sys/dev/mlx4/mlx4_en/mlx4_en_netdev.c

   1 /*
   2  * Copyright (c) 2007, 2014 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/etherdevice.h>
  35 #include <linux/delay.h>
  36 #include <linux/slab.h>
  37 #include <linux/compat.h>
  38 #ifdef CONFIG_NET_RX_BUSY_POLL
  39 #include <net/busy_poll.h>
  40 #endif
  41
  42 #include <linux/list.h>
  43 #include <linux/if_ether.h>
  44
  45 #include <dev/mlx4/driver.h>
  46 #include <dev/mlx4/device.h>
  47 #include <dev/mlx4/cmd.h>
  48 #include <dev/mlx4/cq.h>
  49
  50 #include <sys/eventhandler.h>
  51 #include <sys/sockio.h>
  52 #include <sys/sysctl.h>
  53
  54 #include "en.h"
  55 #include "en_port.h"
  56
  57 DEBUGNET_DEFINE(mlx4_en);
  58
  59 static void mlx4_en_sysctl_stat(struct mlx4_en_priv *priv);
  60 static void mlx4_en_sysctl_conf(struct mlx4_en_priv *priv);
  61
  62 #ifdef CONFIG_NET_RX_BUSY_POLL
  63 /* must be called with local_bh_disable()d */
  64 static int mlx4_en_low_latency_recv(struct napi_struct *napi)
  65 {
  66         struct mlx4_en_cq *cq = container_of(napi, struct mlx4_en_cq, napi);
  67         struct net_device *dev = cq->dev;
  68         struct mlx4_en_priv *priv = netdev_priv(dev);
  69         struct mlx4_en_rx_ring *rx_ring = priv->rx_ring[cq->ring];
  70         int done;
  71
  72         if (!priv->port_up)
  73                 return LL_FLUSH_FAILED;
  74
  75         if (!mlx4_en_cq_lock_poll(cq))
  76                 return LL_FLUSH_BUSY;
  77
  78         done = mlx4_en_process_rx_cq(dev, cq, 4);
  79 #ifdef LL_EXTENDED_STATS
  80         if (likely(done))
  81                 rx_ring->cleaned += done;
  82         else
  83                 rx_ring->misses++;
  84 #endif
  85
  86         mlx4_en_cq_unlock_poll(cq);
  87
  88         return done;
  89 }
  90 #endif  /* CONFIG_NET_RX_BUSY_POLL */
  91
  92 #ifdef CONFIG_RFS_ACCEL
  93
  94 struct mlx4_en_filter {
  95         struct list_head next;
  96         struct work_struct work;
  97
  98         u8     ip_proto;
  99         __be32 src_ip;
 100         __be32 dst_ip;
 101         __be16 src_port;
 102         __be16 dst_port;
 103
 104         int rxq_index;
 105         struct mlx4_en_priv *priv;
 106         u32 flow_id;                    /* RFS infrastructure id */
 107         int id;                         /* mlx4_en driver id */
 108         u64 reg_id;                     /* Flow steering API id */
 109         u8 activated;                   /* Used to prevent expiry before filter
 110                                          * is attached
 111                                          */
 112         struct hlist_node filter_chain;
 113 };
 114
 115 static void mlx4_en_filter_rfs_expire(struct mlx4_en_priv *priv);
 116
 117 static enum mlx4_net_trans_rule_id mlx4_ip_proto_to_trans_rule_id(u8 ip_proto)
 118 {
 119         switch (ip_proto) {
 120         case IPPROTO_UDP:
 121                 return MLX4_NET_TRANS_RULE_ID_UDP;
 122         case IPPROTO_TCP:
 123                 return MLX4_NET_TRANS_RULE_ID_TCP;
 124         default:
 125                 return MLX4_NET_TRANS_RULE_NUM;
 126         }
 127 };
 128
 129 static void mlx4_en_filter_work(struct work_struct *work)
 130 {
 131         struct mlx4_en_filter *filter = container_of(work,
 132                                                      struct mlx4_en_filter,
 133                                                      work);
 134         struct mlx4_en_priv *priv = filter->priv;
 135         struct mlx4_spec_list spec_tcp_udp = {
 136                 .id = mlx4_ip_proto_to_trans_rule_id(filter->ip_proto),
 137                 {
 138                         .tcp_udp = {
 139                                 .dst_port = filter->dst_port,
 140                                 .dst_port_msk = (__force __be16)-1,
 141                                 .src_port = filter->src_port,
 142                                 .src_port_msk = (__force __be16)-1,
 143                         },
 144                 },
 145         };
 146         struct mlx4_spec_list spec_ip = {
 147                 .id = MLX4_NET_TRANS_RULE_ID_IPV4,
 148                 {
 149                         .ipv4 = {
 150                                 .dst_ip = filter->dst_ip,
 151                                 .dst_ip_msk = (__force __be32)-1,
 152                                 .src_ip = filter->src_ip,
 153                                 .src_ip_msk = (__force __be32)-1,
 154                         },
 155                 },
 156         };
 157         struct mlx4_spec_list spec_eth = {
 158                 .id = MLX4_NET_TRANS_RULE_ID_ETH,
 159         };
 160         struct mlx4_net_trans_rule rule = {
 161                 .list = LIST_HEAD_INIT(rule.list),
 162                 .queue_mode = MLX4_NET_TRANS_Q_LIFO,
 163                 .exclusive = 1,
 164                 .allow_loopback = 1,
 165                 .promisc_mode = MLX4_FS_REGULAR,
 166                 .port = priv->port,
 167                 .priority = MLX4_DOMAIN_RFS,
 168         };
 169         int rc;
 170         __be64 mac_mask = cpu_to_be64(MLX4_MAC_MASK << 16);
 171
 172         if (spec_tcp_udp.id >= MLX4_NET_TRANS_RULE_NUM) {
 173                 en_warn(priv, "RFS: ignoring unsupported ip protocol (%d)\n",
 174                         filter->ip_proto);
 175                 goto ignore;
 176         }
 177         list_add_tail(&spec_eth.list, &rule.list);
 178         list_add_tail(&spec_ip.list, &rule.list);
 179         list_add_tail(&spec_tcp_udp.list, &rule.list);
 180
 181         rule.qpn = priv->rss_map.qps[filter->rxq_index].qpn;
 182         memcpy(spec_eth.eth.dst_mac, priv->dev->dev_addr, ETH_ALEN);
 183         memcpy(spec_eth.eth.dst_mac_msk, &mac_mask, ETH_ALEN);
 184
 185         filter->activated = 0;
 186
 187         if (filter->reg_id) {
 188                 rc = mlx4_flow_detach(priv->mdev->dev, filter->reg_id);
 189                 if (rc && rc != -ENOENT)
 190                         en_err(priv, "Error detaching flow. rc = %d\n", rc);
 191         }
 192
 193         rc = mlx4_flow_attach(priv->mdev->dev, &rule, &filter->reg_id);
 194         if (rc)
 195                 en_err(priv, "Error attaching flow. err = %d\n", rc);
 196
 197 ignore:
 198         mlx4_en_filter_rfs_expire(priv);
 199
 200         filter->activated = 1;
 201 }
 202
 203 static inline struct hlist_head *
 204 filter_hash_bucket(struct mlx4_en_priv *priv, __be32 src_ip, __be32 dst_ip,
 205                    __be16 src_port, __be16 dst_port)
 206 {
 207         unsigned long l;
 208         int bucket_idx;
 209
 210         l = (__force unsigned long)src_port |
 211             ((__force unsigned long)dst_port << 2);
 212         l ^= (__force unsigned long)(src_ip ^ dst_ip);
 213
 214         bucket_idx = hash_long(l, MLX4_EN_FILTER_HASH_SHIFT);
 215
 216         return &priv->filter_hash[bucket_idx];
 217 }
 218
 219 static struct mlx4_en_filter *
 220 mlx4_en_filter_alloc(struct mlx4_en_priv *priv, int rxq_index, __be32 src_ip,
 221                      __be32 dst_ip, u8 ip_proto, __be16 src_port,
 222                      __be16 dst_port, u32 flow_id)
 223 {
 224         struct mlx4_en_filter *filter = NULL;
 225
 226         filter = kzalloc(sizeof(struct mlx4_en_filter), GFP_ATOMIC);
 227         if (!filter)
 228                 return NULL;
 229
 230         filter->priv = priv;
 231         filter->rxq_index = rxq_index;
 232         INIT_WORK(&filter->work, mlx4_en_filter_work);
 233
 234         filter->src_ip = src_ip;
 235         filter->dst_ip = dst_ip;
 236         filter->ip_proto = ip_proto;
 237         filter->src_port = src_port;
 238         filter->dst_port = dst_port;
 239
 240         filter->flow_id = flow_id;
 241
 242         filter->id = priv->last_filter_id++ % RPS_NO_FILTER;
 243
 244         list_add_tail(&filter->next, &priv->filters);
 245         hlist_add_head(&filter->filter_chain,
 246                        filter_hash_bucket(priv, src_ip, dst_ip, src_port,
 247                                           dst_port));
 248
 249         return filter;
 250 }
 251
 252 static void mlx4_en_filter_free(struct mlx4_en_filter *filter)
 253 {
 254         struct mlx4_en_priv *priv = filter->priv;
 255         int rc;
 256
 257         list_del(&filter->next);
 258
 259         rc = mlx4_flow_detach(priv->mdev->dev, filter->reg_id);
 260         if (rc && rc != -ENOENT)
 261                 en_err(priv, "Error detaching flow. rc = %d\n", rc);
 262
 263         kfree(filter);
 264 }
 265
 266 static inline struct mlx4_en_filter *
 267 mlx4_en_filter_find(struct mlx4_en_priv *priv, __be32 src_ip, __be32 dst_ip,
 268                     u8 ip_proto, __be16 src_port, __be16 dst_port)
 269 {
 270         struct mlx4_en_filter *filter;
 271         struct mlx4_en_filter *ret = NULL;
 272
 273         hlist_for_each_entry(filter,
 274                              filter_hash_bucket(priv, src_ip, dst_ip,
 275                                                 src_port, dst_port),
 276                              filter_chain) {
 277                 if (filter->src_ip == src_ip &&
 278                     filter->dst_ip == dst_ip &&
 279                     filter->ip_proto == ip_proto &&
 280                     filter->src_port == src_port &&
 281                     filter->dst_port == dst_port) {
 282                         ret = filter;
 283                         break;
 284                 }
 285         }
 286
 287         return ret;
 288 }
 289
 290 static int
 291 mlx4_en_filter_rfs(struct net_device *net_dev, const struct sk_buff *skb,
 292                    u16 rxq_index, u32 flow_id)
 293 {
 294         struct mlx4_en_priv *priv = netdev_priv(net_dev);
 295         struct mlx4_en_filter *filter;
 296         const struct iphdr *ip;
 297         const __be16 *ports;
 298         u8 ip_proto;
 299         __be32 src_ip;
 300         __be32 dst_ip;
 301         __be16 src_port;
 302         __be16 dst_port;
 303         int nhoff = skb_network_offset(skb);
 304         int ret = 0;
 305
 306         if (skb->protocol != htons(ETH_P_IP))
 307                 return -EPROTONOSUPPORT;
 308
 309         ip = (const struct iphdr *)(skb->data + nhoff);
 310         if (ip_is_fragment(ip))
 311                 return -EPROTONOSUPPORT;
 312
 313         if ((ip->protocol != IPPROTO_TCP) && (ip->protocol != IPPROTO_UDP))
 314                 return -EPROTONOSUPPORT;
 315         ports = (const __be16 *)(skb->data + nhoff + 4 * ip->ihl);
 316
 317         ip_proto = ip->protocol;
 318         src_ip = ip->saddr;
 319         dst_ip = ip->daddr;
 320         src_port = ports[0];
 321         dst_port = ports[1];
 322
 323         spin_lock_bh(&priv->filters_lock);
 324         filter = mlx4_en_filter_find(priv, src_ip, dst_ip, ip_proto,
 325                                      src_port, dst_port);
 326         if (filter) {
 327                 if (filter->rxq_index == rxq_index)
 328                         goto out;
 329
 330                 filter->rxq_index = rxq_index;
 331         } else {
 332                 filter = mlx4_en_filter_alloc(priv, rxq_index,
 333                                               src_ip, dst_ip, ip_proto,
 334                                               src_port, dst_port, flow_id);
 335                 if (!filter) {
 336                         ret = -ENOMEM;
 337                         goto err;
 338                 }
 339         }
 340
 341         queue_work(priv->mdev->workqueue, &filter->work);
 342
 343 out:
 344         ret = filter->id;
 345 err:
 346         spin_unlock_bh(&priv->filters_lock);
 347
 348         return ret;
 349 }
 350
 351 void mlx4_en_cleanup_filters(struct mlx4_en_priv *priv)
 352 {
 353         struct mlx4_en_filter *filter, *tmp;
 354         LIST_HEAD(del_list);
 355
 356         spin_lock_bh(&priv->filters_lock);
 357         list_for_each_entry_safe(filter, tmp, &priv->filters, next) {
 358                 list_move(&filter->next, &del_list);
 359                 hlist_del(&filter->filter_chain);
 360         }
 361         spin_unlock_bh(&priv->filters_lock);
 362
 363         list_for_each_entry_safe(filter, tmp, &del_list, next) {
 364                 cancel_work_sync(&filter->work);
 365                 mlx4_en_filter_free(filter);
 366         }
 367 }
 368
 369 static void mlx4_en_filter_rfs_expire(struct mlx4_en_priv *priv)
 370 {
 371         struct mlx4_en_filter *filter = NULL, *tmp, *last_filter = NULL;
 372         LIST_HEAD(del_list);
 373         int i = 0;
 374
 375         spin_lock_bh(&priv->filters_lock);
 376         list_for_each_entry_safe(filter, tmp, &priv->filters, next) {
 377                 if (i > MLX4_EN_FILTER_EXPIRY_QUOTA)
 378                         break;
 379
 380                 if (filter->activated &&
 381                     !work_pending(&filter->work) &&
 382                     rps_may_expire_flow(priv->dev,
 383                                         filter->rxq_index, filter->flow_id,
 384                                         filter->id)) {
 385                         list_move(&filter->next, &del_list);
 386                         hlist_del(&filter->filter_chain);
 387                 } else
 388                         last_filter = filter;
 389
 390                 i++;
 391         }
 392
 393         if (last_filter && (&last_filter->next != priv->filters.next))
 394                 list_move(&priv->filters, &last_filter->next);
 395
 396         spin_unlock_bh(&priv->filters_lock);
 397
 398         list_for_each_entry_safe(filter, tmp, &del_list, next)
 399                 mlx4_en_filter_free(filter);
 400 }
 401 #endif
 402
 403 static void mlx4_en_vlan_rx_add_vid(void *arg, struct net_device *dev, u16 vid)
 404 {
 405         struct mlx4_en_priv *priv = netdev_priv(dev);
 406         struct mlx4_en_dev *mdev = priv->mdev;
 407         int err;
 408         int idx;
 409
 410         if (arg != priv)
 411                 return;
 412
 413         en_dbg(HW, priv, "adding VLAN:%d\n", vid);
 414
 415         set_bit(vid, priv->active_vlans);
 416
 417         /* Add VID to port VLAN filter */
 418         mutex_lock(&mdev->state_lock);
 419         if (mdev->device_up && priv->port_up) {
 420                 err = mlx4_SET_VLAN_FLTR(mdev->dev, priv);
 421                 if (err)
 422                         en_err(priv, "Failed configuring VLAN filter\n");
 423         }
 424         if (mlx4_register_vlan(mdev->dev, priv->port, vid, &idx))
 425                 en_dbg(HW, priv, "failed adding vlan %d\n", vid);
 426         mutex_unlock(&mdev->state_lock);
 427
 428 }
 429
 430 static void mlx4_en_vlan_rx_kill_vid(void *arg, struct net_device *dev, u16 vid)
 431 {
 432         struct mlx4_en_priv *priv = netdev_priv(dev);
 433         struct mlx4_en_dev *mdev = priv->mdev;
 434         int err;
 435
 436         if (arg != priv)
 437                 return;
 438
 439         en_dbg(HW, priv, "Killing VID:%d\n", vid);
 440
 441         clear_bit(vid, priv->active_vlans);
 442
 443         /* Remove VID from port VLAN filter */
 444         mutex_lock(&mdev->state_lock);
 445         mlx4_unregister_vlan(mdev->dev, priv->port, vid);
 446
 447         if (mdev->device_up && priv->port_up) {
 448                 err = mlx4_SET_VLAN_FLTR(mdev->dev, priv);
 449                 if (err)
 450                         en_err(priv, "Failed configuring VLAN filter\n");
 451         }
 452         mutex_unlock(&mdev->state_lock);
 453
 454 }
 455
 456 static int mlx4_en_tunnel_steer_add(struct mlx4_en_priv *priv, unsigned char *addr,
 457                                     int qpn, u64 *reg_id)
 458 {
 459         int err;
 460
 461         if (priv->mdev->dev->caps.tunnel_offload_mode != MLX4_TUNNEL_OFFLOAD_MODE_VXLAN ||
 462             priv->mdev->dev->caps.dmfs_high_steer_mode == MLX4_STEERING_DMFS_A0_STATIC)
 463                 return 0; /* do nothing */
 464
 465         err = mlx4_tunnel_steer_add(priv->mdev->dev, addr, priv->port, qpn,
 466                                     MLX4_DOMAIN_NIC, reg_id);
 467         if (err) {
 468                 en_err(priv, "failed to add vxlan steering rule, err %d\n", err);
 469                 return err;
 470         }
 471         en_dbg(DRV, priv, "added vxlan steering rule, mac %pM reg_id %llx\n", addr, (long long)*reg_id);
 472         return 0;
 473 }
 474
 475 static int mlx4_en_uc_steer_add(struct mlx4_en_priv *priv,
 476                                 unsigned char *mac, int *qpn, u64 *reg_id)
 477 {
 478         struct mlx4_en_dev *mdev = priv->mdev;
 479         struct mlx4_dev *dev = mdev->dev;
 480         int err;
 481
 482         switch (dev->caps.steering_mode) {
 483         case MLX4_STEERING_MODE_B0: {
 484                 struct mlx4_qp qp;
 485                 u8 gid[16] = {0};
 486
 487                 qp.qpn = *qpn;
 488                 memcpy(&gid[10], mac, ETH_ALEN);
 489                 gid[5] = priv->port;
 490
 491                 err = mlx4_unicast_attach(dev, &qp, gid, 0, MLX4_PROT_ETH);
 492                 break;
 493         }
 494         case MLX4_STEERING_MODE_DEVICE_MANAGED: {
 495                 struct mlx4_spec_list spec_eth = { {NULL} };
 496                 __be64 mac_mask = cpu_to_be64(MLX4_MAC_MASK << 16);
 497
 498                 struct mlx4_net_trans_rule rule = {
 499                         .queue_mode = MLX4_NET_TRANS_Q_FIFO,
 500                         .exclusive = 0,
 501                         .allow_loopback = 1,
 502                         .promisc_mode = MLX4_FS_REGULAR,
 503                         .priority = MLX4_DOMAIN_NIC,
 504                 };
 505
 506                 rule.port = priv->port;
 507                 rule.qpn = *qpn;
 508                 INIT_LIST_HEAD(&rule.list);
 509
 510                 spec_eth.id = MLX4_NET_TRANS_RULE_ID_ETH;
 511                 memcpy(spec_eth.eth.dst_mac, mac, ETH_ALEN);
 512                 memcpy(spec_eth.eth.dst_mac_msk, &mac_mask, ETH_ALEN);
 513                 list_add_tail(&spec_eth.list, &rule.list);
 514
 515                 err = mlx4_flow_attach(dev, &rule, reg_id);
 516                 break;
 517         }
 518         default:
 519                 return -EINVAL;
 520         }
 521         if (err)
 522                 en_warn(priv, "Failed Attaching Unicast\n");
 523
 524         return err;
 525 }
 526
 527 static void mlx4_en_uc_steer_release(struct mlx4_en_priv *priv,
 528                                      unsigned char *mac, int qpn, u64 reg_id)
 529 {
 530         struct mlx4_en_dev *mdev = priv->mdev;
 531         struct mlx4_dev *dev = mdev->dev;
 532
 533         switch (dev->caps.steering_mode) {
 534         case MLX4_STEERING_MODE_B0: {
 535                 struct mlx4_qp qp;
 536                 u8 gid[16] = {0};
 537
 538                 qp.qpn = qpn;
 539                 memcpy(&gid[10], mac, ETH_ALEN);
 540                 gid[5] = priv->port;
 541
 542                 mlx4_unicast_detach(dev, &qp, gid, MLX4_PROT_ETH);
 543                 break;
 544         }
 545         case MLX4_STEERING_MODE_DEVICE_MANAGED: {
 546                 mlx4_flow_detach(dev, reg_id);
 547                 break;
 548         }
 549         default:
 550                 en_err(priv, "Invalid steering mode.\n");
 551         }
 552 }
 553
 554 static int mlx4_en_get_qp(struct mlx4_en_priv *priv)
 555 {
 556         struct mlx4_en_dev *mdev = priv->mdev;
 557         struct mlx4_dev *dev = mdev->dev;
 558         int index = 0;
 559         int err = 0;
 560         int *qpn = &priv->base_qpn;
 561         u64 mac = mlx4_mac_to_u64(IF_LLADDR(priv->dev));
 562
 563         en_dbg(DRV, priv, "Registering MAC: %pM for adding\n",
 564                IF_LLADDR(priv->dev));
 565         index = mlx4_register_mac(dev, priv->port, mac);
 566         if (index < 0) {
 567                 err = index;
 568                 en_err(priv, "Failed adding MAC: %pM\n",
 569                        IF_LLADDR(priv->dev));
 570                 return err;
 571         }
 572
 573         if (dev->caps.steering_mode == MLX4_STEERING_MODE_A0) {
 574                 int base_qpn = mlx4_get_base_qpn(dev, priv->port);
 575                 *qpn = base_qpn + index;
 576                 return 0;
 577         }
 578
 579         err = mlx4_qp_reserve_range(dev, 1, 1, qpn, MLX4_RESERVE_A0_QP);
 580         en_dbg(DRV, priv, "Reserved qp %d\n", *qpn);
 581         if (err) {
 582                 en_err(priv, "Failed to reserve qp for mac registration\n");
 583                 mlx4_unregister_mac(dev, priv->port, mac);
 584                 return err;
 585         }
 586
 587         return 0;
 588 }
 589
 590 static void mlx4_en_put_qp(struct mlx4_en_priv *priv)
 591 {
 592         struct mlx4_en_dev *mdev = priv->mdev;
 593         struct mlx4_dev *dev = mdev->dev;
 594         int qpn = priv->base_qpn;
 595
 596         if (dev->caps.steering_mode == MLX4_STEERING_MODE_A0) {
 597                 u64 mac = mlx4_mac_to_u64(IF_LLADDR(priv->dev));
 598                 en_dbg(DRV, priv, "Registering MAC: %pM for deleting\n",
 599                        IF_LLADDR(priv->dev));
 600                 mlx4_unregister_mac(dev, priv->port, mac);
 601         } else {
 602                 en_dbg(DRV, priv, "Releasing qp: port %d, qpn %d\n",
 603                        priv->port, qpn);
 604                 mlx4_qp_release_range(dev, qpn, 1);
 605                 priv->flags &= ~MLX4_EN_FLAG_FORCE_PROMISC;
 606         }
 607 }
 608
 609 static void mlx4_en_clear_uclist(struct net_device *dev)
 610 {
 611         struct mlx4_en_priv *priv = netdev_priv(dev);
 612         struct mlx4_en_addr_list *tmp, *uc_to_del;
 613
 614         list_for_each_entry_safe(uc_to_del, tmp, &priv->uc_list, list) {
 615                 list_del(&uc_to_del->list);
 616                 kfree(uc_to_del);
 617         }
 618 }
 619
 620 static u_int mlx4_copy_addr(void *arg, struct sockaddr_dl *sdl, u_int cnt)
 621 {
 622         struct mlx4_en_priv *priv = arg;
 623         struct mlx4_en_addr_list *tmp;
 624
 625         if (sdl->sdl_alen != ETHER_ADDR_LEN)    /* XXXGL: can that happen? */
 626                 return (0);
 627         tmp = kzalloc(sizeof(struct mlx4_en_addr_list), GFP_ATOMIC);
 628         if (tmp == NULL) {
 629                 en_err(priv, "Failed to allocate address list\n");
 630                 return (0);
 631         }
 632         memcpy(tmp->addr, LLADDR(sdl), ETH_ALEN);
 633         list_add_tail(&tmp->list, &priv->uc_list);
 634
 635         return (1);
 636 }
 637
 638 static void mlx4_en_cache_uclist(struct net_device *dev)
 639 {
 640         struct mlx4_en_priv *priv = netdev_priv(dev);
 641
 642         mlx4_en_clear_uclist(dev);
 643         if_foreach_lladdr(dev, mlx4_copy_addr, priv);
 644 }
 645
 646 static void mlx4_en_clear_mclist(struct net_device *dev)
 647 {
 648         struct mlx4_en_priv *priv = netdev_priv(dev);
 649         struct mlx4_en_addr_list *tmp, *mc_to_del;
 650
 651         list_for_each_entry_safe(mc_to_del, tmp, &priv->mc_list, list) {
 652                 list_del(&mc_to_del->list);
 653                 kfree(mc_to_del);
 654         }
 655 }
 656
 657 static u_int mlx4_copy_maddr(void *arg, struct sockaddr_dl *sdl, u_int count)
 658 {
 659         struct mlx4_en_priv *priv = arg;
 660         struct mlx4_en_addr_list *tmp;
 661
 662         if (sdl->sdl_alen != ETHER_ADDR_LEN)    /* XXXGL: can that happen? */
 663                 return (0);
 664         tmp = kzalloc(sizeof(struct mlx4_en_addr_list), GFP_ATOMIC);
 665         if (tmp == NULL) {
 666                 en_err(priv, "Failed to allocate address list\n");
 667                 return (0);
 668         }
 669         memcpy(tmp->addr, LLADDR(sdl), ETH_ALEN);
 670         list_add_tail(&tmp->list, &priv->mc_list);
 671         return (1);
 672 }
 673
 674 static void mlx4_en_cache_mclist(struct net_device *dev)
 675 {
 676         struct mlx4_en_priv *priv = netdev_priv(dev);
 677
 678         mlx4_en_clear_mclist(dev);
 679         if_foreach_llmaddr(dev, mlx4_copy_maddr, priv);
 680 }
 681
 682 static void update_addr_list_flags(struct mlx4_en_priv *priv,
 683                                 struct list_head *dst,
 684                                 struct list_head *src)
 685 {
 686         struct mlx4_en_addr_list *dst_tmp, *src_tmp, *new_mc;
 687         bool found;
 688
 689         /* Find all the entries that should be removed from dst,
 690          * These are the entries that are not found in src
 691          */
 692         list_for_each_entry(dst_tmp, dst, list) {
 693                 found = false;
 694                 list_for_each_entry(src_tmp, src, list) {
 695                         if (!memcmp(dst_tmp->addr, src_tmp->addr, ETH_ALEN)) {
 696                                 found = true;
 697                                 break;
 698                         }
 699                 }
 700                 if (!found)
 701                         dst_tmp->action = MLX4_ADDR_LIST_REM;
 702         }
 703
 704         /* Add entries that exist in src but not in dst
 705          * mark them as need to add
 706          */
 707         list_for_each_entry(src_tmp, src, list) {
 708                 found = false;
 709                 list_for_each_entry(dst_tmp, dst, list) {
 710                         if (!memcmp(dst_tmp->addr, src_tmp->addr, ETH_ALEN)) {
 711                                 dst_tmp->action = MLX4_ADDR_LIST_NONE;
 712                                 found = true;
 713                                 break;
 714                         }
 715                 }
 716                 if (!found) {
 717                         new_mc = kmalloc(sizeof(struct mlx4_en_addr_list),
 718                                          GFP_KERNEL);
 719                         if (!new_mc) {
 720                                 en_err(priv, "Failed to allocate current multicast list\n");
 721                                 return;
 722                         }
 723                         memcpy(new_mc, src_tmp,
 724                                sizeof(struct mlx4_en_addr_list));
 725                         new_mc->action = MLX4_ADDR_LIST_ADD;
 726                         list_add_tail(&new_mc->list, dst);
 727                 }
 728         }
 729 }
 730
 731 static void mlx4_en_set_rx_mode(struct net_device *dev)
 732 {
 733         struct mlx4_en_priv *priv = netdev_priv(dev);
 734
 735         if (!priv->port_up)
 736                 return;
 737
 738         queue_work(priv->mdev->workqueue, &priv->rx_mode_task);
 739 }
 740
 741 static void mlx4_en_set_promisc_mode(struct mlx4_en_priv *priv,
 742                                      struct mlx4_en_dev *mdev)
 743 {
 744         int err = 0;
 745
 746         if (!(priv->flags & MLX4_EN_FLAG_PROMISC)) {
 747                 priv->flags |= MLX4_EN_FLAG_PROMISC;
 748
 749                 /* Enable promiscouos mode */
 750                 switch (mdev->dev->caps.steering_mode) {
 751                 case MLX4_STEERING_MODE_DEVICE_MANAGED:
 752                         err = mlx4_flow_steer_promisc_add(mdev->dev,
 753                                                           priv->port,
 754                                                           priv->base_qpn,
 755                                                           MLX4_FS_ALL_DEFAULT);
 756                         if (err)
 757                                 en_err(priv, "Failed enabling promiscuous mode\n");
 758                         priv->flags |= MLX4_EN_FLAG_MC_PROMISC;
 759                         break;
 760
 761                 case MLX4_STEERING_MODE_B0:
 762                         err = mlx4_unicast_promisc_add(mdev->dev,
 763                                                        priv->base_qpn,
 764                                                        priv->port);
 765                         if (err)
 766                                 en_err(priv, "Failed enabling unicast promiscuous mode\n");
 767
 768                         /* Add the default qp number as multicast
 769                          * promisc
 770                          */
 771                         if (!(priv->flags & MLX4_EN_FLAG_MC_PROMISC)) {
 772                                 err = mlx4_multicast_promisc_add(mdev->dev,
 773                                                                  priv->base_qpn,
 774                                                                  priv->port);
 775                                 if (err)
 776                                         en_err(priv, "Failed enabling multicast promiscuous mode\n");
 777                                 priv->flags |= MLX4_EN_FLAG_MC_PROMISC;
 778                         }
 779                         break;
 780
 781                 case MLX4_STEERING_MODE_A0:
 782                         err = mlx4_SET_PORT_qpn_calc(mdev->dev,
 783                                                      priv->port,
 784                                                      priv->base_qpn,
 785                                                      1);
 786                         if (err)
 787                                 en_err(priv, "Failed enabling promiscuous mode\n");
 788                         break;
 789                 }
 790
 791                 /* Disable port multicast filter (unconditionally) */
 792                 err = mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, 0,
 793                                           0, MLX4_MCAST_DISABLE);
 794                 if (err)
 795                         en_err(priv, "Failed disabling multicast filter\n");
 796         }
 797 }
 798
 799 static void mlx4_en_clear_promisc_mode(struct mlx4_en_priv *priv,
 800                                        struct mlx4_en_dev *mdev)
 801 {
 802         int err = 0;
 803
 804         priv->flags &= ~MLX4_EN_FLAG_PROMISC;
 805
 806         /* Disable promiscouos mode */
 807         switch (mdev->dev->caps.steering_mode) {
 808         case MLX4_STEERING_MODE_DEVICE_MANAGED:
 809                 err = mlx4_flow_steer_promisc_remove(mdev->dev,
 810                                                      priv->port,
 811                                                      MLX4_FS_ALL_DEFAULT);
 812                 if (err)
 813                         en_err(priv, "Failed disabling promiscuous mode\n");
 814                 priv->flags &= ~MLX4_EN_FLAG_MC_PROMISC;
 815                 break;
 816
 817         case MLX4_STEERING_MODE_B0:
 818                 err = mlx4_unicast_promisc_remove(mdev->dev,
 819                                                   priv->base_qpn,
 820                                                   priv->port);
 821                 if (err)
 822                         en_err(priv, "Failed disabling unicast promiscuous mode\n");
 823                 /* Disable Multicast promisc */
 824                 if (priv->flags & MLX4_EN_FLAG_MC_PROMISC) {
 825                         err = mlx4_multicast_promisc_remove(mdev->dev,
 826                                                             priv->base_qpn,
 827                                                             priv->port);
 828                         if (err)
 829                                 en_err(priv, "Failed disabling multicast promiscuous mode\n");
 830                         priv->flags &= ~MLX4_EN_FLAG_MC_PROMISC;
 831                 }
 832                 break;
 833
 834         case MLX4_STEERING_MODE_A0:
 835                 err = mlx4_SET_PORT_qpn_calc(mdev->dev,
 836                                              priv->port,
 837                                              priv->base_qpn, 0);
 838                 if (err)
 839                         en_err(priv, "Failed disabling promiscuous mode\n");
 840                 break;
 841         }
 842 }
 843
 844 static void mlx4_en_do_multicast(struct mlx4_en_priv *priv,
 845                                  struct net_device *dev,
 846                                  struct mlx4_en_dev *mdev)
 847 {
 848         struct mlx4_en_addr_list *addr_list, *tmp;
 849         u8 mc_list[16] = {0};
 850         int err = 0;
 851         u64 mcast_addr = 0;
 852
 853
 854         /* Enable/disable the multicast filter according to IFF_ALLMULTI */
 855         if (dev->if_flags & IFF_ALLMULTI) {
 856                 err = mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, 0,
 857                                           0, MLX4_MCAST_DISABLE);
 858                 if (err)
 859                         en_err(priv, "Failed disabling multicast filter\n");
 860
 861                 /* Add the default qp number as multicast promisc */
 862                 if (!(priv->flags & MLX4_EN_FLAG_MC_PROMISC)) {
 863                         switch (mdev->dev->caps.steering_mode) {
 864                         case MLX4_STEERING_MODE_DEVICE_MANAGED:
 865                                 err = mlx4_flow_steer_promisc_add(mdev->dev,
 866                                                                   priv->port,
 867                                                                   priv->base_qpn,
 868                                                                   MLX4_FS_MC_DEFAULT);
 869                                 break;
 870
 871                         case MLX4_STEERING_MODE_B0:
 872                                 err = mlx4_multicast_promisc_add(mdev->dev,
 873                                                                  priv->base_qpn,
 874                                                                  priv->port);
 875                                 break;
 876
 877                         case MLX4_STEERING_MODE_A0:
 878                                 break;
 879                         }
 880                         if (err)
 881                                 en_err(priv, "Failed entering multicast promisc mode\n");
 882                         priv->flags |= MLX4_EN_FLAG_MC_PROMISC;
 883                 }
 884         } else {
 885                 /* Disable Multicast promisc */
 886                 if (priv->flags & MLX4_EN_FLAG_MC_PROMISC) {
 887                         switch (mdev->dev->caps.steering_mode) {
 888                         case MLX4_STEERING_MODE_DEVICE_MANAGED:
 889                                 err = mlx4_flow_steer_promisc_remove(mdev->dev,
 890                                                                      priv->port,
 891                                                                      MLX4_FS_MC_DEFAULT);
 892                                 break;
 893
 894                         case MLX4_STEERING_MODE_B0:
 895                                 err = mlx4_multicast_promisc_remove(mdev->dev,
 896                                                                     priv->base_qpn,
 897                                                                     priv->port);
 898                                 break;
 899
 900                         case MLX4_STEERING_MODE_A0:
 901                                 break;
 902                         }
 903                         if (err)
 904                                 en_err(priv, "Failed disabling multicast promiscuous mode\n");
 905                         priv->flags &= ~MLX4_EN_FLAG_MC_PROMISC;
 906                 }
 907
 908                 err = mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, 0,
 909                                           0, MLX4_MCAST_DISABLE);
 910                 if (err)
 911                         en_err(priv, "Failed disabling multicast filter\n");
 912
 913                 /* Flush mcast filter and init it with broadcast address */
 914                 mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, ETH_BCAST,
 915                                     1, MLX4_MCAST_CONFIG);
 916
 917                 /* Update multicast list - we cache all addresses so they won't
 918                  * change while HW is updated holding the command semaphor */
 919                 mlx4_en_cache_mclist(dev);
 920                 list_for_each_entry(addr_list, &priv->mc_list, list) {
 921                         mcast_addr = mlx4_mac_to_u64(addr_list->addr);
 922                         mlx4_SET_MCAST_FLTR(mdev->dev, priv->port,
 923                                         mcast_addr, 0, MLX4_MCAST_CONFIG);
 924                 }
 925                 err = mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, 0,
 926                                           0, MLX4_MCAST_ENABLE);
 927                 if (err)
 928                         en_err(priv, "Failed enabling multicast filter\n");
 929
 930                 update_addr_list_flags(priv, &priv->curr_mc_list, &priv->mc_list);
 931
 932                 list_for_each_entry_safe(addr_list, tmp, &priv->curr_mc_list, list) {
 933                         if (addr_list->action == MLX4_ADDR_LIST_REM) {
 934                                 /* detach this address and delete from list */
 935                                 memcpy(&mc_list[10], addr_list->addr, ETH_ALEN);
 936                                 mc_list[5] = priv->port;
 937                                 err = mlx4_multicast_detach(mdev->dev,
 938                                                             &priv->rss_map.indir_qp,
 939                                                             mc_list,
 940                                                             MLX4_PROT_ETH,
 941                                                             addr_list->reg_id);
 942                                 if (err)
 943                                         en_err(priv, "Fail to detach multicast address\n");
 944
 945                                 if (addr_list->tunnel_reg_id) {
 946                                         err = mlx4_flow_detach(priv->mdev->dev, addr_list->tunnel_reg_id);
 947                                         if (err)
 948                                                 en_err(priv, "Failed to detach multicast address\n");
 949                                 }
 950
 951                                 /* remove from list */
 952                                 list_del(&addr_list->list);
 953                                 kfree(addr_list);
 954                         } else if (addr_list->action == MLX4_ADDR_LIST_ADD) {
 955                                 /* attach the address */
 956                                 memcpy(&mc_list[10], addr_list->addr, ETH_ALEN);
 957                                 /* needed for B0 steering support */
 958                                 mc_list[5] = priv->port;
 959                                 err = mlx4_multicast_attach(mdev->dev,
 960                                                             &priv->rss_map.indir_qp,
 961                                                             mc_list,
 962                                                             priv->port, 0,
 963                                                             MLX4_PROT_ETH,
 964                                                             &addr_list->reg_id);
 965                                 if (err)
 966                                         en_err(priv, "Fail to attach multicast address\n");
 967
 968                                 err = mlx4_en_tunnel_steer_add(priv, &mc_list[10], priv->base_qpn,
 969                                                                &addr_list->tunnel_reg_id);
 970                                 if (err)
 971                                         en_err(priv, "Failed to attach multicast address\n");
 972                         }
 973                 }
 974         }
 975 }
 976
 977 static void mlx4_en_do_unicast(struct mlx4_en_priv *priv,
 978                                struct net_device *dev,
 979                                struct mlx4_en_dev *mdev)
 980 {
 981         struct mlx4_en_addr_list *addr_list, *tmp;
 982         int err;
 983
 984         /* Update unicast list */
 985         mlx4_en_cache_uclist(dev);
 986
 987         update_addr_list_flags(priv, &priv->curr_uc_list, &priv->uc_list);
 988
 989         list_for_each_entry_safe(addr_list, tmp, &priv->curr_uc_list, list) {
 990                 if (addr_list->action == MLX4_ADDR_LIST_REM) {
 991                         mlx4_en_uc_steer_release(priv, addr_list->addr,
 992                                                  priv->rss_map.indir_qp.qpn,
 993                                                  addr_list->reg_id);
 994                         /* remove from list */
 995                         list_del(&addr_list->list);
 996                         kfree(addr_list);
 997                 } else if (addr_list->action == MLX4_ADDR_LIST_ADD) {
 998                         err = mlx4_en_uc_steer_add(priv, addr_list->addr,
 999                                                    &priv->rss_map.indir_qp.qpn,
1000                                                    &addr_list->reg_id);
1001                         if (err)
1002                                 en_err(priv, "Fail to add unicast address\n");
1003                 }
1004         }
1005 }
1006
1007 static void mlx4_en_do_set_rx_mode(struct work_struct *work)
1008 {
1009         struct mlx4_en_priv *priv = container_of(work, struct mlx4_en_priv,
1010                                                  rx_mode_task);
1011         struct mlx4_en_dev *mdev = priv->mdev;
1012         struct net_device *dev = priv->dev;
1013
1014         mutex_lock(&mdev->state_lock);
1015         if (!mdev->device_up) {
1016                 en_dbg(HW, priv, "Card is not up, ignoring rx mode change.\n");
1017                 goto out;
1018         }
1019         if (!priv->port_up) {
1020                 en_dbg(HW, priv, "Port is down, ignoring rx mode change.\n");
1021                 goto out;
1022         }
1023         if (!mlx4_en_QUERY_PORT(mdev, priv->port)) {
1024                 if (priv->port_state.link_state) {
1025                         priv->last_link_state = MLX4_DEV_EVENT_PORT_UP;
1026                         /* update netif baudrate */
1027                         priv->dev->if_baudrate =
1028                             IF_Mbps(priv->port_state.link_speed);
1029                         /* Important note: the following call for if_link_state_change
1030                          * is needed for interface up scenario (start port, link state
1031                          * change) */
1032                         if_link_state_change(priv->dev, LINK_STATE_UP);
1033                         en_dbg(HW, priv, "Link Up\n");
1034                 }
1035         }
1036
1037         /* Set unicast rules */
1038         mlx4_en_do_unicast(priv, dev, mdev);
1039
1040         /* Promsicuous mode: disable all filters */
1041         if ((dev->if_flags & IFF_PROMISC) ||
1042             (priv->flags & MLX4_EN_FLAG_FORCE_PROMISC)) {
1043                 mlx4_en_set_promisc_mode(priv, mdev);
1044         } else if (priv->flags & MLX4_EN_FLAG_PROMISC) {
1045                 /* Not in promiscuous mode */
1046                 mlx4_en_clear_promisc_mode(priv, mdev);
1047         }
1048
1049         /* Set multicast rules */
1050         mlx4_en_do_multicast(priv, dev, mdev);
1051 out:
1052         mutex_unlock(&mdev->state_lock);
1053 }
1054
1055 static void mlx4_en_watchdog_timeout(void *arg)
1056 {
1057         struct mlx4_en_priv *priv = arg;
1058         struct mlx4_en_dev *mdev = priv->mdev;
1059
1060         en_dbg(DRV, priv, "Scheduling watchdog\n");
1061         queue_work(mdev->workqueue, &priv->watchdog_task);
1062         if (priv->port_up)
1063                 callout_reset(&priv->watchdog_timer, MLX4_EN_WATCHDOG_TIMEOUT,
1064                                 mlx4_en_watchdog_timeout, priv);
1065 }
1066
1067
1068
1069 static void mlx4_en_set_default_moderation(struct mlx4_en_priv *priv)
1070 {
1071         struct mlx4_en_cq *cq;
1072         int i;
1073
1074         /* If we haven't received a specific coalescing setting
1075          * (module param), we set the moderation parameters as follows:
1076          * - moder_cnt is set to the number of mtu sized packets to
1077          *   satisfy our coalescing target.
1078          * - moder_time is set to a fixed value.
1079          */
1080         priv->rx_frames = MLX4_EN_RX_COAL_TARGET;
1081         priv->rx_usecs = MLX4_EN_RX_COAL_TIME;
1082         priv->tx_frames = MLX4_EN_TX_COAL_PKTS;
1083         priv->tx_usecs = MLX4_EN_TX_COAL_TIME;
1084         en_dbg(INTR, priv, "Default coalesing params for mtu: %u - "
1085                "rx_frames:%d rx_usecs:%d\n",
1086                (unsigned)priv->dev->if_mtu, priv->rx_frames, priv->rx_usecs);
1087
1088         /* Setup cq moderation params */
1089         for (i = 0; i < priv->rx_ring_num; i++) {
1090                 cq = priv->rx_cq[i];
1091                 cq->moder_cnt = priv->rx_frames;
1092                 cq->moder_time = priv->rx_usecs;
1093                 priv->last_moder_time[i] = MLX4_EN_AUTO_CONF;
1094                 priv->last_moder_packets[i] = 0;
1095                 priv->last_moder_bytes[i] = 0;
1096         }
1097
1098         for (i = 0; i < priv->tx_ring_num; i++) {
1099                 cq = priv->tx_cq[i];
1100                 cq->moder_cnt = priv->tx_frames;
1101                 cq->moder_time = priv->tx_usecs;
1102         }
1103
1104         /* Reset auto-moderation params */
1105         priv->pkt_rate_low = MLX4_EN_RX_RATE_LOW;
1106         priv->rx_usecs_low = MLX4_EN_RX_COAL_TIME_LOW;
1107         priv->pkt_rate_high = MLX4_EN_RX_RATE_HIGH;
1108         priv->rx_usecs_high = MLX4_EN_RX_COAL_TIME_HIGH;
1109         priv->sample_interval = MLX4_EN_SAMPLE_INTERVAL;
1110         priv->adaptive_rx_coal = 1;
1111         priv->last_moder_jiffies = 0;
1112         priv->last_moder_tx_packets = 0;
1113 }
1114
1115 static void mlx4_en_auto_moderation(struct mlx4_en_priv *priv)
1116 {
1117         unsigned long period = (unsigned long) (jiffies - priv->last_moder_jiffies);
1118         struct mlx4_en_cq *cq;
1119         unsigned long packets;
1120         unsigned long rate;
1121         unsigned long avg_pkt_size;
1122         unsigned long rx_packets;
1123         unsigned long rx_bytes;
1124         unsigned long rx_pkt_diff;
1125         int moder_time;
1126         int ring, err;
1127
1128         if (!priv->adaptive_rx_coal || period < priv->sample_interval * HZ)
1129                 return;
1130
1131         for (ring = 0; ring < priv->rx_ring_num; ring++) {
1132                 spin_lock(&priv->stats_lock);
1133                 rx_packets = priv->rx_ring[ring]->packets;
1134                 rx_bytes = priv->rx_ring[ring]->bytes;
1135                 spin_unlock(&priv->stats_lock);
1136
1137                 rx_pkt_diff = ((unsigned long) (rx_packets -
1138                                 priv->last_moder_packets[ring]));
1139                 packets = rx_pkt_diff;
1140                 rate = packets * HZ / period;
1141                 avg_pkt_size = packets ? ((unsigned long) (rx_bytes -
1142                                 priv->last_moder_bytes[ring])) / packets : 0;
1143
1144                 /* Apply auto-moderation only when packet rate
1145                  * exceeds a rate that it matters */
1146                 if (rate > (MLX4_EN_RX_RATE_THRESH / priv->rx_ring_num) &&
1147                     avg_pkt_size > MLX4_EN_AVG_PKT_SMALL) {
1148                         if (rate < priv->pkt_rate_low)
1149                                 moder_time = priv->rx_usecs_low;
1150                         else if (rate > priv->pkt_rate_high)
1151                                 moder_time = priv->rx_usecs_high;
1152                         else
1153                                 moder_time = (rate - priv->pkt_rate_low) *
1154                                         (priv->rx_usecs_high - priv->rx_usecs_low) /
1155                                         (priv->pkt_rate_high - priv->pkt_rate_low) +
1156                                         priv->rx_usecs_low;
1157                 } else {
1158                         moder_time = priv->rx_usecs_low;
1159                 }
1160
1161                 if (moder_time != priv->last_moder_time[ring]) {
1162                         priv->last_moder_time[ring] = moder_time;
1163                         cq = priv->rx_cq[ring];
1164                         cq->moder_time = moder_time;
1165                         cq->moder_cnt = priv->rx_frames;
1166                         err = mlx4_en_set_cq_moder(priv, cq);
1167                         if (err)
1168                                 en_err(priv, "Failed modifying moderation for cq:%d\n",
1169                                        ring);
1170                 }
1171                 priv->last_moder_packets[ring] = rx_packets;
1172                 priv->last_moder_bytes[ring] = rx_bytes;
1173         }
1174
1175         priv->last_moder_jiffies = jiffies;
1176 }
1177
1178 static void mlx4_en_do_get_stats(struct work_struct *work)
1179 {
1180         struct delayed_work *delay = to_delayed_work(work);
1181         struct mlx4_en_priv *priv = container_of(delay, struct mlx4_en_priv,
1182                                                  stats_task);
1183         struct mlx4_en_dev *mdev = priv->mdev;
1184         int err;
1185
1186         mutex_lock(&mdev->state_lock);
1187         if (mdev->device_up) {
1188                 if (priv->port_up) {
1189                         if (mlx4_is_slave(mdev->dev))
1190                                 err = mlx4_en_get_vport_stats(mdev, priv->port);
1191                         else
1192                                 err = mlx4_en_DUMP_ETH_STATS(mdev, priv->port, 0);
1193                         if (err)
1194                                 en_dbg(HW, priv, "Could not update stats\n");
1195
1196                         mlx4_en_auto_moderation(priv);
1197                 }
1198
1199                 queue_delayed_work(mdev->workqueue, &priv->stats_task, STATS_DELAY);
1200         }
1201         mutex_unlock(&mdev->state_lock);
1202 }
1203
1204 /* mlx4_en_service_task - Run service task for tasks that needed to be done
1205  * periodically
1206  */
1207 static void mlx4_en_service_task(struct work_struct *work)
1208 {
1209         struct delayed_work *delay = to_delayed_work(work);
1210         struct mlx4_en_priv *priv = container_of(delay, struct mlx4_en_priv,
1211                                                  service_task);
1212         struct mlx4_en_dev *mdev = priv->mdev;
1213
1214         mutex_lock(&mdev->state_lock);
1215         if (mdev->device_up) {
1216                 queue_delayed_work(mdev->workqueue, &priv->service_task,
1217                                    SERVICE_TASK_DELAY);
1218         }
1219         mutex_unlock(&mdev->state_lock);
1220 }
1221
1222 static void mlx4_en_linkstate(struct work_struct *work)
1223 {
1224         struct mlx4_en_priv *priv = container_of(work, struct mlx4_en_priv,
1225                                                  linkstate_task);
1226         struct mlx4_en_dev *mdev = priv->mdev;
1227         int linkstate = priv->link_state;
1228
1229         mutex_lock(&mdev->state_lock);
1230         /* If observable port state changed set carrier state and
1231          * report to system log */
1232         if (priv->last_link_state != linkstate) {
1233                 if (linkstate == MLX4_DEV_EVENT_PORT_DOWN) {
1234                         en_info(priv, "Link Down\n");
1235                         if_link_state_change(priv->dev, LINK_STATE_DOWN);
1236                         /* update netif baudrate */
1237                         priv->dev->if_baudrate = 0;
1238
1239                 /* make sure the port is up before notifying the OS.
1240                  * This is tricky since we get here on INIT_PORT and
1241                  * in such case we can't tell the OS the port is up.
1242                  * To solve this there is a call to if_link_state_change
1243                  * in set_rx_mode.
1244                  * */
1245                 } else if (priv->port_up && (linkstate == MLX4_DEV_EVENT_PORT_UP)){
1246                         if (mlx4_en_QUERY_PORT(priv->mdev, priv->port))
1247                                 en_info(priv, "Query port failed\n");
1248                         priv->dev->if_baudrate =
1249                             IF_Mbps(priv->port_state.link_speed);
1250                         en_info(priv, "Link Up\n");
1251                         if_link_state_change(priv->dev, LINK_STATE_UP);
1252                 }
1253         }
1254         priv->last_link_state = linkstate;
1255         mutex_unlock(&mdev->state_lock);
1256 }
1257
1258
1259 int mlx4_en_start_port(struct net_device *dev)
1260 {
1261         struct mlx4_en_priv *priv = netdev_priv(dev);
1262         struct mlx4_en_dev *mdev = priv->mdev;
1263         struct mlx4_en_cq *cq;
1264         struct mlx4_en_tx_ring *tx_ring;
1265         int rx_index = 0;
1266         int tx_index = 0;
1267         int err = 0;
1268         int i;
1269         int j;
1270         u8 mc_list[16] = {0};
1271
1272
1273         if (priv->port_up) {
1274                 en_dbg(DRV, priv, "start port called while port already up\n");
1275                 return 0;
1276         }
1277
1278         INIT_LIST_HEAD(&priv->mc_list);
1279         INIT_LIST_HEAD(&priv->uc_list);
1280         INIT_LIST_HEAD(&priv->curr_mc_list);
1281         INIT_LIST_HEAD(&priv->curr_uc_list);
1282         INIT_LIST_HEAD(&priv->ethtool_list);
1283
1284         /* Calculate Rx buf size */
1285         dev->if_mtu = min(dev->if_mtu, priv->max_mtu);
1286         mlx4_en_calc_rx_buf(dev);
1287         en_dbg(DRV, priv, "Rx buf size:%d\n", priv->rx_mb_size);
1288
1289         /* Configure rx cq's and rings */
1290         err = mlx4_en_activate_rx_rings(priv);
1291         if (err) {
1292                 en_err(priv, "Failed to activate RX rings\n");
1293                 return err;
1294         }
1295         for (i = 0; i < priv->rx_ring_num; i++) {
1296                 cq = priv->rx_cq[i];
1297
1298                 mlx4_en_cq_init_lock(cq);
1299                 err = mlx4_en_activate_cq(priv, cq, i);
1300                 if (err) {
1301                         en_err(priv, "Failed activating Rx CQ\n");
1302                         goto cq_err;
1303                 }
1304                 for (j = 0; j < cq->size; j++)
1305                         cq->buf[j].owner_sr_opcode = MLX4_CQE_OWNER_MASK;
1306                 err = mlx4_en_set_cq_moder(priv, cq);
1307                 if (err) {
1308                         en_err(priv, "Failed setting cq moderation parameters");
1309                         mlx4_en_deactivate_cq(priv, cq);
1310                         goto cq_err;
1311                 }
1312                 mlx4_en_arm_cq(priv, cq);
1313                 priv->rx_ring[i]->cqn = cq->mcq.cqn;
1314                 ++rx_index;
1315         }
1316
1317         /* Set qp number */
1318         en_dbg(DRV, priv, "Getting qp number for port %d\n", priv->port);
1319         err = mlx4_en_get_qp(priv);
1320         if (err) {
1321                 en_err(priv, "Failed getting eth qp\n");
1322                 goto cq_err;
1323         }
1324         mdev->mac_removed[priv->port] = 0;
1325
1326         priv->counter_index =
1327                         mlx4_get_default_counter_index(mdev->dev, priv->port);
1328
1329         err = mlx4_en_config_rss_steer(priv);
1330         if (err) {
1331                 en_err(priv, "Failed configuring rss steering\n");
1332                 goto mac_err;
1333         }
1334
1335         err = mlx4_en_create_drop_qp(priv);
1336         if (err)
1337                 goto rss_err;
1338
1339         /* Configure tx cq's and rings */
1340         for (i = 0; i < priv->tx_ring_num; i++) {
1341                 /* Configure cq */
1342                 cq = priv->tx_cq[i];
1343                 err = mlx4_en_activate_cq(priv, cq, i);
1344                 if (err) {
1345                         en_err(priv, "Failed activating Tx CQ\n");
1346                         goto tx_err;
1347                 }
1348                 err = mlx4_en_set_cq_moder(priv, cq);
1349                 if (err) {
1350                         en_err(priv, "Failed setting cq moderation parameters");
1351                         mlx4_en_deactivate_cq(priv, cq);
1352                         goto tx_err;
1353                 }
1354                 en_dbg(DRV, priv, "Resetting index of collapsed CQ:%d to -1\n", i);
1355                 cq->buf->wqe_index = cpu_to_be16(0xffff);
1356
1357                 /* Configure ring */
1358                 tx_ring = priv->tx_ring[i];
1359
1360                 err = mlx4_en_activate_tx_ring(priv, tx_ring, cq->mcq.cqn,
1361                                                i / priv->num_tx_rings_p_up);
1362                 if (err) {
1363                         en_err(priv, "Failed activating Tx ring %d\n", i);
1364                         mlx4_en_deactivate_cq(priv, cq);
1365                         goto tx_err;
1366                 }
1367
1368                 /* Arm CQ for TX completions */
1369                 mlx4_en_arm_cq(priv, cq);
1370
1371                 /* Set initial ownership of all Tx TXBBs to SW (1) */
1372                 for (j = 0; j < tx_ring->buf_size; j += STAMP_STRIDE)
1373                         *((u32 *) (tx_ring->buf + j)) = INIT_OWNER_BIT;
1374                 ++tx_index;
1375         }
1376
1377         /* Configure port */
1378         err = mlx4_SET_PORT_general(mdev->dev, priv->port,
1379                                     priv->rx_mb_size,
1380                                     priv->prof->tx_pause,
1381                                     priv->prof->tx_ppp,
1382                                     priv->prof->rx_pause,
1383                                     priv->prof->rx_ppp);
1384         if (err) {
1385                 en_err(priv, "Failed setting port general configurations for port %d, with error %d\n",
1386                        priv->port, err);
1387                 goto tx_err;
1388         }
1389         /* Set default qp number */
1390         err = mlx4_SET_PORT_qpn_calc(mdev->dev, priv->port, priv->base_qpn, 0);
1391         if (err) {
1392                 en_err(priv, "Failed setting default qp numbers\n");
1393                 goto tx_err;
1394         }
1395
1396         /* Init port */
1397         en_dbg(HW, priv, "Initializing port\n");
1398         err = mlx4_INIT_PORT(mdev->dev, priv->port);
1399         if (err) {
1400                 en_err(priv, "Failed Initializing port\n");
1401                 goto tx_err;
1402         }
1403
1404         /* Attach rx QP to bradcast address */
1405         memset(&mc_list[10], 0xff, ETH_ALEN);
1406         mc_list[5] = priv->port; /* needed for B0 steering support */
1407         if (mlx4_multicast_attach(mdev->dev, &priv->rss_map.indir_qp, mc_list,
1408                                   priv->port, 0, MLX4_PROT_ETH,
1409                                   &priv->broadcast_id))
1410                 mlx4_warn(mdev, "Failed Attaching Broadcast\n");
1411
1412         /* Must redo promiscuous mode setup. */
1413         priv->flags &= ~(MLX4_EN_FLAG_PROMISC | MLX4_EN_FLAG_MC_PROMISC);
1414
1415         /* Schedule multicast task to populate multicast list */
1416         queue_work(mdev->workqueue, &priv->rx_mode_task);
1417
1418         priv->port_up = true;
1419
1420         /* Enable the queues. */
1421         dev->if_drv_flags &= ~IFF_DRV_OACTIVE;
1422         dev->if_drv_flags |= IFF_DRV_RUNNING;
1423 #ifdef CONFIG_DEBUG_FS
1424         mlx4_en_create_debug_files(priv);
1425 #endif
1426         callout_reset(&priv->watchdog_timer, MLX4_EN_WATCHDOG_TIMEOUT,
1427                     mlx4_en_watchdog_timeout, priv);
1428
1429
1430         return 0;
1431
1432 tx_err:
1433         while (tx_index--) {
1434                 mlx4_en_deactivate_tx_ring(priv, priv->tx_ring[tx_index]);
1435                 mlx4_en_deactivate_cq(priv, priv->tx_cq[tx_index]);
1436         }
1437         mlx4_en_destroy_drop_qp(priv);
1438 rss_err:
1439         mlx4_en_release_rss_steer(priv);
1440 mac_err:
1441         mlx4_en_put_qp(priv);
1442 cq_err:
1443         while (rx_index--)
1444                 mlx4_en_deactivate_cq(priv, priv->rx_cq[rx_index]);
1445         for (i = 0; i < priv->rx_ring_num; i++)
1446                 mlx4_en_deactivate_rx_ring(priv, priv->rx_ring[i]);
1447
1448         return err; /* need to close devices */
1449 }
1450
1451
1452 void mlx4_en_stop_port(struct net_device *dev)
1453 {
1454         struct mlx4_en_priv *priv = netdev_priv(dev);
1455         struct mlx4_en_dev *mdev = priv->mdev;
1456         struct mlx4_en_addr_list *addr_list, *tmp;
1457         int i;
1458         u8 mc_list[16] = {0};
1459
1460         if (!priv->port_up) {
1461                 en_dbg(DRV, priv, "stop port called while port already down\n");
1462                 return;
1463         }
1464
1465 #ifdef CONFIG_DEBUG_FS
1466         mlx4_en_delete_debug_files(priv);
1467 #endif
1468
1469         /* close port*/
1470         mlx4_CLOSE_PORT(mdev->dev, priv->port);
1471
1472         /* Set port as not active */
1473         priv->port_up = false;
1474         priv->counter_index = MLX4_SINK_COUNTER_INDEX(mdev->dev);
1475
1476         /* Promsicuous mode */
1477         if (mdev->dev->caps.steering_mode ==
1478             MLX4_STEERING_MODE_DEVICE_MANAGED) {
1479                 priv->flags &= ~(MLX4_EN_FLAG_PROMISC |
1480                                  MLX4_EN_FLAG_MC_PROMISC);
1481                 mlx4_flow_steer_promisc_remove(mdev->dev,
1482                                                priv->port,
1483                                                MLX4_FS_ALL_DEFAULT);
1484                 mlx4_flow_steer_promisc_remove(mdev->dev,
1485                                                priv->port,
1486                                                MLX4_FS_MC_DEFAULT);
1487         } else if (priv->flags & MLX4_EN_FLAG_PROMISC) {
1488                 priv->flags &= ~MLX4_EN_FLAG_PROMISC;
1489
1490                 /* Disable promiscouos mode */
1491                 mlx4_unicast_promisc_remove(mdev->dev, priv->base_qpn,
1492                                             priv->port);
1493
1494                 /* Disable Multicast promisc */
1495                 if (priv->flags & MLX4_EN_FLAG_MC_PROMISC) {
1496                         mlx4_multicast_promisc_remove(mdev->dev, priv->base_qpn,
1497                                                       priv->port);
1498                         priv->flags &= ~MLX4_EN_FLAG_MC_PROMISC;
1499                 }
1500         }
1501
1502         /* Detach All unicasts */
1503         list_for_each_entry(addr_list, &priv->curr_uc_list, list) {
1504                 mlx4_en_uc_steer_release(priv, addr_list->addr,
1505                                          priv->rss_map.indir_qp.qpn,
1506                                          addr_list->reg_id);
1507         }
1508         mlx4_en_clear_uclist(dev);
1509         list_for_each_entry_safe(addr_list, tmp, &priv->curr_uc_list, list) {
1510                 list_del(&addr_list->list);
1511                 kfree(addr_list);
1512         }
1513
1514         /* Detach All multicasts */
1515         memset(&mc_list[10], 0xff, ETH_ALEN);
1516         mc_list[5] = priv->port; /* needed for B0 steering support */
1517         mlx4_multicast_detach(mdev->dev, &priv->rss_map.indir_qp, mc_list,
1518                               MLX4_PROT_ETH, priv->broadcast_id);
1519         list_for_each_entry(addr_list, &priv->curr_mc_list, list) {
1520                 memcpy(&mc_list[10], addr_list->addr, ETH_ALEN);
1521                 mc_list[5] = priv->port;
1522                 mlx4_multicast_detach(mdev->dev, &priv->rss_map.indir_qp,
1523                                       mc_list, MLX4_PROT_ETH, addr_list->reg_id);
1524         }
1525         mlx4_en_clear_mclist(dev);
1526         list_for_each_entry_safe(addr_list, tmp, &priv->curr_mc_list, list) {
1527                 list_del(&addr_list->list);
1528                 kfree(addr_list);
1529         }
1530
1531         /* Flush multicast filter */
1532         mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, 0, 1, MLX4_MCAST_CONFIG);
1533         mlx4_en_destroy_drop_qp(priv);
1534
1535         /* Free TX Rings */
1536         for (i = 0; i < priv->tx_ring_num; i++) {
1537                 mlx4_en_deactivate_tx_ring(priv, priv->tx_ring[i]);
1538                 mlx4_en_deactivate_cq(priv, priv->tx_cq[i]);
1539         }
1540         msleep(10);
1541
1542         for (i = 0; i < priv->tx_ring_num; i++)
1543                 mlx4_en_free_tx_buf(dev, priv->tx_ring[i]);
1544
1545         /* Free RSS qps */
1546         mlx4_en_release_rss_steer(priv);
1547
1548         /* Unregister Mac address for the port */
1549         mlx4_en_put_qp(priv);
1550         mdev->mac_removed[priv->port] = 1;
1551
1552         /* Free RX Rings */
1553         for (i = 0; i < priv->rx_ring_num; i++) {
1554                 struct mlx4_en_cq *cq = priv->rx_cq[i];
1555                 mlx4_en_deactivate_rx_ring(priv, priv->rx_ring[i]);
1556                 mlx4_en_deactivate_cq(priv, cq);
1557         }
1558
1559         callout_stop(&priv->watchdog_timer);
1560
1561         dev->if_drv_flags &= ~(IFF_DRV_RUNNING | IFF_DRV_OACTIVE);
1562 }
1563
1564 static void mlx4_en_restart(struct work_struct *work)
1565 {
1566         struct mlx4_en_priv *priv = container_of(work, struct mlx4_en_priv,
1567                                                  watchdog_task);
1568         struct mlx4_en_dev *mdev = priv->mdev;
1569         struct net_device *dev = priv->dev;
1570         struct mlx4_en_tx_ring *ring;
1571         int i;
1572
1573
1574         if (priv->blocked == 0 || priv->port_up == 0)
1575                 return;
1576         for (i = 0; i < priv->tx_ring_num; i++) {
1577                 int watchdog_time;
1578
1579                 ring = priv->tx_ring[i];
1580                 watchdog_time = READ_ONCE(ring->watchdog_time);
1581                 if (watchdog_time != 0 &&
1582                     time_after(ticks, ring->watchdog_time))
1583                         goto reset;
1584         }
1585         return;
1586
1587 reset:
1588         priv->port_stats.tx_timeout++;
1589         en_dbg(DRV, priv, "Watchdog task called for port %d\n", priv->port);
1590
1591         mutex_lock(&mdev->state_lock);
1592         if (priv->port_up) {
1593                 mlx4_en_stop_port(dev);
1594                 //for (i = 0; i < priv->tx_ring_num; i++)
1595                 //        netdev_tx_reset_queue(priv->tx_ring[i]->tx_queue);
1596                 if (mlx4_en_start_port(dev))
1597                         en_err(priv, "Failed restarting port %d\n", priv->port);
1598         }
1599         mutex_unlock(&mdev->state_lock);
1600 }
1601
1602 static void mlx4_en_clear_stats(struct net_device *dev)
1603 {
1604         struct mlx4_en_priv *priv = netdev_priv(dev);
1605         struct mlx4_en_dev *mdev = priv->mdev;
1606         int i;
1607
1608         if (!mlx4_is_slave(mdev->dev))
1609                 if (mlx4_en_DUMP_ETH_STATS(mdev, priv->port, 1))
1610                         en_dbg(HW, priv, "Failed dumping statistics\n");
1611
1612         memset(&priv->pstats, 0, sizeof(priv->pstats));
1613         memset(&priv->pkstats, 0, sizeof(priv->pkstats));
1614         memset(&priv->port_stats, 0, sizeof(priv->port_stats));
1615         memset(&priv->vport_stats, 0, sizeof(priv->vport_stats));
1616
1617         for (i = 0; i < priv->tx_ring_num; i++) {
1618                 priv->tx_ring[i]->bytes = 0;
1619                 priv->tx_ring[i]->packets = 0;
1620                 priv->tx_ring[i]->tx_csum = 0;
1621                 priv->tx_ring[i]->oversized_packets = 0;
1622         }
1623         for (i = 0; i < priv->rx_ring_num; i++) {
1624                 priv->rx_ring[i]->bytes = 0;
1625                 priv->rx_ring[i]->packets = 0;
1626                 priv->rx_ring[i]->csum_ok = 0;
1627                 priv->rx_ring[i]->csum_none = 0;
1628         }
1629 }
1630
1631 static void mlx4_en_open(void* arg)
1632 {
1633
1634         struct mlx4_en_priv *priv;
1635         struct mlx4_en_dev *mdev;
1636         struct net_device *dev;
1637         int err = 0;
1638
1639         priv = arg;
1640         mdev = priv->mdev;
1641         dev = priv->dev;
1642
1643
1644         mutex_lock(&mdev->state_lock);
1645
1646         if (!mdev->device_up) {
1647                 en_err(priv, "Cannot open - device down/disabled\n");
1648                 goto out;
1649         }
1650
1651         /* Reset HW statistics and SW counters */
1652         mlx4_en_clear_stats(dev);
1653
1654         err = mlx4_en_start_port(dev);
1655         if (err)
1656                 en_err(priv, "Failed starting port:%d\n", priv->port);
1657
1658 out:
1659         mutex_unlock(&mdev->state_lock);
1660         return;
1661 }
1662
1663 void mlx4_en_free_resources(struct mlx4_en_priv *priv)
1664 {
1665         int i;
1666
1667 #ifdef CONFIG_RFS_ACCEL
1668         if (priv->dev->rx_cpu_rmap) {
1669                 free_irq_cpu_rmap(priv->dev->rx_cpu_rmap);
1670                 priv->dev->rx_cpu_rmap = NULL;
1671         }
1672 #endif
1673
1674         for (i = 0; i < priv->tx_ring_num; i++) {
1675                 if (priv->tx_ring && priv->tx_ring[i])
1676                         mlx4_en_destroy_tx_ring(priv, &priv->tx_ring[i]);
1677                 if (priv->tx_cq && priv->tx_cq[i])
1678                         mlx4_en_destroy_cq(priv, &priv->tx_cq[i]);
1679         }
1680
1681         for (i = 0; i < priv->rx_ring_num; i++) {
1682                 if (priv->rx_ring[i])
1683                         mlx4_en_destroy_rx_ring(priv, &priv->rx_ring[i],
1684                                 priv->prof->rx_ring_size);
1685                 if (priv->rx_cq[i])
1686                         mlx4_en_destroy_cq(priv, &priv->rx_cq[i]);
1687         }
1688
1689         if (priv->stat_sysctl != NULL)
1690                 sysctl_ctx_free(&priv->stat_ctx);
1691 }
1692
1693 int mlx4_en_alloc_resources(struct mlx4_en_priv *priv)
1694 {
1695         struct mlx4_en_port_profile *prof = priv->prof;
1696         int i;
1697         int node = 0;
1698
1699         /* Create rx Rings */
1700         for (i = 0; i < priv->rx_ring_num; i++) {
1701                 if (mlx4_en_create_cq(priv, &priv->rx_cq[i],
1702                                       prof->rx_ring_size, i, RX, node))
1703                         goto err;
1704
1705                 if (mlx4_en_create_rx_ring(priv, &priv->rx_ring[i],
1706                                            prof->rx_ring_size, node))
1707                         goto err;
1708         }
1709
1710         /* Create tx Rings */
1711         for (i = 0; i < priv->tx_ring_num; i++) {
1712                 if (mlx4_en_create_cq(priv, &priv->tx_cq[i],
1713                                       prof->tx_ring_size, i, TX, node))
1714                         goto err;
1715
1716                 if (mlx4_en_create_tx_ring(priv, &priv->tx_ring[i],
1717                                            prof->tx_ring_size, TXBB_SIZE, node, i))
1718                         goto err;
1719         }
1720
1721 #ifdef CONFIG_RFS_ACCEL
1722         priv->dev->rx_cpu_rmap = alloc_irq_cpu_rmap(priv->rx_ring_num);
1723         if (!priv->dev->rx_cpu_rmap)
1724                 goto err;
1725 #endif
1726         /* Re-create stat sysctls in case the number of rings changed. */
1727         mlx4_en_sysctl_stat(priv);
1728         return 0;
1729
1730 err:
1731         en_err(priv, "Failed to allocate NIC resources\n");
1732         for (i = 0; i < priv->rx_ring_num; i++) {
1733                 if (priv->rx_ring[i])
1734                         mlx4_en_destroy_rx_ring(priv, &priv->rx_ring[i],
1735                                                 prof->rx_ring_size);
1736                 if (priv->rx_cq[i])
1737                         mlx4_en_destroy_cq(priv, &priv->rx_cq[i]);
1738         }
1739         for (i = 0; i < priv->tx_ring_num; i++) {
1740                 if (priv->tx_ring[i])
1741                         mlx4_en_destroy_tx_ring(priv, &priv->tx_ring[i]);
1742                 if (priv->tx_cq[i])
1743                         mlx4_en_destroy_cq(priv, &priv->tx_cq[i]);
1744         }
1745         priv->port_up = false;
1746         return -ENOMEM;
1747 }
1748
1749 struct en_port_attribute {
1750         struct attribute attr;
1751         ssize_t (*show)(struct en_port *, struct en_port_attribute *, char *buf);
1752         ssize_t (*store)(struct en_port *, struct en_port_attribute *, char *buf, size_t count);
1753 };
1754
1755 #define PORT_ATTR_RO(_name) \
1756 struct en_port_attribute en_port_attr_##_name = __ATTR_RO(_name)
1757
1758 #define EN_PORT_ATTR(_name, _mode, _show, _store) \
1759 struct en_port_attribute en_port_attr_##_name = __ATTR(_name, _mode, _show, _store)
1760
1761 void mlx4_en_destroy_netdev(struct net_device *dev)
1762 {
1763         struct mlx4_en_priv *priv = netdev_priv(dev);
1764         struct mlx4_en_dev *mdev = priv->mdev;
1765
1766         en_dbg(DRV, priv, "Destroying netdev on port:%d\n", priv->port);
1767
1768         /* don't allow more IOCTLs */
1769         priv->gone = 1;
1770
1771         /* XXX wait a bit to allow IOCTL handlers to complete */
1772         pause("W", hz);
1773
1774         if (priv->vlan_attach != NULL)
1775                 EVENTHANDLER_DEREGISTER(vlan_config, priv->vlan_attach);
1776         if (priv->vlan_detach != NULL)
1777                 EVENTHANDLER_DEREGISTER(vlan_unconfig, priv->vlan_detach);
1778
1779         mutex_lock(&mdev->state_lock);
1780         mlx4_en_stop_port(dev);
1781         mutex_unlock(&mdev->state_lock);
1782
1783         /* Unregister device - this will close the port if it was up */
1784         if (priv->registered)
1785                 ether_ifdetach(dev);
1786
1787         if (priv->allocated)
1788                 mlx4_free_hwq_res(mdev->dev, &priv->res, MLX4_EN_PAGE_SIZE);
1789
1790         cancel_delayed_work(&priv->stats_task);
1791         cancel_delayed_work(&priv->service_task);
1792         /* flush any pending task for this netdev */
1793         flush_workqueue(mdev->workqueue);
1794         callout_drain(&priv->watchdog_timer);
1795
1796         /* Detach the netdev so tasks would not attempt to access it */
1797         mutex_lock(&mdev->state_lock);
1798         mdev->pndev[priv->port] = NULL;
1799         mutex_unlock(&mdev->state_lock);
1800
1801
1802         mlx4_en_free_resources(priv);
1803
1804         /* freeing the sysctl conf cannot be called from within mlx4_en_free_resources */
1805         if (priv->conf_sysctl != NULL)
1806                 sysctl_ctx_free(&priv->conf_ctx);
1807
1808         kfree(priv->tx_ring);
1809         kfree(priv->tx_cq);
1810
1811         kfree(priv);
1812         if_free(dev);
1813
1814 }
1815
1816 static int mlx4_en_change_mtu(struct net_device *dev, int new_mtu)
1817 {
1818         struct mlx4_en_priv *priv = netdev_priv(dev);
1819         struct mlx4_en_dev *mdev = priv->mdev;
1820         int err = 0;
1821
1822         en_dbg(DRV, priv, "Change MTU called - current:%u new:%u\n",
1823                (unsigned)dev->if_mtu, (unsigned)new_mtu);
1824
1825         if ((new_mtu < MLX4_EN_MIN_MTU) || (new_mtu > priv->max_mtu)) {
1826                 en_err(priv, "Bad MTU size:%d, max %u.\n", new_mtu,
1827                     priv->max_mtu);
1828                 return -EPERM;
1829         }
1830         mutex_lock(&mdev->state_lock);
1831         dev->if_mtu = new_mtu;
1832         if (dev->if_drv_flags & IFF_DRV_RUNNING) {
1833                 if (!mdev->device_up) {
1834                         /* NIC is probably restarting - let watchdog task reset
1835                          *                          * the port */
1836                         en_dbg(DRV, priv, "Change MTU called with card down!?\n");
1837                 } else {
1838                         mlx4_en_stop_port(dev);
1839                         err = mlx4_en_start_port(dev);
1840                         if (err) {
1841                                 en_err(priv, "Failed restarting port:%d\n",
1842                                                 priv->port);
1843                                 queue_work(mdev->workqueue, &priv->watchdog_task);
1844                         }
1845                 }
1846         }
1847         mutex_unlock(&mdev->state_lock);
1848         return 0;
1849 }
1850
1851 static int mlx4_en_calc_media(struct mlx4_en_priv *priv)
1852 {
1853         int trans_type;
1854         int active;
1855
1856         active = IFM_ETHER;
1857         if (priv->last_link_state == MLX4_DEV_EVENT_PORT_DOWN)
1858                 return (active);
1859         active |= IFM_FDX;
1860         trans_type = priv->port_state.transceiver;
1861         /* XXX I don't know all of the transceiver values. */
1862         switch (priv->port_state.link_speed) {
1863         case 100:
1864                 active |= IFM_100_T;
1865                 break;
1866         case 1000:
1867                 active |= IFM_1000_T;
1868                 break;
1869         case 10000:
1870                 if (trans_type > 0 && trans_type <= 0xC)
1871                         active |= IFM_10G_SR;
1872                 else if (trans_type == 0x80 || trans_type == 0)
1873                         active |= IFM_10G_CX4;
1874                 break;
1875         case 40000:
1876                 active |= IFM_40G_CR4;
1877                 break;
1878         }
1879         if (priv->prof->tx_pause)
1880                 active |= IFM_ETH_TXPAUSE;
1881         if (priv->prof->rx_pause)
1882                 active |= IFM_ETH_RXPAUSE;
1883
1884         return (active);
1885 }
1886
1887 static void mlx4_en_media_status(struct ifnet *dev, struct ifmediareq *ifmr)
1888 {
1889         struct mlx4_en_priv *priv;
1890
1891         priv = dev->if_softc;
1892         ifmr->ifm_status = IFM_AVALID;
1893         if (priv->last_link_state != MLX4_DEV_EVENT_PORT_DOWN)
1894                 ifmr->ifm_status |= IFM_ACTIVE;
1895         ifmr->ifm_active = mlx4_en_calc_media(priv);
1896
1897         return;
1898 }
1899
1900 static int mlx4_en_media_change(struct ifnet *dev)
1901 {
1902         struct mlx4_en_priv *priv;
1903         struct ifmedia *ifm;
1904         int rxpause;
1905         int txpause;
1906         int error;
1907
1908         priv = dev->if_softc;
1909         ifm = &priv->media;
1910         rxpause = txpause = 0;
1911         error = 0;
1912
1913         if (IFM_TYPE(ifm->ifm_media) != IFM_ETHER)
1914                 return (EINVAL);
1915         switch (IFM_SUBTYPE(ifm->ifm_media)) {
1916         case IFM_AUTO:
1917                 break;
1918         case IFM_10G_SR:
1919         case IFM_10G_CX4:
1920         case IFM_1000_T:
1921         case IFM_40G_CR4:
1922                 if ((IFM_SUBTYPE(ifm->ifm_media)
1923                         == IFM_SUBTYPE(mlx4_en_calc_media(priv)))
1924                         && (ifm->ifm_media & IFM_FDX))
1925                         break;
1926                 /* Fallthrough */
1927         default:
1928                 printf("%s: Only auto media type\n", if_name(dev));
1929                 return (EINVAL);
1930         }
1931         /* Allow user to set/clear pause */
1932         if (IFM_OPTIONS(ifm->ifm_media) & IFM_ETH_RXPAUSE)
1933                 rxpause = 1;
1934         if (IFM_OPTIONS(ifm->ifm_media) & IFM_ETH_TXPAUSE)
1935                 txpause = 1;
1936         if (priv->prof->tx_pause != txpause || priv->prof->rx_pause != rxpause) {
1937                 priv->prof->tx_pause = txpause;
1938                 priv->prof->rx_pause = rxpause;
1939                 error = -mlx4_SET_PORT_general(priv->mdev->dev, priv->port,
1940                      priv->rx_mb_size + ETHER_CRC_LEN, priv->prof->tx_pause,
1941                      priv->prof->tx_ppp, priv->prof->rx_pause,
1942                      priv->prof->rx_ppp);
1943         }
1944         return (error);
1945 }
1946
1947 static int mlx4_en_ioctl(struct ifnet *dev, u_long command, caddr_t data)
1948 {
1949         struct mlx4_en_priv *priv;
1950         struct mlx4_en_dev *mdev;
1951         struct ifreq *ifr;
1952         int error;
1953         int mask;
1954         struct ifrsskey *ifrk;
1955         const u32 *key;
1956         struct ifrsshash *ifrh;
1957         u8 rss_mask;
1958
1959         error = 0;
1960         mask = 0;
1961         priv = dev->if_softc;
1962
1963         /* check if detaching */
1964         if (priv == NULL || priv->gone != 0)
1965                 return (ENXIO);
1966
1967         mdev = priv->mdev;
1968         ifr = (struct ifreq *) data;
1969
1970         switch (command) {
1971         case SIOCSIFMTU:
1972                 error = -mlx4_en_change_mtu(dev, ifr->ifr_mtu);
1973                 break;
1974         case SIOCSIFFLAGS:
1975                 if (dev->if_flags & IFF_UP) {
1976                         if ((dev->if_drv_flags & IFF_DRV_RUNNING) == 0) {
1977                                 mutex_lock(&mdev->state_lock);
1978                                 mlx4_en_start_port(dev);
1979                                 mutex_unlock(&mdev->state_lock);
1980                         } else {
1981                                 mlx4_en_set_rx_mode(dev);
1982                         }
1983                 } else {
1984                         mutex_lock(&mdev->state_lock);
1985                         if (dev->if_drv_flags & IFF_DRV_RUNNING) {
1986                                 mlx4_en_stop_port(dev);
1987                                 if_link_state_change(dev, LINK_STATE_DOWN);
1988                         }
1989                         mutex_unlock(&mdev->state_lock);
1990                 }
1991                 break;
1992         case SIOCADDMULTI:
1993         case SIOCDELMULTI:
1994                 mlx4_en_set_rx_mode(dev);
1995                 break;
1996         case SIOCSIFMEDIA:
1997         case SIOCGIFMEDIA:
1998                 error = ifmedia_ioctl(dev, ifr, &priv->media, command);
1999                 break;
2000         case SIOCSIFCAP:
2001                 mutex_lock(&mdev->state_lock);
2002                 mask = ifr->ifr_reqcap ^ dev->if_capenable;
2003                 if (mask & IFCAP_TXCSUM) {
2004                         dev->if_capenable ^= IFCAP_TXCSUM;
2005                         dev->if_hwassist ^= (CSUM_TCP | CSUM_UDP | CSUM_IP);
2006
2007                         if (IFCAP_TSO4 & dev->if_capenable &&
2008                             !(IFCAP_TXCSUM & dev->if_capenable)) {
2009                                 dev->if_capenable &= ~IFCAP_TSO4;
2010                                 dev->if_hwassist &= ~CSUM_IP_TSO;
2011                                 if_printf(dev,
2012                                     "tso4 disabled due to -txcsum.\n");
2013                         }
2014                 }
2015                 if (mask & IFCAP_TXCSUM_IPV6) {
2016                         dev->if_capenable ^= IFCAP_TXCSUM_IPV6;
2017                         dev->if_hwassist ^= (CSUM_UDP_IPV6 | CSUM_TCP_IPV6);
2018
2019                         if (IFCAP_TSO6 & dev->if_capenable &&
2020                             !(IFCAP_TXCSUM_IPV6 & dev->if_capenable)) {
2021                                 dev->if_capenable &= ~IFCAP_TSO6;
2022                                 dev->if_hwassist &= ~CSUM_IP6_TSO;
2023                                 if_printf(dev,
2024                                     "tso6 disabled due to -txcsum6.\n");
2025                         }
2026                 }
2027                 if (mask & IFCAP_RXCSUM)
2028                         dev->if_capenable ^= IFCAP_RXCSUM;
2029                 if (mask & IFCAP_RXCSUM_IPV6)
2030                         dev->if_capenable ^= IFCAP_RXCSUM_IPV6;
2031
2032                 if (mask & IFCAP_TSO4) {
2033                         if (!(IFCAP_TSO4 & dev->if_capenable) &&
2034                             !(IFCAP_TXCSUM & dev->if_capenable)) {
2035                                 if_printf(dev, "enable txcsum first.\n");
2036                                 error = EAGAIN;
2037                                 goto out;
2038                         }
2039                         dev->if_capenable ^= IFCAP_TSO4;
2040                         dev->if_hwassist ^= CSUM_IP_TSO;
2041                 }
2042                 if (mask & IFCAP_TSO6) {
2043                         if (!(IFCAP_TSO6 & dev->if_capenable) &&
2044                             !(IFCAP_TXCSUM_IPV6 & dev->if_capenable)) {
2045                                 if_printf(dev, "enable txcsum6 first.\n");
2046                                 error = EAGAIN;
2047                                 goto out;
2048                         }
2049                         dev->if_capenable ^= IFCAP_TSO6;
2050                         dev->if_hwassist ^= CSUM_IP6_TSO;
2051                 }
2052                 if (mask & IFCAP_LRO)
2053                         dev->if_capenable ^= IFCAP_LRO;
2054                 if (mask & IFCAP_VLAN_HWTAGGING)
2055                         dev->if_capenable ^= IFCAP_VLAN_HWTAGGING;
2056                 if (mask & IFCAP_VLAN_HWFILTER)
2057                         dev->if_capenable ^= IFCAP_VLAN_HWFILTER;
2058                 if (mask & IFCAP_WOL_MAGIC)
2059                         dev->if_capenable ^= IFCAP_WOL_MAGIC;
2060                 if (dev->if_drv_flags & IFF_DRV_RUNNING)
2061                         mlx4_en_start_port(dev);
2062 out:
2063                 mutex_unlock(&mdev->state_lock);
2064                 VLAN_CAPABILITIES(dev);
2065                 break;
2066 #if __FreeBSD_version >= 1100036
2067         case SIOCGI2C: {
2068                 struct ifi2creq i2c;
2069
2070                 error = copyin(ifr_data_get_ptr(ifr), &i2c, sizeof(i2c));
2071                 if (error)
2072                         break;
2073                 if (i2c.len > sizeof(i2c.data)) {
2074                         error = EINVAL;
2075                         break;
2076                 }
2077                 /*
2078                  * Note that we ignore i2c.addr here. The driver hardcodes
2079                  * the address to 0x50, while standard expects it to be 0xA0.
2080                  */
2081                 error = mlx4_get_module_info(mdev->dev, priv->port,
2082                     i2c.offset, i2c.len, i2c.data);
2083                 if (error < 0) {
2084                         error = -error;
2085                         break;
2086                 }
2087                 error = copyout(&i2c, ifr_data_get_ptr(ifr), sizeof(i2c));
2088                 break;
2089         }
2090 #endif
2091         case SIOCGIFRSSKEY:
2092                 ifrk = (struct ifrsskey *)data;
2093                 ifrk->ifrk_func = RSS_FUNC_TOEPLITZ;
2094                 mutex_lock(&mdev->state_lock);
2095                 key = mlx4_en_get_rss_key(priv, &ifrk->ifrk_keylen);
2096                 if (ifrk->ifrk_keylen > RSS_KEYLEN)
2097                         error = EINVAL;
2098                 else
2099                         memcpy(ifrk->ifrk_key, key, ifrk->ifrk_keylen);
2100                 mutex_unlock(&mdev->state_lock);
2101                 break;
2102
2103         case SIOCGIFRSSHASH:
2104                 mutex_lock(&mdev->state_lock);
2105                 rss_mask = mlx4_en_get_rss_mask(priv);
2106                 mutex_unlock(&mdev->state_lock);
2107                 ifrh = (struct ifrsshash *)data;
2108                 ifrh->ifrh_func = RSS_FUNC_TOEPLITZ;
2109                 ifrh->ifrh_types = 0;
2110                 if (rss_mask & MLX4_RSS_IPV4)
2111                         ifrh->ifrh_types |= RSS_TYPE_IPV4;
2112                 if (rss_mask & MLX4_RSS_TCP_IPV4)
2113                         ifrh->ifrh_types |= RSS_TYPE_TCP_IPV4;
2114                 if (rss_mask & MLX4_RSS_IPV6)
2115                         ifrh->ifrh_types |= RSS_TYPE_IPV6;
2116                 if (rss_mask & MLX4_RSS_TCP_IPV6)
2117                         ifrh->ifrh_types |= RSS_TYPE_TCP_IPV6;
2118                 if (rss_mask & MLX4_RSS_UDP_IPV4)
2119                         ifrh->ifrh_types |= RSS_TYPE_UDP_IPV4;
2120                 if (rss_mask & MLX4_RSS_UDP_IPV6)
2121                         ifrh->ifrh_types |= RSS_TYPE_UDP_IPV6;
2122                 break;
2123
2124         default:
2125                 error = ether_ioctl(dev, command, data);
2126                 break;
2127         }
2128
2129         return (error);
2130 }
2131
2132
2133 int mlx4_en_init_netdev(struct mlx4_en_dev *mdev, int port,
2134                         struct mlx4_en_port_profile *prof)
2135 {
2136         struct net_device *dev;
2137         struct mlx4_en_priv *priv;
2138         uint8_t dev_addr[ETHER_ADDR_LEN];
2139         int err;
2140         int i;
2141
2142         priv = kzalloc(sizeof(*priv), GFP_KERNEL);
2143         dev = priv->dev = if_alloc(IFT_ETHER);
2144         if (dev == NULL) {
2145                 en_err(priv, "Net device allocation failed\n");
2146                 kfree(priv);
2147                 return -ENOMEM;
2148         }
2149         dev->if_softc = priv;
2150         if_initname(dev, "mlxen", (device_get_unit(
2151             mdev->pdev->dev.bsddev) * MLX4_MAX_PORTS) + port - 1);
2152         dev->if_mtu = ETHERMTU;
2153         dev->if_init = mlx4_en_open;
2154         dev->if_flags = IFF_BROADCAST | IFF_SIMPLEX | IFF_MULTICAST;
2155         dev->if_ioctl = mlx4_en_ioctl;
2156         dev->if_transmit = mlx4_en_transmit;
2157         dev->if_qflush = mlx4_en_qflush;
2158         dev->if_snd.ifq_maxlen = prof->tx_ring_size;
2159
2160         /*
2161          * Initialize driver private data
2162          */
2163         priv->counter_index = 0xff;
2164         spin_lock_init(&priv->stats_lock);
2165         INIT_WORK(&priv->rx_mode_task, mlx4_en_do_set_rx_mode);
2166         INIT_WORK(&priv->watchdog_task, mlx4_en_restart);
2167         INIT_WORK(&priv->linkstate_task, mlx4_en_linkstate);
2168         INIT_DELAYED_WORK(&priv->stats_task, mlx4_en_do_get_stats);
2169         INIT_DELAYED_WORK(&priv->service_task, mlx4_en_service_task);
2170         callout_init(&priv->watchdog_timer, 1);
2171 #ifdef CONFIG_RFS_ACCEL
2172         INIT_LIST_HEAD(&priv->filters);
2173         spin_lock_init(&priv->filters_lock);
2174 #endif
2175
2176         priv->msg_enable = MLX4_EN_MSG_LEVEL;
2177         priv->dev = dev;
2178         priv->mdev = mdev;
2179         priv->ddev = &mdev->pdev->dev;
2180         priv->prof = prof;
2181         priv->port = port;
2182         priv->port_up = false;
2183         priv->flags = prof->flags;
2184
2185         priv->num_tx_rings_p_up = mdev->profile.num_tx_rings_p_up;
2186         priv->tx_ring_num = prof->tx_ring_num;
2187         priv->tx_ring = kcalloc(MAX_TX_RINGS,
2188                                 sizeof(struct mlx4_en_tx_ring *), GFP_KERNEL);
2189         if (!priv->tx_ring) {
2190                 err = -ENOMEM;
2191                 goto out;
2192         }
2193         priv->tx_cq = kcalloc(sizeof(struct mlx4_en_cq *), MAX_TX_RINGS,
2194                         GFP_KERNEL);
2195         if (!priv->tx_cq) {
2196                 err = -ENOMEM;
2197                 goto out;
2198         }
2199
2200         priv->rx_ring_num = prof->rx_ring_num;
2201         priv->cqe_factor = (mdev->dev->caps.cqe_size == 64) ? 1 : 0;
2202         priv->mac_index = -1;
2203         priv->last_ifq_jiffies = 0;
2204         priv->if_counters_rx_errors = 0;
2205         priv->if_counters_rx_no_buffer = 0;
2206 #ifdef CONFIG_MLX4_EN_DCB
2207         if (!mlx4_is_slave(priv->mdev->dev)) {
2208                 priv->dcbx_cap = DCB_CAP_DCBX_HOST;
2209                 priv->flags |= MLX4_EN_FLAG_DCB_ENABLED;
2210                 if (mdev->dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_ETS_CFG) {
2211                         dev->dcbnl_ops = &mlx4_en_dcbnl_ops;
2212                 } else {
2213                         en_info(priv, "QoS disabled - no HW support\n");
2214                         dev->dcbnl_ops = &mlx4_en_dcbnl_pfc_ops;
2215                 }
2216         }
2217 #endif
2218
2219         /* Query for default mac and max mtu */
2220         priv->max_mtu = mdev->dev->caps.eth_mtu_cap[priv->port];
2221         priv->mac = mdev->dev->caps.def_mac[priv->port];
2222         if (ILLEGAL_MAC(priv->mac)) {
2223 #if BITS_PER_LONG == 64
2224                 en_err(priv, "Port: %d, invalid mac burned: 0x%lx, quiting\n",
2225                                 priv->port, priv->mac);
2226 #elif BITS_PER_LONG == 32
2227                 en_err(priv, "Port: %d, invalid mac burned: 0x%llx, quiting\n",
2228                                 priv->port, priv->mac);
2229 #endif
2230                 err = -EINVAL;
2231                 goto out;
2232         }
2233
2234         mlx4_en_sysctl_conf(priv);
2235
2236         err = mlx4_en_alloc_resources(priv);
2237         if (err)
2238                 goto out;
2239
2240         /* Allocate page for receive rings */
2241         err = mlx4_alloc_hwq_res(mdev->dev, &priv->res,
2242                                 MLX4_EN_PAGE_SIZE, MLX4_EN_PAGE_SIZE);
2243         if (err) {
2244                 en_err(priv, "Failed to allocate page for rx qps\n");
2245                 goto out;
2246         }
2247         priv->allocated = 1;
2248
2249         /*
2250          * Set driver features
2251          */
2252         dev->if_capabilities |= IFCAP_HWCSUM | IFCAP_HWCSUM_IPV6;
2253         dev->if_capabilities |= IFCAP_VLAN_MTU | IFCAP_VLAN_HWTAGGING;
2254         dev->if_capabilities |= IFCAP_VLAN_HWCSUM | IFCAP_VLAN_HWFILTER;
2255         dev->if_capabilities |= IFCAP_LINKSTATE | IFCAP_JUMBO_MTU;
2256         dev->if_capabilities |= IFCAP_LRO;
2257         dev->if_capabilities |= IFCAP_HWSTATS;
2258
2259         if (mdev->LSO_support)
2260                 dev->if_capabilities |= IFCAP_TSO4 | IFCAP_TSO6 | IFCAP_VLAN_HWTSO;
2261
2262 #if __FreeBSD_version >= 1100000
2263         /* set TSO limits so that we don't have to drop TX packets */
2264         dev->if_hw_tsomax = MLX4_EN_TX_MAX_PAYLOAD_SIZE - (ETHER_HDR_LEN + ETHER_VLAN_ENCAP_LEN) /* hdr */;
2265         dev->if_hw_tsomaxsegcount = MLX4_EN_TX_MAX_MBUF_FRAGS - 1 /* hdr */;
2266         dev->if_hw_tsomaxsegsize = MLX4_EN_TX_MAX_MBUF_SIZE;
2267 #endif
2268
2269         dev->if_capenable = dev->if_capabilities;
2270
2271         dev->if_hwassist = 0;
2272         if (dev->if_capenable & (IFCAP_TSO4 | IFCAP_TSO6))
2273                 dev->if_hwassist |= CSUM_TSO;
2274         if (dev->if_capenable & IFCAP_TXCSUM)
2275                 dev->if_hwassist |= (CSUM_TCP | CSUM_UDP | CSUM_IP);
2276         if (dev->if_capenable & IFCAP_TXCSUM_IPV6)
2277                 dev->if_hwassist |= (CSUM_UDP_IPV6 | CSUM_TCP_IPV6);
2278
2279
2280         /* Register for VLAN events */
2281         priv->vlan_attach = EVENTHANDLER_REGISTER(vlan_config,
2282             mlx4_en_vlan_rx_add_vid, priv, EVENTHANDLER_PRI_FIRST);
2283         priv->vlan_detach = EVENTHANDLER_REGISTER(vlan_unconfig,
2284             mlx4_en_vlan_rx_kill_vid, priv, EVENTHANDLER_PRI_FIRST);
2285
2286         mdev->pndev[priv->port] = dev;
2287
2288         priv->last_link_state = MLX4_DEV_EVENT_PORT_DOWN;
2289         mlx4_en_set_default_moderation(priv);
2290
2291         /* Set default MAC */
2292         for (i = 0; i < ETHER_ADDR_LEN; i++)
2293                 dev_addr[ETHER_ADDR_LEN - 1 - i] = (u8) (priv->mac >> (8 * i));
2294
2295
2296         ether_ifattach(dev, dev_addr);
2297         if_link_state_change(dev, LINK_STATE_DOWN);
2298         ifmedia_init(&priv->media, IFM_IMASK | IFM_ETH_FMASK,
2299             mlx4_en_media_change, mlx4_en_media_status);
2300         ifmedia_add(&priv->media, IFM_ETHER | IFM_FDX | IFM_1000_T, 0, NULL);
2301         ifmedia_add(&priv->media, IFM_ETHER | IFM_FDX | IFM_10G_SR, 0, NULL);
2302         ifmedia_add(&priv->media, IFM_ETHER | IFM_FDX | IFM_10G_CX4, 0, NULL);
2303         ifmedia_add(&priv->media, IFM_ETHER | IFM_FDX | IFM_40G_CR4, 0, NULL);
2304         ifmedia_add(&priv->media, IFM_ETHER | IFM_AUTO, 0, NULL);
2305         ifmedia_set(&priv->media, IFM_ETHER | IFM_AUTO);
2306
2307         DEBUGNET_SET(dev, mlx4_en);
2308
2309         en_warn(priv, "Using %d TX rings\n", prof->tx_ring_num);
2310         en_warn(priv, "Using %d RX rings\n", prof->rx_ring_num);
2311
2312         priv->registered = 1;
2313
2314         en_warn(priv, "Using %d TX rings\n", prof->tx_ring_num);
2315         en_warn(priv, "Using %d RX rings\n", prof->rx_ring_num);
2316
2317
2318         priv->rx_mb_size = dev->if_mtu + ETH_HLEN + VLAN_HLEN + ETH_FCS_LEN;
2319         err = mlx4_SET_PORT_general(mdev->dev, priv->port,
2320                                     priv->rx_mb_size,
2321                                     prof->tx_pause, prof->tx_ppp,
2322                                     prof->rx_pause, prof->rx_ppp);
2323         if (err) {
2324                 en_err(priv, "Failed setting port general configurations "
2325                        "for port %d, with error %d\n", priv->port, err);
2326                 goto out;
2327         }
2328
2329         /* Init port */
2330         en_warn(priv, "Initializing port\n");
2331         err = mlx4_INIT_PORT(mdev->dev, priv->port);
2332         if (err) {
2333                 en_err(priv, "Failed Initializing port\n");
2334                 goto out;
2335         }
2336
2337         queue_delayed_work(mdev->workqueue, &priv->stats_task, STATS_DELAY);
2338
2339         if (mdev->dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_TS)
2340                 queue_delayed_work(mdev->workqueue, &priv->service_task, SERVICE_TASK_DELAY);
2341
2342         return 0;
2343
2344 out:
2345         mlx4_en_destroy_netdev(dev);
2346         return err;
2347 }
2348
2349 static int mlx4_en_set_ring_size(struct net_device *dev,
2350     int rx_size, int tx_size)
2351 {
2352         struct mlx4_en_priv *priv = netdev_priv(dev);
2353         struct mlx4_en_dev *mdev = priv->mdev;
2354         int port_up = 0;
2355         int err = 0;
2356
2357         rx_size = roundup_pow_of_two(rx_size);
2358         rx_size = max_t(u32, rx_size, MLX4_EN_MIN_RX_SIZE);
2359         rx_size = min_t(u32, rx_size, MLX4_EN_MAX_RX_SIZE);
2360         tx_size = roundup_pow_of_two(tx_size);
2361         tx_size = max_t(u32, tx_size, MLX4_EN_MIN_TX_SIZE);
2362         tx_size = min_t(u32, tx_size, MLX4_EN_MAX_TX_SIZE);
2363
2364         if (rx_size == (priv->port_up ?
2365             priv->rx_ring[0]->actual_size : priv->rx_ring[0]->size) &&
2366             tx_size == priv->tx_ring[0]->size)
2367                 return 0;
2368         mutex_lock(&mdev->state_lock);
2369         if (priv->port_up) {
2370                 port_up = 1;
2371                 mlx4_en_stop_port(dev);
2372         }
2373         mlx4_en_free_resources(priv);
2374         priv->prof->tx_ring_size = tx_size;
2375         priv->prof->rx_ring_size = rx_size;
2376         err = mlx4_en_alloc_resources(priv);
2377         if (err) {
2378                 en_err(priv, "Failed reallocating port resources\n");
2379                 goto out;
2380         }
2381         if (port_up) {
2382                 err = mlx4_en_start_port(dev);
2383                 if (err)
2384                         en_err(priv, "Failed starting port\n");
2385         }
2386 out:
2387         mutex_unlock(&mdev->state_lock);
2388         return err;
2389 }
2390 static int mlx4_en_set_rx_ring_size(SYSCTL_HANDLER_ARGS)
2391 {
2392         struct mlx4_en_priv *priv;
2393         int size;
2394         int error;
2395
2396         priv = arg1;
2397         size = priv->prof->rx_ring_size;
2398         error = sysctl_handle_int(oidp, &size, 0, req);
2399         if (error || !req->newptr)
2400                 return (error);
2401         error = -mlx4_en_set_ring_size(priv->dev, size,
2402             priv->prof->tx_ring_size);
2403         return (error);
2404 }
2405
2406 static int mlx4_en_set_tx_ring_size(SYSCTL_HANDLER_ARGS)
2407 {
2408         struct mlx4_en_priv *priv;
2409         int size;
2410         int error;
2411
2412         priv = arg1;
2413         size = priv->prof->tx_ring_size;
2414         error = sysctl_handle_int(oidp, &size, 0, req);
2415         if (error || !req->newptr)
2416                 return (error);
2417         error = -mlx4_en_set_ring_size(priv->dev, priv->prof->rx_ring_size,
2418             size);
2419
2420         return (error);
2421 }
2422
2423 static int mlx4_en_get_module_info(struct net_device *dev,
2424                                    struct ethtool_modinfo *modinfo)
2425 {
2426         struct mlx4_en_priv *priv = netdev_priv(dev);
2427         struct mlx4_en_dev *mdev = priv->mdev;
2428         int ret;
2429         u8 data[4];
2430
2431         /* Read first 2 bytes to get Module & REV ID */
2432         ret = mlx4_get_module_info(mdev->dev, priv->port,
2433                                    0/*offset*/, 2/*size*/, data);
2434
2435         if (ret < 2) {
2436                 en_err(priv, "Failed to read eeprom module first two bytes, error: 0x%x\n", -ret);
2437                 return -EIO;
2438         }
2439
2440         switch (data[0] /* identifier */) {
2441         case MLX4_MODULE_ID_QSFP:
2442                 modinfo->type = ETH_MODULE_SFF_8436;
2443                 modinfo->eeprom_len = ETH_MODULE_SFF_8436_LEN;
2444                 break;
2445         case MLX4_MODULE_ID_QSFP_PLUS:
2446                 if (data[1] >= 0x3) { /* revision id */
2447                         modinfo->type = ETH_MODULE_SFF_8636;
2448                         modinfo->eeprom_len = ETH_MODULE_SFF_8636_LEN;
2449                 } else {
2450                         modinfo->type = ETH_MODULE_SFF_8436;
2451                         modinfo->eeprom_len = ETH_MODULE_SFF_8436_LEN;
2452                 }
2453                 break;
2454         case MLX4_MODULE_ID_QSFP28:
2455                 modinfo->type = ETH_MODULE_SFF_8636;
2456                 modinfo->eeprom_len = ETH_MODULE_SFF_8636_LEN;
2457                 break;
2458         case MLX4_MODULE_ID_SFP:
2459                 modinfo->type = ETH_MODULE_SFF_8472;
2460                 modinfo->eeprom_len = ETH_MODULE_SFF_8472_LEN;
2461                 break;
2462         default:
2463                 en_err(priv, "mlx4_en_get_module_info :  Not recognized cable type\n");
2464                 return -EINVAL;
2465         }
2466
2467         return 0;
2468 }
2469
2470 static int mlx4_en_get_module_eeprom(struct net_device *dev,
2471                                      struct ethtool_eeprom *ee,
2472                                      u8 *data)
2473 {
2474         struct mlx4_en_priv *priv = netdev_priv(dev);
2475         struct mlx4_en_dev *mdev = priv->mdev;
2476         int offset = ee->offset;
2477         int i = 0, ret;
2478
2479         if (ee->len == 0)
2480                 return -EINVAL;
2481
2482         memset(data, 0, ee->len);
2483
2484         while (i < ee->len) {
2485                 en_dbg(DRV, priv,
2486                        "mlx4_get_module_info i(%d) offset(%d) len(%d)\n",
2487                        i, offset, ee->len - i);
2488
2489                 ret = mlx4_get_module_info(mdev->dev, priv->port,
2490                                            offset, ee->len - i, data + i);
2491
2492                 if (!ret) /* Done reading */
2493                         return 0;
2494
2495                 if (ret < 0) {
2496                         en_err(priv,
2497                                "mlx4_get_module_info i(%d) offset(%d) bytes_to_read(%d) - FAILED (0x%x)\n",
2498                                i, offset, ee->len - i, ret);
2499                         return -1;
2500                 }
2501
2502                 i += ret;
2503                 offset += ret;
2504         }
2505         return 0;
2506 }
2507
2508 static void mlx4_en_print_eeprom(u8 *data, __u32 len)
2509 {
2510         int             i;
2511         int             j = 0;
2512         int             row = 0;
2513         const int       NUM_OF_BYTES = 16;
2514
2515         printf("\nOffset\t\tValues\n");
2516         printf("------\t\t------\n");
2517         while(row < len){
2518                 printf("0x%04x\t\t",row);
2519                 for(i=0; i < NUM_OF_BYTES; i++){
2520                         printf("%02x ", data[j]);
2521                         row++;
2522                         j++;
2523                 }
2524                 printf("\n");
2525         }
2526 }
2527
2528 /* Read cable EEPROM module information by first inspecting the first
2529  * two bytes to get the length and then read the rest of the information.
2530  * The information is printed to dmesg. */
2531 static int mlx4_en_read_eeprom(SYSCTL_HANDLER_ARGS)
2532 {
2533
2534         u8*             data;
2535         int             error;
2536         int             result = 0;
2537         struct          mlx4_en_priv *priv;
2538         struct          net_device *dev;
2539         struct          ethtool_modinfo modinfo;
2540         struct          ethtool_eeprom ee;
2541
2542         error = sysctl_handle_int(oidp, &result, 0, req);
2543         if (error || !req->newptr)
2544                 return (error);
2545
2546         if (result == 1) {
2547                 priv = arg1;
2548                 dev = priv->dev;
2549                 data = kmalloc(PAGE_SIZE, GFP_KERNEL);
2550
2551                 error = mlx4_en_get_module_info(dev, &modinfo);
2552                 if (error) {
2553                         en_err(priv,
2554                                "mlx4_en_get_module_info returned with error - FAILED (0x%x)\n",
2555                                -error);
2556                         goto out;
2557                 }
2558
2559                 ee.len = modinfo.eeprom_len;
2560                 ee.offset = 0;
2561
2562                 error = mlx4_en_get_module_eeprom(dev, &ee, data);
2563                 if (error) {
2564                         en_err(priv,
2565                                "mlx4_en_get_module_eeprom returned with error - FAILED (0x%x)\n",
2566                                -error);
2567                         /* Continue printing partial information in case of an error */
2568                 }
2569
2570                 /* EEPROM information will be printed in dmesg */
2571                 mlx4_en_print_eeprom(data, ee.len);
2572 out:
2573                 kfree(data);
2574         }
2575         /* Return zero to prevent sysctl failure. */
2576         return (0);
2577 }
2578
2579 static int mlx4_en_set_tx_ppp(SYSCTL_HANDLER_ARGS)
2580 {
2581         struct mlx4_en_priv *priv;
2582         int ppp;
2583         int error;
2584
2585         priv = arg1;
2586         ppp = priv->prof->tx_ppp;
2587         error = sysctl_handle_int(oidp, &ppp, 0, req);
2588         if (error || !req->newptr)
2589                 return (error);
2590         if (ppp > 0xff || ppp < 0)
2591                 return (-EINVAL);
2592         priv->prof->tx_ppp = ppp;
2593         error = -mlx4_SET_PORT_general(priv->mdev->dev, priv->port,
2594                                        priv->rx_mb_size + ETHER_CRC_LEN,
2595                                        priv->prof->tx_pause,
2596                                        priv->prof->tx_ppp,
2597                                        priv->prof->rx_pause,
2598                                        priv->prof->rx_ppp);
2599
2600         return (error);
2601 }
2602
2603 static int mlx4_en_set_rx_ppp(SYSCTL_HANDLER_ARGS)
2604 {
2605         struct mlx4_en_priv *priv;
2606         struct mlx4_en_dev *mdev;
2607         int ppp;
2608         int error;
2609         int port_up;
2610
2611         port_up = 0;
2612         priv = arg1;
2613         mdev = priv->mdev;
2614         ppp = priv->prof->rx_ppp;
2615         error = sysctl_handle_int(oidp, &ppp, 0, req);
2616         if (error || !req->newptr)
2617                 return (error);
2618         if (ppp > 0xff || ppp < 0)
2619                 return (-EINVAL);
2620         /* See if we have to change the number of tx queues. */
2621         if (!ppp != !priv->prof->rx_ppp) {
2622                 mutex_lock(&mdev->state_lock);
2623                 if (priv->port_up) {
2624                         port_up = 1;
2625                         mlx4_en_stop_port(priv->dev);
2626                 }
2627                 mlx4_en_free_resources(priv);
2628                 priv->prof->rx_ppp = ppp;
2629                 error = -mlx4_en_alloc_resources(priv);
2630                 if (error)
2631                         en_err(priv, "Failed reallocating port resources\n");
2632                 if (error == 0 && port_up) {
2633                         error = -mlx4_en_start_port(priv->dev);
2634                         if (error)
2635                                 en_err(priv, "Failed starting port\n");
2636                 }
2637                 mutex_unlock(&mdev->state_lock);
2638                 return (error);
2639
2640         }
2641         priv->prof->rx_ppp = ppp;
2642         error = -mlx4_SET_PORT_general(priv->mdev->dev, priv->port,
2643                                        priv->rx_mb_size + ETHER_CRC_LEN,
2644                                        priv->prof->tx_pause,
2645                                        priv->prof->tx_ppp,
2646                                        priv->prof->rx_pause,
2647                                        priv->prof->rx_ppp);
2648
2649         return (error);
2650 }
2651
2652 static void mlx4_en_sysctl_conf(struct mlx4_en_priv *priv)
2653 {
2654         struct net_device *dev;
2655         struct sysctl_ctx_list *ctx;
2656         struct sysctl_oid *node;
2657         struct sysctl_oid_list *node_list;
2658         struct sysctl_oid *coal;
2659         struct sysctl_oid_list *coal_list;
2660         const char *pnameunit;
2661         dev = priv->dev;
2662         ctx = &priv->conf_ctx;
2663         pnameunit = device_get_nameunit(priv->mdev->pdev->dev.bsddev);
2664
2665         sysctl_ctx_init(ctx);
2666         priv->conf_sysctl = SYSCTL_ADD_NODE(ctx, SYSCTL_STATIC_CHILDREN(_hw),
2667             OID_AUTO, dev->if_xname, CTLFLAG_RD, 0, "mlx4 10gig ethernet");
2668         node = SYSCTL_ADD_NODE(ctx, SYSCTL_CHILDREN(priv->conf_sysctl), OID_AUTO,
2669             "conf", CTLFLAG_RD, NULL, "Configuration");
2670         node_list = SYSCTL_CHILDREN(node);
2671
2672         SYSCTL_ADD_UINT(ctx, node_list, OID_AUTO, "msg_enable",
2673             CTLFLAG_RW, &priv->msg_enable, 0,
2674             "Driver message enable bitfield");
2675         SYSCTL_ADD_UINT(ctx, node_list, OID_AUTO, "rx_rings",
2676             CTLFLAG_RD, &priv->rx_ring_num, 0,
2677             "Number of receive rings");
2678         SYSCTL_ADD_UINT(ctx, node_list, OID_AUTO, "tx_rings",
2679             CTLFLAG_RD, &priv->tx_ring_num, 0,
2680             "Number of transmit rings");
2681         SYSCTL_ADD_PROC(ctx, node_list, OID_AUTO, "rx_size",
2682             CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, priv, 0,
2683             mlx4_en_set_rx_ring_size, "I", "Receive ring size");
2684         SYSCTL_ADD_PROC(ctx, node_list, OID_AUTO, "tx_size",
2685             CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, priv, 0,
2686             mlx4_en_set_tx_ring_size, "I", "Transmit ring size");
2687         SYSCTL_ADD_PROC(ctx, node_list, OID_AUTO, "tx_ppp",
2688             CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, priv, 0,
2689             mlx4_en_set_tx_ppp, "I", "TX Per-priority pause");
2690         SYSCTL_ADD_PROC(ctx, node_list, OID_AUTO, "rx_ppp",
2691             CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, priv, 0,
2692             mlx4_en_set_rx_ppp, "I", "RX Per-priority pause");
2693         SYSCTL_ADD_UINT(ctx, node_list, OID_AUTO, "port_num",
2694             CTLFLAG_RD, &priv->port, 0,
2695             "Port Number");
2696         SYSCTL_ADD_STRING(ctx, node_list, OID_AUTO, "device_name",
2697             CTLFLAG_RD, __DECONST(void *, pnameunit), 0,
2698             "PCI device name");
2699         /* Add coalescer configuration. */
2700         coal = SYSCTL_ADD_NODE(ctx, node_list, OID_AUTO,
2701             "coalesce", CTLFLAG_RD, NULL, "Interrupt coalesce configuration");
2702         coal_list = SYSCTL_CHILDREN(coal);
2703         SYSCTL_ADD_UINT(ctx, coal_list, OID_AUTO, "pkt_rate_low",
2704             CTLFLAG_RW, &priv->pkt_rate_low, 0,
2705             "Packets per-second for minimum delay");
2706         SYSCTL_ADD_UINT(ctx, coal_list, OID_AUTO, "rx_usecs_low",
2707             CTLFLAG_RW, &priv->rx_usecs_low, 0,
2708             "Minimum RX delay in micro-seconds");
2709         SYSCTL_ADD_UINT(ctx, coal_list, OID_AUTO, "pkt_rate_high",
2710             CTLFLAG_RW, &priv->pkt_rate_high, 0,
2711             "Packets per-second for maximum delay");
2712         SYSCTL_ADD_UINT(ctx, coal_list, OID_AUTO, "rx_usecs_high",
2713             CTLFLAG_RW, &priv->rx_usecs_high, 0,
2714             "Maximum RX delay in micro-seconds");
2715         SYSCTL_ADD_UINT(ctx, coal_list, OID_AUTO, "sample_interval",
2716             CTLFLAG_RW, &priv->sample_interval, 0,
2717             "adaptive frequency in units of HZ ticks");
2718         SYSCTL_ADD_UINT(ctx, coal_list, OID_AUTO, "adaptive_rx_coal",
2719             CTLFLAG_RW, &priv->adaptive_rx_coal, 0,
2720             "Enable adaptive rx coalescing");
2721         /* EEPROM support */
2722         SYSCTL_ADD_PROC(ctx, node_list, OID_AUTO, "eeprom_info",
2723             CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, priv, 0,
2724             mlx4_en_read_eeprom, "I", "EEPROM information");
2725 }
2726
2727 static void mlx4_en_sysctl_stat(struct mlx4_en_priv *priv)
2728 {
2729         struct sysctl_ctx_list *ctx;
2730         struct sysctl_oid_list *node_list;
2731         struct sysctl_oid *ring_node;
2732         struct sysctl_oid_list *ring_list;
2733         struct mlx4_en_tx_ring *tx_ring;
2734         struct mlx4_en_rx_ring *rx_ring;
2735         char namebuf[128];
2736         int i;
2737
2738         ctx = &priv->stat_ctx;
2739         sysctl_ctx_init(ctx);
2740         priv->stat_sysctl = SYSCTL_ADD_NODE(ctx, SYSCTL_CHILDREN(priv->conf_sysctl), OID_AUTO,
2741             "stat", CTLFLAG_RD, NULL, "Statistics");
2742         node_list = SYSCTL_CHILDREN(priv->stat_sysctl);
2743
2744 #ifdef MLX4_EN_PERF_STAT
2745         SYSCTL_ADD_UINT(ctx, node_list, OID_AUTO, "tx_poll", CTLFLAG_RD,
2746             &priv->pstats.tx_poll, "TX Poll calls");
2747         SYSCTL_ADD_QUAD(ctx, node_list, OID_AUTO, "tx_pktsz_avg", CTLFLAG_RD,
2748             &priv->pstats.tx_pktsz_avg, "TX average packet size");
2749         SYSCTL_ADD_UINT(ctx, node_list, OID_AUTO, "inflight_avg", CTLFLAG_RD,
2750             &priv->pstats.inflight_avg, "TX average packets in-flight");
2751         SYSCTL_ADD_UINT(ctx, node_list, OID_AUTO, "tx_coal_avg", CTLFLAG_RD,
2752             &priv->pstats.tx_coal_avg, "TX average coalesced completions");
2753         SYSCTL_ADD_UINT(ctx, node_list, OID_AUTO, "rx_coal_avg", CTLFLAG_RD,
2754             &priv->pstats.rx_coal_avg, "RX average coalesced completions");
2755 #endif
2756
2757         SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "tso_packets", CTLFLAG_RD,
2758             &priv->port_stats.tso_packets, 0, "TSO packets sent");
2759         SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "queue_stopped", CTLFLAG_RD,
2760             &priv->port_stats.queue_stopped, 0, "Queue full");
2761         SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "wake_queue", CTLFLAG_RD,
2762             &priv->port_stats.wake_queue, 0, "Queue resumed after full");
2763         SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "tx_timeout", CTLFLAG_RD,
2764             &priv->port_stats.tx_timeout, 0, "Transmit timeouts");
2765         SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "tx_oversized_packets", CTLFLAG_RD,
2766             &priv->port_stats.oversized_packets, 0, "TX oversized packets, m_defrag failed");
2767         SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_alloc_failed", CTLFLAG_RD,
2768             &priv->port_stats.rx_alloc_failed, 0, "RX failed to allocate mbuf");
2769         SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_chksum_good", CTLFLAG_RD,
2770             &priv->port_stats.rx_chksum_good, 0, "RX checksum offload success");
2771         SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_chksum_none", CTLFLAG_RD,
2772             &priv->port_stats.rx_chksum_none, 0, "RX without checksum offload");
2773         SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "tx_chksum_offload",
2774             CTLFLAG_RD, &priv->port_stats.tx_chksum_offload, 0,
2775             "TX checksum offloads");
2776         SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "defrag_attempts",
2777             CTLFLAG_RD, &priv->port_stats.defrag_attempts, 0,
2778             "Oversized chains defragged");
2779
2780         /* Could strdup the names and add in a loop.  This is simpler. */
2781         SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_bytes", CTLFLAG_RD,
2782             &priv->pkstats.rx_bytes, 0, "RX Bytes");
2783         SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_packets", CTLFLAG_RD,
2784             &priv->pkstats.rx_packets, 0, "RX packets");
2785         SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_multicast_packets", CTLFLAG_RD,
2786             &priv->pkstats.rx_multicast_packets, 0, "RX Multicast Packets");
2787         SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_broadcast_packets", CTLFLAG_RD,
2788             &priv->pkstats.rx_broadcast_packets, 0, "RX Broadcast Packets");
2789         SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_errors", CTLFLAG_RD,
2790             &priv->pkstats.rx_errors, 0, "RX Errors");
2791         SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_dropped", CTLFLAG_RD,
2792             &priv->pkstats.rx_dropped, 0, "RX Dropped");
2793         SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_length_errors", CTLFLAG_RD,
2794             &priv->pkstats.rx_length_errors, 0, "RX Length Errors");
2795         SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_over_errors", CTLFLAG_RD,
2796             &priv->pkstats.rx_over_errors, 0, "RX Over Errors");
2797         SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_crc_errors", CTLFLAG_RD,
2798             &priv->pkstats.rx_crc_errors, 0, "RX CRC Errors");
2799         SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_jabbers", CTLFLAG_RD,
2800             &priv->pkstats.rx_jabbers, 0, "RX Jabbers");
2801
2802         SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_in_range_length_error", CTLFLAG_RD,
2803             &priv->pkstats.rx_in_range_length_error, 0, "RX IN_Range Length Error");
2804         SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_out_range_length_error",
2805             CTLFLAG_RD, &priv->pkstats.rx_out_range_length_error, 0,
2806             "RX Out Range Length Error");
2807         SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_lt_64_bytes_packets", CTLFLAG_RD,
2808             &priv->pkstats.rx_lt_64_bytes_packets, 0, "RX Lt 64 Bytes Packets");
2809         SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_127_bytes_packets", CTLFLAG_RD,
2810             &priv->pkstats.rx_127_bytes_packets, 0, "RX 127 bytes Packets");
2811         SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_255_bytes_packets", CTLFLAG_RD,
2812             &priv->pkstats.rx_255_bytes_packets, 0, "RX 255 bytes Packets");
2813         SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_511_bytes_packets", CTLFLAG_RD,
2814             &priv->pkstats.rx_511_bytes_packets, 0, "RX 511 bytes Packets");
2815         SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_1023_bytes_packets", CTLFLAG_RD,
2816             &priv->pkstats.rx_1023_bytes_packets, 0, "RX 1023 bytes Packets");
2817         SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_1518_bytes_packets", CTLFLAG_RD,
2818             &priv->pkstats.rx_1518_bytes_packets, 0, "RX 1518 bytes Packets");
2819         SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_1522_bytes_packets", CTLFLAG_RD,
2820             &priv->pkstats.rx_1522_bytes_packets, 0, "RX 1522 bytes Packets");
2821         SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_1548_bytes_packets", CTLFLAG_RD,
2822             &priv->pkstats.rx_1548_bytes_packets, 0, "RX 1548 bytes Packets");
2823         SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_gt_1548_bytes_packets", CTLFLAG_RD,
2824             &priv->pkstats.rx_gt_1548_bytes_packets, 0,
2825             "RX Greater Then 1548 bytes Packets");
2826
2827         SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "tx_packets", CTLFLAG_RD,
2828             &priv->pkstats.tx_packets, 0, "TX packets");
2829         SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "tx_bytes", CTLFLAG_RD,
2830             &priv->pkstats.tx_bytes, 0, "TX Bytes");
2831         SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "tx_multicast_packets", CTLFLAG_RD,
2832             &priv->pkstats.tx_multicast_packets, 0, "TX Multicast Packets");
2833         SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "tx_broadcast_packets", CTLFLAG_RD,
2834             &priv->pkstats.tx_broadcast_packets, 0, "TX Broadcast Packets");
2835         SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "tx_errors", CTLFLAG_RD,
2836             &priv->pkstats.tx_errors, 0, "TX Errors");
2837         SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "tx_dropped", CTLFLAG_RD,
2838             &priv->pkstats.tx_dropped, 0, "TX Dropped");
2839         SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "tx_lt_64_bytes_packets", CTLFLAG_RD,
2840             &priv->pkstats.tx_lt_64_bytes_packets, 0, "TX Less Then 64 Bytes Packets");
2841         SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "tx_127_bytes_packets", CTLFLAG_RD,
2842             &priv->pkstats.tx_127_bytes_packets, 0, "TX 127 Bytes Packets");
2843         SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "tx_255_bytes_packets", CTLFLAG_RD,
2844             &priv->pkstats.tx_255_bytes_packets, 0, "TX 255 Bytes Packets");
2845         SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "tx_511_bytes_packets", CTLFLAG_RD,
2846             &priv->pkstats.tx_511_bytes_packets, 0, "TX 511 Bytes Packets");
2847         SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "tx_1023_bytes_packets", CTLFLAG_RD,
2848             &priv->pkstats.tx_1023_bytes_packets, 0, "TX 1023 Bytes Packets");
2849         SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "tx_1518_bytes_packets", CTLFLAG_RD,
2850             &priv->pkstats.tx_1518_bytes_packets, 0, "TX 1518 Bytes Packets");
2851         SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "tx_1522_bytes_packets", CTLFLAG_RD,
2852             &priv->pkstats.tx_1522_bytes_packets, 0, "TX 1522 Bytes Packets");
2853         SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "tx_1548_bytes_packets", CTLFLAG_RD,
2854             &priv->pkstats.tx_1548_bytes_packets, 0, "TX 1548 Bytes Packets");
2855         SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "tx_gt_1548_bytes_packets", CTLFLAG_RD,
2856             &priv->pkstats.tx_gt_1548_bytes_packets, 0,
2857             "TX Greater Then 1548 Bytes Packets");
2858
2859         for (i = 0; i < priv->tx_ring_num; i++) {
2860                 tx_ring = priv->tx_ring[i];
2861                 snprintf(namebuf, sizeof(namebuf), "tx_ring%d", i);
2862                 ring_node = SYSCTL_ADD_NODE(ctx, node_list, OID_AUTO, namebuf,
2863                     CTLFLAG_RD, NULL, "TX Ring");
2864                 ring_list = SYSCTL_CHILDREN(ring_node);
2865                 SYSCTL_ADD_U64(ctx, ring_list, OID_AUTO, "packets",
2866                     CTLFLAG_RD, &tx_ring->packets, 0, "TX packets");
2867                 SYSCTL_ADD_U64(ctx, ring_list, OID_AUTO, "bytes",
2868                     CTLFLAG_RD, &tx_ring->bytes, 0, "TX bytes");
2869                 SYSCTL_ADD_U64(ctx, ring_list, OID_AUTO, "tso_packets",
2870                     CTLFLAG_RD, &tx_ring->tso_packets, 0, "TSO packets");
2871                 SYSCTL_ADD_U64(ctx, ring_list, OID_AUTO, "defrag_attempts",
2872                     CTLFLAG_RD, &tx_ring->defrag_attempts, 0,
2873                     "Oversized chains defragged");
2874         }
2875
2876         for (i = 0; i < priv->rx_ring_num; i++) {
2877                 rx_ring = priv->rx_ring[i];
2878                 snprintf(namebuf, sizeof(namebuf), "rx_ring%d", i);
2879                 ring_node = SYSCTL_ADD_NODE(ctx, node_list, OID_AUTO, namebuf,
2880                     CTLFLAG_RD, NULL, "RX Ring");
2881                 ring_list = SYSCTL_CHILDREN(ring_node);
2882                 SYSCTL_ADD_U64(ctx, ring_list, OID_AUTO, "packets",
2883                     CTLFLAG_RD, &rx_ring->packets, 0, "RX packets");
2884                 SYSCTL_ADD_U64(ctx, ring_list, OID_AUTO, "bytes",
2885                     CTLFLAG_RD, &rx_ring->bytes, 0, "RX bytes");
2886                 SYSCTL_ADD_U64(ctx, ring_list, OID_AUTO, "error",
2887                     CTLFLAG_RD, &rx_ring->errors, 0, "RX soft errors");
2888         }
2889 }
2890
2891 #ifdef DEBUGNET
2892 static void
2893 mlx4_en_debugnet_init(struct ifnet *dev, int *nrxr, int *ncl, int *clsize)
2894 {
2895         struct mlx4_en_priv *priv;
2896
2897         priv = if_getsoftc(dev);
2898         mutex_lock(&priv->mdev->state_lock);
2899         *nrxr = priv->rx_ring_num;
2900         *ncl = DEBUGNET_MAX_IN_FLIGHT;
2901         *clsize = priv->rx_mb_size;
2902         mutex_unlock(&priv->mdev->state_lock);
2903 }
2904
2905 static void
2906 mlx4_en_debugnet_event(struct ifnet *dev, enum debugnet_ev event)
2907 {
2908 }
2909
2910 static int
2911 mlx4_en_debugnet_transmit(struct ifnet *dev, struct mbuf *m)
2912 {
2913         struct mlx4_en_priv *priv;
2914         int err;
2915
2916         priv = if_getsoftc(dev);
2917         if ((if_getdrvflags(dev) & (IFF_DRV_RUNNING | IFF_DRV_OACTIVE)) !=
2918             IFF_DRV_RUNNING || !priv->link_state)
2919                 return (ENOENT);
2920
2921         err = mlx4_en_xmit(priv, 0, &m);
2922         if (err != 0 && m != NULL)
2923                 m_freem(m);
2924         return (err);
2925 }
2926
2927 static int
2928 mlx4_en_debugnet_poll(struct ifnet *dev, int count)
2929 {
2930         struct mlx4_en_priv *priv;
2931
2932         priv = if_getsoftc(dev);
2933         if ((if_getdrvflags(dev) & IFF_DRV_RUNNING) == 0 || !priv->link_state)
2934                 return (ENOENT);
2935
2936         mlx4_poll_interrupts(priv->mdev->dev);
2937
2938         return (0);
2939 }
2940 #endif /* DEBUGNET */