sys/ofed/drivers/net/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 #ifdef CONFIG_NET_RX_BUSY_POLL
  38 #include <net/busy_poll.h>
  39 #endif
  40
  41 #include <linux/list.h>
  42 #include <linux/if_ether.h>
  43
  44 #include <linux/mlx4/driver.h>
  45 #include <linux/mlx4/device.h>
  46 #include <linux/mlx4/cmd.h>
  47 #include <linux/mlx4/cq.h>
  48
  49 #include <sys/sockio.h>
  50 #include <sys/sysctl.h>
  51
  52 #include "mlx4_en.h"
  53 #include "en_port.h"
  54
  55 static void mlx4_en_sysctl_stat(struct mlx4_en_priv *priv);
  56 static void mlx4_en_sysctl_conf(struct mlx4_en_priv *priv);
  57 static int mlx4_en_unit;
  58
  59 #ifdef CONFIG_NET_RX_BUSY_POLL
  60 /* must be called with local_bh_disable()d */
  61 static int mlx4_en_low_latency_recv(struct napi_struct *napi)
  62 {
  63         struct mlx4_en_cq *cq = container_of(napi, struct mlx4_en_cq, napi);
  64         struct net_device *dev = cq->dev;
  65         struct mlx4_en_priv *priv = netdev_priv(dev);
  66         struct mlx4_en_rx_ring *rx_ring = priv->rx_ring[cq->ring];
  67         int done;
  68
  69         if (!priv->port_up)
  70                 return LL_FLUSH_FAILED;
  71
  72         if (!mlx4_en_cq_lock_poll(cq))
  73                 return LL_FLUSH_BUSY;
  74
  75         done = mlx4_en_process_rx_cq(dev, cq, 4);
  76 #ifdef LL_EXTENDED_STATS
  77         if (done)
  78                 rx_ring->cleaned += done;
  79         else
  80                 rx_ring->misses++;
  81 #endif
  82
  83         mlx4_en_cq_unlock_poll(cq);
  84
  85         return done;
  86 }
  87 #endif  /* CONFIG_NET_RX_BUSY_POLL */
  88
  89 #ifdef CONFIG_RFS_ACCEL
  90
  91 struct mlx4_en_filter {
  92         struct list_head next;
  93         struct work_struct work;
  94
  95         u8     ip_proto;
  96         __be32 src_ip;
  97         __be32 dst_ip;
  98         __be16 src_port;
  99         __be16 dst_port;
 100
 101         int rxq_index;
 102         struct mlx4_en_priv *priv;
 103         u32 flow_id;                    /* RFS infrastructure id */
 104         int id;                         /* mlx4_en driver id */
 105         u64 reg_id;                     /* Flow steering API id */
 106         u8 activated;                   /* Used to prevent expiry before filter
 107                                          * is attached
 108                                          */
 109         struct hlist_node filter_chain;
 110 };
 111
 112 static void mlx4_en_filter_rfs_expire(struct mlx4_en_priv *priv);
 113
 114 static enum mlx4_net_trans_rule_id mlx4_ip_proto_to_trans_rule_id(u8 ip_proto)
 115 {
 116         switch (ip_proto) {
 117         case IPPROTO_UDP:
 118                 return MLX4_NET_TRANS_RULE_ID_UDP;
 119         case IPPROTO_TCP:
 120                 return MLX4_NET_TRANS_RULE_ID_TCP;
 121         default:
 122                 return -EPROTONOSUPPORT;
 123         }
 124 };
 125
 126 static void mlx4_en_filter_work(struct work_struct *work)
 127 {
 128         struct mlx4_en_filter *filter = container_of(work,
 129                                                      struct mlx4_en_filter,
 130                                                      work);
 131         struct mlx4_en_priv *priv = filter->priv;
 132         struct mlx4_spec_list spec_tcp_udp = {
 133                 .id = mlx4_ip_proto_to_trans_rule_id(filter->ip_proto),
 134                 {
 135                         .tcp_udp = {
 136                                 .dst_port = filter->dst_port,
 137                                 .dst_port_msk = (__force __be16)-1,
 138                                 .src_port = filter->src_port,
 139                                 .src_port_msk = (__force __be16)-1,
 140                         },
 141                 },
 142         };
 143         struct mlx4_spec_list spec_ip = {
 144                 .id = MLX4_NET_TRANS_RULE_ID_IPV4,
 145                 {
 146                         .ipv4 = {
 147                                 .dst_ip = filter->dst_ip,
 148                                 .dst_ip_msk = (__force __be32)-1,
 149                                 .src_ip = filter->src_ip,
 150                                 .src_ip_msk = (__force __be32)-1,
 151                         },
 152                 },
 153         };
 154         struct mlx4_spec_list spec_eth = {
 155                 .id = MLX4_NET_TRANS_RULE_ID_ETH,
 156         };
 157         struct mlx4_net_trans_rule rule = {
 158                 .list = LIST_HEAD_INIT(rule.list),
 159                 .queue_mode = MLX4_NET_TRANS_Q_LIFO,
 160                 .exclusive = 1,
 161                 .allow_loopback = 1,
 162                 .promisc_mode = MLX4_FS_REGULAR,
 163                 .port = priv->port,
 164                 .priority = MLX4_DOMAIN_RFS,
 165         };
 166         int rc;
 167         __be64 mac_mask = cpu_to_be64(MLX4_MAC_MASK << 16);
 168
 169         if (spec_tcp_udp.id < 0) {
 170                 en_warn(priv, "RFS: ignoring unsupported ip protocol (%d)\n",
 171                         filter->ip_proto);
 172                 goto ignore;
 173         }
 174         list_add_tail(&spec_eth.list, &rule.list);
 175         list_add_tail(&spec_ip.list, &rule.list);
 176         list_add_tail(&spec_tcp_udp.list, &rule.list);
 177
 178         rule.qpn = priv->rss_map.qps[filter->rxq_index].qpn;
 179         memcpy(spec_eth.eth.dst_mac, priv->dev->dev_addr, ETH_ALEN);
 180         memcpy(spec_eth.eth.dst_mac_msk, &mac_mask, ETH_ALEN);
 181
 182         filter->activated = 0;
 183
 184         if (filter->reg_id) {
 185                 rc = mlx4_flow_detach(priv->mdev->dev, filter->reg_id);
 186                 if (rc && rc != -ENOENT)
 187                         en_err(priv, "Error detaching flow. rc = %d\n", rc);
 188         }
 189
 190         rc = mlx4_flow_attach(priv->mdev->dev, &rule, &filter->reg_id);
 191         if (rc)
 192                 en_err(priv, "Error attaching flow. err = %d\n", rc);
 193
 194 ignore:
 195         mlx4_en_filter_rfs_expire(priv);
 196
 197         filter->activated = 1;
 198 }
 199
 200 static inline struct hlist_head *
 201 filter_hash_bucket(struct mlx4_en_priv *priv, __be32 src_ip, __be32 dst_ip,
 202                    __be16 src_port, __be16 dst_port)
 203 {
 204         unsigned long l;
 205         int bucket_idx;
 206
 207         l = (__force unsigned long)src_port |
 208             ((__force unsigned long)dst_port << 2);
 209         l ^= (__force unsigned long)(src_ip ^ dst_ip);
 210
 211         bucket_idx = hash_long(l, MLX4_EN_FILTER_HASH_SHIFT);
 212
 213         return &priv->filter_hash[bucket_idx];
 214 }
 215
 216 static struct mlx4_en_filter *
 217 mlx4_en_filter_alloc(struct mlx4_en_priv *priv, int rxq_index, __be32 src_ip,
 218                      __be32 dst_ip, u8 ip_proto, __be16 src_port,
 219                      __be16 dst_port, u32 flow_id)
 220 {
 221         struct mlx4_en_filter *filter = NULL;
 222
 223         filter = kzalloc(sizeof(struct mlx4_en_filter), GFP_ATOMIC);
 224         if (!filter)
 225                 return NULL;
 226
 227         filter->priv = priv;
 228         filter->rxq_index = rxq_index;
 229         INIT_WORK(&filter->work, mlx4_en_filter_work);
 230
 231         filter->src_ip = src_ip;
 232         filter->dst_ip = dst_ip;
 233         filter->ip_proto = ip_proto;
 234         filter->src_port = src_port;
 235         filter->dst_port = dst_port;
 236
 237         filter->flow_id = flow_id;
 238
 239         filter->id = priv->last_filter_id++ % RPS_NO_FILTER;
 240
 241         list_add_tail(&filter->next, &priv->filters);
 242         hlist_add_head(&filter->filter_chain,
 243                        filter_hash_bucket(priv, src_ip, dst_ip, src_port,
 244                                           dst_port));
 245
 246         return filter;
 247 }
 248
 249 static void mlx4_en_filter_free(struct mlx4_en_filter *filter)
 250 {
 251         struct mlx4_en_priv *priv = filter->priv;
 252         int rc;
 253
 254         list_del(&filter->next);
 255
 256         rc = mlx4_flow_detach(priv->mdev->dev, filter->reg_id);
 257         if (rc && rc != -ENOENT)
 258                 en_err(priv, "Error detaching flow. rc = %d\n", rc);
 259
 260         kfree(filter);
 261 }
 262
 263 static inline struct mlx4_en_filter *
 264 mlx4_en_filter_find(struct mlx4_en_priv *priv, __be32 src_ip, __be32 dst_ip,
 265                     u8 ip_proto, __be16 src_port, __be16 dst_port)
 266 {
 267         struct hlist_node *elem;
 268         struct mlx4_en_filter *filter;
 269         struct mlx4_en_filter *ret = NULL;
 270
 271         hlist_for_each_entry(filter, elem,
 272                              filter_hash_bucket(priv, src_ip, dst_ip,
 273                                                 src_port, dst_port),
 274                              filter_chain) {
 275                 if (filter->src_ip == src_ip &&
 276                     filter->dst_ip == dst_ip &&
 277                     filter->ip_proto == ip_proto &&
 278                     filter->src_port == src_port &&
 279                     filter->dst_port == dst_port) {
 280                         ret = filter;
 281                         break;
 282                 }
 283         }
 284
 285         return ret;
 286 }
 287
 288 static int
 289 mlx4_en_filter_rfs(struct net_device *net_dev, const struct sk_buff *skb,
 290                    u16 rxq_index, u32 flow_id)
 291 {
 292         struct mlx4_en_priv *priv = netdev_priv(net_dev);
 293         struct mlx4_en_filter *filter;
 294         const struct iphdr *ip;
 295         const __be16 *ports;
 296         u8 ip_proto;
 297         __be32 src_ip;
 298         __be32 dst_ip;
 299         __be16 src_port;
 300         __be16 dst_port;
 301         int nhoff = skb_network_offset(skb);
 302         int ret = 0;
 303
 304         if (skb->protocol != htons(ETH_P_IP))
 305                 return -EPROTONOSUPPORT;
 306
 307         ip = (const struct iphdr *)(skb->data + nhoff);
 308         if (ip_is_fragment(ip))
 309                 return -EPROTONOSUPPORT;
 310
 311         if ((ip->protocol != IPPROTO_TCP) && (ip->protocol != IPPROTO_UDP))
 312                 return -EPROTONOSUPPORT;
 313         ports = (const __be16 *)(skb->data + nhoff + 4 * ip->ihl);
 314
 315         ip_proto = ip->protocol;
 316         src_ip = ip->saddr;
 317         dst_ip = ip->daddr;
 318         src_port = ports[0];
 319         dst_port = ports[1];
 320
 321         spin_lock_bh(&priv->filters_lock);
 322         filter = mlx4_en_filter_find(priv, src_ip, dst_ip, ip_proto,
 323                                      src_port, dst_port);
 324         if (filter) {
 325                 if (filter->rxq_index == rxq_index)
 326                         goto out;
 327
 328                 filter->rxq_index = rxq_index;
 329         } else {
 330                 filter = mlx4_en_filter_alloc(priv, rxq_index,
 331                                               src_ip, dst_ip, ip_proto,
 332                                               src_port, dst_port, flow_id);
 333                 if (!filter) {
 334                         ret = -ENOMEM;
 335                         goto err;
 336                 }
 337         }
 338
 339         queue_work(priv->mdev->workqueue, &filter->work);
 340
 341 out:
 342         ret = filter->id;
 343 err:
 344         spin_unlock_bh(&priv->filters_lock);
 345
 346         return ret;
 347 }
 348
 349 void mlx4_en_cleanup_filters(struct mlx4_en_priv *priv,
 350                              struct mlx4_en_rx_ring *rx_ring)
 351 {
 352         struct mlx4_en_filter *filter, *tmp;
 353         LIST_HEAD(del_list);
 354
 355         spin_lock_bh(&priv->filters_lock);
 356         list_for_each_entry_safe(filter, tmp, &priv->filters, next) {
 357                 list_move(&filter->next, &del_list);
 358                 hlist_del(&filter->filter_chain);
 359         }
 360         spin_unlock_bh(&priv->filters_lock);
 361
 362         list_for_each_entry_safe(filter, tmp, &del_list, next) {
 363                 cancel_work_sync(&filter->work);
 364                 mlx4_en_filter_free(filter);
 365         }
 366 }
 367
 368 static void mlx4_en_filter_rfs_expire(struct mlx4_en_priv *priv)
 369 {
 370         struct mlx4_en_filter *filter = NULL, *tmp, *last_filter = NULL;
 371         LIST_HEAD(del_list);
 372         int i = 0;
 373
 374         spin_lock_bh(&priv->filters_lock);
 375         list_for_each_entry_safe(filter, tmp, &priv->filters, next) {
 376                 if (i > MLX4_EN_FILTER_EXPIRY_QUOTA)
 377                         break;
 378
 379                 if (filter->activated &&
 380                     !work_pending(&filter->work) &&
 381                     rps_may_expire_flow(priv->dev,
 382                                         filter->rxq_index, filter->flow_id,
 383                                         filter->id)) {
 384                         list_move(&filter->next, &del_list);
 385                         hlist_del(&filter->filter_chain);
 386                 } else
 387                         last_filter = filter;
 388
 389                 i++;
 390         }
 391
 392         if (last_filter && (&last_filter->next != priv->filters.next))
 393                 list_move(&priv->filters, &last_filter->next);
 394
 395         spin_unlock_bh(&priv->filters_lock);
 396
 397         list_for_each_entry_safe(filter, tmp, &del_list, next)
 398                 mlx4_en_filter_free(filter);
 399 }
 400 #endif
 401
 402 static void mlx4_en_vlan_rx_add_vid(void *arg, struct net_device *dev, u16 vid)
 403 {
 404         struct mlx4_en_priv *priv = netdev_priv(dev);
 405         struct mlx4_en_dev *mdev = priv->mdev;
 406         int err;
 407         int idx;
 408
 409         if (arg != priv)
 410                 return;
 411
 412         en_dbg(HW, priv, "adding VLAN:%d\n", vid);
 413
 414         set_bit(vid, priv->active_vlans);
 415
 416         /* Add VID to port VLAN filter */
 417         mutex_lock(&mdev->state_lock);
 418         if (mdev->device_up && priv->port_up) {
 419                 err = mlx4_SET_VLAN_FLTR(mdev->dev, priv);
 420                 if (err)
 421                         en_err(priv, "Failed configuring VLAN filter\n");
 422         }
 423         if (mlx4_register_vlan(mdev->dev, priv->port, vid, &idx))
 424                 en_dbg(HW, priv, "failed adding vlan %d\n", vid);
 425         mutex_unlock(&mdev->state_lock);
 426
 427 }
 428
 429 static void mlx4_en_vlan_rx_kill_vid(void *arg, struct net_device *dev, u16 vid)
 430 {
 431         struct mlx4_en_priv *priv = netdev_priv(dev);
 432         struct mlx4_en_dev *mdev = priv->mdev;
 433         int err;
 434
 435         if (arg != priv)
 436                 return;
 437
 438         en_dbg(HW, priv, "Killing VID:%d\n", vid);
 439
 440         clear_bit(vid, priv->active_vlans);
 441
 442         /* Remove VID from port VLAN filter */
 443         mutex_lock(&mdev->state_lock);
 444         mlx4_unregister_vlan(mdev->dev, priv->port, vid);
 445
 446         if (mdev->device_up && priv->port_up) {
 447                 err = mlx4_SET_VLAN_FLTR(mdev->dev, priv);
 448                 if (err)
 449                         en_err(priv, "Failed configuring VLAN filter\n");
 450         }
 451         mutex_unlock(&mdev->state_lock);
 452
 453 }
 454
 455 static int mlx4_en_uc_steer_add(struct mlx4_en_priv *priv,
 456                                 unsigned char *mac, int *qpn, u64 *reg_id)
 457 {
 458         struct mlx4_en_dev *mdev = priv->mdev;
 459         struct mlx4_dev *dev = mdev->dev;
 460         int err;
 461
 462         switch (dev->caps.steering_mode) {
 463         case MLX4_STEERING_MODE_B0: {
 464                 struct mlx4_qp qp;
 465                 u8 gid[16] = {0};
 466
 467                 qp.qpn = *qpn;
 468                 memcpy(&gid[10], mac, ETH_ALEN);
 469                 gid[5] = priv->port;
 470
 471                 err = mlx4_unicast_attach(dev, &qp, gid, 0, MLX4_PROT_ETH);
 472                 break;
 473         }
 474         case MLX4_STEERING_MODE_DEVICE_MANAGED: {
 475                 struct mlx4_spec_list spec_eth = { {NULL} };
 476                 __be64 mac_mask = cpu_to_be64(MLX4_MAC_MASK << 16);
 477
 478                 struct mlx4_net_trans_rule rule = {
 479                         .queue_mode = MLX4_NET_TRANS_Q_FIFO,
 480                         .exclusive = 0,
 481                         .allow_loopback = 1,
 482                         .promisc_mode = MLX4_FS_REGULAR,
 483                         .priority = MLX4_DOMAIN_NIC,
 484                 };
 485
 486                 rule.port = priv->port;
 487                 rule.qpn = *qpn;
 488                 INIT_LIST_HEAD(&rule.list);
 489
 490                 spec_eth.id = MLX4_NET_TRANS_RULE_ID_ETH;
 491                 memcpy(spec_eth.eth.dst_mac, mac, ETH_ALEN);
 492                 memcpy(spec_eth.eth.dst_mac_msk, &mac_mask, ETH_ALEN);
 493                 list_add_tail(&spec_eth.list, &rule.list);
 494
 495                 err = mlx4_flow_attach(dev, &rule, reg_id);
 496                 break;
 497         }
 498         default:
 499                 return -EINVAL;
 500         }
 501         if (err)
 502                 en_warn(priv, "Failed Attaching Unicast\n");
 503
 504         return err;
 505 }
 506
 507 static void mlx4_en_uc_steer_release(struct mlx4_en_priv *priv,
 508                                      unsigned char *mac, int qpn, u64 reg_id)
 509 {
 510         struct mlx4_en_dev *mdev = priv->mdev;
 511         struct mlx4_dev *dev = mdev->dev;
 512
 513         switch (dev->caps.steering_mode) {
 514         case MLX4_STEERING_MODE_B0: {
 515                 struct mlx4_qp qp;
 516                 u8 gid[16] = {0};
 517
 518                 qp.qpn = qpn;
 519                 memcpy(&gid[10], mac, ETH_ALEN);
 520                 gid[5] = priv->port;
 521
 522                 mlx4_unicast_detach(dev, &qp, gid, MLX4_PROT_ETH);
 523                 break;
 524         }
 525         case MLX4_STEERING_MODE_DEVICE_MANAGED: {
 526                 mlx4_flow_detach(dev, reg_id);
 527                 break;
 528         }
 529         default:
 530                 en_err(priv, "Invalid steering mode.\n");
 531         }
 532 }
 533
 534 static int mlx4_en_get_qp(struct mlx4_en_priv *priv)
 535 {
 536         struct mlx4_en_dev *mdev = priv->mdev;
 537         struct mlx4_dev *dev = mdev->dev;
 538         struct mlx4_mac_entry *entry;
 539         int index = 0;
 540         int err = 0;
 541         u64 reg_id;
 542         int *qpn = &priv->base_qpn;
 543         u64 mac = mlx4_mac_to_u64(IF_LLADDR(priv->dev));
 544
 545         en_dbg(DRV, priv, "Registering MAC: %pM for adding\n",
 546                IF_LLADDR(priv->dev));
 547         index = mlx4_register_mac(dev, priv->port, mac);
 548         if (index < 0) {
 549                 err = index;
 550                 en_err(priv, "Failed adding MAC: %pM\n",
 551                        IF_LLADDR(priv->dev));
 552                 return err;
 553         }
 554
 555         if (dev->caps.steering_mode == MLX4_STEERING_MODE_A0) {
 556                 int base_qpn = mlx4_get_base_qpn(dev, priv->port);
 557                 *qpn = base_qpn + index;
 558                 return 0;
 559         }
 560
 561         err = mlx4_qp_reserve_range(dev, 1, 1, qpn, 0);
 562         en_dbg(DRV, priv, "Reserved qp %d\n", *qpn);
 563         if (err) {
 564                 en_err(priv, "Failed to reserve qp for mac registration\n");
 565                 goto qp_err;
 566         }
 567
 568         err = mlx4_en_uc_steer_add(priv, IF_LLADDR(priv->dev), qpn, &reg_id);
 569         if (err)
 570                 goto steer_err;
 571
 572         entry = kmalloc(sizeof(*entry), GFP_KERNEL);
 573         if (!entry) {
 574                 err = -ENOMEM;
 575                 goto alloc_err;
 576         }
 577         memcpy(entry->mac, IF_LLADDR(priv->dev), sizeof(entry->mac));
 578         entry->reg_id = reg_id;
 579
 580         hlist_add_head(&entry->hlist,
 581                            &priv->mac_hash[entry->mac[MLX4_EN_MAC_HASH_IDX]]);
 582
 583         return 0;
 584
 585 alloc_err:
 586         mlx4_en_uc_steer_release(priv, IF_LLADDR(priv->dev), *qpn, reg_id);
 587
 588 steer_err:
 589         mlx4_qp_release_range(dev, *qpn, 1);
 590
 591 qp_err:
 592         mlx4_unregister_mac(dev, priv->port, mac);
 593         return err;
 594 }
 595
 596 static void mlx4_en_put_qp(struct mlx4_en_priv *priv)
 597 {
 598         struct mlx4_en_dev *mdev = priv->mdev;
 599         struct mlx4_dev *dev = mdev->dev;
 600         int qpn = priv->base_qpn;
 601         u64 mac;
 602
 603         if (dev->caps.steering_mode == MLX4_STEERING_MODE_A0) {
 604                 mac = mlx4_mac_to_u64(IF_LLADDR(priv->dev));
 605                 en_dbg(DRV, priv, "Registering MAC: %pM for deleting\n",
 606                        IF_LLADDR(priv->dev));
 607                 mlx4_unregister_mac(dev, priv->port, mac);
 608         } else {
 609                 struct mlx4_mac_entry *entry;
 610                 struct hlist_node *n, *tmp;
 611                 struct hlist_head *bucket;
 612                 unsigned int i;
 613
 614                 for (i = 0; i < MLX4_EN_MAC_HASH_SIZE; ++i) {
 615                         bucket = &priv->mac_hash[i];
 616                         hlist_for_each_entry_safe(entry, n, tmp, bucket, hlist) {
 617                                 mac = mlx4_mac_to_u64(entry->mac);
 618                                 en_dbg(DRV, priv, "Registering MAC: %pM for deleting\n",
 619                                        entry->mac);
 620                                 mlx4_en_uc_steer_release(priv, entry->mac,
 621                                                          qpn, entry->reg_id);
 622
 623                                 mlx4_unregister_mac(dev, priv->port, mac);
 624                                 hlist_del(&entry->hlist);
 625                                 kfree(entry);
 626                         }
 627                 }
 628
 629                 en_dbg(DRV, priv, "Releasing qp: port %d, qpn %d\n",
 630                        priv->port, qpn);
 631                 mlx4_qp_release_range(dev, qpn, 1);
 632                 priv->flags &= ~MLX4_EN_FLAG_FORCE_PROMISC;
 633         }
 634 }
 635
 636 static void mlx4_en_clear_list(struct net_device *dev)
 637 {
 638         struct mlx4_en_priv *priv = netdev_priv(dev);
 639         struct mlx4_en_mc_list *tmp, *mc_to_del;
 640
 641         list_for_each_entry_safe(mc_to_del, tmp, &priv->mc_list, list) {
 642                 list_del(&mc_to_del->list);
 643                 kfree(mc_to_del);
 644         }
 645 }
 646
 647 static void mlx4_en_cache_mclist(struct net_device *dev)
 648 {
 649         struct ifmultiaddr *ifma;
 650         struct mlx4_en_mc_list *tmp;
 651         struct mlx4_en_priv *priv = netdev_priv(dev);
 652
 653         if_maddr_rlock(dev);
 654         TAILQ_FOREACH(ifma, &dev->if_multiaddrs, ifma_link) {
 655                 if (ifma->ifma_addr->sa_family != AF_LINK)
 656                         continue;
 657                 if (((struct sockaddr_dl *)ifma->ifma_addr)->sdl_alen !=
 658                                 ETHER_ADDR_LEN)
 659                         continue;
 660                 /* Make sure the list didn't grow. */
 661                 tmp = kzalloc(sizeof(struct mlx4_en_mc_list), GFP_ATOMIC);
 662                 if (tmp == NULL) {
 663                         en_err(priv, "Failed to allocate multicast list\n");
 664                         break;
 665                 }
 666                 memcpy(tmp->addr,
 667                         LLADDR((struct sockaddr_dl *)ifma->ifma_addr), ETH_ALEN);
 668                 list_add_tail(&tmp->list, &priv->mc_list);
 669         }
 670         if_maddr_runlock(dev);
 671 }
 672
 673 static void update_mclist_flags(struct mlx4_en_priv *priv,
 674                                 struct list_head *dst,
 675                                 struct list_head *src)
 676 {
 677         struct mlx4_en_mc_list *dst_tmp, *src_tmp, *new_mc;
 678         bool found;
 679
 680         /* Find all the entries that should be removed from dst,
 681          * These are the entries that are not found in src
 682          */
 683         list_for_each_entry(dst_tmp, dst, list) {
 684                 found = false;
 685                 list_for_each_entry(src_tmp, src, list) {
 686                         if (!memcmp(dst_tmp->addr, src_tmp->addr, ETH_ALEN)) {
 687                                 found = true;
 688                                 break;
 689                         }
 690                 }
 691                 if (!found)
 692                         dst_tmp->action = MCLIST_REM;
 693         }
 694
 695         /* Add entries that exist in src but not in dst
 696          * mark them as need to add
 697          */
 698         list_for_each_entry(src_tmp, src, list) {
 699                 found = false;
 700                 list_for_each_entry(dst_tmp, dst, list) {
 701                         if (!memcmp(dst_tmp->addr, src_tmp->addr, ETH_ALEN)) {
 702                                 dst_tmp->action = MCLIST_NONE;
 703                                 found = true;
 704                                 break;
 705                         }
 706                 }
 707                 if (!found) {
 708                         new_mc = kmalloc(sizeof(struct mlx4_en_mc_list),
 709                                          GFP_KERNEL);
 710                         if (!new_mc) {
 711                                 en_err(priv, "Failed to allocate current multicast list\n");
 712                                 return;
 713                         }
 714                         memcpy(new_mc, src_tmp,
 715                                sizeof(struct mlx4_en_mc_list));
 716                         new_mc->action = MCLIST_ADD;
 717                         list_add_tail(&new_mc->list, dst);
 718                 }
 719         }
 720 }
 721
 722 static void mlx4_en_set_rx_mode(struct net_device *dev)
 723 {
 724         struct mlx4_en_priv *priv = netdev_priv(dev);
 725
 726         if (!priv->port_up)
 727                 return;
 728
 729         queue_work(priv->mdev->workqueue, &priv->rx_mode_task);
 730 }
 731
 732 static void mlx4_en_set_promisc_mode(struct mlx4_en_priv *priv,
 733                                      struct mlx4_en_dev *mdev)
 734 {
 735         int err = 0;
 736         if (!(priv->flags & MLX4_EN_FLAG_PROMISC)) {
 737                 priv->flags |= MLX4_EN_FLAG_PROMISC;
 738
 739                 /* Enable promiscouos mode */
 740                 switch (mdev->dev->caps.steering_mode) {
 741                 case MLX4_STEERING_MODE_DEVICE_MANAGED:
 742                         err = mlx4_flow_steer_promisc_add(mdev->dev,
 743                                                           priv->port,
 744                                                           priv->base_qpn,
 745                                                           MLX4_FS_ALL_DEFAULT);
 746                         if (err)
 747                                 en_err(priv, "Failed enabling promiscuous mode\n");
 748                         priv->flags |= MLX4_EN_FLAG_MC_PROMISC;
 749                         break;
 750
 751                 case MLX4_STEERING_MODE_B0:
 752                         err = mlx4_unicast_promisc_add(mdev->dev,
 753                                                        priv->base_qpn,
 754                                                        priv->port);
 755                         if (err)
 756                                 en_err(priv, "Failed enabling unicast promiscuous mode\n");
 757
 758                         /* Add the default qp number as multicast
 759                          * promisc
 760                          */
 761                         if (!(priv->flags & MLX4_EN_FLAG_MC_PROMISC)) {
 762                                 err = mlx4_multicast_promisc_add(mdev->dev,
 763                                                                  priv->base_qpn,
 764                                                                  priv->port);
 765                                 if (err)
 766                                         en_err(priv, "Failed enabling multicast promiscuous mode\n");
 767                                 priv->flags |= MLX4_EN_FLAG_MC_PROMISC;
 768                         }
 769                         break;
 770
 771                 case MLX4_STEERING_MODE_A0:
 772                         err = mlx4_SET_PORT_qpn_calc(mdev->dev,
 773                                                      priv->port,
 774                                                      priv->base_qpn,
 775                                                      1);
 776                         if (err)
 777                                 en_err(priv, "Failed enabling promiscuous mode\n");
 778                         break;
 779                 }
 780
 781                 /* Disable port multicast filter (unconditionally) */
 782                 err = mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, 0,
 783                                           0, MLX4_MCAST_DISABLE);
 784                 if (err)
 785                         en_err(priv, "Failed disabling multicast filter\n");
 786         }
 787 }
 788
 789 static void mlx4_en_clear_promisc_mode(struct mlx4_en_priv *priv,
 790                                        struct mlx4_en_dev *mdev)
 791 {
 792         int err = 0;
 793
 794         priv->flags &= ~MLX4_EN_FLAG_PROMISC;
 795
 796         /* Disable promiscouos mode */
 797         switch (mdev->dev->caps.steering_mode) {
 798         case MLX4_STEERING_MODE_DEVICE_MANAGED:
 799                 err = mlx4_flow_steer_promisc_remove(mdev->dev,
 800                                                      priv->port,
 801                                                      MLX4_FS_ALL_DEFAULT);
 802                 if (err)
 803                         en_err(priv, "Failed disabling promiscuous mode\n");
 804                 priv->flags &= ~MLX4_EN_FLAG_MC_PROMISC;
 805                 break;
 806
 807         case MLX4_STEERING_MODE_B0:
 808                 err = mlx4_unicast_promisc_remove(mdev->dev,
 809                                                   priv->base_qpn,
 810                                                   priv->port);
 811                 if (err)
 812                         en_err(priv, "Failed disabling unicast promiscuous mode\n");
 813                 /* Disable Multicast promisc */
 814                 if (priv->flags & MLX4_EN_FLAG_MC_PROMISC) {
 815                         err = mlx4_multicast_promisc_remove(mdev->dev,
 816                                                             priv->base_qpn,
 817                                                             priv->port);
 818                         if (err)
 819                                 en_err(priv, "Failed disabling multicast promiscuous mode\n");
 820                         priv->flags &= ~MLX4_EN_FLAG_MC_PROMISC;
 821                 }
 822                 break;
 823
 824         case MLX4_STEERING_MODE_A0:
 825                 err = mlx4_SET_PORT_qpn_calc(mdev->dev,
 826                                              priv->port,
 827                                              priv->base_qpn, 0);
 828                 if (err)
 829                         en_err(priv, "Failed disabling promiscuous mode\n");
 830                 break;
 831         }
 832 }
 833
 834 static void mlx4_en_do_multicast(struct mlx4_en_priv *priv,
 835                                  struct net_device *dev,
 836                                  struct mlx4_en_dev *mdev)
 837 {
 838         struct mlx4_en_mc_list *mclist, *tmp;
 839         u8 mc_list[16] = {0};
 840         int err = 0;
 841         u64 mcast_addr = 0;
 842
 843
 844         /* Enable/disable the multicast filter according to IFF_ALLMULTI */
 845         if (dev->if_flags & IFF_ALLMULTI) {
 846                 err = mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, 0,
 847                                           0, MLX4_MCAST_DISABLE);
 848                 if (err)
 849                         en_err(priv, "Failed disabling multicast filter\n");
 850
 851                 /* Add the default qp number as multicast promisc */
 852                 if (!(priv->flags & MLX4_EN_FLAG_MC_PROMISC)) {
 853                         switch (mdev->dev->caps.steering_mode) {
 854                         case MLX4_STEERING_MODE_DEVICE_MANAGED:
 855                                 err = mlx4_flow_steer_promisc_add(mdev->dev,
 856                                                                   priv->port,
 857                                                                   priv->base_qpn,
 858                                                                   MLX4_FS_MC_DEFAULT);
 859                                 break;
 860
 861                         case MLX4_STEERING_MODE_B0:
 862                                 err = mlx4_multicast_promisc_add(mdev->dev,
 863                                                                  priv->base_qpn,
 864                                                                  priv->port);
 865                                 break;
 866
 867                         case MLX4_STEERING_MODE_A0:
 868                                 break;
 869                         }
 870                         if (err)
 871                                 en_err(priv, "Failed entering multicast promisc mode\n");
 872                         priv->flags |= MLX4_EN_FLAG_MC_PROMISC;
 873                 }
 874         } else {
 875                 /* Disable Multicast promisc */
 876                 if (priv->flags & MLX4_EN_FLAG_MC_PROMISC) {
 877                         switch (mdev->dev->caps.steering_mode) {
 878                         case MLX4_STEERING_MODE_DEVICE_MANAGED:
 879                                 err = mlx4_flow_steer_promisc_remove(mdev->dev,
 880                                                                      priv->port,
 881                                                                      MLX4_FS_MC_DEFAULT);
 882                                 break;
 883
 884                         case MLX4_STEERING_MODE_B0:
 885                                 err = mlx4_multicast_promisc_remove(mdev->dev,
 886                                                                     priv->base_qpn,
 887                                                                     priv->port);
 888                                 break;
 889
 890                         case MLX4_STEERING_MODE_A0:
 891                                 break;
 892                         }
 893                         if (err)
 894                                 en_err(priv, "Failed disabling multicast promiscuous mode\n");
 895                         priv->flags &= ~MLX4_EN_FLAG_MC_PROMISC;
 896                 }
 897
 898                 err = mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, 0,
 899                                           0, MLX4_MCAST_DISABLE);
 900                 if (err)
 901                         en_err(priv, "Failed disabling multicast filter\n");
 902
 903                 /* Flush mcast filter and init it with broadcast address */
 904                 mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, ETH_BCAST,
 905                                     1, MLX4_MCAST_CONFIG);
 906
 907                 /* Update multicast list - we cache all addresses so they won't
 908                  * change while HW is updated holding the command semaphor */
 909                 mlx4_en_cache_mclist(dev);
 910                 list_for_each_entry(mclist, &priv->mc_list, list) {
 911                         mcast_addr = mlx4_mac_to_u64(mclist->addr);
 912                         mlx4_SET_MCAST_FLTR(mdev->dev, priv->port,
 913                                         mcast_addr, 0, MLX4_MCAST_CONFIG);
 914                 }
 915                 err = mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, 0,
 916                                           0, MLX4_MCAST_ENABLE);
 917                 if (err)
 918                         en_err(priv, "Failed enabling multicast filter\n");
 919
 920                 update_mclist_flags(priv, &priv->curr_list, &priv->mc_list);
 921                 list_for_each_entry_safe(mclist, tmp, &priv->curr_list, list) {
 922                         if (mclist->action == MCLIST_REM) {
 923                                 /* detach this address and delete from list */
 924                                 memcpy(&mc_list[10], mclist->addr, ETH_ALEN);
 925                                 mc_list[5] = priv->port;
 926                                 err = mlx4_multicast_detach(mdev->dev,
 927                                                             &priv->rss_map.indir_qp,
 928                                                             mc_list,
 929                                                             MLX4_PROT_ETH,
 930                                                             mclist->reg_id);
 931                                 if (err)
 932                                         en_err(priv, "Fail to detach multicast address\n");
 933
 934                                 /* remove from list */
 935                                 list_del(&mclist->list);
 936                                 kfree(mclist);
 937                         } else if (mclist->action == MCLIST_ADD) {
 938                                 /* attach the address */
 939                                 memcpy(&mc_list[10], mclist->addr, ETH_ALEN);
 940                                 /* needed for B0 steering support */
 941                                 mc_list[5] = priv->port;
 942                                 err = mlx4_multicast_attach(mdev->dev,
 943                                                             &priv->rss_map.indir_qp,
 944                                                             mc_list,
 945                                                             priv->port, 0,
 946                                                             MLX4_PROT_ETH,
 947                                                             &mclist->reg_id);
 948                                 if (err)
 949                                         en_err(priv, "Fail to attach multicast address\n");
 950
 951                         }
 952                 }
 953         }
 954 }
 955
 956 static void mlx4_en_do_set_rx_mode(struct work_struct *work)
 957 {
 958         struct mlx4_en_priv *priv = container_of(work, struct mlx4_en_priv,
 959                                                  rx_mode_task);
 960         struct mlx4_en_dev *mdev = priv->mdev;
 961         struct net_device *dev = priv->dev;
 962
 963
 964         mutex_lock(&mdev->state_lock);
 965         if (!mdev->device_up) {
 966                 en_dbg(HW, priv, "Card is not up, ignoring rx mode change.\n");
 967                 goto out;
 968         }
 969         if (!priv->port_up) {
 970                 en_dbg(HW, priv, "Port is down, ignoring rx mode change.\n");
 971                 goto out;
 972         }
 973         if (!mlx4_en_QUERY_PORT(mdev, priv->port)) {
 974                 if (priv->port_state.link_state) {
 975                         priv->last_link_state = MLX4_DEV_EVENT_PORT_UP;
 976                         /* update netif baudrate */
 977                         priv->dev->if_baudrate =
 978                             IF_Mbps(priv->port_state.link_speed);
 979                         /* Important note: the following call for if_link_state_change
 980                          * is needed for interface up scenario (start port, link state
 981                          * change) */
 982                         if_link_state_change(priv->dev, LINK_STATE_UP);
 983                         en_dbg(HW, priv, "Link Up\n");
 984                 }
 985         }
 986
 987         /* Promsicuous mode: disable all filters */
 988         if ((dev->if_flags & IFF_PROMISC) ||
 989             (priv->flags & MLX4_EN_FLAG_FORCE_PROMISC)) {
 990                 mlx4_en_set_promisc_mode(priv, mdev);
 991                 goto out;
 992         }
 993
 994         /* Not in promiscuous mode */
 995         if (priv->flags & MLX4_EN_FLAG_PROMISC)
 996                 mlx4_en_clear_promisc_mode(priv, mdev);
 997
 998         mlx4_en_do_multicast(priv, dev, mdev);
 999 out:
1000         mutex_unlock(&mdev->state_lock);
1001 }
1002
1003 #ifdef CONFIG_NET_POLL_CONTROLLER
1004 static void mlx4_en_netpoll(struct net_device *dev)
1005 {
1006         struct mlx4_en_priv *priv = netdev_priv(dev);
1007         struct mlx4_en_cq *cq;
1008         unsigned long flags;
1009         int i;
1010
1011         for (i = 0; i < priv->rx_ring_num; i++) {
1012                 cq = priv->rx_cq[i];
1013                 spin_lock_irqsave(&cq->lock, flags);
1014                 napi_synchronize(&cq->napi);
1015                 mlx4_en_process_rx_cq(dev, cq, 0);
1016                 spin_unlock_irqrestore(&cq->lock, flags);
1017         }
1018 }
1019 #endif
1020
1021 static void mlx4_en_watchdog_timeout(void *arg)
1022 {
1023         struct mlx4_en_priv *priv = arg;
1024         struct mlx4_en_dev *mdev = priv->mdev;
1025
1026         en_dbg(DRV, priv, "Scheduling watchdog\n");
1027         queue_work(mdev->workqueue, &priv->watchdog_task);
1028         if (priv->port_up)
1029                 callout_reset(&priv->watchdog_timer, MLX4_EN_WATCHDOG_TIMEOUT,
1030                                 mlx4_en_watchdog_timeout, priv);
1031 }
1032
1033
1034
1035 static void mlx4_en_set_default_moderation(struct mlx4_en_priv *priv)
1036 {
1037         struct mlx4_en_cq *cq;
1038         int i;
1039
1040         /* If we haven't received a specific coalescing setting
1041          * (module param), we set the moderation parameters as follows:
1042          * - moder_cnt is set to the number of mtu sized packets to
1043          *   satisfy our coelsing target.
1044          * - moder_time is set to a fixed value.
1045          */
1046         priv->rx_frames = MLX4_EN_RX_COAL_TARGET / priv->dev->if_mtu + 1;
1047         priv->rx_usecs = MLX4_EN_RX_COAL_TIME;
1048         priv->tx_frames = MLX4_EN_TX_COAL_PKTS;
1049         priv->tx_usecs = MLX4_EN_TX_COAL_TIME;
1050         en_dbg(INTR, priv, "Default coalesing params for mtu: %u - "
1051                "rx_frames:%d rx_usecs:%d\n",
1052                (unsigned)priv->dev->if_mtu, priv->rx_frames, priv->rx_usecs);
1053
1054         /* Setup cq moderation params */
1055         for (i = 0; i < priv->rx_ring_num; i++) {
1056                 cq = priv->rx_cq[i];
1057                 cq->moder_cnt = priv->rx_frames;
1058                 cq->moder_time = priv->rx_usecs;
1059                 priv->last_moder_time[i] = MLX4_EN_AUTO_CONF;
1060                 priv->last_moder_packets[i] = 0;
1061                 priv->last_moder_bytes[i] = 0;
1062         }
1063
1064         for (i = 0; i < priv->tx_ring_num; i++) {
1065                 cq = priv->tx_cq[i];
1066                 cq->moder_cnt = priv->tx_frames;
1067                 cq->moder_time = priv->tx_usecs;
1068         }
1069
1070         /* Reset auto-moderation params */
1071         priv->pkt_rate_low = MLX4_EN_RX_RATE_LOW;
1072         priv->rx_usecs_low = MLX4_EN_RX_COAL_TIME_LOW;
1073         priv->pkt_rate_high = MLX4_EN_RX_RATE_HIGH;
1074         priv->rx_usecs_high = MLX4_EN_RX_COAL_TIME_HIGH;
1075         priv->sample_interval = MLX4_EN_SAMPLE_INTERVAL;
1076         priv->adaptive_rx_coal = 1;
1077         priv->last_moder_jiffies = 0;
1078         priv->last_moder_tx_packets = 0;
1079 }
1080
1081 static void mlx4_en_auto_moderation(struct mlx4_en_priv *priv)
1082 {
1083         unsigned long period = (unsigned long) (jiffies - priv->last_moder_jiffies);
1084         struct mlx4_en_cq *cq;
1085         unsigned long packets;
1086         unsigned long rate;
1087         unsigned long avg_pkt_size;
1088         unsigned long rx_packets;
1089         unsigned long rx_bytes;
1090         unsigned long rx_pkt_diff;
1091         int moder_time;
1092         int ring, err;
1093
1094         if (!priv->adaptive_rx_coal || period < priv->sample_interval * HZ)
1095                 return;
1096
1097         for (ring = 0; ring < priv->rx_ring_num; ring++) {
1098                 spin_lock(&priv->stats_lock);
1099                 rx_packets = priv->rx_ring[ring]->packets;
1100                 rx_bytes = priv->rx_ring[ring]->bytes;
1101                 spin_unlock(&priv->stats_lock);
1102
1103                 rx_pkt_diff = ((unsigned long) (rx_packets -
1104                                 priv->last_moder_packets[ring]));
1105                 packets = rx_pkt_diff;
1106                 rate = packets * HZ / period;
1107                 avg_pkt_size = packets ? ((unsigned long) (rx_bytes -
1108                                 priv->last_moder_bytes[ring])) / packets : 0;
1109
1110                 /* Apply auto-moderation only when packet rate
1111                  * exceeds a rate that it matters */
1112                 if (rate > (MLX4_EN_RX_RATE_THRESH / priv->rx_ring_num) &&
1113                     avg_pkt_size > MLX4_EN_AVG_PKT_SMALL) {
1114                         if (rate < priv->pkt_rate_low)
1115                                 moder_time = priv->rx_usecs_low;
1116                         else if (rate > priv->pkt_rate_high)
1117                                 moder_time = priv->rx_usecs_high;
1118                         else
1119                                 moder_time = (rate - priv->pkt_rate_low) *
1120                                         (priv->rx_usecs_high - priv->rx_usecs_low) /
1121                                         (priv->pkt_rate_high - priv->pkt_rate_low) +
1122                                         priv->rx_usecs_low;
1123                 } else {
1124                         moder_time = priv->rx_usecs_low;
1125                 }
1126
1127                 if (moder_time != priv->last_moder_time[ring]) {
1128                         priv->last_moder_time[ring] = moder_time;
1129                         cq = priv->rx_cq[ring];
1130                         cq->moder_time = moder_time;
1131                         err = mlx4_en_set_cq_moder(priv, cq);
1132                         if (err)
1133                                 en_err(priv, "Failed modifying moderation for cq:%d\n",
1134                                        ring);
1135                 }
1136                 priv->last_moder_packets[ring] = rx_packets;
1137                 priv->last_moder_bytes[ring] = rx_bytes;
1138         }
1139
1140         priv->last_moder_jiffies = jiffies;
1141 }
1142
1143 static void mlx4_en_do_get_stats(struct work_struct *work)
1144 {
1145         struct delayed_work *delay = to_delayed_work(work);
1146         struct mlx4_en_priv *priv = container_of(delay, struct mlx4_en_priv,
1147                                                  stats_task);
1148         struct mlx4_en_dev *mdev = priv->mdev;
1149         int err;
1150
1151         mutex_lock(&mdev->state_lock);
1152         if (mdev->device_up) {
1153                 if (priv->port_up) {
1154                         err = mlx4_en_DUMP_ETH_STATS(mdev, priv->port, 0);
1155                         if (err)
1156                                 en_dbg(HW, priv, "Could not update stats\n");
1157
1158                         mlx4_en_auto_moderation(priv);
1159                 }
1160
1161                 queue_delayed_work(mdev->workqueue, &priv->stats_task, STATS_DELAY);
1162         }
1163         mutex_unlock(&mdev->state_lock);
1164 }
1165
1166 /* mlx4_en_service_task - Run service task for tasks that needed to be done
1167  * periodically
1168  */
1169 static void mlx4_en_service_task(struct work_struct *work)
1170 {
1171         struct delayed_work *delay = to_delayed_work(work);
1172         struct mlx4_en_priv *priv = container_of(delay, struct mlx4_en_priv,
1173                                                  service_task);
1174         struct mlx4_en_dev *mdev = priv->mdev;
1175
1176         mutex_lock(&mdev->state_lock);
1177         if (mdev->device_up) {
1178                 queue_delayed_work(mdev->workqueue, &priv->service_task,
1179                                    SERVICE_TASK_DELAY);
1180         }
1181         mutex_unlock(&mdev->state_lock);
1182 }
1183
1184 static void mlx4_en_linkstate(struct work_struct *work)
1185 {
1186         struct mlx4_en_priv *priv = container_of(work, struct mlx4_en_priv,
1187                                                  linkstate_task);
1188         struct mlx4_en_dev *mdev = priv->mdev;
1189         int linkstate = priv->link_state;
1190
1191         mutex_lock(&mdev->state_lock);
1192         /* If observable port state changed set carrier state and
1193          * report to system log */
1194         if (priv->last_link_state != linkstate) {
1195                 if (linkstate == MLX4_DEV_EVENT_PORT_DOWN) {
1196                         en_info(priv, "Link Down\n");
1197                         if_link_state_change(priv->dev, LINK_STATE_DOWN);
1198                         /* update netif baudrate */
1199                         priv->dev->if_baudrate = 0;
1200
1201                 /* make sure the port is up before notifying the OS.
1202                  * This is tricky since we get here on INIT_PORT and
1203                  * in such case we can't tell the OS the port is up.
1204                  * To solve this there is a call to if_link_state_change
1205                  * in set_rx_mode.
1206                  * */
1207                 } else if (priv->port_up && (linkstate == MLX4_DEV_EVENT_PORT_UP)){
1208                         if (mlx4_en_QUERY_PORT(priv->mdev, priv->port))
1209                                 en_info(priv, "Query port failed\n");
1210                         priv->dev->if_baudrate =
1211                             IF_Mbps(priv->port_state.link_speed);
1212                         en_info(priv, "Link Up\n");
1213                         if_link_state_change(priv->dev, LINK_STATE_UP);
1214                 }
1215         }
1216         priv->last_link_state = linkstate;
1217         mutex_unlock(&mdev->state_lock);
1218 }
1219
1220
1221 int mlx4_en_start_port(struct net_device *dev)
1222 {
1223         struct mlx4_en_priv *priv = netdev_priv(dev);
1224         struct mlx4_en_dev *mdev = priv->mdev;
1225         struct mlx4_en_cq *cq;
1226         struct mlx4_en_tx_ring *tx_ring;
1227         int rx_index = 0;
1228         int tx_index = 0;
1229         int err = 0;
1230         int i;
1231         int j;
1232         u8 mc_list[16] = {0};
1233
1234
1235         if (priv->port_up) {
1236                 en_dbg(DRV, priv, "start port called while port already up\n");
1237                 return 0;
1238         }
1239
1240         INIT_LIST_HEAD(&priv->mc_list);
1241         INIT_LIST_HEAD(&priv->curr_list);
1242         INIT_LIST_HEAD(&priv->ethtool_list);
1243
1244         /* Calculate Rx buf size */
1245         dev->if_mtu = min(dev->if_mtu, priv->max_mtu);
1246         mlx4_en_calc_rx_buf(dev);
1247         en_dbg(DRV, priv, "Rx buf size:%d\n", priv->rx_mb_size);
1248
1249         /* Configure rx cq's and rings */
1250         err = mlx4_en_activate_rx_rings(priv);
1251         if (err) {
1252                 en_err(priv, "Failed to activate RX rings\n");
1253                 return err;
1254         }
1255         for (i = 0; i < priv->rx_ring_num; i++) {
1256                 cq = priv->rx_cq[i];
1257
1258                 mlx4_en_cq_init_lock(cq);
1259                 err = mlx4_en_activate_cq(priv, cq, i);
1260                 if (err) {
1261                         en_err(priv, "Failed activating Rx CQ\n");
1262                         goto cq_err;
1263                 }
1264                 for (j = 0; j < cq->size; j++)
1265                         cq->buf[j].owner_sr_opcode = MLX4_CQE_OWNER_MASK;
1266                 err = mlx4_en_set_cq_moder(priv, cq);
1267                 if (err) {
1268                         en_err(priv, "Failed setting cq moderation parameters");
1269                         mlx4_en_deactivate_cq(priv, cq);
1270                         goto cq_err;
1271                 }
1272                 mlx4_en_arm_cq(priv, cq);
1273                 priv->rx_ring[i]->cqn = cq->mcq.cqn;
1274                 ++rx_index;
1275         }
1276
1277         /* Set qp number */
1278         en_dbg(DRV, priv, "Getting qp number for port %d\n", priv->port);
1279         err = mlx4_en_get_qp(priv);
1280         if (err) {
1281                 en_err(priv, "Failed getting eth qp\n");
1282                 goto cq_err;
1283         }
1284         mdev->mac_removed[priv->port] = 0;
1285
1286         /* gets default allocated counter index from func cap */
1287         /* or sink counter index if no resources */
1288         priv->counter_index = mdev->dev->caps.def_counter_index[priv->port - 1];
1289
1290         en_dbg(DRV, priv, "%s: default counter index %d for port %d\n",
1291                __func__, priv->counter_index, priv->port);
1292
1293         err = mlx4_en_config_rss_steer(priv);
1294         if (err) {
1295                 en_err(priv, "Failed configuring rss steering\n");
1296                 goto mac_err;
1297         }
1298
1299         err = mlx4_en_create_drop_qp(priv);
1300         if (err)
1301                 goto rss_err;
1302
1303         /* Configure tx cq's and rings */
1304         for (i = 0; i < priv->tx_ring_num; i++) {
1305                 /* Configure cq */
1306                 cq = priv->tx_cq[i];
1307                 err = mlx4_en_activate_cq(priv, cq, i);
1308                 if (err) {
1309                         en_err(priv, "Failed activating Tx CQ\n");
1310                         goto tx_err;
1311                 }
1312                 err = mlx4_en_set_cq_moder(priv, cq);
1313                 if (err) {
1314                         en_err(priv, "Failed setting cq moderation parameters");
1315                         mlx4_en_deactivate_cq(priv, cq);
1316                         goto tx_err;
1317                 }
1318                 en_dbg(DRV, priv, "Resetting index of collapsed CQ:%d to -1\n", i);
1319                 cq->buf->wqe_index = cpu_to_be16(0xffff);
1320
1321                 /* Configure ring */
1322                 tx_ring = priv->tx_ring[i];
1323
1324                 err = mlx4_en_activate_tx_ring(priv, tx_ring, cq->mcq.cqn,
1325                                                i / priv->num_tx_rings_p_up);
1326                 if (err) {
1327                         en_err(priv, "Failed activating Tx ring %d\n", i);
1328                         mlx4_en_deactivate_cq(priv, cq);
1329                         goto tx_err;
1330                 }
1331
1332                 /* Arm CQ for TX completions */
1333                 mlx4_en_arm_cq(priv, cq);
1334
1335                 /* Set initial ownership of all Tx TXBBs to SW (1) */
1336                 for (j = 0; j < tx_ring->buf_size; j += STAMP_STRIDE)
1337                         *((u32 *) (tx_ring->buf + j)) = 0xffffffff;
1338                 ++tx_index;
1339         }
1340
1341         /* Configure port */
1342         err = mlx4_SET_PORT_general(mdev->dev, priv->port,
1343                                     priv->rx_mb_size,
1344                                     priv->prof->tx_pause,
1345                                     priv->prof->tx_ppp,
1346                                     priv->prof->rx_pause,
1347                                     priv->prof->rx_ppp);
1348         if (err) {
1349                 en_err(priv, "Failed setting port general configurations for port %d, with error %d\n",
1350                        priv->port, err);
1351                 goto tx_err;
1352         }
1353         /* Set default qp number */
1354         err = mlx4_SET_PORT_qpn_calc(mdev->dev, priv->port, priv->base_qpn, 0);
1355         if (err) {
1356                 en_err(priv, "Failed setting default qp numbers\n");
1357                 goto tx_err;
1358         }
1359
1360         /* Init port */
1361         en_dbg(HW, priv, "Initializing port\n");
1362         err = mlx4_INIT_PORT(mdev->dev, priv->port);
1363         if (err) {
1364                 en_err(priv, "Failed Initializing port\n");
1365                 goto tx_err;
1366         }
1367
1368         /* Attach rx QP to bradcast address */
1369         memset(&mc_list[10], 0xff, ETH_ALEN);
1370         mc_list[5] = priv->port; /* needed for B0 steering support */
1371         if (mlx4_multicast_attach(mdev->dev, &priv->rss_map.indir_qp, mc_list,
1372                                   priv->port, 0, MLX4_PROT_ETH,
1373                                   &priv->broadcast_id))
1374                 mlx4_warn(mdev, "Failed Attaching Broadcast\n");
1375
1376         /* Must redo promiscuous mode setup. */
1377         priv->flags &= ~(MLX4_EN_FLAG_PROMISC | MLX4_EN_FLAG_MC_PROMISC);
1378
1379         /* Schedule multicast task to populate multicast list */
1380         queue_work(mdev->workqueue, &priv->rx_mode_task);
1381
1382         mlx4_set_stats_bitmap(mdev->dev, priv->stats_bitmap);
1383
1384         priv->port_up = true;
1385
1386         /* Enable the queues. */
1387         dev->if_drv_flags &= ~IFF_DRV_OACTIVE;
1388         dev->if_drv_flags |= IFF_DRV_RUNNING;
1389 #ifdef CONFIG_DEBUG_FS
1390         mlx4_en_create_debug_files(priv);
1391 #endif
1392         callout_reset(&priv->watchdog_timer, MLX4_EN_WATCHDOG_TIMEOUT,
1393                     mlx4_en_watchdog_timeout, priv);
1394
1395
1396         return 0;
1397
1398 tx_err:
1399         while (tx_index--) {
1400                 mlx4_en_deactivate_tx_ring(priv, priv->tx_ring[tx_index]);
1401                 mlx4_en_deactivate_cq(priv, priv->tx_cq[tx_index]);
1402         }
1403         mlx4_en_destroy_drop_qp(priv);
1404 rss_err:
1405         mlx4_en_release_rss_steer(priv);
1406 mac_err:
1407         mlx4_en_put_qp(priv);
1408 cq_err:
1409         while (rx_index--)
1410                 mlx4_en_deactivate_cq(priv, priv->rx_cq[rx_index]);
1411         for (i = 0; i < priv->rx_ring_num; i++)
1412                 mlx4_en_deactivate_rx_ring(priv, priv->rx_ring[i]);
1413
1414         return err; /* need to close devices */
1415 }
1416
1417
1418 void mlx4_en_stop_port(struct net_device *dev)
1419 {
1420         struct mlx4_en_priv *priv = netdev_priv(dev);
1421         struct mlx4_en_dev *mdev = priv->mdev;
1422         struct mlx4_en_mc_list *mclist, *tmp;
1423         int i;
1424         u8 mc_list[16] = {0};
1425
1426         if (!priv->port_up) {
1427                 en_dbg(DRV, priv, "stop port called while port already down\n");
1428                 return;
1429         }
1430
1431 #ifdef CONFIG_DEBUG_FS
1432         mlx4_en_delete_debug_files(priv);
1433 #endif
1434
1435         /* close port*/
1436         mlx4_CLOSE_PORT(mdev->dev, priv->port);
1437
1438         /* Set port as not active */
1439         priv->port_up = false;
1440         if (priv->counter_index != 0xff) {
1441                 mlx4_counter_free(mdev->dev, priv->port, priv->counter_index);
1442                 priv->counter_index = 0xff;
1443         }
1444
1445         /* Promsicuous mode */
1446         if (mdev->dev->caps.steering_mode ==
1447             MLX4_STEERING_MODE_DEVICE_MANAGED) {
1448                 priv->flags &= ~(MLX4_EN_FLAG_PROMISC |
1449                                  MLX4_EN_FLAG_MC_PROMISC);
1450                 mlx4_flow_steer_promisc_remove(mdev->dev,
1451                                                priv->port,
1452                                                MLX4_FS_ALL_DEFAULT);
1453                 mlx4_flow_steer_promisc_remove(mdev->dev,
1454                                                priv->port,
1455                                                MLX4_FS_MC_DEFAULT);
1456         } else if (priv->flags & MLX4_EN_FLAG_PROMISC) {
1457                 priv->flags &= ~MLX4_EN_FLAG_PROMISC;
1458
1459                 /* Disable promiscouos mode */
1460                 mlx4_unicast_promisc_remove(mdev->dev, priv->base_qpn,
1461                                             priv->port);
1462
1463                 /* Disable Multicast promisc */
1464                 if (priv->flags & MLX4_EN_FLAG_MC_PROMISC) {
1465                         mlx4_multicast_promisc_remove(mdev->dev, priv->base_qpn,
1466                                                       priv->port);
1467                         priv->flags &= ~MLX4_EN_FLAG_MC_PROMISC;
1468                 }
1469         }
1470
1471         /* Detach All multicasts */
1472         memset(&mc_list[10], 0xff, ETH_ALEN);
1473         mc_list[5] = priv->port; /* needed for B0 steering support */
1474         mlx4_multicast_detach(mdev->dev, &priv->rss_map.indir_qp, mc_list,
1475                               MLX4_PROT_ETH, priv->broadcast_id);
1476         list_for_each_entry(mclist, &priv->curr_list, list) {
1477                 memcpy(&mc_list[10], mclist->addr, ETH_ALEN);
1478                 mc_list[5] = priv->port;
1479                 mlx4_multicast_detach(mdev->dev, &priv->rss_map.indir_qp,
1480                                       mc_list, MLX4_PROT_ETH, mclist->reg_id);
1481         }
1482         mlx4_en_clear_list(dev);
1483         list_for_each_entry_safe(mclist, tmp, &priv->curr_list, list) {
1484                 list_del(&mclist->list);
1485                 kfree(mclist);
1486         }
1487
1488         /* Flush multicast filter */
1489         mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, 0, 1, MLX4_MCAST_CONFIG);
1490         mlx4_en_destroy_drop_qp(priv);
1491
1492         /* Free TX Rings */
1493         for (i = 0; i < priv->tx_ring_num; i++) {
1494                 mlx4_en_deactivate_tx_ring(priv, priv->tx_ring[i]);
1495                 mlx4_en_deactivate_cq(priv, priv->tx_cq[i]);
1496         }
1497         msleep(10);
1498
1499         for (i = 0; i < priv->tx_ring_num; i++)
1500                 mlx4_en_free_tx_buf(dev, priv->tx_ring[i]);
1501
1502         /* Free RSS qps */
1503         mlx4_en_release_rss_steer(priv);
1504
1505         /* Unregister Mac address for the port */
1506         mlx4_en_put_qp(priv);
1507         mdev->mac_removed[priv->port] = 1;
1508
1509         /* Free RX Rings */
1510         for (i = 0; i < priv->rx_ring_num; i++) {
1511                 struct mlx4_en_cq *cq = priv->rx_cq[i];
1512                 mlx4_en_deactivate_rx_ring(priv, priv->rx_ring[i]);
1513                 mlx4_en_deactivate_cq(priv, cq);
1514         }
1515
1516         callout_stop(&priv->watchdog_timer);
1517
1518         dev->if_drv_flags &= ~(IFF_DRV_RUNNING | IFF_DRV_OACTIVE);
1519 }
1520
1521 static void mlx4_en_restart(struct work_struct *work)
1522 {
1523         struct mlx4_en_priv *priv = container_of(work, struct mlx4_en_priv,
1524                                                  watchdog_task);
1525         struct mlx4_en_dev *mdev = priv->mdev;
1526         struct net_device *dev = priv->dev;
1527         struct mlx4_en_tx_ring *ring;
1528         int i;
1529
1530
1531         if (priv->blocked == 0 || priv->port_up == 0)
1532                 return;
1533         for (i = 0; i < priv->tx_ring_num; i++) {
1534                 ring = priv->tx_ring[i];
1535                 if (ring->blocked &&
1536                                 ring->watchdog_time + MLX4_EN_WATCHDOG_TIMEOUT < ticks)
1537                         goto reset;
1538         }
1539         return;
1540
1541 reset:
1542         priv->port_stats.tx_timeout++;
1543         en_dbg(DRV, priv, "Watchdog task called for port %d\n", priv->port);
1544
1545         mutex_lock(&mdev->state_lock);
1546         if (priv->port_up) {
1547                 mlx4_en_stop_port(dev);
1548                 //for (i = 0; i < priv->tx_ring_num; i++)
1549                 //        netdev_tx_reset_queue(priv->tx_ring[i]->tx_queue);
1550                 if (mlx4_en_start_port(dev))
1551                         en_err(priv, "Failed restarting port %d\n", priv->port);
1552         }
1553         mutex_unlock(&mdev->state_lock);
1554 }
1555
1556 static void mlx4_en_clear_stats(struct net_device *dev)
1557 {
1558         struct mlx4_en_priv *priv = netdev_priv(dev);
1559         struct mlx4_en_dev *mdev = priv->mdev;
1560         int i;
1561
1562         if (!mlx4_is_slave(mdev->dev))
1563                 if (mlx4_en_DUMP_ETH_STATS(mdev, priv->port, 1))
1564                         en_dbg(HW, priv, "Failed dumping statistics\n");
1565
1566         memset(&priv->pstats, 0, sizeof(priv->pstats));
1567         memset(&priv->pkstats, 0, sizeof(priv->pkstats));
1568         memset(&priv->port_stats, 0, sizeof(priv->port_stats));
1569         memset(&priv->vport_stats, 0, sizeof(priv->vport_stats));
1570
1571         for (i = 0; i < priv->tx_ring_num; i++) {
1572                 priv->tx_ring[i]->bytes = 0;
1573                 priv->tx_ring[i]->packets = 0;
1574                 priv->tx_ring[i]->tx_csum = 0;
1575                 priv->tx_ring[i]->oversized_packets = 0;
1576         }
1577         for (i = 0; i < priv->rx_ring_num; i++) {
1578                 priv->rx_ring[i]->bytes = 0;
1579                 priv->rx_ring[i]->packets = 0;
1580                 priv->rx_ring[i]->csum_ok = 0;
1581                 priv->rx_ring[i]->csum_none = 0;
1582         }
1583 }
1584
1585 static void mlx4_en_open(void* arg)
1586 {
1587
1588         struct mlx4_en_priv *priv;
1589         struct mlx4_en_dev *mdev;
1590         struct net_device *dev;
1591         int err = 0;
1592
1593         priv = arg;
1594         mdev = priv->mdev;
1595         dev = priv->dev;
1596
1597
1598         mutex_lock(&mdev->state_lock);
1599
1600         if (!mdev->device_up) {
1601                 en_err(priv, "Cannot open - device down/disabled\n");
1602                 goto out;
1603         }
1604
1605         /* Reset HW statistics and SW counters */
1606         mlx4_en_clear_stats(dev);
1607
1608         err = mlx4_en_start_port(dev);
1609         if (err)
1610                 en_err(priv, "Failed starting port:%d\n", priv->port);
1611
1612 out:
1613         mutex_unlock(&mdev->state_lock);
1614         return;
1615 }
1616
1617 void mlx4_en_free_resources(struct mlx4_en_priv *priv)
1618 {
1619         int i;
1620
1621 #ifdef CONFIG_RFS_ACCEL
1622         if (priv->dev->rx_cpu_rmap) {
1623                 free_irq_cpu_rmap(priv->dev->rx_cpu_rmap);
1624                 priv->dev->rx_cpu_rmap = NULL;
1625         }
1626 #endif
1627
1628         for (i = 0; i < priv->tx_ring_num; i++) {
1629                 if (priv->tx_ring && priv->tx_ring[i])
1630                         mlx4_en_destroy_tx_ring(priv, &priv->tx_ring[i]);
1631                 if (priv->tx_cq && priv->tx_cq[i])
1632                         mlx4_en_destroy_cq(priv, &priv->tx_cq[i]);
1633         }
1634
1635         for (i = 0; i < priv->rx_ring_num; i++) {
1636                 if (priv->rx_ring[i])
1637                         mlx4_en_destroy_rx_ring(priv, &priv->rx_ring[i],
1638                                 priv->prof->rx_ring_size, priv->stride);
1639                 if (priv->rx_cq[i])
1640                         mlx4_en_destroy_cq(priv, &priv->rx_cq[i]);
1641         }
1642
1643         if (priv->stat_sysctl != NULL)
1644                 sysctl_ctx_free(&priv->stat_ctx);
1645 }
1646
1647 int mlx4_en_alloc_resources(struct mlx4_en_priv *priv)
1648 {
1649         struct mlx4_en_port_profile *prof = priv->prof;
1650         int i;
1651         int node = 0;
1652
1653         /* Create rx Rings */
1654         for (i = 0; i < priv->rx_ring_num; i++) {
1655                 if (mlx4_en_create_cq(priv, &priv->rx_cq[i],
1656                                       prof->rx_ring_size, i, RX, node))
1657                         goto err;
1658
1659                 if (mlx4_en_create_rx_ring(priv, &priv->rx_ring[i],
1660                                            prof->rx_ring_size, node))
1661                         goto err;
1662         }
1663
1664         /* Create tx Rings */
1665         for (i = 0; i < priv->tx_ring_num; i++) {
1666                 if (mlx4_en_create_cq(priv, &priv->tx_cq[i],
1667                                       prof->tx_ring_size, i, TX, node))
1668                         goto err;
1669
1670                 if (mlx4_en_create_tx_ring(priv, &priv->tx_ring[i],
1671                                            prof->tx_ring_size, TXBB_SIZE, node, i))
1672                         goto err;
1673         }
1674
1675 #ifdef CONFIG_RFS_ACCEL
1676         priv->dev->rx_cpu_rmap = alloc_irq_cpu_rmap(priv->rx_ring_num);
1677         if (!priv->dev->rx_cpu_rmap)
1678                 goto err;
1679 #endif
1680         /* Re-create stat sysctls in case the number of rings changed. */
1681         mlx4_en_sysctl_stat(priv);
1682         return 0;
1683
1684 err:
1685         en_err(priv, "Failed to allocate NIC resources\n");
1686         for (i = 0; i < priv->rx_ring_num; i++) {
1687                 if (priv->rx_ring[i])
1688                         mlx4_en_destroy_rx_ring(priv, &priv->rx_ring[i],
1689                                                 prof->rx_ring_size,
1690                                                 priv->stride);
1691                 if (priv->rx_cq[i])
1692                         mlx4_en_destroy_cq(priv, &priv->rx_cq[i]);
1693         }
1694         for (i = 0; i < priv->tx_ring_num; i++) {
1695                 if (priv->tx_ring[i])
1696                         mlx4_en_destroy_tx_ring(priv, &priv->tx_ring[i]);
1697                 if (priv->tx_cq[i])
1698                         mlx4_en_destroy_cq(priv, &priv->tx_cq[i]);
1699         }
1700         priv->port_up = false;
1701         return -ENOMEM;
1702 }
1703
1704 struct en_port_attribute {
1705         struct attribute attr;
1706         ssize_t (*show)(struct en_port *, struct en_port_attribute *, char *buf);
1707         ssize_t (*store)(struct en_port *, struct en_port_attribute *, char *buf, size_t count);
1708 };
1709
1710 #define PORT_ATTR_RO(_name) \
1711 struct en_port_attribute en_port_attr_##_name = __ATTR_RO(_name)
1712
1713 #define EN_PORT_ATTR(_name, _mode, _show, _store) \
1714 struct en_port_attribute en_port_attr_##_name = __ATTR(_name, _mode, _show, _store)
1715
1716 void mlx4_en_destroy_netdev(struct net_device *dev)
1717 {
1718         struct mlx4_en_priv *priv = netdev_priv(dev);
1719         struct mlx4_en_dev *mdev = priv->mdev;
1720
1721         en_dbg(DRV, priv, "Destroying netdev on port:%d\n", priv->port);
1722
1723         if (priv->vlan_attach != NULL)
1724                 EVENTHANDLER_DEREGISTER(vlan_config, priv->vlan_attach);
1725         if (priv->vlan_detach != NULL)
1726                 EVENTHANDLER_DEREGISTER(vlan_unconfig, priv->vlan_detach);
1727
1728         /* Unregister device - this will close the port if it was up */
1729         if (priv->registered) {
1730                 mutex_lock(&mdev->state_lock);
1731                 ether_ifdetach(dev);
1732                 mutex_unlock(&mdev->state_lock);
1733         }
1734
1735         mutex_lock(&mdev->state_lock);
1736         mlx4_en_stop_port(dev);
1737         mutex_unlock(&mdev->state_lock);
1738
1739         if (priv->allocated)
1740                 mlx4_free_hwq_res(mdev->dev, &priv->res, MLX4_EN_PAGE_SIZE);
1741
1742         cancel_delayed_work(&priv->stats_task);
1743         cancel_delayed_work(&priv->service_task);
1744         /* flush any pending task for this netdev */
1745         flush_workqueue(mdev->workqueue);
1746         callout_drain(&priv->watchdog_timer);
1747
1748         /* Detach the netdev so tasks would not attempt to access it */
1749         mutex_lock(&mdev->state_lock);
1750         mdev->pndev[priv->port] = NULL;
1751         mutex_unlock(&mdev->state_lock);
1752
1753
1754         mlx4_en_free_resources(priv);
1755
1756         /* freeing the sysctl conf cannot be called from within mlx4_en_free_resources */
1757         if (priv->conf_sysctl != NULL)
1758                 sysctl_ctx_free(&priv->conf_ctx);
1759
1760         kfree(priv->tx_ring);
1761         kfree(priv->tx_cq);
1762
1763         kfree(priv);
1764         if_free(dev);
1765
1766 }
1767
1768 static int mlx4_en_change_mtu(struct net_device *dev, int new_mtu)
1769 {
1770         struct mlx4_en_priv *priv = netdev_priv(dev);
1771         struct mlx4_en_dev *mdev = priv->mdev;
1772         int err = 0;
1773
1774         en_dbg(DRV, priv, "Change MTU called - current:%u new:%u\n",
1775                (unsigned)dev->if_mtu, (unsigned)new_mtu);
1776
1777         if ((new_mtu < MLX4_EN_MIN_MTU) || (new_mtu > priv->max_mtu)) {
1778                 en_err(priv, "Bad MTU size:%d.\n", new_mtu);
1779                 return -EPERM;
1780         }
1781         mutex_lock(&mdev->state_lock);
1782         dev->if_mtu = new_mtu;
1783         if (dev->if_drv_flags & IFF_DRV_RUNNING) {
1784                 if (!mdev->device_up) {
1785                         /* NIC is probably restarting - let watchdog task reset
1786                          *                          * the port */
1787                         en_dbg(DRV, priv, "Change MTU called with card down!?\n");
1788                 } else {
1789                         mlx4_en_stop_port(dev);
1790                         err = mlx4_en_start_port(dev);
1791                         if (err) {
1792                                 en_err(priv, "Failed restarting port:%d\n",
1793                                                 priv->port);
1794                                 queue_work(mdev->workqueue, &priv->watchdog_task);
1795                         }
1796                 }
1797         }
1798         mutex_unlock(&mdev->state_lock);
1799         return 0;
1800 }
1801
1802 static int mlx4_en_calc_media(struct mlx4_en_priv *priv)
1803 {
1804         int trans_type;
1805         int active;
1806
1807         active = IFM_ETHER;
1808         if (priv->last_link_state == MLX4_DEV_EVENT_PORT_DOWN)
1809                 return (active);
1810         active |= IFM_FDX;
1811         trans_type = priv->port_state.transciver;
1812         /* XXX I don't know all of the transceiver values. */
1813         switch (priv->port_state.link_speed) {
1814         case 1000:
1815                 active |= IFM_1000_T;
1816                 break;
1817         case 10000:
1818                 if (trans_type > 0 && trans_type <= 0xC)
1819                         active |= IFM_10G_SR;
1820                 else if (trans_type == 0x80 || trans_type == 0)
1821                         active |= IFM_10G_CX4;
1822                 break;
1823         case 40000:
1824                 active |= IFM_40G_CR4;
1825                 break;
1826         }
1827         if (priv->prof->tx_pause)
1828                 active |= IFM_ETH_TXPAUSE;
1829         if (priv->prof->rx_pause)
1830                 active |= IFM_ETH_RXPAUSE;
1831
1832         return (active);
1833 }
1834
1835 static void mlx4_en_media_status(struct ifnet *dev, struct ifmediareq *ifmr)
1836 {
1837         struct mlx4_en_priv *priv;
1838
1839         priv = dev->if_softc;
1840         ifmr->ifm_status = IFM_AVALID;
1841         if (priv->last_link_state != MLX4_DEV_EVENT_PORT_DOWN)
1842                 ifmr->ifm_status |= IFM_ACTIVE;
1843         ifmr->ifm_active = mlx4_en_calc_media(priv);
1844
1845         return;
1846 }
1847
1848 static int mlx4_en_media_change(struct ifnet *dev)
1849 {
1850         struct mlx4_en_priv *priv;
1851         struct ifmedia *ifm;
1852         int rxpause;
1853         int txpause;
1854         int error;
1855
1856         priv = dev->if_softc;
1857         ifm = &priv->media;
1858         rxpause = txpause = 0;
1859         error = 0;
1860
1861         if (IFM_TYPE(ifm->ifm_media) != IFM_ETHER)
1862                 return (EINVAL);
1863         switch (IFM_SUBTYPE(ifm->ifm_media)) {
1864         case IFM_AUTO:
1865                 break;
1866         case IFM_10G_SR:
1867         case IFM_10G_CX4:
1868         case IFM_1000_T:
1869         case IFM_40G_CR4:
1870                 if ((IFM_SUBTYPE(ifm->ifm_media)
1871                         == IFM_SUBTYPE(mlx4_en_calc_media(priv)))
1872                         && (ifm->ifm_media & IFM_FDX))
1873                         break;
1874                 /* Fallthrough */
1875         default:
1876                 printf("%s: Only auto media type\n", if_name(dev));
1877                 return (EINVAL);
1878         }
1879         /* Allow user to set/clear pause */
1880         if (IFM_OPTIONS(ifm->ifm_media) & IFM_ETH_RXPAUSE)
1881                 rxpause = 1;
1882         if (IFM_OPTIONS(ifm->ifm_media) & IFM_ETH_TXPAUSE)
1883                 txpause = 1;
1884         if (priv->prof->tx_pause != txpause || priv->prof->rx_pause != rxpause) {
1885                 priv->prof->tx_pause = txpause;
1886                 priv->prof->rx_pause = rxpause;
1887                 error = -mlx4_SET_PORT_general(priv->mdev->dev, priv->port,
1888                      priv->rx_mb_size + ETHER_CRC_LEN, priv->prof->tx_pause,
1889                      priv->prof->tx_ppp, priv->prof->rx_pause,
1890                      priv->prof->rx_ppp);
1891         }
1892         return (error);
1893 }
1894
1895 static int mlx4_en_ioctl(struct ifnet *dev, u_long command, caddr_t data)
1896 {
1897         struct mlx4_en_priv *priv;
1898         struct mlx4_en_dev *mdev;
1899         struct ifreq *ifr;
1900         int error;
1901         int mask;
1902
1903         error = 0;
1904         mask = 0;
1905         priv = dev->if_softc;
1906         mdev = priv->mdev;
1907         ifr = (struct ifreq *) data;
1908         switch (command) {
1909
1910         case SIOCSIFMTU:
1911                 error = -mlx4_en_change_mtu(dev, ifr->ifr_mtu);
1912                 break;
1913         case SIOCSIFFLAGS:
1914                 if (dev->if_flags & IFF_UP) {
1915                         if ((dev->if_drv_flags & IFF_DRV_RUNNING) == 0) {
1916                                 mutex_lock(&mdev->state_lock);
1917                                 mlx4_en_start_port(dev);
1918                                 mutex_unlock(&mdev->state_lock);
1919                         } else {
1920                                 mlx4_en_set_rx_mode(dev);
1921                         }
1922                 } else {
1923                         mutex_lock(&mdev->state_lock);
1924                         if (dev->if_drv_flags & IFF_DRV_RUNNING) {
1925                                 mlx4_en_stop_port(dev);
1926                                 if_link_state_change(dev, LINK_STATE_DOWN);
1927                         }
1928                         mutex_unlock(&mdev->state_lock);
1929                 }
1930                 break;
1931         case SIOCADDMULTI:
1932         case SIOCDELMULTI:
1933                 mlx4_en_set_rx_mode(dev);
1934                 break;
1935         case SIOCSIFMEDIA:
1936         case SIOCGIFMEDIA:
1937                 error = ifmedia_ioctl(dev, ifr, &priv->media, command);
1938                 break;
1939         case SIOCSIFCAP:
1940                 mutex_lock(&mdev->state_lock);
1941                 mask = ifr->ifr_reqcap ^ dev->if_capenable;
1942                 if (mask & IFCAP_TXCSUM) {
1943                         dev->if_capenable ^= IFCAP_TXCSUM;
1944                         dev->if_hwassist ^= (CSUM_TCP | CSUM_UDP | CSUM_IP);
1945
1946                         if (IFCAP_TSO4 & dev->if_capenable &&
1947                             !(IFCAP_TXCSUM & dev->if_capenable)) {
1948                                 dev->if_capenable &= ~IFCAP_TSO4;
1949                                 dev->if_hwassist &= ~CSUM_TSO;
1950                                 if_printf(dev,
1951                                     "tso4 disabled due to -txcsum.\n");
1952                         }
1953                 }
1954                 if (mask & IFCAP_TXCSUM_IPV6) {
1955                         dev->if_capenable ^= IFCAP_TXCSUM_IPV6;
1956                         dev->if_hwassist ^= (CSUM_UDP_IPV6 | CSUM_TCP_IPV6);
1957
1958                         if (IFCAP_TSO6 & dev->if_capenable &&
1959                             !(IFCAP_TXCSUM_IPV6 & dev->if_capenable)) {
1960                                 dev->if_capenable &= ~IFCAP_TSO6;
1961                                 dev->if_hwassist &= ~CSUM_TSO;
1962                                 if_printf(dev,
1963                                     "tso6 disabled due to -txcsum6.\n");
1964                         }
1965                 }
1966                 if (mask & IFCAP_RXCSUM)
1967                         dev->if_capenable ^= IFCAP_RXCSUM;
1968                 if (mask & IFCAP_RXCSUM_IPV6)
1969                         dev->if_capenable ^= IFCAP_RXCSUM_IPV6;
1970
1971                 if (mask & IFCAP_TSO4) {
1972                         if (!(IFCAP_TSO4 & dev->if_capenable) &&
1973                             !(IFCAP_TXCSUM & dev->if_capenable)) {
1974                                 if_printf(dev, "enable txcsum first.\n");
1975                                 error = EAGAIN;
1976                                 goto out;
1977                         }
1978                         dev->if_capenable ^= IFCAP_TSO4;
1979                 }
1980                 if (mask & IFCAP_TSO6) {
1981                         if (!(IFCAP_TSO6 & dev->if_capenable) &&
1982                             !(IFCAP_TXCSUM_IPV6 & dev->if_capenable)) {
1983                                 if_printf(dev, "enable txcsum6 first.\n");
1984                                 error = EAGAIN;
1985                                 goto out;
1986                         }
1987                         dev->if_capenable ^= IFCAP_TSO6;
1988                 }
1989                 if (dev->if_capenable & (IFCAP_TSO4 | IFCAP_TSO6))
1990                         dev->if_hwassist |= CSUM_TSO;
1991                 else
1992                         dev->if_hwassist &= ~CSUM_TSO;
1993                 if (mask & IFCAP_LRO)
1994                         dev->if_capenable ^= IFCAP_LRO;
1995                 if (mask & IFCAP_VLAN_HWTAGGING)
1996                         dev->if_capenable ^= IFCAP_VLAN_HWTAGGING;
1997                 if (mask & IFCAP_VLAN_HWFILTER)
1998                         dev->if_capenable ^= IFCAP_VLAN_HWFILTER;
1999                 if (mask & IFCAP_WOL_MAGIC)
2000                         dev->if_capenable ^= IFCAP_WOL_MAGIC;
2001                 if (dev->if_drv_flags & IFF_DRV_RUNNING)
2002                         mlx4_en_start_port(dev);
2003 out:
2004                 mutex_unlock(&mdev->state_lock);
2005                 VLAN_CAPABILITIES(dev);
2006                 break;
2007 #if __FreeBSD_version >= 1100036
2008         case SIOCGI2C: {
2009                 struct ifi2creq i2c;
2010
2011                 error = copyin(ifr->ifr_data, &i2c, sizeof(i2c));
2012                 if (error)
2013                         break;
2014                 if (i2c.len > sizeof(i2c.data)) {
2015                         error = EINVAL;
2016                         break;
2017                 }
2018                 /*
2019                  * Note that we ignore i2c.addr here. The driver hardcodes
2020                  * the address to 0x50, while standard expects it to be 0xA0.
2021                  */
2022                 error = mlx4_get_module_info(mdev->dev, priv->port,
2023                     i2c.offset, i2c.len, i2c.data);
2024                 if (error < 0) {
2025                         error = -error;
2026                         break;
2027                 }
2028                 error = copyout(&i2c, ifr->ifr_data, sizeof(i2c));
2029                 break;
2030         }
2031 #endif
2032         default:
2033                 error = ether_ioctl(dev, command, data);
2034                 break;
2035         }
2036
2037         return (error);
2038 }
2039
2040
2041 int mlx4_en_init_netdev(struct mlx4_en_dev *mdev, int port,
2042                         struct mlx4_en_port_profile *prof)
2043 {
2044         struct net_device *dev;
2045         struct mlx4_en_priv *priv;
2046         uint8_t dev_addr[ETHER_ADDR_LEN];
2047         int err;
2048         int i;
2049
2050         priv = kzalloc(sizeof(*priv), GFP_KERNEL);
2051         dev = priv->dev = if_alloc(IFT_ETHER);
2052         if (dev == NULL) {
2053                 en_err(priv, "Net device allocation failed\n");
2054                 kfree(priv);
2055                 return -ENOMEM;
2056         }
2057         dev->if_softc = priv;
2058         if_initname(dev, "mlxen", atomic_fetchadd_int(&mlx4_en_unit, 1));
2059         dev->if_mtu = ETHERMTU;
2060         dev->if_init = mlx4_en_open;
2061         dev->if_flags = IFF_BROADCAST | IFF_SIMPLEX | IFF_MULTICAST;
2062         dev->if_ioctl = mlx4_en_ioctl;
2063         dev->if_transmit = mlx4_en_transmit;
2064         dev->if_qflush = mlx4_en_qflush;
2065         dev->if_snd.ifq_maxlen = prof->tx_ring_size;
2066
2067         /*
2068          * Initialize driver private data
2069          */
2070         priv->counter_index = 0xff;
2071         spin_lock_init(&priv->stats_lock);
2072         INIT_WORK(&priv->rx_mode_task, mlx4_en_do_set_rx_mode);
2073         INIT_WORK(&priv->watchdog_task, mlx4_en_restart);
2074         INIT_WORK(&priv->linkstate_task, mlx4_en_linkstate);
2075         INIT_DELAYED_WORK(&priv->stats_task, mlx4_en_do_get_stats);
2076         INIT_DELAYED_WORK(&priv->service_task, mlx4_en_service_task);
2077         callout_init(&priv->watchdog_timer, 1);
2078 #ifdef CONFIG_RFS_ACCEL
2079         INIT_LIST_HEAD(&priv->filters);
2080         spin_lock_init(&priv->filters_lock);
2081 #endif
2082
2083         priv->msg_enable = MLX4_EN_MSG_LEVEL;
2084         priv->dev = dev;
2085         priv->mdev = mdev;
2086         priv->ddev = &mdev->pdev->dev;
2087         priv->prof = prof;
2088         priv->port = port;
2089         priv->port_up = false;
2090         priv->flags = prof->flags;
2091
2092         priv->num_tx_rings_p_up = mdev->profile.num_tx_rings_p_up;
2093         priv->tx_ring_num = prof->tx_ring_num;
2094         priv->tx_ring = kcalloc(MAX_TX_RINGS,
2095                                 sizeof(struct mlx4_en_tx_ring *), GFP_KERNEL);
2096         if (!priv->tx_ring) {
2097                 err = -ENOMEM;
2098                 goto out;
2099         }
2100         priv->tx_cq = kcalloc(sizeof(struct mlx4_en_cq *), MAX_TX_RINGS,
2101                         GFP_KERNEL);
2102         if (!priv->tx_cq) {
2103                 err = -ENOMEM;
2104                 goto out;
2105         }
2106
2107         priv->rx_ring_num = prof->rx_ring_num;
2108         priv->cqe_factor = (mdev->dev->caps.cqe_size == 64) ? 1 : 0;
2109         priv->mac_index = -1;
2110         priv->last_ifq_jiffies = 0;
2111         priv->if_counters_rx_errors = 0;
2112         priv->if_counters_rx_no_buffer = 0;
2113 #ifdef CONFIG_MLX4_EN_DCB
2114         if (!mlx4_is_slave(priv->mdev->dev)) {
2115                 priv->dcbx_cap = DCB_CAP_DCBX_HOST;
2116                 priv->flags |= MLX4_EN_FLAG_DCB_ENABLED;
2117                 if (mdev->dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_ETS_CFG) {
2118                         dev->dcbnl_ops = &mlx4_en_dcbnl_ops;
2119                 } else {
2120                         en_info(priv, "QoS disabled - no HW support\n");
2121                         dev->dcbnl_ops = &mlx4_en_dcbnl_pfc_ops;
2122                 }
2123         }
2124 #endif
2125
2126         for (i = 0; i < MLX4_EN_MAC_HASH_SIZE; ++i)
2127                 INIT_HLIST_HEAD(&priv->mac_hash[i]);
2128
2129         /* Query for default mac and max mtu */
2130         priv->max_mtu = mdev->dev->caps.eth_mtu_cap[priv->port];
2131         priv->mac = mdev->dev->caps.def_mac[priv->port];
2132         if (ILLEGAL_MAC(priv->mac)) {
2133 #if BITS_PER_LONG == 64
2134                 en_err(priv, "Port: %d, invalid mac burned: 0x%lx, quiting\n",
2135                                 priv->port, priv->mac);
2136 #elif BITS_PER_LONG == 32
2137                 en_err(priv, "Port: %d, invalid mac burned: 0x%llx, quiting\n",
2138                                 priv->port, priv->mac);
2139 #endif
2140                 err = -EINVAL;
2141                 goto out;
2142         }
2143
2144         priv->stride = roundup_pow_of_two(sizeof(struct mlx4_en_rx_desc) +
2145                                           DS_SIZE);
2146
2147         mlx4_en_sysctl_conf(priv);
2148
2149         err = mlx4_en_alloc_resources(priv);
2150         if (err)
2151                 goto out;
2152
2153         /* Allocate page for receive rings */
2154         err = mlx4_alloc_hwq_res(mdev->dev, &priv->res,
2155                                 MLX4_EN_PAGE_SIZE, MLX4_EN_PAGE_SIZE);
2156         if (err) {
2157                 en_err(priv, "Failed to allocate page for rx qps\n");
2158                 goto out;
2159         }
2160         priv->allocated = 1;
2161
2162         /*
2163          * Set driver features
2164          */
2165         dev->if_capabilities |= IFCAP_HWCSUM | IFCAP_HWCSUM_IPV6;
2166         dev->if_capabilities |= IFCAP_VLAN_MTU | IFCAP_VLAN_HWTAGGING;
2167         dev->if_capabilities |= IFCAP_VLAN_HWCSUM | IFCAP_VLAN_HWFILTER;
2168         dev->if_capabilities |= IFCAP_LINKSTATE | IFCAP_JUMBO_MTU;
2169         dev->if_capabilities |= IFCAP_LRO;
2170
2171         if (mdev->LSO_support)
2172                 dev->if_capabilities |= IFCAP_TSO4 | IFCAP_TSO6 | IFCAP_VLAN_HWTSO;
2173
2174         /* set TSO limits so that we don't have to drop TX packets */
2175         dev->if_hw_tsomax = MLX4_EN_TX_MAX_PAYLOAD_SIZE - (ETHER_HDR_LEN + ETHER_VLAN_ENCAP_LEN) /* hdr */;
2176         dev->if_hw_tsomaxsegcount = MLX4_EN_TX_MAX_MBUF_FRAGS - 1 /* hdr */;
2177         dev->if_hw_tsomaxsegsize = MLX4_EN_TX_MAX_MBUF_SIZE;
2178
2179         dev->if_capenable = dev->if_capabilities;
2180
2181         dev->if_hwassist = 0;
2182         if (dev->if_capenable & (IFCAP_TSO4 | IFCAP_TSO6))
2183                 dev->if_hwassist |= CSUM_TSO;
2184         if (dev->if_capenable & IFCAP_TXCSUM)
2185                 dev->if_hwassist |= (CSUM_TCP | CSUM_UDP | CSUM_IP);
2186         if (dev->if_capenable & IFCAP_TXCSUM_IPV6)
2187                 dev->if_hwassist |= (CSUM_UDP_IPV6 | CSUM_TCP_IPV6);
2188
2189
2190         /* Register for VLAN events */
2191         priv->vlan_attach = EVENTHANDLER_REGISTER(vlan_config,
2192             mlx4_en_vlan_rx_add_vid, priv, EVENTHANDLER_PRI_FIRST);
2193         priv->vlan_detach = EVENTHANDLER_REGISTER(vlan_unconfig,
2194             mlx4_en_vlan_rx_kill_vid, priv, EVENTHANDLER_PRI_FIRST);
2195
2196         mdev->pndev[priv->port] = dev;
2197
2198         priv->last_link_state = MLX4_DEV_EVENT_PORT_DOWN;
2199         mlx4_en_set_default_moderation(priv);
2200
2201         /* Set default MAC */
2202         for (i = 0; i < ETHER_ADDR_LEN; i++)
2203                 dev_addr[ETHER_ADDR_LEN - 1 - i] = (u8) (priv->mac >> (8 * i));
2204
2205
2206         ether_ifattach(dev, dev_addr);
2207         if_link_state_change(dev, LINK_STATE_DOWN);
2208         ifmedia_init(&priv->media, IFM_IMASK | IFM_ETH_FMASK,
2209             mlx4_en_media_change, mlx4_en_media_status);
2210         ifmedia_add(&priv->media, IFM_ETHER | IFM_FDX | IFM_1000_T, 0, NULL);
2211         ifmedia_add(&priv->media, IFM_ETHER | IFM_FDX | IFM_10G_SR, 0, NULL);
2212         ifmedia_add(&priv->media, IFM_ETHER | IFM_FDX | IFM_10G_CX4, 0, NULL);
2213         ifmedia_add(&priv->media, IFM_ETHER | IFM_FDX | IFM_40G_CR4, 0, NULL);
2214         ifmedia_add(&priv->media, IFM_ETHER | IFM_AUTO, 0, NULL);
2215         ifmedia_set(&priv->media, IFM_ETHER | IFM_AUTO);
2216
2217         en_warn(priv, "Using %d TX rings\n", prof->tx_ring_num);
2218         en_warn(priv, "Using %d RX rings\n", prof->rx_ring_num);
2219
2220         priv->registered = 1;
2221
2222         en_warn(priv, "Using %d TX rings\n", prof->tx_ring_num);
2223         en_warn(priv, "Using %d RX rings\n", prof->rx_ring_num);
2224
2225
2226         priv->rx_mb_size = dev->if_mtu + ETH_HLEN + VLAN_HLEN + ETH_FCS_LEN;
2227         err = mlx4_SET_PORT_general(mdev->dev, priv->port,
2228                                     priv->rx_mb_size,
2229                                     prof->tx_pause, prof->tx_ppp,
2230                                     prof->rx_pause, prof->rx_ppp);
2231         if (err) {
2232                 en_err(priv, "Failed setting port general configurations "
2233                        "for port %d, with error %d\n", priv->port, err);
2234                 goto out;
2235         }
2236
2237         /* Init port */
2238         en_warn(priv, "Initializing port\n");
2239         err = mlx4_INIT_PORT(mdev->dev, priv->port);
2240         if (err) {
2241                 en_err(priv, "Failed Initializing port\n");
2242                 goto out;
2243         }
2244
2245         queue_delayed_work(mdev->workqueue, &priv->stats_task, STATS_DELAY);
2246
2247         if (mdev->dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_TS)
2248                 queue_delayed_work(mdev->workqueue, &priv->service_task, SERVICE_TASK_DELAY);
2249
2250         return 0;
2251
2252 out:
2253         mlx4_en_destroy_netdev(dev);
2254         return err;
2255 }
2256
2257 static int mlx4_en_set_ring_size(struct net_device *dev,
2258     int rx_size, int tx_size)
2259 {
2260         struct mlx4_en_priv *priv = netdev_priv(dev);
2261         struct mlx4_en_dev *mdev = priv->mdev;
2262         int port_up = 0;
2263         int err = 0;
2264
2265         rx_size = roundup_pow_of_two(rx_size);
2266         rx_size = max_t(u32, rx_size, MLX4_EN_MIN_RX_SIZE);
2267         rx_size = min_t(u32, rx_size, MLX4_EN_MAX_RX_SIZE);
2268         tx_size = roundup_pow_of_two(tx_size);
2269         tx_size = max_t(u32, tx_size, MLX4_EN_MIN_TX_SIZE);
2270         tx_size = min_t(u32, tx_size, MLX4_EN_MAX_TX_SIZE);
2271
2272         if (rx_size == (priv->port_up ?
2273             priv->rx_ring[0]->actual_size : priv->rx_ring[0]->size) &&
2274             tx_size == priv->tx_ring[0]->size)
2275                 return 0;
2276         mutex_lock(&mdev->state_lock);
2277         if (priv->port_up) {
2278                 port_up = 1;
2279                 mlx4_en_stop_port(dev);
2280         }
2281         mlx4_en_free_resources(priv);
2282         priv->prof->tx_ring_size = tx_size;
2283         priv->prof->rx_ring_size = rx_size;
2284         err = mlx4_en_alloc_resources(priv);
2285         if (err) {
2286                 en_err(priv, "Failed reallocating port resources\n");
2287                 goto out;
2288         }
2289         if (port_up) {
2290                 err = mlx4_en_start_port(dev);
2291                 if (err)
2292                         en_err(priv, "Failed starting port\n");
2293         }
2294 out:
2295         mutex_unlock(&mdev->state_lock);
2296         return err;
2297 }
2298 static int mlx4_en_set_rx_ring_size(SYSCTL_HANDLER_ARGS)
2299 {
2300         struct mlx4_en_priv *priv;
2301         int size;
2302         int error;
2303
2304         priv = arg1;
2305         size = priv->prof->rx_ring_size;
2306         error = sysctl_handle_int(oidp, &size, 0, req);
2307         if (error || !req->newptr)
2308                 return (error);
2309         error = -mlx4_en_set_ring_size(priv->dev, size,
2310             priv->prof->tx_ring_size);
2311         return (error);
2312 }
2313
2314 static int mlx4_en_set_tx_ring_size(SYSCTL_HANDLER_ARGS)
2315 {
2316         struct mlx4_en_priv *priv;
2317         int size;
2318         int error;
2319
2320         priv = arg1;
2321         size = priv->prof->tx_ring_size;
2322         error = sysctl_handle_int(oidp, &size, 0, req);
2323         if (error || !req->newptr)
2324                 return (error);
2325         error = -mlx4_en_set_ring_size(priv->dev, priv->prof->rx_ring_size,
2326             size);
2327
2328         return (error);
2329 }
2330
2331 static int mlx4_en_get_module_info(struct net_device *dev,
2332                                    struct ethtool_modinfo *modinfo)
2333 {
2334         struct mlx4_en_priv *priv = netdev_priv(dev);
2335         struct mlx4_en_dev *mdev = priv->mdev;
2336         int ret;
2337         u8 data[4];
2338
2339         /* Read first 2 bytes to get Module & REV ID */
2340         ret = mlx4_get_module_info(mdev->dev, priv->port,
2341                                    0/*offset*/, 2/*size*/, data);
2342
2343         if (ret < 2) {
2344                 en_err(priv, "Failed to read eeprom module first two bytes, error: 0x%x\n", -ret);
2345                 return -EIO;
2346         }
2347
2348         switch (data[0] /* identifier */) {
2349         case MLX4_MODULE_ID_QSFP:
2350                 modinfo->type = ETH_MODULE_SFF_8436;
2351                 modinfo->eeprom_len = ETH_MODULE_SFF_8436_LEN;
2352                 break;
2353         case MLX4_MODULE_ID_QSFP_PLUS:
2354                 if (data[1] >= 0x3) { /* revision id */
2355                         modinfo->type = ETH_MODULE_SFF_8636;
2356                         modinfo->eeprom_len = ETH_MODULE_SFF_8636_LEN;
2357                 } else {
2358                         modinfo->type = ETH_MODULE_SFF_8436;
2359                         modinfo->eeprom_len = ETH_MODULE_SFF_8436_LEN;
2360                 }
2361                 break;
2362         case MLX4_MODULE_ID_QSFP28:
2363                 modinfo->type = ETH_MODULE_SFF_8636;
2364                 modinfo->eeprom_len = ETH_MODULE_SFF_8636_LEN;
2365                 break;
2366         case MLX4_MODULE_ID_SFP:
2367                 modinfo->type = ETH_MODULE_SFF_8472;
2368                 modinfo->eeprom_len = ETH_MODULE_SFF_8472_LEN;
2369                 break;
2370         default:
2371                 en_err(priv, "mlx4_en_get_module_info :  Not recognized cable type\n");
2372                 return -EINVAL;
2373         }
2374
2375         return 0;
2376 }
2377
2378 static int mlx4_en_get_module_eeprom(struct net_device *dev,
2379                                      struct ethtool_eeprom *ee,
2380                                      u8 *data)
2381 {
2382         struct mlx4_en_priv *priv = netdev_priv(dev);
2383         struct mlx4_en_dev *mdev = priv->mdev;
2384         int offset = ee->offset;
2385         int i = 0, ret;
2386
2387         if (ee->len == 0)
2388                 return -EINVAL;
2389
2390         memset(data, 0, ee->len);
2391
2392         while (i < ee->len) {
2393                 en_dbg(DRV, priv,
2394                        "mlx4_get_module_info i(%d) offset(%d) len(%d)\n",
2395                        i, offset, ee->len - i);
2396
2397                 ret = mlx4_get_module_info(mdev->dev, priv->port,
2398                                            offset, ee->len - i, data + i);
2399
2400                 if (!ret) /* Done reading */
2401                         return 0;
2402
2403                 if (ret < 0) {
2404                         en_err(priv,
2405                                "mlx4_get_module_info i(%d) offset(%d) bytes_to_read(%d) - FAILED (0x%x)\n",
2406                                i, offset, ee->len - i, ret);
2407                         return -1;
2408                 }
2409
2410                 i += ret;
2411                 offset += ret;
2412         }
2413         return 0;
2414 }
2415
2416 static void mlx4_en_print_eeprom(u8 *data, __u32 len)
2417 {
2418         int             i;
2419         int             j = 0;
2420         int             row = 0;
2421         const int       NUM_OF_BYTES = 16;
2422
2423         printf("\nOffset\t\tValues\n");
2424         printf("------\t\t------\n");
2425         while(row < len){
2426                 printf("0x%04x\t\t",row);
2427                 for(i=0; i < NUM_OF_BYTES; i++){
2428                         printf("%02x ", data[j]);
2429                         row++;
2430                         j++;
2431                 }
2432                 printf("\n");
2433         }
2434 }
2435
2436 /* Read cable EEPROM module information by first inspecting the first
2437  * two bytes to get the length and then read the rest of the information.
2438  * The information is printed to dmesg. */
2439 static int mlx4_en_read_eeprom(SYSCTL_HANDLER_ARGS)
2440 {
2441
2442         u8*             data;
2443         int             error;
2444         int             result = 0;
2445         struct          mlx4_en_priv *priv;
2446         struct          net_device *dev;
2447         struct          ethtool_modinfo modinfo;
2448         struct          ethtool_eeprom ee;
2449
2450         error = sysctl_handle_int(oidp, &result, 0, req);
2451         if (error || !req->newptr)
2452                 return (error);
2453
2454         if (result == 1) {
2455                 priv = arg1;
2456                 dev = priv->dev;
2457                 data = kmalloc(PAGE_SIZE, GFP_KERNEL);
2458
2459                 error = mlx4_en_get_module_info(dev, &modinfo);
2460                 if (error) {
2461                         en_err(priv,
2462                                "mlx4_en_get_module_info returned with error - FAILED (0x%x)\n",
2463                                -error);
2464                         goto out;
2465                 }
2466
2467                 ee.len = modinfo.eeprom_len;
2468                 ee.offset = 0;
2469
2470                 error = mlx4_en_get_module_eeprom(dev, &ee, data);
2471                 if (error) {
2472                         en_err(priv,
2473                                "mlx4_en_get_module_eeprom returned with error - FAILED (0x%x)\n",
2474                                -error);
2475                         /* Continue printing partial information in case of an error */
2476                 }
2477
2478                 /* EEPROM information will be printed in dmesg */
2479                 mlx4_en_print_eeprom(data, ee.len);
2480 out:
2481                 kfree(data);
2482         }
2483         /* Return zero to prevent sysctl failure. */
2484         return (0);
2485 }
2486
2487 static int mlx4_en_set_tx_ppp(SYSCTL_HANDLER_ARGS)
2488 {
2489         struct mlx4_en_priv *priv;
2490         int ppp;
2491         int error;
2492
2493         priv = arg1;
2494         ppp = priv->prof->tx_ppp;
2495         error = sysctl_handle_int(oidp, &ppp, 0, req);
2496         if (error || !req->newptr)
2497                 return (error);
2498         if (ppp > 0xff || ppp < 0)
2499                 return (-EINVAL);
2500         priv->prof->tx_ppp = ppp;
2501         error = -mlx4_SET_PORT_general(priv->mdev->dev, priv->port,
2502                                        priv->rx_mb_size + ETHER_CRC_LEN,
2503                                        priv->prof->tx_pause,
2504                                        priv->prof->tx_ppp,
2505                                        priv->prof->rx_pause,
2506                                        priv->prof->rx_ppp);
2507
2508         return (error);
2509 }
2510
2511 static int mlx4_en_set_rx_ppp(SYSCTL_HANDLER_ARGS)
2512 {
2513         struct mlx4_en_priv *priv;
2514         struct mlx4_en_dev *mdev;
2515         int ppp;
2516         int error;
2517         int port_up;
2518
2519         port_up = 0;
2520         priv = arg1;
2521         mdev = priv->mdev;
2522         ppp = priv->prof->rx_ppp;
2523         error = sysctl_handle_int(oidp, &ppp, 0, req);
2524         if (error || !req->newptr)
2525                 return (error);
2526         if (ppp > 0xff || ppp < 0)
2527                 return (-EINVAL);
2528         /* See if we have to change the number of tx queues. */
2529         if (!ppp != !priv->prof->rx_ppp) {
2530                 mutex_lock(&mdev->state_lock);
2531                 if (priv->port_up) {
2532                         port_up = 1;
2533                         mlx4_en_stop_port(priv->dev);
2534                 }
2535                 mlx4_en_free_resources(priv);
2536                 priv->prof->rx_ppp = ppp;
2537                 error = -mlx4_en_alloc_resources(priv);
2538                 if (error)
2539                         en_err(priv, "Failed reallocating port resources\n");
2540                 if (error == 0 && port_up) {
2541                         error = -mlx4_en_start_port(priv->dev);
2542                         if (error)
2543                                 en_err(priv, "Failed starting port\n");
2544                 }
2545                 mutex_unlock(&mdev->state_lock);
2546                 return (error);
2547
2548         }
2549         priv->prof->rx_ppp = ppp;
2550         error = -mlx4_SET_PORT_general(priv->mdev->dev, priv->port,
2551                                        priv->rx_mb_size + ETHER_CRC_LEN,
2552                                        priv->prof->tx_pause,
2553                                        priv->prof->tx_ppp,
2554                                        priv->prof->rx_pause,
2555                                        priv->prof->rx_ppp);
2556
2557         return (error);
2558 }
2559
2560 static void mlx4_en_sysctl_conf(struct mlx4_en_priv *priv)
2561 {
2562         struct net_device *dev;
2563         struct sysctl_ctx_list *ctx;
2564         struct sysctl_oid *node;
2565         struct sysctl_oid_list *node_list;
2566         struct sysctl_oid *coal;
2567         struct sysctl_oid_list *coal_list;
2568         const char *pnameunit;
2569
2570         dev = priv->dev;
2571         ctx = &priv->conf_ctx;
2572         pnameunit = device_get_nameunit(priv->mdev->pdev->dev.bsddev);
2573
2574         sysctl_ctx_init(ctx);
2575         priv->conf_sysctl = SYSCTL_ADD_NODE(ctx, SYSCTL_STATIC_CHILDREN(_hw),
2576             OID_AUTO, dev->if_xname, CTLFLAG_RD, 0, "mlx4 10gig ethernet");
2577         node = SYSCTL_ADD_NODE(ctx, SYSCTL_CHILDREN(priv->conf_sysctl), OID_AUTO,
2578             "conf", CTLFLAG_RD, NULL, "Configuration");
2579         node_list = SYSCTL_CHILDREN(node);
2580
2581         SYSCTL_ADD_UINT(ctx, node_list, OID_AUTO, "msg_enable",
2582             CTLFLAG_RW, &priv->msg_enable, 0,
2583             "Driver message enable bitfield");
2584         SYSCTL_ADD_UINT(ctx, node_list, OID_AUTO, "rx_rings",
2585             CTLFLAG_RD, &priv->rx_ring_num, 0,
2586             "Number of receive rings");
2587         SYSCTL_ADD_UINT(ctx, node_list, OID_AUTO, "tx_rings",
2588             CTLFLAG_RD, &priv->tx_ring_num, 0,
2589             "Number of transmit rings");
2590         SYSCTL_ADD_PROC(ctx, node_list, OID_AUTO, "rx_size",
2591             CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, priv, 0,
2592             mlx4_en_set_rx_ring_size, "I", "Receive ring size");
2593         SYSCTL_ADD_PROC(ctx, node_list, OID_AUTO, "tx_size",
2594             CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, priv, 0,
2595             mlx4_en_set_tx_ring_size, "I", "Transmit ring size");
2596         SYSCTL_ADD_PROC(ctx, node_list, OID_AUTO, "tx_ppp",
2597             CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, priv, 0,
2598             mlx4_en_set_tx_ppp, "I", "TX Per-priority pause");
2599         SYSCTL_ADD_PROC(ctx, node_list, OID_AUTO, "rx_ppp",
2600             CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, priv, 0,
2601             mlx4_en_set_rx_ppp, "I", "RX Per-priority pause");
2602         SYSCTL_ADD_UINT(ctx, node_list, OID_AUTO, "port_num",
2603             CTLFLAG_RD, &priv->port, 0,
2604             "Port Number");
2605         SYSCTL_ADD_STRING(ctx, node_list, OID_AUTO, "device_name",
2606             CTLFLAG_RD, __DECONST(void *, pnameunit), 0,
2607             "PCI device name");
2608
2609         /* Add coalescer configuration. */
2610         coal = SYSCTL_ADD_NODE(ctx, node_list, OID_AUTO,
2611             "coalesce", CTLFLAG_RD, NULL, "Interrupt coalesce configuration");
2612         coal_list = SYSCTL_CHILDREN(coal);
2613         SYSCTL_ADD_UINT(ctx, coal_list, OID_AUTO, "pkt_rate_low",
2614             CTLFLAG_RW, &priv->pkt_rate_low, 0,
2615             "Packets per-second for minimum delay");
2616         SYSCTL_ADD_UINT(ctx, coal_list, OID_AUTO, "rx_usecs_low",
2617             CTLFLAG_RW, &priv->rx_usecs_low, 0,
2618             "Minimum RX delay in micro-seconds");
2619         SYSCTL_ADD_UINT(ctx, coal_list, OID_AUTO, "pkt_rate_high",
2620             CTLFLAG_RW, &priv->pkt_rate_high, 0,
2621             "Packets per-second for maximum delay");
2622         SYSCTL_ADD_UINT(ctx, coal_list, OID_AUTO, "rx_usecs_high",
2623             CTLFLAG_RW, &priv->rx_usecs_high, 0,
2624             "Maximum RX delay in micro-seconds");
2625         SYSCTL_ADD_UINT(ctx, coal_list, OID_AUTO, "sample_interval",
2626             CTLFLAG_RW, &priv->sample_interval, 0,
2627             "adaptive frequency in units of HZ ticks");
2628         SYSCTL_ADD_UINT(ctx, coal_list, OID_AUTO, "adaptive_rx_coal",
2629             CTLFLAG_RW, &priv->adaptive_rx_coal, 0,
2630             "Enable adaptive rx coalescing");
2631         /* EEPROM support */
2632         SYSCTL_ADD_PROC(ctx, node_list, OID_AUTO, "eeprom_info",
2633             CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, priv, 0,
2634             mlx4_en_read_eeprom, "I", "EEPROM information");
2635 }
2636
2637 static void mlx4_en_sysctl_stat(struct mlx4_en_priv *priv)
2638 {
2639         struct sysctl_ctx_list *ctx;
2640         struct sysctl_oid_list *node_list;
2641         struct sysctl_oid *ring_node;
2642         struct sysctl_oid_list *ring_list;
2643         struct mlx4_en_tx_ring *tx_ring;
2644         struct mlx4_en_rx_ring *rx_ring;
2645         char namebuf[128];
2646         int i;
2647
2648         ctx = &priv->stat_ctx;
2649         sysctl_ctx_init(ctx);
2650         priv->stat_sysctl = SYSCTL_ADD_NODE(ctx, SYSCTL_CHILDREN(priv->conf_sysctl), OID_AUTO,
2651             "stat", CTLFLAG_RD, NULL, "Statistics");
2652         node_list = SYSCTL_CHILDREN(priv->stat_sysctl);
2653
2654 #ifdef MLX4_EN_PERF_STAT
2655         SYSCTL_ADD_UINT(ctx, node_list, OID_AUTO, "tx_poll", CTLFLAG_RD,
2656             &priv->pstats.tx_poll, "TX Poll calls");
2657         SYSCTL_ADD_QUAD(ctx, node_list, OID_AUTO, "tx_pktsz_avg", CTLFLAG_RD,
2658             &priv->pstats.tx_pktsz_avg, "TX average packet size");
2659         SYSCTL_ADD_UINT(ctx, node_list, OID_AUTO, "inflight_avg", CTLFLAG_RD,
2660             &priv->pstats.inflight_avg, "TX average packets in-flight");
2661         SYSCTL_ADD_UINT(ctx, node_list, OID_AUTO, "tx_coal_avg", CTLFLAG_RD,
2662             &priv->pstats.tx_coal_avg, "TX average coalesced completions");
2663         SYSCTL_ADD_UINT(ctx, node_list, OID_AUTO, "rx_coal_avg", CTLFLAG_RD,
2664             &priv->pstats.rx_coal_avg, "RX average coalesced completions");
2665 #endif
2666
2667         SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "tso_packets", CTLFLAG_RD,
2668             &priv->port_stats.tso_packets, "TSO packets sent");
2669         SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "queue_stopped", CTLFLAG_RD,
2670             &priv->port_stats.queue_stopped, "Queue full");
2671         SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "wake_queue", CTLFLAG_RD,
2672             &priv->port_stats.wake_queue, "Queue resumed after full");
2673         SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "tx_timeout", CTLFLAG_RD,
2674             &priv->port_stats.tx_timeout, "Transmit timeouts");
2675         SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "tx_oversized_packets", CTLFLAG_RD,
2676             &priv->port_stats.oversized_packets, "TX oversized packets, m_defrag failed");
2677         SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "rx_alloc_failed", CTLFLAG_RD,
2678             &priv->port_stats.rx_alloc_failed, "RX failed to allocate mbuf");
2679         SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "rx_chksum_good", CTLFLAG_RD,
2680             &priv->port_stats.rx_chksum_good, "RX checksum offload success");
2681         SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "rx_chksum_none", CTLFLAG_RD,
2682             &priv->port_stats.rx_chksum_none, "RX without checksum offload");
2683         SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "tx_chksum_offload",
2684             CTLFLAG_RD, &priv->port_stats.tx_chksum_offload,
2685             "TX checksum offloads");
2686
2687         /* Could strdup the names and add in a loop.  This is simpler. */
2688         SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "rx_bytes", CTLFLAG_RD,
2689             &priv->pkstats.rx_bytes, "RX Bytes");
2690         SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "rx_packets", CTLFLAG_RD,
2691             &priv->pkstats.rx_packets, "RX packets");
2692         SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "rx_multicast_packets", CTLFLAG_RD,
2693             &priv->pkstats.rx_multicast_packets, "RX Multicast Packets");
2694         SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "rx_broadcast_packets", CTLFLAG_RD,
2695             &priv->pkstats.rx_broadcast_packets, "RX Broadcast Packets");
2696         SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "rx_errors", CTLFLAG_RD,
2697             &priv->pkstats.rx_errors, "RX Errors");
2698         SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "rx_dropped", CTLFLAG_RD,
2699             &priv->pkstats.rx_dropped, "RX Dropped");
2700         SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "rx_length_errors", CTLFLAG_RD,
2701             &priv->pkstats.rx_length_errors, "RX Length Errors");
2702         SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "rx_over_errors", CTLFLAG_RD,
2703             &priv->pkstats.rx_over_errors, "RX Over Errors");
2704         SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "rx_crc_errors", CTLFLAG_RD,
2705             &priv->pkstats.rx_crc_errors, "RX CRC Errors");
2706         SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "rx_jabbers", CTLFLAG_RD,
2707             &priv->pkstats.rx_jabbers, "RX Jabbers");
2708
2709
2710         SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "rx_in_range_length_error", CTLFLAG_RD,
2711             &priv->pkstats.rx_in_range_length_error, "RX IN_Range Length Error");
2712         SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "rx_out_range_length_error",
2713                 CTLFLAG_RD, &priv->pkstats.rx_out_range_length_error,
2714                 "RX Out Range Length Error");
2715         SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "rx_lt_64_bytes_packets", CTLFLAG_RD,
2716             &priv->pkstats.rx_lt_64_bytes_packets, "RX Lt 64 Bytes Packets");
2717         SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "rx_127_bytes_packets", CTLFLAG_RD,
2718             &priv->pkstats.rx_127_bytes_packets, "RX 127 bytes Packets");
2719         SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "rx_255_bytes_packets", CTLFLAG_RD,
2720             &priv->pkstats.rx_255_bytes_packets, "RX 255 bytes Packets");
2721         SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "rx_511_bytes_packets", CTLFLAG_RD,
2722             &priv->pkstats.rx_511_bytes_packets, "RX 511 bytes Packets");
2723         SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "rx_1023_bytes_packets", CTLFLAG_RD,
2724             &priv->pkstats.rx_1023_bytes_packets, "RX 1023 bytes Packets");
2725         SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "rx_1518_bytes_packets", CTLFLAG_RD,
2726             &priv->pkstats.rx_1518_bytes_packets, "RX 1518 bytes Packets");
2727         SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "rx_1522_bytes_packets", CTLFLAG_RD,
2728             &priv->pkstats.rx_1522_bytes_packets, "RX 1522 bytes Packets");
2729         SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "rx_1548_bytes_packets", CTLFLAG_RD,
2730             &priv->pkstats.rx_1548_bytes_packets, "RX 1548 bytes Packets");
2731         SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "rx_gt_1548_bytes_packets", CTLFLAG_RD,
2732             &priv->pkstats.rx_gt_1548_bytes_packets,
2733             "RX Greater Then 1548 bytes Packets");
2734
2735 struct mlx4_en_pkt_stats {
2736         unsigned long tx_packets;
2737         unsigned long tx_bytes;
2738         unsigned long tx_multicast_packets;
2739         unsigned long tx_broadcast_packets;
2740         unsigned long tx_errors;
2741         unsigned long tx_dropped;
2742         unsigned long tx_lt_64_bytes_packets;
2743         unsigned long tx_127_bytes_packets;
2744         unsigned long tx_255_bytes_packets;
2745         unsigned long tx_511_bytes_packets;
2746         unsigned long tx_1023_bytes_packets;
2747         unsigned long tx_1518_bytes_packets;
2748         unsigned long tx_1522_bytes_packets;
2749         unsigned long tx_1548_bytes_packets;
2750         unsigned long tx_gt_1548_bytes_packets;
2751         unsigned long rx_prio[NUM_PRIORITIES][NUM_PRIORITY_STATS];
2752         unsigned long tx_prio[NUM_PRIORITIES][NUM_PRIORITY_STATS];
2753 #define NUM_PKT_STATS           72
2754 };
2755
2756
2757         SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "tx_packets", CTLFLAG_RD,
2758             &priv->pkstats.tx_packets, "TX packets");
2759         SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "tx_bytes", CTLFLAG_RD,
2760             &priv->pkstats.tx_bytes, "TX Bytes");
2761         SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "tx_multicast_packets", CTLFLAG_RD,
2762             &priv->pkstats.tx_multicast_packets, "TX Multicast Packets");
2763         SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "tx_broadcast_packets", CTLFLAG_RD,
2764             &priv->pkstats.tx_broadcast_packets, "TX Broadcast Packets");
2765         SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "tx_errors", CTLFLAG_RD,
2766             &priv->pkstats.tx_errors, "TX Errors");
2767         SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "tx_dropped", CTLFLAG_RD,
2768             &priv->pkstats.tx_dropped, "TX Dropped");
2769         SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "tx_lt_64_bytes_packets", CTLFLAG_RD,
2770             &priv->pkstats.tx_lt_64_bytes_packets, "TX Less Then 64 Bytes Packets");
2771         SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "tx_127_bytes_packets", CTLFLAG_RD,
2772             &priv->pkstats.tx_127_bytes_packets, "TX 127 Bytes Packets");
2773         SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "tx_255_bytes_packets", CTLFLAG_RD,
2774             &priv->pkstats.tx_255_bytes_packets, "TX 255 Bytes Packets");
2775         SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "tx_511_bytes_packets", CTLFLAG_RD,
2776             &priv->pkstats.tx_511_bytes_packets, "TX 511 Bytes Packets");
2777         SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "tx_1023_bytes_packets", CTLFLAG_RD,
2778             &priv->pkstats.tx_1023_bytes_packets, "TX 1023 Bytes Packets");
2779         SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "tx_1518_bytes_packets", CTLFLAG_RD,
2780             &priv->pkstats.tx_1518_bytes_packets, "TX 1518 Bytes Packets");
2781         SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "tx_1522_bytes_packets", CTLFLAG_RD,
2782             &priv->pkstats.tx_1522_bytes_packets, "TX 1522 Bytes Packets");
2783         SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "tx_1548_bytes_packets", CTLFLAG_RD,
2784             &priv->pkstats.tx_1548_bytes_packets, "TX 1548 Bytes Packets");
2785         SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "tx_gt_1548_bytes_packets", CTLFLAG_RD,
2786             &priv->pkstats.tx_gt_1548_bytes_packets,
2787             "TX Greater Then 1548 Bytes Packets");
2788
2789
2790
2791         for (i = 0; i < priv->tx_ring_num; i++) {
2792                 tx_ring = priv->tx_ring[i];
2793                 snprintf(namebuf, sizeof(namebuf), "tx_ring%d", i);
2794                 ring_node = SYSCTL_ADD_NODE(ctx, node_list, OID_AUTO, namebuf,
2795                     CTLFLAG_RD, NULL, "TX Ring");
2796                 ring_list = SYSCTL_CHILDREN(ring_node);
2797                 SYSCTL_ADD_ULONG(ctx, ring_list, OID_AUTO, "packets",
2798                     CTLFLAG_RD, &tx_ring->packets, "TX packets");
2799                 SYSCTL_ADD_ULONG(ctx, ring_list, OID_AUTO, "bytes",
2800                     CTLFLAG_RD, &tx_ring->bytes, "TX bytes");
2801         }
2802
2803         for (i = 0; i < priv->rx_ring_num; i++) {
2804                 rx_ring = priv->rx_ring[i];
2805                 snprintf(namebuf, sizeof(namebuf), "rx_ring%d", i);
2806                 ring_node = SYSCTL_ADD_NODE(ctx, node_list, OID_AUTO, namebuf,
2807                     CTLFLAG_RD, NULL, "RX Ring");
2808                 ring_list = SYSCTL_CHILDREN(ring_node);
2809                 SYSCTL_ADD_ULONG(ctx, ring_list, OID_AUTO, "packets",
2810                     CTLFLAG_RD, &rx_ring->packets, "RX packets");
2811                 SYSCTL_ADD_ULONG(ctx, ring_list, OID_AUTO, "bytes",
2812                     CTLFLAG_RD, &rx_ring->bytes, "RX bytes");
2813                 SYSCTL_ADD_ULONG(ctx, ring_list, OID_AUTO, "error",
2814                     CTLFLAG_RD, &rx_ring->errors, "RX soft errors");
2815         }
2816 }