_en_priv,
1493                                                  linkstate_task);
1494         struct mlx4_en_dev *mdev = priv->mdev;
1495         int linkstate = priv->link_state;
1496
1497         mutex_lock(&mdev->state_lock);
1498         /* If observable port state changed set carrier state and
1499          * report to system log */
1500         if (priv->last_link_state != linkstate) {
1501                 if (linkstate == MLX4_DEV_EVENT_PORT_DOWN) {
1502                         en_info(priv, "Link Down\n");
1503                         netif_carrier_off(priv->dev);
1504                 } else {
1505                         en_info(priv, "Link Up\n");
1506                         netif_carrier_on(priv->dev);
1507                 }
1508         }
1509         priv->last_link_state = linkstate;
1510         mutex_unlock(&mdev->state_lock);
1511 }
1512
1513 static int mlx4_en_init_affinity_hint(struct mlx4_en_priv *priv, int ring_idx)
1514 {
1515         struct mlx4_en_rx_ring *ring = priv->rx_ring[ring_idx];
1516         int numa_node = priv->mdev->dev->numa_node;
1517
1518         if (!zalloc_cpumask_var(&ring->affinity_mask, GFP_KERNEL))
1519                 return -ENOMEM;
1520
1521         cpumask_set_cpu(cpumask_local_spread(ring_idx, numa_node),
1522                         ring->affinity_mask);
1523         return 0;
1524 }
1525
1526 static void mlx4_en_free_affinity_hint(struct mlx4_en_priv *priv, int ring_idx)
1527 {
1528         free_cpumask_var(priv->rx_ring[ring_idx]->affinity_mask);
1529 }
1530
1531 static void mlx4_en_init_recycle_ring(struct mlx4_en_priv *priv,
1532                                       int tx_ring_idx)
1533 {
1534         struct mlx4_en_tx_ring *tx_ring = priv->tx_ring[tx_ring_idx];
1535         int rr_index;
1536
1537         rr_index = (priv->xdp_ring_num - priv->tx_ring_num) + tx_ring_idx;
1538         if (rr_index >= 0) {
1539                 tx_ring->free_tx_desc = mlx4_en_recycle_tx_desc;
1540                 tx_ring->recycle_ring = priv->rx_ring[rr_index];
1541                 en_dbg(DRV, priv,
1542                        "Set tx_ring[%d]->recycle_ring = rx_ring[%d]\n",
1543                        tx_ring_idx, rr_index);
1544         } else {
1545                 tx_ring->recycle_ring = NULL;
1546         }
1547 }
1548
1549 int mlx4_en_start_port(struct net_device *dev)
1550 {
1551         struct mlx4_en_priv *priv = netdev_priv(dev);
1552         struct mlx4_en_dev *mdev = priv->mdev;
1553         struct mlx4_en_cq *cq;
1554         struct mlx4_en_tx_ring *tx_ring;
1555         int rx_index = 0;
1556         int tx_index = 0;
1557         int err = 0;
1558         int i;
1559         int j;
1560         u8 mc_list[16] = {0};
1561
1562         if (priv->port_up) {
1563                 en_dbg(DRV, priv, "start port called while port already up\n");
1564                 return 0;
1565         }
1566
1567         INIT_LIST_HEAD(&priv->mc_list);
1568         INIT_LIST_HEAD(&priv->curr_list);
1569         INIT_LIST_HEAD(&priv->ethtool_list);
1570         memset(&priv->ethtool_rules[0], 0,
1571                sizeof(struct ethtool_flow_id) * MAX_NUM_OF_FS_RULES);
1572
1573         /* Calculate Rx buf size */
1574         dev->mtu = min(dev->mtu, priv->max_mtu);
1575         mlx4_en_calc_rx_buf(dev);
1576         en_dbg(DRV, priv, "Rx buf size:%d\n", priv->rx_skb_size);
1577
1578         /* Configure rx cq's and rings */
1579         err = mlx4_en_activate_rx_rings(priv);
1580         if (err) {
1581                 en_err(priv, "Failed to activate RX rings\n");
1582                 return err;
1583         }
1584         for (i = 0; i < priv->rx_ring_num; i++) {
1585                 cq = priv->rx_cq[i];
1586
1587                 err = mlx4_en_init_affinity_hint(priv, i);
1588                 if (err) {
1589                         en_err(priv, "Failed preparing IRQ affinity hint\n");
1590                         goto cq_err;
1591                 }
1592
1593                 err = mlx4_en_activate_cq(priv, cq, i);
1594                 if (err) {
1595                         en_err(priv, "Failed activating Rx CQ\n");
1596                         mlx4_en_free_affinity_hint(priv, i);
1597                         goto cq_err;
1598                 }
1599
1600                 for (j = 0; j < cq->size; j++) {
1601                         struct mlx4_cqe *cqe = NULL;
1602
1603                         cqe = mlx4_en_get_cqe(cq->buf, j, priv->cqe_size) +
1604                               priv->cqe_factor;
1605                         cqe->owner_sr_opcode = MLX4_CQE_OWNER_MASK;
1606                 }
1607
1608                 err = mlx4_en_set_cq_moder(priv, cq);
1609                 if (err) {
1610                         en_err(priv, "Failed setting cq moderation parameters\n");
1611                         mlx4_en_deactivate_cq(priv, cq);
1612                         mlx4_en_free_affinity_hint(priv, i);
1613                         goto cq_err;
1614                 }
1615                 mlx4_en_arm_cq(priv, cq);
1616                 priv->rx_ring[i]->cqn = cq->mcq.cqn;
1617                 ++rx_index;
1618         }
1619
1620         /* Set qp number */
1621         en_dbg(DRV, priv, "Getting qp number for port %d\n", priv->port);
1622         err = mlx4_en_get_qp(priv);
1623         if (err) {
1624                 en_err(priv, "Failed getting eth qp\n");
1625                 goto cq_err;
1626         }
1627         mdev->mac_removed[priv->port] = 0;
1628
1629         priv->counter_index =
1630                         mlx4_get_default_counter_index(mdev->dev, priv->port);
1631
1632         err = mlx4_en_config_rss_steer(priv);
1633         if (err) {
1634                 en_err(priv, "Failed configuring rss steering\n");
1635                 goto mac_err;
1636         }
1637
1638         err = mlx4_en_create_drop_qp(priv);
1639         if (err)
1640                 goto rss_err;
1641
1642         /* Configure tx cq's and rings */
1643         for (i = 0; i < priv->tx_ring_num; i++) {
1644                 /* Configure cq */
1645                 cq = priv->tx_cq[i];
1646                 err = mlx4_en_activate_cq(priv, cq, i);
1647                 if (err) {
1648                         en_err(priv, "Failed allocating Tx CQ\n");
1649                         goto tx_err;
1650                 }
1651                 err = mlx4_en_set_cq_moder(priv, cq);
1652                 if (err) {
1653                         en_err(priv, "Failed setting cq moderation parameters\n");
1654                         mlx4_en_deactivate_cq(priv, cq);
1655                         goto tx_err;
1656                 }
1657                 en_dbg(DRV, priv, "Resetting index of collapsed CQ:%d to -1\n", i);
1658                 cq->buf->wqe_index = cpu_to_be16(0xffff);
1659
1660                 /* Configure ring */
1661                 tx_ring = priv->tx_ring[i];
1662                 err = mlx4_en_activate_tx_ring(priv, tx_ring, cq->mcq.cqn,
1663                         i / priv->num_tx_rings_p_up);
1664                 if (err) {
1665                         en_err(priv, "Failed allocating Tx ring\n");
1666                         mlx4_en_deactivate_cq(priv, cq);
1667                         goto tx_err;
1668                 }
1669                 tx_ring->tx_queue = netdev_get_tx_queue(dev, i);
1670
1671                 mlx4_en_init_recycle_ring(priv, i);
1672
1673                 /* Arm CQ for TX completions */
1674                 mlx4_en_arm_cq(priv, cq);
1675
1676                 /* Set initial ownership of all Tx TXBBs to SW (1) */
1677                 for (j = 0; j < tx_ring->buf_size; j += STAMP_STRIDE)
1678                         *((u32 *) (tx_ring->buf + j)) = 0xffffffff;
1679                 ++tx_index;
1680         }
1681
1682         /* Configure port */
1683         err = mlx4_SET_PORT_general(mdev->dev, priv->port,
1684                                     priv->rx_skb_size + ETH_FCS_LEN,
1685                                     priv->prof->tx_pause,
1686                                     priv->prof->tx_ppp,
1687                                     priv->prof->rx_pause,
1688                                     priv->prof->rx_ppp);
1689         if (err) {
1690                 en_err(priv, "Failed setting port general configurations for port %d, with error %d\n",
1691                        priv->port, err);
1692                 goto tx_err;
1693         }
1694         /* Set default qp number */
1695         err = mlx4_SET_PORT_qpn_calc(mdev->dev, priv->port, priv->base_qpn, 0);
1696         if (err) {
1697                 en_err(priv, "Failed setting default qp numbers\n");
1698                 goto tx_err;
1699         }
1700
1701         if (mdev->dev->caps.tunnel_offload_mode == MLX4_TUNNEL_OFFLOAD_MODE_VXLAN) {
1702                 err = mlx4_SET_PORT_VXLAN(mdev->dev, priv->port, VXLAN_STEER_BY_OUTER_MAC, 1);
1703                 if (err) {
1704                         en_err(priv, "Failed setting port L2 tunnel configuration, err %d\n",
1705                                err);
1706                         goto tx_err;
1707                 }
1708         }
1709
1710         /* Init port */
1711         en_dbg(HW, priv, "Initializing port\n");
1712         err = mlx4_INIT_PORT(mdev->dev, priv->port);
1713         if (err) {
1714                 en_err(priv, "Failed Initializing port\n");
1715                 goto tx_err;
1716         }
1717
1718         /* Set Unicast and VXLAN steering rules */
1719         if (mdev->dev->caps.steering_mode != MLX4_STEERING_MODE_A0 &&
1720             mlx4_en_set_rss_steer_rules(priv))
1721                 mlx4_warn(mdev, "Failed setting steering rules\n");
1722
1723         /* Attach rx QP to bradcast address */
1724         eth_broadcast_addr(&mc_list[10]);
1725         mc_list[5] = priv->port; /* needed for B0 steering support */
1726         if (mlx4_multicast_attach(mdev->dev, &priv->rss_map.indir_qp, mc_list,
1727                                   priv->port, 0, MLX4_PROT_ETH,
1728                                   &priv->broadcast_id))
1729                 mlx4_warn(mdev, "Failed Attaching Broadcast\n");
1730
1731         /* Must redo promiscuous mode setup. */
1732         priv->flags &= ~(MLX4_EN_FLAG_PROMISC | MLX4_EN_FLAG_MC_PROMISC);
1733
1734         /* Schedule multicast task to populate multicast list */
1735         queue_work(mdev->workqueue, &priv->rx_mode_task);
1736
1737         if (priv->mdev->dev->caps.tunnel_offload_mode == MLX4_TUNNEL_OFFLOAD_MODE_VXLAN)
1738                 udp_tunnel_get_rx_info(dev);
1739
1740         priv->port_up = true;
1741
1742         /* Process all completions if exist to prevent
1743          * the queues freezing if they are full
1744          */
1745         for (i = 0; i < priv->rx_ring_num; i++) {
1746                 local_bh_disable();
1747                 napi_schedule(&priv->rx_cq[i]->napi);
1748                 local_bh_enable();
1749         }
1750
1751         clear_bit(MLX4_EN_STATE_FLAG_RESTARTING, &priv->state);
1752         netif_tx_start_all_queues(dev);
1753         netif_device_attach(dev);
1754
1755         return 0;
1756
1757 tx_err:
1758         while (tx_index--) {
1759                 mlx4_en_deactivate_tx_ring(priv, priv->tx_ring[tx_index]);
1760                 mlx4_en_deactivate_cq(priv, priv->tx_cq[tx_index]);
1761         }
1762         mlx4_en_destroy_drop_qp(priv);
1763 rss_err:
1764         mlx4_en_release_rss_steer(priv);
1765 mac_err:
1766         mlx4_en_put_qp(priv);
1767 cq_err:
1768         while (rx_index--) {
1769                 mlx4_en_deactivate_cq(priv, priv->rx_cq[rx_index]);
1770                 mlx4_en_free_affinity_hint(priv, rx_index);
1771         }
1772         for (i = 0; i < priv->rx_ring_num; i++)
1773                 mlx4_en_deactivate_rx_ring(priv, priv->rx_ring[i]);
1774
1775         return err; /* need to close devices */
1776 }
1777
1778
1779 void mlx4_en_stop_port(struct net_device *dev, int detach)
1780 {
1781         struct mlx4_en_priv *priv = netdev_priv(dev);
1782         struct mlx4_en_dev *mdev = priv->mdev;
1783         struct mlx4_en_mc_list *mclist, *tmp;
1784         struct ethtool_flow_id *flow, *tmp_flow;
1785         int i;
1786         u8 mc_list[16] = {0};
1787
1788         if (!priv->port_up) {
1789                 en_dbg(DRV, priv, "stop port called while port already down\n");
1790                 return;
1791         }
1792
1793         /* close port*/
1794         mlx4_CLOSE_PORT(mdev->dev, priv->port);
1795
1796         /* Synchronize with tx routine */
1797         netif_tx_lock_bh(dev);
1798         if (detach)
1799                 netif_device_detach(dev);
1800         netif_tx_stop_all_queues(dev);
1801         netif_tx_unlock_bh(dev);
1802
1803         netif_tx_disable(dev);
1804
1805         /* Set port as not active */
1806         priv->port_up = false;
1807         priv->counter_index = MLX4_SINK_COUNTER_INDEX(mdev->dev);
1808
1809         /* Promsicuous mode */
1810         if (mdev->dev->caps.steering_mode ==
1811             MLX4_STEERING_MODE_DEVICE_MANAGED) {
1812                 priv->flags &= ~(MLX4_EN_FLAG_PROMISC |
1813                                  MLX4_EN_FLAG_MC_PROMISC);
1814                 mlx4_flow_steer_promisc_remove(mdev->dev,
1815                                                priv->port,
1816                                                MLX4_FS_ALL_DEFAULT);
1817                 mlx4_flow_steer_promisc_remove(mdev->dev,
1818                                                priv->port,
1819                                                MLX4_FS_MC_DEFAULT);
1820         } else if (priv->flags & MLX4_EN_FLAG_PROMISC) {
1821                 priv->flags &= ~MLX4_EN_FLAG_PROMISC;
1822
1823                 /* Disable promiscouos mode */
1824                 mlx4_unicast_promisc_remove(mdev->dev, priv->base_qpn,
1825                                             priv->port);
1826
1827                 /* Disable Multicast promisc */
1828                 if (priv->flags & MLX4_EN_FLAG_MC_PROMISC) {
1829                         mlx4_multicast_promisc_remove(mdev->dev, priv->base_qpn,
1830                                                       priv->port);
1831                         priv->flags &= ~MLX4_EN_FLAG_MC_PROMISC;
1832                 }
1833         }
1834
1835         /* Detach All multicasts */
1836         eth_broadcast_addr(&mc_list[10]);
1837         mc_list[5] = priv->port; /* needed for B0 steering support */
1838         mlx4_multicast_detach(mdev->dev, &priv->rss_map.indir_qp, mc_list,
1839                               MLX4_PROT_ETH, priv->broadcast_id);
1840         list_for_each_entry(mclist, &priv->curr_list, list) {
1841                 memcpy(&mc_list[10], mclist->addr, ETH_ALEN);
1842                 mc_list[5] = priv->port;
1843                 mlx4_multicast_detach(mdev->dev, &priv->rss_map.indir_qp,
1844                                       mc_list, MLX4_PROT_ETH, mclist->reg_id);
1845                 if (mclist->tunnel_reg_id)
1846                         mlx4_flow_detach(mdev->dev, mclist->tunnel_reg_id);
1847         }
1848         mlx4_en_clear_list(dev);
1849         list_for_each_entry_safe(mclist, tmp, &priv->curr_list, list) {
1850                 list_del(&mclist->list);
1851                 kfree(mclist);
1852         }
1853
1854         /* Flush multicast filter */
1855         mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, 0, 1, MLX4_MCAST_CONFIG);
1856
1857         /* Remove flow steering rules for the port*/
1858         if (mdev->dev->caps.steering_mode ==
1859             MLX4_STEERING_MODE_DEVICE_MANAGED) {
1860                 ASSERT_RTNL();
1861                 list_for_each_entry_safe(flow, tmp_flow,
1862                                          &priv->ethtool_list, list) {
1863                         mlx4_flow_detach(mdev->dev, flow->id);
1864                         list_del(&flow->list);
1865                 }
1866         }
1867
1868         mlx4_en_destroy_drop_qp(priv);
1869
1870         /* Free TX Rings */
1871         for (i = 0; i < priv->tx_ring_num; i++) {
1872                 mlx4_en_deactivate_tx_ring(priv, priv->tx_ring[i]);
1873                 mlx4_en_deactivate_cq(priv, priv->tx_cq[i]);
1874         }
1875         msleep(10);
1876
1877         for (i = 0; i < priv->tx_ring_num; i++)
1878                 mlx4_en_free_tx_buf(dev, priv->tx_ring[i]);
1879
1880         if (mdev->dev->caps.steering_mode != MLX4_STEERING_MODE_A0)
1881                 mlx4_en_delete_rss_steer_rules(priv);
1882
1883         /* Free RSS qps */
1884         mlx4_en_release_rss_steer(priv);
1885
1886         /* Unregister Mac address for the port */
1887         mlx4_en_put_qp(priv);
1888         if (!(mdev->dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_REASSIGN_MAC_EN))
1889                 mdev->mac_removed[priv->port] = 1;
1890
1891         /* Free RX Rings */
1892         for (i = 0; i < priv->rx_ring_num; i++) {
1893                 struct mlx4_en_cq *cq = priv->rx_cq[i];
1894
1895                 napi_synchronize(&cq->napi);
1896                 mlx4_en_deactivate_rx_ring(priv, priv->rx_ring[i]);
1897                 mlx4_en_deactivate_cq(priv, cq);
1898
1899                 mlx4_en_free_affinity_hint(priv, i);
1900         }
1901 }
1902
1903 static void mlx4_en_restart(struct work_struct *work)
1904 {
1905         struct mlx4_en_priv *priv = container_of(work, struct mlx4_en_priv,
1906                                                  restart_task);
1907         struct mlx4_en_dev *mdev = priv->mdev;
1908         struct net_device *dev = priv->dev;
1909
1910         en_dbg(DRV, priv, "Watchdog task called for port %d\n", priv->port);
1911
1912         rtnl_lock();
1913         mutex_lock(&mdev->state_lock);
1914         if (priv->port_up) {
1915                 mlx4_en_stop_port(dev, 1);
1916                 if (mlx4_en_start_port(dev))
1917                         en_err(priv, "Failed restarting port %d\n", priv->port);
1918         }
1919         mutex_unlock(&mdev->state_lock);
1920         rtnl_unlock();
1921 }
1922
1923 static void mlx4_en_clear_stats(struct net_device *dev)
1924 {
1925         struct mlx4_en_priv *priv = netdev_priv(dev);
1926         struct mlx4_en_dev *mdev = priv->mdev;
1927         int i;
1928
1929         if (!mlx4_is_slave(mdev->dev))
1930                 if (mlx4_en_DUMP_ETH_STATS(mdev, priv->port, 1))
1931                         en_dbg(HW, priv, "Failed dumping statistics\n");
1932
1933         memset(&priv->pstats, 0, sizeof(priv->pstats));
1934         memset(&priv->pkstats, 0, sizeof(priv->pkstats));
1935         memset(&priv->port_stats, 0, sizeof(priv->port_stats));
1936         memset(&priv->rx_flowstats, 0, sizeof(priv->rx_flowstats));
1937         memset(&priv->tx_flowstats, 0, sizeof(priv->tx_flowstats));
1938         memset(&priv->rx_priority_flowstats, 0,
1939                sizeof(priv->rx_priority_flowstats));
1940         memset(&priv->tx_priority_flowstats, 0,
1941                sizeof(priv->tx_priority_flowstats));
1942         memset(&priv->pf_stats, 0, sizeof(priv->pf_stats));
1943
1944         for (i = 0; i < priv->tx_ring_num; i++) {
1945                 priv->tx_ring[i]->bytes = 0;
1946                 priv->tx_ring[i]->packets = 0;
1947                 priv->tx_ring[i]->tx_csum = 0;
1948                 priv->tx_ring[i]->tx_dropped = 0;
1949                 priv->tx_ring[i]->queue_stopped = 0;
1950                 priv->tx_ring[i]->wake_queue = 0;
1951                 priv->tx_ring[i]->tso_packets = 0;
1952                 priv->tx_ring[i]->xmit_more = 0;
1953         }
1954         for (i = 0; i < priv->rx_ring_num; i++) {
1955                 priv->rx_ring[i]->bytes = 0;
1956                 priv->rx_ring[i]->packets = 0;
1957                 priv->rx_ring[i]->csum_ok = 0;
1958                 priv->rx_ring[i]->csum_none = 0;
1959                 priv->rx_ring[i]->csum_complete = 0;
1960         }
1961 }
1962
1963 static int mlx4_en_open(struct net_device *dev)
1964 {
1965         struct mlx4_en_priv *priv = netdev_priv(dev);
1966         struct mlx4_en_dev *mdev = priv->mdev;
1967         int err = 0;
1968
1969         mutex_lock(&mdev->state_lock);
1970
1971         if (!mdev->device_up) {
1972                 en_err(priv, "Cannot open - device down/disabled\n");
1973                 err = -EBUSY;
1974                 goto out;
1975         }
1976
1977         /* Reset HW statistics and SW counters */
1978         mlx4_en_clear_stats(dev);
1979
1980         err = mlx4_en_start_port(dev);
1981         if (err)
1982                 en_err(priv, "Failed starting port:%d\n", priv->port);
1983
1984 out:
1985         mutex_unlock(&mdev->state_lock);
1986         return err;
1987 }
1988
1989
1990 static int mlx4_en_close(struct net_device *dev)
1991 {
1992         struct mlx4_en_priv *priv = netdev_priv(dev);
1993         struct mlx4_en_dev *mdev = priv->mdev;
1994
1995         en_dbg(IFDOWN, priv, "Close port called\n");
1996
1997         mutex_lock(&mdev->state_lock);
1998
1999         mlx4_en_stop_port(dev, 0);
2000         netif_carrier_off(dev);
2001
2002         mutex_unlock(&mdev->state_lock);
2003         return 0;
2004 }
2005
2006 static void mlx4_en_free_resources(struct mlx4_en_priv *priv)
2007 {
2008         int i;
2009
2010 #ifdef CONFIG_RFS_ACCEL
2011         priv->dev->rx_cpu_rmap = NULL;
2012 #endif
2013
2014         for (i = 0; i < priv->tx_ring_num; i++) {
2015                 if (priv->tx_ring && priv->tx_ring[i])
2016                         mlx4_en_destroy_tx_ring(priv, &priv->tx_ring[i]);
2017                 if (priv->tx_cq && priv->tx_cq[i])
2018                         mlx4_en_destroy_cq(priv, &priv->tx_cq[i]);
2019         }
2020
2021         for (i = 0; i < priv->rx_ring_num; i++) {
2022                 if (priv->rx_ring[i])
2023                         mlx4_en_destroy_rx_ring(priv, &priv->rx_ring[i],
2024                                 priv->prof->rx_ring_size, priv->stride);
2025                 if (priv->rx_cq[i])
2026                         mlx4_en_destroy_cq(priv, &priv->rx_cq[i]);
2027         }
2028
2029 }
2030
2031 static int mlx4_en_alloc_resources(struct mlx4_en_priv *priv)
2032 {
2033         struct mlx4_en_port_profile *prof = priv->prof;
2034         int i;
2035         int node;
2036
2037         /* Create tx Rings */
2038         for (i = 0; i < priv->tx_ring_num; i++) {
2039                 node = cpu_to_node(i % num_online_cpus());
2040                 if (mlx4_en_create_cq(priv, &priv->tx_cq[i],
2041                                       prof->tx_ring_size, i, TX, node))
2042                         goto err;
2043
2044                 if (mlx4_en_create_tx_ring(priv, &priv->tx_ring[i],
2045                                            prof->tx_ring_size, TXBB_SIZE,
2046                                            node, i))
2047                         goto err;
2048         }
2049
2050         /* Create rx Rings */
2051         for (i = 0; i < priv->rx_ring_num; i++) {
2052                 node = cpu_to_node(i % num_online_cpus());
2053                 if (mlx4_en_create_cq(priv, &priv->rx_cq[i],
2054                                       prof->rx_ring_size, i, RX, node))
2055                         goto err;
2056
2057                 if (mlx4_en_create_rx_ring(priv, &priv->rx_ring[i],
2058                                            prof->rx_ring_size, priv->stride,
2059                                            node))
2060                         goto err;
2061         }
2062
2063 #ifdef CONFIG_RFS_ACCEL
2064         priv->dev->rx_cpu_rmap = mlx4_get_cpu_rmap(priv->mdev->dev, priv->port);
2065 #endif
2066
2067         return 0;
2068
2069 err:
2070         en_err(priv, "Failed to allocate NIC resources\n");
2071         for (i = 0; i < priv->rx_ring_num; i++) {
2072                 if (priv->rx_ring[i])
2073                         mlx4_en_destroy_rx_ring(priv, &priv->rx_ring[i],
2074                                                 prof->rx_ring_size,
2075                                                 priv->stride);
2076                 if (priv->rx_cq[i])
2077                         mlx4_en_destroy_cq(priv, &priv->rx_cq[i]);
2078         }
2079         for (i = 0; i < priv->tx_ring_num; i++) {
2080                 if (priv->tx_ring[i])
2081                         mlx4_en_destroy_tx_ring(priv, &priv->tx_ring[i]);
2082                 if (priv->tx_cq[i])
2083                         mlx4_en_destroy_cq(priv, &priv->tx_cq[i]);
2084         }
2085         return -ENOMEM;
2086 }
2087
2088
2089 static int mlx4_en_copy_priv(struct mlx4_en_priv *dst,
2090                              struct mlx4_en_priv *src,
2091                              struct mlx4_en_port_profile *prof)
2092 {
2093         memcpy(&dst->hwtstamp_config, &prof->hwtstamp_config,
2094                sizeof(dst->hwtstamp_config));
2095         dst->num_tx_rings_p_up = src->mdev->profile.num_tx_rings_p_up;
2096         dst->tx_ring_num = prof->tx_ring_num;
2097         dst->rx_ring_num = prof->rx_ring_num;
2098         dst->flags = prof->flags;
2099         dst->mdev = src->mdev;
2100         dst->port = src->port;
2101         dst->dev = src->dev;
2102         dst->prof = prof;
2103         dst->stride = roundup_pow_of_two(sizeof(struct mlx4_en_rx_desc) +
2104                                          DS_SIZE * MLX4_EN_MAX_RX_FRAGS);
2105
2106         dst->tx_ring = kzalloc(sizeof(struct mlx4_en_tx_ring *) * MAX_TX_RINGS,
2107                                 GFP_KERNEL);
2108         if (!dst->tx_ring)
2109                 return -ENOMEM;
2110
2111         dst->tx_cq = kzalloc(sizeof(struct mlx4_en_cq *) * MAX_TX_RINGS,
2112                               GFP_KERNEL);
2113         if (!dst->tx_cq) {
2114                 kfree(dst->tx_ring);
2115                 return -ENOMEM;
2116         }
2117         return 0;
2118 }
2119
2120 static void mlx4_en_update_priv(struct mlx4_en_priv *dst,
2121                                 struct mlx4_en_priv *src)
2122 {
2123         memcpy(dst->rx_ring, src->rx_ring,
2124                sizeof(struct mlx4_en_rx_ring *) * src->rx_ring_num);
2125         memcpy(dst->rx_cq, src->rx_cq,
2126                sizeof(struct mlx4_en_cq *) * src->rx_ring_num);
2127         memcpy(&dst->hwtstamp_config, &src->hwtstamp_config,
2128                sizeof(dst->hwtstamp_config));
2129         dst->tx_ring_num = src->tx_ring_num;
2130         dst->rx_ring_num = src->rx_ring_num;
2131         dst->tx_ring = src->tx_ring;
2132         dst->tx_cq = src->tx_cq;
2133         memcpy(dst->prof, src->prof, sizeof(struct mlx4_en_port_profile));
2134 }
2135
2136 int mlx4_en_try_alloc_resources(struct mlx4_en_priv *priv,
2137                                 struct mlx4_en_priv *tmp,
2138                                 struct mlx4_en_port_profile *prof)
2139 {
2140         mlx4_en_copy_priv(tmp, priv, prof);
2141
2142         if (mlx4_en_alloc_resources(tmp)) {
2143                 en_warn(priv,
2144                         "%s: Resource allocation failed, using previous configuration\n",
2145                         __func__);
2146                 kfree(tmp->tx_ring);
2147                 kfree(tmp->tx_cq);
2148                 return -ENOMEM;
2149         }
2150         return 0;
2151 }
2152
2153 void mlx4_en_safe_replace_resources(struct mlx4_en_priv *priv,
2154                                     struct mlx4_en_priv *tmp)
2155 {
2156         mlx4_en_free_resources(priv);
2157         mlx4_en_update_priv(priv, tmp);
2158 }
2159
2160 void mlx4_en_destroy_netdev(struct net_device *dev)
2161 {
2162         struct mlx4_en_priv *priv = netdev_priv(dev);
2163         struct mlx4_en_dev *mdev = priv->mdev;
2164
2165         en_dbg(DRV, priv, "Destroying netdev on port:%d\n", priv->port);
2166
2167         /* Unregister device - this will close the port if it was up */
2168         if (priv->registered) {
2169                 devlink_port_type_clear(mlx4_get_devlink_port(mdev->dev,
2170                                                               priv->port));
2171                 unregister_netdev(dev);
2172         }
2173
2174         if (priv->allocated)
2175                 mlx4_free_hwq_res(mdev->dev, &priv->res, MLX4_EN_PAGE_SIZE);
2176
2177         cancel_delayed_work(&priv->stats_task);
2178         cancel_delayed_work(&priv->service_task);
2179         /* flush any pending task for this netdev */
2180         flush_workqueue(mdev->workqueue);
2181
2182         if (mdev->dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_TS)
2183                 mlx4_en_remove_timestamp(mdev);
2184
2185         /* Detach the netdev so tasks would not attempt to access it */
2186         mutex_lock(&mdev->state_lock);
2187         mdev->pndev[priv->port] = NULL;
2188         mdev->upper[priv->port] = NULL;
2189
2190 #ifdef CONFIG_RFS_ACCEL
2191         mlx4_en_cleanup_filters(priv);
2192 #endif
2193
2194         mlx4_en_free_resources(priv);
2195         mutex_unlock(&mdev->state_lock);
2196
2197         kfree(priv->tx_ring);
2198         kfree(priv->tx_cq);
2199
2200         free_netdev(dev);
2201 }
2202
2203 static int mlx4_en_change_mtu(struct net_device *dev, int new_mtu)
2204 {
2205         struct mlx4_en_priv *priv = netdev_priv(dev);
2206         struct mlx4_en_dev *mdev = priv->mdev;
2207         int err = 0;
2208
2209         en_dbg(DRV, priv, "Change MTU called - current:%d new:%d\n",
2210                  dev->mtu, new_mtu);
2211
2212         if ((new_mtu < MLX4_EN_MIN_MTU) || (new_mtu > priv->max_mtu)) {
2213                 en_err(priv, "Bad MTU size:%d.\n", new_mtu);
2214                 return -EPERM;
2215         }
2216         if (priv->xdp_ring_num && MLX4_EN_EFF_MTU(new_mtu) > FRAG_SZ0) {
2217                 en_err(priv, "MTU size:%d requires frags but XDP running\n",
2218                        new_mtu);
2219                 return -EOPNOTSUPP;
2220         }
2221         dev->mtu = new_mtu;
2222
2223         if (netif_running(dev)) {
2224                 mutex_lock(&mdev->state_lock);
2225                 if (!mdev->device_up) {
2226                         /* NIC is probably restarting - let restart task reset
2227                          * the port */
2228                         en_dbg(DRV, priv, "Change MTU called with card down!?\n");
2229                 } else {
2230                         mlx4_en_stop_port(dev, 1);
2231                         err = mlx4_en_start_port(dev);
2232                         if (err) {
2233                                 en_err(priv, "Failed restarting port:%d\n",
2234                                          priv->port);
2235                                 if (!test_and_set_bit(MLX4_EN_STATE_FLAG_RESTARTING,
2236                                                       &priv->state))
2237                                         queue_work(mdev->workqueue, &priv->restart_task);
2238                         }
2239                 }
2240                 mutex_unlock(&mdev->state_lock);
2241         }
2242         return 0;
2243 }
2244
2245 static int mlx4_en_hwtstamp_set(struct net_device *dev, struct ifreq *ifr)
2246 {
2247         struct mlx4_en_priv *priv = netdev_priv(dev);
2248         struct mlx4_en_dev *mdev = priv->mdev;
2249         struct hwtstamp_config config;
2250
2251         if (copy_from_user(&config, ifr->ifr_data, sizeof(config)))
2252                 return -EFAULT;
2253
2254         /* reserved for future extensions */
2255         if (config.flags)
2256                 return -EINVAL;
2257
2258         /* device doesn't support time stamping */
2259         if (!(mdev->dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_TS))
2260                 return -EINVAL;
2261
2262         /* TX HW timestamp */
2263         switch (config.tx_type) {
2264         case HWTSTAMP_TX_OFF:
2265         case HWTSTAMP_TX_ON:
2266                 break;
2267         default:
2268                 return -ERANGE;
2269         }
2270
2271         /* RX HW timestamp */
2272         switch (config.rx_filter) {
2273         case HWTSTAMP_FILTER_NONE:
2274                 break;
2275         case HWTSTAMP_FILTER_ALL:
2276         case HWTSTAMP_FILTER_SOME:
2277         case HWTSTAMP_FILTER_PTP_V1_L4_EVENT:
2278         case HWTSTAMP_FILTER_PTP_V1_L4_SYNC:
2279         case HWTSTAMP_FILTER_PTP_V1_L4_DELAY_REQ:
2280         case HWTSTAMP_FILTER_PTP_V2_L4_EVENT:
2281         case HWTSTAMP_FILTER_PTP_V2_L4_SYNC:
2282         case HWTSTAMP_FILTER_PTP_V2_L4_DELAY_REQ:
2283         case HWTSTAMP_FILTER_PTP_V2_L2_EVENT:
2284         case HWTSTAMP_FILTER_PTP_V2_L2_SYNC:
2285         case HWTSTAMP_FILTER_PTP_V2_L2_DELAY_REQ:
2286         case HWTSTAMP_FILTER_PTP_V2_EVENT:
2287         case HWTSTAMP_FILTER_PTP_V2_SYNC:
2288         case HWTSTAMP_FILTER_PTP_V2_DELAY_REQ:
2289                 config.rx_filter = HWTSTAMP_FILTER_ALL;
2290                 break;
2291         default:
2292                 return -ERANGE;
2293         }
2294
2295         if (mlx4_en_reset_config(dev, config, dev->features)) {
2296                 config.tx_type = HWTSTAMP_TX_OFF;
2297                 config.rx_filter = HWTSTAMP_FILTER_NONE;
2298         }
2299
2300         return copy_to_user(ifr->ifr_data, &config,
2301                             sizeof(config)) ? -EFAULT : 0;
2302 }
2303
2304 static int mlx4_en_hwtstamp_get(struct net_device *dev, struct ifreq *ifr)
2305 {
2306         struct mlx4_en_priv *priv = netdev_priv(dev);
2307
2308         return copy_to_user(ifr->ifr_data, &priv->hwtstamp_config,
2309                             sizeof(priv->hwtstamp_config)) ? -EFAULT : 0;
2310 }
2311
2312 static int mlx4_en_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
2313 {
2314         switch (cmd) {
2315         case SIOCSHWTSTAMP:
2316                 return mlx4_en_hwtstamp_set(dev, ifr);
2317         case SIOCGHWTSTAMP:
2318                 return mlx4_en_hwtstamp_get(dev, ifr);
2319         default:
2320                 return -EOPNOTSUPP;
2321         }
2322 }
2323
2324 static netdev_features_t mlx4_en_fix_features(struct net_device *netdev,
2325                                               netdev_features_t features)
2326 {
2327         struct mlx4_en_priv *en_priv = netdev_priv(netdev);
2328         struct mlx4_en_dev *mdev = en_priv->mdev;
2329
2330         /* Since there is no support for separate RX C-TAG/S-TAG vlan accel
2331          * enable/disable make sure S-TAG flag is always in same state as
2332          * C-TAG.
2333          */
2334         if (features & NETIF_F_HW_VLAN_CTAG_RX &&
2335             !(mdev->dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_SKIP_OUTER_VLAN))
2336                 features |= NETIF_F_HW_VLAN_STAG_RX;
2337         else
2338                 features &= ~NETIF_F_HW_VLAN_STAG_RX;
2339
2340         return features;
2341 }
2342
2343 static int mlx4_en_set_features(struct net_device *netdev,
2344                 netdev_features_t features)
2345 {
2346         struct mlx4_en_priv *priv = netdev_priv(netdev);
2347         bool reset = false;
2348         int ret = 0;
2349
2350         if (DEV_FEATURE_CHANGED(netdev, features, NETIF_F_RXFCS)) {
2351                 en_info(priv, "Turn %s RX-FCS\n",
2352                         (features & NETIF_F_RXFCS) ? "ON" : "OFF");
2353                 reset = true;
2354         }
2355
2356         if (DEV_FEATURE_CHANGED(netdev, features, NETIF_F_RXALL)) {
2357                 u8 ignore_fcs_value = (features & NETIF_F_RXALL) ? 1 : 0;
2358
2359                 en_info(priv, "Turn %s RX-ALL\n",
2360                         ignore_fcs_value ? "ON" : "OFF");
2361                 ret = mlx4_SET_PORT_fcs_check(priv->mdev->dev,
2362                                               priv->port, ignore_fcs_value);
2363                 if (ret)
2364                         return ret;
2365         }
2366
2367         if (DEV_FEATURE_CHANGED(netdev, features, NETIF_F_HW_VLAN_CTAG_RX)) {
2368                 en_info(priv, "Turn %s RX vlan strip offload\n",
2369                         (features & NETIF_F_HW_VLAN_CTAG_RX) ? "ON" : "OFF");
2370                 reset = true;
2371         }
2372
2373         if (DEV_FEATURE_CHANGED(netdev, features, NETIF_F_HW_VLAN_CTAG_TX))
2374                 en_info(priv, "Turn %s TX vlan strip offload\n",
2375                         (features & NETIF_F_HW_VLAN_CTAG_TX) ? "ON" : "OFF");
2376
2377         if (DEV_FEATURE_CHANGED(netdev, features, NETIF_F_HW_VLAN_STAG_TX))
2378                 en_info(priv, "Turn %s TX S-VLAN strip offload\n",
2379                         (features & NETIF_F_HW_VLAN_STAG_TX) ? "ON" : "OFF");
2380
2381         if (DEV_FEATURE_CHANGED(netdev, features, NETIF_F_LOOPBACK)) {
2382                 en_info(priv, "Turn %s loopback\n",
2383                         (features & NETIF_F_LOOPBACK) ? "ON" : "OFF");
2384                 mlx4_en_update_loopback_state(netdev, features);
2385         }
2386
2387         if (reset) {
2388                 ret = mlx4_en_reset_config(netdev, priv->hwtstamp_config,
2389                                            features);
2390                 if (ret)
2391                         return ret;
2392         }
2393
2394         return 0;
2395 }
2396
2397 static int mlx4_en_set_vf_mac(struct net_device *dev, int queue, u8 *mac)
2398 {
2399         struct mlx4_en_priv *en_priv = netdev_priv(dev);
2400         struct mlx4_en_dev *mdev = en_priv->mdev;
2401         u64 mac_u64 = mlx4_mac_to_u64(mac);
2402
2403         if (is_multicast_ether_addr(mac))
2404                 return -EINVAL;
2405
2406         return mlx4_set_vf_mac(mdev->dev, en_priv->port, queue, mac_u64);
2407 }
2408
2409 static int mlx4_en_set_vf_vlan(struct net_device *dev, int vf, u16 vlan, u8 qos,
2410                                __be16 vlan_proto)
2411 {
2412         struct mlx4_en_priv *en_priv = netdev_priv(dev);
2413         struct mlx4_en_dev *mdev = en_priv->mdev;
2414
2415         return mlx4_set_vf_vlan(mdev->dev, en_priv->port, vf, vlan, qos,
2416                                 vlan_proto);
2417 }
2418
2419 static int mlx4_en_set_vf_rate(struct net_device *dev, int vf, int min_tx_rate,
2420                                int max_tx_rate)
2421 {
2422         struct mlx4_en_priv *en_priv = netdev_priv(dev);
2423         struct mlx4_en_dev *mdev = en_priv->mdev;
2424
2425         return mlx4_set_vf_rate(mdev->dev, en_priv->port, vf, min_tx_rate,
2426                                 max_tx_rate);
2427 }
2428
2429 static int mlx4_en_set_vf_spoofchk(struct net_device *dev, int vf, bool setting)
2430 {
2431         struct mlx4_en_priv *en_priv = netdev_priv(dev);
2432         struct mlx4_en_dev *mdev = en_priv->mdev;
2433
2434         return mlx4_set_vf_spoofchk(mdev->dev, en_priv->port, vf, setting);
2435 }
2436
2437 static int mlx4_en_get_vf_config(struct net_device *dev, int vf, struct ifla_vf_info *ivf)
2438 {
2439         struct mlx4_en_priv *en_priv = netdev_priv(dev);
2440         struct mlx4_en_dev *mdev = en_priv->mdev;
2441
2442         return mlx4_get_vf_config(mdev->dev, en_priv->port, vf, ivf);
2443 }
2444
2445 static int mlx4_en_set_vf_link_state(struct net_device *dev, int vf, int link_state)
2446 {
2447         struct mlx4_en_priv *en_priv = netdev_priv(dev);
2448         struct mlx4_en_dev *mdev = en_priv->mdev;
2449
2450         return mlx4_set_vf_link_state(mdev->dev, en_priv->port, vf, link_state);
2451 }
2452
2453 static int mlx4_en_get_vf_stats(struct net_device *dev, int vf,
2454                                 struct ifla_vf_stats *vf_stats)
2455 {
2456         struct mlx4_en_priv *en_priv = netdev_priv(dev);
2457         struct mlx4_en_dev *mdev = en_priv->mdev;
2458
2459         return mlx4_get_vf_stats(mdev->dev, en_priv->port, vf, vf_stats);
2460 }
2461
2462 #define PORT_ID_BYTE_LEN 8
2463 static int mlx4_en_get_phys_port_id(struct net_device *dev,
2464                                     struct netdev_phys_item_id *ppid)
2465 {
2466         struct mlx4_en_priv *priv = netdev_priv(dev);
2467         struct mlx4_dev *mdev = priv->mdev->dev;
2468         int i;
2469         u64 phys_port_id = mdev->caps.phys_port_id[priv->port];
2470
2471         if (!phys_port_id)
2472                 return -EOPNOTSUPP;
2473
2474         ppid->id_len = sizeof(phys_port_id);
2475         for (i = PORT_ID_BYTE_LEN - 1; i >= 0; --i) {
2476                 ppid->id[i] =  phys_port_id & 0xff;
2477                 phys_port_id >>= 8;
2478         }
2479         return 0;
2480 }
2481
2482 static void mlx4_en_add_vxlan_offloads(struct work_struct *work)
2483 {
2484         int ret;
2485         struct mlx4_en_priv *priv = container_of(work, struct mlx4_en_priv,
2486                                                  vxlan_add_task);
2487
2488         ret = mlx4_config_vxlan_port(priv->mdev->dev, priv->vxlan_port);
2489         if (ret)
2490                 goto out;
2491
2492         ret = mlx4_SET_PORT_VXLAN(priv->mdev->dev, priv->port,
2493                                   VXLAN_STEER_BY_OUTER_MAC, 1);
2494 out:
2495         if (ret) {
2496                 en_err(priv, "failed setting L2 tunnel configuration ret %d\n", ret);
2497                 return;
2498         }
2499
2500         /* set offloads */
2501         priv->dev->hw_enc_features |= NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM |
2502                                       NETIF_F_RXCSUM |
2503                                       NETIF_F_TSO | NETIF_F_TSO6 |
2504                                       NETIF_F_GSO_UDP_TUNNEL |
2505                                       NETIF_F_GSO_UDP_TUNNEL_CSUM |
2506                                       NETIF_F_GSO_PARTIAL;
2507 }
2508
2509 static void mlx4_en_del_vxlan_offloads(struct work_struct *work)
2510 {
2511         int ret;
2512         struct mlx4_en_priv *priv = container_of(work, struct mlx4_en_priv,
2513                                                  vxlan_del_task);
2514         /* unset offloads */
2515         priv->dev->hw_enc_features &= ~(NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM |
2516                                         NETIF_F_RXCSUM |
2517                                         NETIF_F_TSO | NETIF_F_TSO6 |
2518                                         NETIF_F_GSO_UDP_TUNNEL |
2519                                         NETIF_F_GSO_UDP_TUNNEL_CSUM |
2520                                         NETIF_F_GSO_PARTIAL);
2521
2522         ret = mlx4_SET_PORT_VXLAN(priv->mdev->dev, priv->port,
2523                                   VXLAN_STEER_BY_OUTER_MAC, 0);
2524         if (ret)
2525                 en_err(priv, "failed setting L2 tunnel configuration ret %d\n", ret);
2526
2527         priv->vxlan_port = 0;
2528 }
2529
2530 static void mlx4_en_add_vxlan_port(struct  net_device *dev,
2531                                    struct udp_tunnel_info *ti)
2532 {
2533         struct mlx4_en_priv *priv = netdev_priv(dev);
2534         __be16 port = ti->port;
2535         __be16 current_port;
2536
2537         if (ti->type != UDP_TUNNEL_TYPE_VXLAN)
2538                 return;
2539
2540         if (ti->sa_family != AF_INET)
2541                 return;
2542
2543         if (priv->mdev->dev->caps.tunnel_offload_mode != MLX4_TUNNEL_OFFLOAD_MODE_VXLAN)
2544                 return;
2545
2546         current_port = priv->vxlan_port;
2547         if (current_port && current_port != port) {
2548                 en_warn(priv, "vxlan port %d configured, can't add port %d\n",
2549                         ntohs(current_port), ntohs(port));
2550                 return;
2551         }
2552
2553         priv->vxlan_port = port;
2554         queue_work(priv->mdev->workqueue, &priv->vxlan_add_task);
2555 }
2556
2557 static void mlx4_en_del_vxlan_port(struct  net_device *dev,
2558                                    struct udp_tunnel_info *ti)
2559 {
2560         struct mlx4_en_priv *priv = netdev_priv(dev);
2561         __be16 port = ti->port;
2562         __be16 current_port;
2563
2564         if (ti->type != UDP_TUNNEL_TYPE_VXLAN)
2565                 return;
2566
2567         if (ti->sa_family != AF_INET)
2568                 return;
2569
2570         if (priv->mdev->dev->caps.tunnel_offload_mode != MLX4_TUNNEL_OFFLOAD_MODE_VXLAN)
2571                 return;
2572
2573         current_port = priv->vxlan_port;
2574         if (current_port != port) {
2575                 en_dbg(DRV, priv, "vxlan port %d isn't configured, ignoring\n", ntohs(port));
2576                 return;
2577         }
2578
2579         queue_work(priv->mdev->workqueue, &priv->vxlan_del_task);
2580 }
2581
2582 static netdev_features_t mlx4_en_features_check(struct sk_buff *skb,
2583                                                 struct net_device *dev,
2584                                                 netdev_features_t features)
2585 {
2586         features = vlan_features_check(skb, features);
2587         features = vxlan_features_check(skb, features);
2588
2589         /* The ConnectX-3 doesn't support outer IPv6 checksums but it does
2590          * support inner IPv6 checksums and segmentation so  we need to
2591          * strip that feature if this is an IPv6 encapsulated frame.
2592          */
2593         if (skb->encapsulation &&
2594             (skb->ip_summed == CHECKSUM_PARTIAL)) {
2595                 struct mlx4_en_priv *priv = netdev_priv(dev);
2596
2597                 if (!priv->vxlan_port ||
2598                     (ip_hdr(skb)->version != 4) ||
2599                     (udp_hdr(skb)->dest != priv->vxlan_port))
2600                         features &= ~(NETIF_F_CSUM_MASK | NETIF_F_GSO_MASK);
2601         }
2602
2603         return features;
2604 }
2605
2606 static int mlx4_en_set_tx_maxrate(struct net_device *dev, int queue_index, u32 maxrate)
2607 {
2608         struct mlx4_en_priv *priv = netdev_priv(dev);
2609         struct mlx4_en_tx_ring *tx_ring = priv->tx_ring[queue_index];
2610         struct mlx4_update_qp_params params;
2611         int err;
2612
2613         if (!(priv->mdev->dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_QP_RATE_LIMIT))
2614                 return -EOPNOTSUPP;
2615
2616         /* rate provided to us in Mbs, check if it fits into 12 bits, if not use Gbs */
2617         if (maxrate >> 12) {
2618                 params.rate_unit = MLX4_QP_RATE_LIMIT_GBS;
2619                 params.rate_val  = maxrate / 1000;
2620         } else if (maxrate) {
2621                 params.rate_unit = MLX4_QP_RATE_LIMIT_MBS;
2622                 params.rate_val  = maxrate;
2623         } else { /* zero serves to revoke the QP rate-limitation */
2624                 params.rate_unit = 0;
2625                 params.rate_val  = 0;
2626         }
2627
2628         err = mlx4_update_qp(priv->mdev->dev, tx_ring->qpn, MLX4_UPDATE_QP_RATE_LIMIT,
2629                              &params);
2630         return err;
2631 }
2632
2633 static int mlx4_xdp_set(struct net_device *dev, struct bpf_prog *prog)
2634 {
2635         struct mlx4_en_priv *priv = netdev_priv(dev);
2636         struct mlx4_en_dev *mdev = priv->mdev;
2637         struct bpf_prog *old_prog;
2638         int xdp_ring_num;
2639         int port_up = 0;
2640         int err;
2641         int i;
2642
2643         xdp_ring_num = prog ? ALIGN(priv->rx_ring_num, MLX4_EN_NUM_UP) : 0;
2644
2645         /* No need to reconfigure buffers when simply swapping the
2646          * program for a new one.
2647          */
2648         if (priv->xdp_ring_num == xdp_ring_num) {
2649                 if (prog) {
2650                         prog = bpf_prog_add(prog, priv->rx_ring_num - 1);
2651                         if (IS_ERR(prog))
2652                                 return PTR_ERR(prog);
2653                 }
2654                 mutex_lock(&mdev->state_lock);
2655                 for (i = 0; i < priv->rx_ring_num; i++) {
2656                         old_prog = rcu_dereference_protected(
2657                                         priv->rx_ring[i]->xdp_prog,
2658                                         lockdep_is_held(&mdev->state_lock));
2659                         rcu_assign_pointer(priv->rx_ring[i]->xdp_prog, prog);
2660                         if (old_prog)
2661                                 bpf_prog_put(old_prog);
2662                 }
2663                 mutex_unlock(&mdev->state_lock);
2664                 return 0;
2665         }
2666
2667         if (priv->num_frags > 1) {
2668                 en_err(priv, "Cannot set XDP if MTU requires multiple frags\n");
2669                 return -EOPNOTSUPP;
2670         }
2671
2672         if (priv->tx_ring_num < xdp_ring_num + MLX4_EN_NUM_UP) {
2673                 en_err(priv,
2674                        "Minimum %d tx channels required to run XDP\n",
2675                        (xdp_ring_num + MLX4_EN_NUM_UP) / MLX4_EN_NUM_UP);
2676                 return -EINVAL;
2677         }
2678
2679         if (prog) {
2680                 prog = bpf_prog_add(prog, priv->rx_ring_num - 1);
2681                 if (IS_ERR(prog))
2682                         return PTR_ERR(prog);
2683         }
2684
2685         mutex_lock(&mdev->state_lock);
2686         if (priv->port_up) {
2687                 port_up = 1;
2688                 mlx4_en_stop_port(dev, 1);
2689         }
2690
2691         priv->xdp_ring_num = xdp_ring_num;
2692         netif_set_real_num_tx_queues(dev, priv->tx_ring_num -
2693                                                         priv->xdp_ring_num);
2694
2695         for (i = 0; i < priv->rx_ring_num; i++) {
2696                 old_prog = rcu_dereference_protected(
2697                                         priv->rx_ring[i]->xdp_prog,
2698                                         lockdep_is_held(&mdev->state_lock));
2699                 rcu_assign_pointer(priv->rx_ring[i]->xdp_prog, prog);
2700                 if (old_prog)
2701                         bpf_prog_put(old_prog);
2702         }
2703
2704         if (port_up) {
2705                 err = mlx4_en_start_port(dev);
2706                 if (err) {
2707                         en_err(priv, "Failed starting port %d for XDP change\n",
2708                                priv->port);
2709                         if (!test_and_set_bit(MLX4_EN_STATE_FLAG_RESTARTING, &priv->state))
2710                                 queue_work(mdev->workqueue, &priv->restart_task);
2711                 }
2712         }
2713
2714         mutex_unlock(&mdev->state_lock);
2715         return 0;
2716 }
2717
2718 static bool mlx4_xdp_attached(struct net_device *dev)
2719 {
2720         struct mlx4_en_priv *priv = netdev_priv(dev);
2721
2722         return !!priv->xdp_ring_num;
2723 }
2724
2725 static int mlx4_xdp(struct net_device *dev, struct netdev_xdp *xdp)
2726 {
2727         switch (xdp->command) {
2728         case XDP_SETUP_PROG:
2729                 return mlx4_xdp_set(dev, xdp->prog);
2730         case XDP_QUERY_PROG:
2731                 xdp->prog_attached = mlx4_xdp_attached(dev);
2732                 return 0;
2733         default:
2734                 return -EINVAL;
2735         }
2736 }
2737
2738 static const struct net_device_ops mlx4_netdev_ops = {
2739         .ndo_open               = mlx4_en_open,
2740         .ndo_stop               = mlx4_en_close,
2741         .ndo_start_xmit         = mlx4_en_xmit,
2742         .ndo_select_queue       = mlx4_en_select_queue,
2743         .ndo_get_stats64        = mlx4_en_get_stats64,
2744         .ndo_set_rx_mode        = mlx4_en_set_rx_mode,
2745         .ndo_set_mac_address    = mlx4_en_set_mac,
2746         .ndo_validate_addr      = eth_validate_addr,
2747         .ndo_change_mtu         = mlx4_en_change_mtu,
2748         .ndo_do_ioctl           = mlx4_en_ioctl,
2749         .ndo_tx_timeout         = mlx4_en_tx_timeout,
2750         .ndo_vlan_rx_add_vid    = mlx4_en_vlan_rx_add_vid,
2751         .ndo_vlan_rx_kill_vid   = mlx4_en_vlan_rx_kill_vid,
2752 #ifdef CONFIG_NET_POLL_CONTROLLER
2753         .ndo_poll_controller    = mlx4_en_netpoll,
2754 #endif
2755         .ndo_set_features       = mlx4_en_set_features,
2756         .ndo_fix_features       = mlx4_en_fix_features,
2757         .ndo_setup_tc           = __mlx4_en_setup_tc,
2758 #ifdef CONFIG_RFS_ACCEL
2759         .ndo_rx_flow_steer      = mlx4_en_filter_rfs,
2760 #endif
2761         .ndo_get_phys_port_id   = mlx4_en_get_phys_port_id,
2762         .ndo_udp_tunnel_add     = mlx4_en_add_vxlan_port,
2763         .ndo_udp_tunnel_del     = mlx4_en_del_vxlan_port,
2764         .ndo_features_check     = mlx4_en_features_check,
2765         .ndo_set_tx_maxrate     = mlx4_en_set_tx_maxrate,
2766         .ndo_xdp                = mlx4_xdp,
2767 };
2768
2769 static const struct net_device_ops mlx4_netdev_ops_master = {
2770         .ndo_open               = mlx4_en_open,
2771         .ndo_stop               = mlx4_en_close,
2772         .ndo_start_xmit         = mlx4_en_xmit,
2773         .ndo_select_queue       = mlx4_en_select_queue,
2774         .ndo_get_stats64        = mlx4_en_get_stats64,
2775         .ndo_set_rx_mode        = mlx4_en_set_rx_mode,
2776         .ndo_set_mac_address    = mlx4_en_set_mac,
2777         .ndo_validate_addr      = eth_validate_addr,
2778         .ndo_change_mtu         = mlx4_en_change_mtu,
2779         .ndo_tx_timeout         = mlx4_en_tx_timeout,
2780         .ndo_vlan_rx_add_vid    = mlx4_en_vlan_rx_add_vid,
2781         .ndo_vlan_rx_kill_vid   = mlx4_en_vlan_rx_kill_vid,
2782         .ndo_set_vf_mac         = mlx4_en_set_vf_mac,
2783         .ndo_set_vf_vlan        = mlx4_en_set_vf_vlan,
2784         .ndo_set_vf_rate        = mlx4_en_set_vf_rate,
2785         .ndo_set_vf_spoofchk    = mlx4_en_set_vf_spoofchk,
2786         .ndo_set_vf_link_state  = mlx4_en_set_vf_link_state,
2787         .ndo_get_vf_stats       = mlx4_en_get_vf_stats,
2788         .ndo_get_vf_config      = mlx4_en_get_vf_config,
2789 #ifdef CONFIG_NET_POLL_CONTROLLER
2790         .ndo_poll_controller    = mlx4_en_netpoll,
2791 #endif
2792         .ndo_set_features       = mlx4_en_set_features,
2793         .ndo_fix_features       = mlx4_en_fix_features,
2794         .ndo_setup_tc           = __mlx4_en_setup_tc,
2795 #ifdef CONFIG_RFS_ACCEL
2796         .ndo_rx_flow_steer      = mlx4_en_filter_rfs,
2797 #endif
2798         .ndo_get_phys_port_id   = mlx4_en_get_phys_port_id,
2799         .ndo_udp_tunnel_add     = mlx4_en_add_vxlan_port,
2800         .ndo_udp_tunnel_del     = mlx4_en_del_vxlan_port,
2801         .ndo_features_check     = mlx4_en_features_check,
2802         .ndo_set_tx_maxrate     = mlx4_en_set_tx_maxrate,
2803         .ndo_xdp                = mlx4_xdp,
2804 };
2805
2806 struct mlx4_en_bond {
2807         struct work_struct work;
2808         struct mlx4_en_priv *priv;
2809         int is_bonded;
2810         struct mlx4_port_map port_map;
2811 };
2812
2813 static void mlx4_en_bond_work(struct work_struct *work)
2814 {
2815         struct mlx4_en_bond *bond = container_of(work,
2816                                                      struct mlx4_en_bond,
2817                                                      work);
2818         int err = 0;
2819         struct mlx4_dev *dev = bond->priv->mdev->dev;
2820
2821         if (bond->is_bonded) {
2822                 if (!mlx4_is_bonded(dev)) {
2823                         err = mlx4_bond(dev);
2824                         if (err)
2825                                 en_err(bond->priv, "Fail to bond device\n");
2826                 }
2827                 if (!err) {
2828                         err = mlx4_port_map_set(dev, &bond->port_map);
2829                         if (err)
2830                                 en_err(bond->priv, "Fail to set port map [%d][%d]: %d\n",
2831                                        bond->port_map.port1,
2832                                        bond->port_map.port2,
2833                                        err);
2834                 }
2835         } else if (mlx4_is_bonded(dev)) {
2836                 err = mlx4_unbond(dev);
2837                 if (err)
2838                         en_err(bond->priv, "Fail to unbond device\n");
2839         }
2840         dev_put(bond->priv->dev);
2841         kfree(bond);
2842 }
2843
2844 static int mlx4_en_queue_bond_work(struct mlx4_en_priv *priv, int is_bonded,
2845                                    u8 v2p_p1, u8 v2p_p2)
2846 {
2847         struct mlx4_en_bond *bond = NULL;
2848
2849         bond = kzalloc(sizeof(*bond), GFP_ATOMIC);
2850         if (!bond)
2851                 return -ENOMEM;
2852
2853         INIT_WORK(&bond->work, mlx4_en_bond_work);
2854         bond->priv = priv;
2855         bond->is_bonded = is_bonded;
2856         bond->port_map.port1 = v2p_p1;
2857         bond->port_map.port2 = v2p_p2;
2858         dev_hold(priv->dev);
2859         queue_work(priv->mdev->workqueue, &bond->work);
2860         return 0;
2861 }
2862
2863 int mlx4_en_netdev_event(struct notifier_block *this,
2864                          unsigned long event, void *ptr)
2865 {
2866         struct net_device *ndev = netdev_notifier_info_to_dev(ptr);
2867         u8 port = 0;
2868         struct mlx4_en_dev *mdev;
2869         struct mlx4_dev *dev;
2870         int i, num_eth_ports = 0;
2871         bool do_bond = true;
2872         struct mlx4_en_priv *priv;
2873         u8 v2p_port1 = 0;
2874         u8 v2p_port2 = 0;
2875
2876         if (!net_eq(dev_net(ndev), &init_net))
2877                 return NOTIFY_DONE;
2878
2879         mdev = container_of(this, struct mlx4_en_dev, nb);
2880         dev = mdev->dev;
2881
2882         /* Go into this mode only when two network devices set on two ports
2883          * of the same mlx4 device are slaves of the same bonding master
2884          */
2885         mlx4_foreach_port(i, dev, MLX4_PORT_TYPE_ETH) {
2886                 ++num_eth_ports;
2887                 if (!port && (mdev->pndev[i] == ndev))
2888                         port = i;
2889                 mdev->upper[i] = mdev->pndev[i] ?
2890                         netdev_master_upper_dev_get(mdev->pndev[i]) : NULL;
2891                 /* condition not met: network device is a slave */
2892                 if (!mdev->upper[i])
2893                         do_bond = false;
2894                 if (num_eth_ports < 2)
2895                         continue;
2896                 /* condition not met: same master */
2897                 if (mdev->upper[i] != mdev->upper[i-1])
2898                         do_bond = false;
2899         }
2900         /* condition not met: 2 salves */
2901         do_bond = (num_eth_ports ==  2) ? do_bond : false;
2902
2903         /* handle only events that come with enough info */
2904         if ((do_bond && (event != NETDEV_BONDING_INFO)) || !port)
2905                 return NOTIFY_DONE;
2906
2907         priv = netdev_priv(ndev);
2908         if (do_bond) {
2909                 struct netdev_notifier_bonding_info *notifier_info = ptr;
2910                 struct netdev_bonding_info *bonding_info =
2911                         &notifier_info->bonding_info;
2912
2913                 /* required mode 1, 2 or 4 */
2914                 if ((bonding_info->master.bond_mode != BOND_MODE_ACTIVEBACKUP) &&
2915                     (bonding_info->master.bond_mode != BOND_MODE_XOR) &&
2916                     (bonding_info->master.bond_mode != BOND_MODE_8023AD))
2917                         do_bond = false;
2918
2919                 /* require exactly 2 slaves */
2920                 if (bonding_info->master.num_slaves != 2)
2921                         do_bond = false;
2922
2923                 /* calc v2p */
2924                 if (do_bond) {
2925                         if (bonding_info->master.bond_mode ==
2926                             BOND_MODE_ACTIVEBACKUP) {
2927                                 /* in active-backup mode virtual ports are
2928                                  * mapped to the physical port of the active
2929                                  * slave */
2930                                 if (bonding_info->slave.state ==
2931                                     BOND_STATE_BACKUP) {
2932                                         if (port == 1) {
2933                                                 v2p_port1 = 2;
2934                                                 v2p_port2 = 2;
2935                                         } else {
2936                                                 v2p_port1 = 1;
2937                                                 v2p_port2 = 1;
2938                                         }
2939                                 } else { /* BOND_STATE_ACTIVE */
2940                                         if (port == 1) {
2941                                                 v2p_port1 = 1;
2942                                                 v2p_port2 = 1;
2943                                         } else {
2944                                                 v2p_port1 = 2;
2945                                                 v2p_port2 = 2;
2946                                         }
2947                                 }
2948                         } else { /* Active-Active */
2949                                 /* in active-active mode a virtual port is
2950                                  * mapped to the native physical port if and only
2951                                  * if the physical port is up */
2952                                 __s8 link = bonding_info->slave.link;
2953
2954                                 if (port == 1)
2955                                         v2p_port2 = 2;
2956                                 else
2957                                         v2p_port1 = 1;
2958                                 if ((link == BOND_LINK_UP) ||
2959                                     (link == BOND_LINK_FAIL)) {
2960                                         if (port == 1)
2961                                                 v2p_port1 = 1;
2962                                         else
2963                                                 v2p_port2 = 2;
2964                                 } else { /* BOND_LINK_DOWN || BOND_LINK_BACK */
2965                                         if (port == 1)
2966                                                 v2p_port1 = 2;
2967                                         else
2968                                                 v2p_port2 = 1;
2969                                 }
2970                         }
2971                 }
2972         }
2973
2974         mlx4_en_queue_bond_work(priv, do_bond,
2975                                 v2p_port1, v2p_port2);
2976
2977         return NOTIFY_DONE;
2978 }
2979
2980 void mlx4_en_update_pfc_stats_bitmap(struct mlx4_dev *dev,
2981                                      struct mlx4_en_stats_bitmap *stats_bitmap,
2982                                      u8 rx_ppp, u8 rx_pause,
2983                                      u8 tx_ppp, u8 tx_pause)
2984 {
2985         int last_i = NUM_MAIN_STATS + NUM_PORT_STATS + NUM_PF_STATS;
2986
2987         if (!mlx4_is_slave(dev) &&
2988             (dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_FLOWSTATS_EN)) {
2989                 mutex_lock(&stats_bitmap->mutex);
2990                 bitmap_clear(stats_bitmap->bitmap, last_i, NUM_FLOW_STATS);
2991
2992                 if (rx_ppp)
2993                         bitmap_set(stats_bitmap->bitmap, last_i,
2994                                    NUM_FLOW_PRIORITY_STATS_RX);
2995                 last_i += NUM_FLOW_PRIORITY_STATS_RX;
2996
2997                 if (rx_pause && !(rx_ppp))
2998                         bitmap_set(stats_bitmap->bitmap, last_i,
2999                                    NUM_FLOW_STATS_RX);
3000                 last_i += NUM_FLOW_STATS_RX;
3001
3002                 if (tx_ppp)
3003                         bitmap_set(stats_bitmap->bitmap, last_i,
3004                                    NUM_FLOW_PRIORITY_STATS_TX);
3005                 last_i += NUM_FLOW_PRIORITY_STATS_TX;
3006
3007                 if (tx_pause && !(tx_ppp))
3008                         bitmap_set(stats_bitmap->bitmap, last_i,
3009                                    NUM_FLOW_STATS_TX);
3010                 last_i += NUM_FLOW_STATS_TX;
3011
3012                 mutex_unlock(&stats_bitmap->mutex);
3013         }
3014 }
3015
3016 void mlx4_en_set_stats_bitmap(struct mlx4_dev *dev,
3017                               struct mlx4_en_stats_bitmap *stats_bitmap,
3018                               u8 rx_ppp, u8 rx_pause,
3019                               u8 tx_ppp, u8 tx_pause)
3020 {
3021         int last_i = 0;
3022
3023         mutex_init(&stats_bitmap->mutex);
3024         bitmap_zero(stats_bitmap->bitmap, NUM_ALL_STATS);
3025
3026         if (mlx4_is_slave(dev)) {
3027                 bitmap_set(stats_bitmap->bitmap, last_i +
3028                                          MLX4_FIND_NETDEV_STAT(rx_packets), 1);
3029                 bitmap_set(stats_bitmap->bitmap, last_i +
3030                                          MLX4_FIND_NETDEV_STAT(tx_packets), 1);
3031                 bitmap_set(stats_bitmap->bitmap, last_i +
3032                                          MLX4_FIND_NETDEV_STAT(rx_bytes), 1);
3033                 bitmap_set(stats_bitmap->bitmap, last_i +
3034                                          MLX4_FIND_NETDEV_STAT(tx_bytes), 1);
3035                 bitmap_set(stats_bitmap->bitmap, last_i +
3036                                          MLX4_FIND_NETDEV_STAT(rx_dropped), 1);
3037                 bitmap_set(stats_bitmap->bitmap, last_i +
3038                                          MLX4_FIND_NETDEV_STAT(tx_dropped), 1);
3039         } else {
3040                 bitmap_set(stats_bitmap->bitmap, last_i, NUM_MAIN_STATS);
3041         }
3042         last_i += NUM_MAIN_STATS;
3043
3044         bitmap_set(stats_bitmap->bitmap, last_i, NUM_PORT_STATS);
3045         last_i += NUM_PORT_STATS;
3046
3047         if (mlx4_is_master(dev))
3048                 bitmap_set(stats_bitmap->bitmap, last_i,
3049                            NUM_PF_STATS);
3050         last_i += NUM_PF_STATS;
3051
3052         mlx4_en_update_pfc_stats_bitmap(dev, stats_bitmap,
3053                                         rx_ppp, rx_pause,
3054                                         tx_ppp, tx_pause);
3055         last_i += NUM_FLOW_STATS;
3056
3057         if (!mlx4_is_slave(dev))
3058                 bitmap_set(stats_bitmap->bitmap, last_i, NUM_PKT_STATS);
3059 }
3060
3061 int mlx4_en_init_netdev(struct mlx4_en_dev *mdev, int port,
3062                         struct mlx4_en_port_profile *prof)
3063 {
3064         struct net_device *dev;
3065         struct mlx4_en_priv *priv;
3066         int i;
3067         int err;
3068
3069         dev = alloc_etherdev_mqs(sizeof(struct mlx4_en_priv),
3070                                  MAX_TX_RINGS, MAX_RX_RINGS);
3071         if (dev == NULL)
3072                 return -ENOMEM;
3073
3074         netif_set_real_num_tx_queues(dev, prof->tx_ring_num);
3075         netif_set_real_num_rx_queues(dev, prof->rx_ring_num);
3076
3077         SET_NETDEV_DEV(dev, &mdev->dev->persist->pdev->dev);
3078         dev->dev_port = port - 1;
3079
3080         /*
3081          * Initialize driver private data
3082          */
3083
3084         priv = netdev_priv(dev);
3085         memset(priv, 0, sizeof(struct mlx4_en_priv));
3086         priv->counter_index = MLX4_SINK_COUNTER_INDEX(mdev->dev);
3087         spin_lock_init(&priv->stats_lock);
3088         INIT_WORK(&priv->rx_mode_task, mlx4_en_do_set_rx_mode);
3089         INIT_WORK(&priv->restart_task, mlx4_en_restart);
3090         INIT_WORK(&priv->linkstate_task, mlx4_en_linkstate);
3091         INIT_DELAYED_WORK(&priv->stats_task, mlx4_en_do_get_stats);
3092         INIT_DELAYED_WORK(&priv->service_task, mlx4_en_service_task);
3093         INIT_WORK(&priv->vxlan_add_task, mlx4_en_add_vxlan_offloads);
3094         INIT_WORK(&priv->vxlan_del_task, mlx4_en_del_vxlan_offloads);
3095 #ifdef CONFIG_RFS_ACCEL
3096         INIT_LIST_HEAD(&priv->filters);
3097         spin_lock_init(&priv->filters_lock);
3098 #endif
3099
3100         priv->dev = dev;
3101         priv->mdev = mdev;
3102         priv->ddev = &mdev->pdev->dev;
3103         priv->prof = prof;
3104         priv->port = port;
3105         priv->port_up = false;
3106         priv->flags = prof->flags;
3107         priv->pflags = MLX4_EN_PRIV_FLAGS_BLUEFLAME;
3108         priv->ctrl_flags = cpu_to_be32(MLX4_WQE_CTRL_CQ_UPDATE |
3109                         MLX4_WQE_CTRL_SOLICITED);
3110         priv->num_tx_rings_p_up = mdev->profile.num_tx_rings_p_up;
3111         priv->tx_ring_num = prof->tx_ring_num;
3112         priv->tx_work_limit = MLX4_EN_DEFAULT_TX_WORK;
3113         netdev_rss_key_fill(priv->rss_key, sizeof(priv->rss_key));
3114
3115         priv->tx_ring = kzalloc(sizeof(struct mlx4_en_tx_ring *) * MAX_TX_RINGS,
3116                                 GFP_KERNEL);
3117         if (!priv->tx_ring) {
3118                 err = -ENOMEM;
3119                 goto out;
3120         }
3121         priv->tx_cq = kzalloc(sizeof(struct mlx4_en_cq *) * MAX_TX_RINGS,
3122                               GFP_KERNEL);
3123         if (!priv->tx_cq) {
3124                 err = -ENOMEM;
3125                 goto out;
3126         }
3127         priv->rx_ring_num = prof->rx_ring_num;
3128         priv->cqe_factor = (mdev->dev->caps.cqe_size == 64) ? 1 : 0;
3129         priv->cqe_size = mdev->dev->caps.cqe_size;
3130         priv->mac_index = -1;
3131         priv->msg_enable = MLX4_EN_MSG_LEVEL;
3132 #ifdef CONFIG_MLX4_EN_DCB
3133         if (!mlx4_is_slave(priv->mdev->dev)) {
3134                 u8 prio;
3135
3136                 for (prio = 0; prio < IEEE_8021QAZ_MAX_TCS; ++prio) {
3137                         priv->ets.prio_tc[prio] = prio;
3138                         priv->ets.tc_tsa[prio]  = IEEE_8021QAZ_TSA_VENDOR;
3139                 }
3140
3141                 priv->dcbx_cap = DCB_CAP_DCBX_VER_CEE | DCB_CAP_DCBX_HOST |
3142                         DCB_CAP_DCBX_VER_IEEE;
3143                 priv->flags |= MLX4_EN_DCB_ENABLED;
3144                 priv->cee_config.pfc_state = false;
3145
3146                 for (i = 0; i < MLX4_EN_NUM_UP; i++)
3147                         priv->cee_config.dcb_pfc[i] = pfc_disabled;
3148
3149                 if (mdev->dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_ETS_CFG) {
3150                         dev->dcbnl_ops = &mlx4_en_dcbnl_ops;
3151                 } else {
3152                         en_info(priv, "enabling only PFC DCB ops\n");
3153                         dev->dcbnl_ops = &mlx4_en_dcbnl_pfc_ops;
3154                 }
3155         }
3156 #endif
3157
3158         for (i = 0; i < MLX4_EN_MAC_HASH_SIZE; ++i)
3159                 INIT_HLIST_HEAD(&priv->mac_hash[i]);
3160
3161         /* Query for default mac and max mtu */
3162         priv->max_mtu = mdev->dev->caps.eth_mtu_cap[priv->port];
3163
3164         if (mdev->dev->caps.rx_checksum_flags_port[priv->port] &
3165             MLX4_RX_CSUM_MODE_VAL_NON_TCP_UDP)
3166                 priv->flags |= MLX4_EN_FLAG_RX_CSUM_NON_TCP_UDP;
3167
3168         /* Set default MAC */
3169         dev->addr_len = ETH_ALEN;
3170         mlx4_en_u64_to_mac(dev->dev_addr, mdev->dev->caps.def_mac[priv->port]);
3171         if (!is_valid_ether_addr(dev->dev_addr)) {
3172                 en_err(priv, "Port: %d, invalid mac burned: %pM, quiting\n",
3173                        priv->port, dev->dev_addr);
3174                 err = -EINVAL;
3175                 goto out;
3176         } else if (mlx4_is_slave(priv->mdev->dev) &&
3177                    (priv->mdev->dev->port_random_macs & 1 << priv->port)) {
3178                 /* Random MAC was assigned in mlx4_slave_cap
3179                  * in mlx4_core module
3180                  */
3181                 dev->addr_assign_type |= NET_ADDR_RANDOM;
3182                 en_warn(priv, "Assigned random MAC address %pM\n", dev->dev_addr);
3183         }
3184
3185         memcpy(priv->current_mac, dev->dev_addr, sizeof(priv->current_mac));
3186
3187         priv->stride = roundup_pow_of_two(sizeof(struct mlx4_en_rx_desc) +
3188                                           DS_SIZE * MLX4_EN_MAX_RX_FRAGS);
3189         err = mlx4_en_alloc_resources(priv);
3190         if (err)
3191                 goto out;
3192
3193         /* Initialize time stamping config */
3194         priv->hwtstamp_config.flags = 0;
3195         priv->hwtstamp_config.tx_type = HWTSTAMP_TX_OFF;
3196         priv->hwtstamp_config.rx_filter = HWTSTAMP_FILTER_NONE;
3197
3198         /* Allocate page for receive rings */
3199         err = mlx4_alloc_hwq_res(mdev->dev, &priv->res,
3200                                 MLX4_EN_PAGE_SIZE);
3201         if (err) {
3202                 en_err(priv, "Failed to allocate page for rx qps\n");
3203                 goto out;
3204         }
3205         priv->allocated = 1;
3206
3207         /*
3208          * Initialize netdev entry points
3209          */
3210         if (mlx4_is_master(priv->mdev->dev))
3211                 dev->netdev_ops = &mlx4_netdev_ops_master;
3212         else
3213                 dev->netdev_ops = &mlx4_netdev_ops;
3214         dev->watchdog_timeo = MLX4_EN_WATCHDOG_TIMEOUT;
3215         netif_set_real_num_tx_queues(dev, priv->tx_ring_num);
3216         netif_set_real_num_rx_queues(dev, priv->rx_ring_num);
3217
3218         dev->ethtool_ops = &mlx4_en_ethtool_ops;
3219
3220         /*
3221          * Set driver features
3222          */
3223         dev->hw_features = NETIF_F_SG | NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM;
3224         if (mdev->LSO_support)
3225                 dev->hw_features |= NETIF_F_TSO | NETIF_F_TSO6;
3226
3227         dev->vlan_features = dev->hw_features;
3228
3229         dev->hw_features |= NETIF_F_RXCSUM | NETIF_F_RXHASH;
3230         dev->features = dev->hw_features | NETIF_F_HIGHDMA |
3231                         NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_CTAG_RX |
3232                         NETIF_F_HW_VLAN_CTAG_FILTER;
3233         dev->hw_features |= NETIF_F_LOOPBACK |
3234                         NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_CTAG_RX;
3235
3236         if (!(mdev->dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_SKIP_OUTER_VLAN)) {
3237                 dev->features |= NETIF_F_HW_VLAN_STAG_RX |
3238                         NETIF_F_HW_VLAN_STAG_FILTER;
3239                 dev->hw_features |= NETIF_F_HW_VLAN_STAG_RX;
3240         }
3241
3242         if (mlx4_is_slave(mdev->dev)) {
3243                 bool vlan_offload_disabled;
3244                 int phv;
3245
3246                 err = get_phv_bit(mdev->dev, port, &phv);
3247                 if (!err && phv) {
3248                         dev->hw_features |= NETIF_F_HW_VLAN_STAG_TX;
3249                         priv->pflags |= MLX4_EN_PRIV_FLAGS_PHV;
3250                 }
3251                 err = mlx4_get_is_vlan_offload_disabled(mdev->dev, port,
3252                                                         &vlan_offload_disabled);
3253                 if (!err && vlan_offload_disabled) {
3254                         dev->hw_features &= ~(NETIF_F_HW_VLAN_CTAG_TX |
3255                                               NETIF_F_HW_VLAN_CTAG_RX |
3256                                               NETIF_F_HW_VLAN_STAG_TX |
3257                                               NETIF_F_HW_VLAN_STAG_RX);
3258                         dev->features &= ~(NETIF_F_HW_VLAN_CTAG_TX |
3259                                            NETIF_F_HW_VLAN_CTAG_RX |
3260                                            NETIF_F_HW_VLAN_STAG_TX |
3261                                            NETIF_F_HW_VLAN_STAG_RX);
3262                 }
3263         } else {
3264                 if (mdev->dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_PHV_EN &&
3265                     !(mdev->dev->caps.flags2 &
3266                       MLX4_DEV_CAP_FLAG2_SKIP_OUTER_VLAN))
3267                         dev->hw_features |= NETIF_F_HW_VLAN_STAG_TX;
3268         }
3269
3270         if (mdev->dev->caps.flags & MLX4_DEV_CAP_FLAG_FCS_KEEP)
3271                 dev->hw_features |= NETIF_F_RXFCS;
3272
3273         if (mdev->dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_IGNORE_FCS)
3274                 dev->hw_features |= NETIF_F_RXALL;
3275
3276         if (mdev->dev->caps.steering_mode ==
3277             MLX4_STEERING_MODE_DEVICE_MANAGED &&
3278             mdev->dev->caps.dmfs_high_steer_mode != MLX4_STEERING_DMFS_A0_STATIC)
3279                 dev->hw_features |= NETIF_F_NTUPLE;
3280
3281         if (mdev->dev->caps.steering_mode != MLX4_STEERING_MODE_A0)
3282                 dev->priv_flags |= IFF_UNICAST_FLT;
3283
3284         /* Setting a default hash function value */
3285         if (mdev->dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_RSS_TOP) {
3286                 priv->rss_hash_fn = ETH_RSS_HASH_TOP;
3287         } else if (mdev->dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_RSS_XOR) {
3288                 priv->rss_hash_fn = ETH_RSS_HASH_XOR;
3289         } else {
3290                 en_warn(priv,
3291                         "No RSS hash capabilities exposed, using Toeplitz\n");
3292                 priv->rss_hash_fn = ETH_RSS_HASH_TOP;
3293         }
3294
3295         if (mdev->dev->caps.tunnel_offload_mode == MLX4_TUNNEL_OFFLOAD_MODE_VXLAN) {
3296                 dev->hw_features |= NETIF_F_GSO_UDP_TUNNEL |
3297                                     NETIF_F_GSO_UDP_TUNNEL_CSUM |
3298                                     NETIF_F_GSO_PARTIAL;
3299                 dev->features    |= NETIF_F_GSO_UDP_TUNNEL |
3300                                     NETIF_F_GSO_UDP_TUNNEL_CSUM |
3301                                     NETIF_F_GSO_PARTIAL;
3302                 dev->gso_partial_features = NETIF_F_GSO_UDP_TUNNEL_CSUM;
3303         }
3304
3305         mdev->pndev[port] = dev;
3306         mdev->upper[port] = NULL;
3307
3308         netif_carrier_off(dev);
3309         mlx4_en_set_default_moderation(priv);
3310
3311         en_warn(priv, "Using %d TX rings\n", prof->tx_ring_num);
3312         en_warn(priv, "Using %d RX rings\n", prof->rx_ring_num);
3313
3314         mlx4_en_update_loopback_state(priv->dev, priv->dev->features);
3315
3316         /* Configure port */
3317         mlx4_en_calc_rx_buf(dev);
3318         err = mlx4_SET_PORT_general(mdev->dev, priv->port,
3319                                     priv->rx_skb_size + ETH_FCS_LEN,
3320                                     prof->tx_pause, prof->tx_ppp,
3321                                     prof->rx_pause, prof->rx_ppp);
3322         if (err) {
3323                 en_err(priv, "Failed setting port general configurations for port %d, with error %d\n",
3324                        priv->port, err);
3325                 goto out;
3326         }
3327
3328         if (mdev->dev->caps.tunnel_offload_mode == MLX4_TUNNEL_OFFLOAD_MODE_VXLAN) {
3329                 err = mlx4_SET_PORT_VXLAN(mdev->dev, priv->port, VXLAN_STEER_BY_OUTER_MAC, 1);
3330                 if (err) {
3331                         en_err(priv, "Failed setting port L2 tunnel configuration, err %d\n",
3332                                err);
3333                         goto out;
3334                 }
3335         }
3336
3337         /* Init port */
3338         en_warn(priv, "Initializing port\n");
3339         err = mlx4_INIT_PORT(mdev->dev, priv->port);
3340         if (err) {
3341                 en_err(priv, "Failed Initializing port\n");
3342                 goto out;
3343         }
3344         queue_delayed_work(mdev->workqueue, &priv->stats_task, STATS_DELAY);
3345
3346         /* Initialize time stamp mechanism */
3347         if (mdev->dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_TS)
3348                 mlx4_en_init_timestamp(mdev);
3349
3350         queue_delayed_work(mdev->workqueue, &priv->service_task,
3351                            SERVICE_TASK_DELAY);
3352
3353         mlx4_en_set_stats_bitmap(mdev->dev, &priv->stats_bitmap,
3354                                  mdev->profile.prof[priv->port].rx_ppp,
3355                                  mdev->profile.prof[priv->port].rx_pause,
3356                                  mdev->profile.prof[priv->port].tx_ppp,
3357                                  mdev->profile.prof[priv->port].tx_pause);
3358
3359         err = register_netdev(dev);
3360         if (err) {
3361                 en_err(priv, "Netdev registration failed for port %d\n", port);
3362                 goto out;
3363         }
3364
3365         priv->registered = 1;
3366         devlink_port_type_eth_set(mlx4_get_devlink_port(mdev->dev, priv->port),
3367                                   dev);
3368
3369         return 0;
3370
3371 out:
3372         mlx4_en_destroy_netdev(dev);
3373         return err;
3374 }
3375
3376 int mlx4_en_reset_config(struct net_device *dev,
3377                          struct hwtstamp_config ts_config,
3378                          netdev_features_t features)
3379 {
3380         struct mlx4_en_priv *priv = netdev_priv(dev);
3381         struct mlx4_en_dev *mdev = priv->mdev;
3382         struct mlx4_en_port_profile new_prof;
3383         struct mlx4_en_priv *tmp;
3384         int port_up = 0;
3385         int err = 0;
3386
3387         if (priv->hwtstamp_config.tx_type == ts_config.tx_type &&
3388             priv->hwtstamp_config.rx_filter == ts_config.rx_filter &&
3389             !DEV_FEATURE_CHANGED(dev, features, NETIF_F_HW_VLAN_CTAG_RX) &&
3390             !DEV_FEATURE_CHANGED(dev, features, NETIF_F_RXFCS))
3391                 return 0; /* Nothing to change */
3392
3393         if (DEV_FEATURE_CHANGED(dev, features, NETIF_F_HW_VLAN_CTAG_RX) &&
3394             (features & NETIF_F_HW_VLAN_CTAG_RX) &&
3395             (priv->hwtstamp_config.rx_filter != HWTSTAMP_FILTER_NONE)) {
3396                 en_warn(priv, "Can't turn ON rx vlan offload while time-stamping rx filter is ON\n");
3397                 return -EINVAL;
3398         }
3399
3400         tmp = kzalloc(sizeof(*tmp), GFP_KERNEL);
3401         if (!tmp)
3402                 return -ENOMEM;
3403
3404         mutex_lock(&mdev->state_lock);
3405
3406         memcpy(&new_prof, priv->prof, sizeof(struct mlx4_en_port_profile));
3407         memcpy(&new_prof.hwtstamp_config, &ts_config, sizeof(ts_config));
3408
3409         err = mlx4_en_try_alloc_resources(priv, tmp, &new_prof);
3410         if (err)
3411                 goto out;
3412
3413         if (priv->port_up) {
3414                 port_up = 1;
3415                 mlx4_en_stop_port(dev, 1);
3416         }
3417
3418         en_warn(priv, "Changing device configuration rx filter(%x) rx vlan(%x)\n",
3419                 ts_config.rx_filter,
3420                 !!(features & NETIF_F_HW_VLAN_CTAG_RX));
3421
3422         mlx4_en_safe_replace_resources(priv, tmp);
3423
3424         if (DEV_FEATURE_CHANGED(dev, features, NETIF_F_HW_VLAN_CTAG_RX)) {
3425                 if (features & NETIF_F_HW_VLAN_CTAG_RX)
3426                         dev->features |= NETIF_F_HW_VLAN_CTAG_RX;
3427                 else
3428                         dev->features &= ~NETIF_F_HW_VLAN_CTAG_RX;
3429         } else if (ts_config.rx_filter == HWTSTAMP_FILTER_NONE) {
3430                 /* RX time-stamping is OFF, update the RX vlan offload
3431                  * to the latest wanted state
3432                  */
3433                 if (dev->wanted_features & NETIF_F_HW_VLAN_CTAG_RX)
3434                         dev->features |= NETIF_F_HW_VLAN_CTAG_RX;
3435                 else
3436                         dev->features &= ~NETIF_F_HW_VLAN_CTAG_RX;
3437         }
3438
3439         if (DEV_FEATURE_CHANGED(dev, features, NETIF_F_RXFCS)) {
3440                 if (features & NETIF_F_RXFCS)
3441                         dev->features |= NETIF_F_RXFCS;
3442                 else
3443                         dev->features &= ~NETIF_F_RXFCS;
3444         }
3445
3446         /* RX vlan offload and RX time-stamping can't co-exist !
3447          * Regardless of the caller's choice,
3448          * Turn Off RX vlan offload in case of time-stamping is ON
3449          */
3450         if (ts_config.rx_filter != HWTSTAMP_FILTER_NONE) {
3451                 if (dev->features & NETIF_F_HW_VLAN_CTAG_RX)
3452                         en_warn(priv, "Turning off RX vlan offload since RX time-stamping is ON\n");
3453                 dev->features &= ~NETIF_F_HW_VLAN_CTAG_RX;
3454         }
3455
3456         if (port_up) {
3457                 err = mlx4_en_start_port(dev);
3458                 if (err)
3459                         en_err(priv, "Failed starting port\n");
3460         }
3461
3462         if (!err)
3463                 err = mlx4_en_moderation_update(priv);
3464 out:
3465         mutex_unlock(&mdev->state_lock);
3466         kfree(tmp);
3467         if (!err)
3468                 netdev_features_change(dev);
3469         return err;
3470 }