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         priv->rx_alloc_size = max_t(int, 2 * roundup_pow_of_two(priv->rx_mb_size),
1248                                     PAGE_SIZE);
1249         priv->rx_alloc_order = get_order(priv->rx_alloc_size);
1250         priv->rx_buf_size = roundup_pow_of_two(priv->rx_mb_size);
1251         en_dbg(DRV, priv, "Rx buf size:%d\n", priv->rx_mb_size);
1252
1253         /* Configure rx cq's and rings */
1254         err = mlx4_en_activate_rx_rings(priv);
1255         if (err) {
1256                 en_err(priv, "Failed to activate RX rings\n");
1257                 return err;
1258         }
1259         for (i = 0; i < priv->rx_ring_num; i++) {
1260                 cq = priv->rx_cq[i];
1261
1262                 mlx4_en_cq_init_lock(cq);
1263                 err = mlx4_en_activate_cq(priv, cq, i);
1264                 if (err) {
1265                         en_err(priv, "Failed activating Rx CQ\n");
1266                         goto cq_err;
1267                 }
1268                 for (j = 0; j < cq->size; j++)
1269                         cq->buf[j].owner_sr_opcode = MLX4_CQE_OWNER_MASK;
1270                 err = mlx4_en_set_cq_moder(priv, cq);
1271                 if (err) {
1272                         en_err(priv, "Failed setting cq moderation parameters");
1273                         mlx4_en_deactivate_cq(priv, cq);
1274                         goto cq_err;
1275                 }
1276                 mlx4_en_arm_cq(priv, cq);
1277                 priv->rx_ring[i]->cqn = cq->mcq.cqn;
1278                 ++rx_index;
1279         }
1280
1281         /* Set qp number */
1282         en_dbg(DRV, priv, "Getting qp number for port %d\n", priv->port);
1283         err = mlx4_en_get_qp(priv);
1284         if (err) {
1285                 en_err(priv, "Failed getting eth qp\n");
1286                 goto cq_err;
1287         }
1288         mdev->mac_removed[priv->port] = 0;
1289
1290         /* gets default allocated counter index from func cap */
1291         /* or sink counter index if no resources */
1292         priv->counter_index = mdev->dev->caps.def_counter_index[priv->port - 1];
1293
1294         en_dbg(DRV, priv, "%s: default counter index %d for port %d\n",
1295                __func__, priv->counter_index, priv->port);
1296
1297         err = mlx4_en_config_rss_steer(priv);
1298         if (err) {
1299                 en_err(priv, "Failed configuring rss steering\n");
1300                 goto mac_err;
1301         }
1302
1303         err = mlx4_en_create_drop_qp(priv);
1304         if (err)
1305                 goto rss_err;
1306
1307         /* Configure tx cq's and rings */
1308         for (i = 0; i < priv->tx_ring_num; i++) {
1309                 /* Configure cq */
1310                 cq = priv->tx_cq[i];
1311                 err = mlx4_en_activate_cq(priv, cq, i);
1312                 if (err) {
1313                         en_err(priv, "Failed activating Tx CQ\n");
1314                         goto tx_err;
1315                 }
1316                 err = mlx4_en_set_cq_moder(priv, cq);
1317                 if (err) {
1318                         en_err(priv, "Failed setting cq moderation parameters");
1319                         mlx4_en_deactivate_cq(priv, cq);
1320                         goto tx_err;
1321                 }
1322                 en_dbg(DRV, priv, "Resetting index of collapsed CQ:%d to -1\n", i);
1323                 cq->buf->wqe_index = cpu_to_be16(0xffff);
1324
1325                 /* Configure ring */
1326                 tx_ring = priv->tx_ring[i];
1327
1328                 err = mlx4_en_activate_tx_ring(priv, tx_ring, cq->mcq.cqn,
1329                                                i / priv->num_tx_rings_p_up);
1330                 if (err) {
1331                         en_err(priv, "Failed activating Tx ring %d\n", i);
1332                         mlx4_en_deactivate_cq(priv, cq);
1333                         goto tx_err;
1334                 }
1335
1336                 /* Arm CQ for TX completions */
1337                 mlx4_en_arm_cq(priv, cq);
1338
1339                 /* Set initial ownership of all Tx TXBBs to SW (1) */
1340                 for (j = 0; j < tx_ring->buf_size; j += STAMP_STRIDE)
1341                         *((u32 *) (tx_ring->buf + j)) = 0xffffffff;
1342                 ++tx_index;
1343         }
1344
1345         /* Configure port */
1346         err = mlx4_SET_PORT_general(mdev->dev, priv->port,
1347                                     priv->rx_mb_size,
1348                                     priv->prof->tx_pause,
1349                                     priv->prof->tx_ppp,
1350                                     priv->prof->rx_pause,
1351                                     priv->prof->rx_ppp);
1352         if (err) {
1353                 en_err(priv, "Failed setting port general configurations for port %d, with error %d\n",
1354                        priv->port, err);
1355                 goto tx_err;
1356         }
1357         /* Set default qp number */
1358         err = mlx4_SET_PORT_qpn_calc(mdev->dev, priv->port, priv->base_qpn, 0);
1359         if (err) {
1360                 en_err(priv, "Failed setting default qp numbers\n");
1361                 goto tx_err;
1362         }
1363
1364         /* Init port */
1365         en_dbg(HW, priv, "Initializing port\n");
1366         err = mlx4_INIT_PORT(mdev->dev, priv->port);
1367         if (err) {
1368                 en_err(priv, "Failed Initializing port\n");
1369                 goto tx_err;
1370         }
1371
1372         /* Attach rx QP to bradcast address */
1373         memset(&mc_list[10], 0xff, ETH_ALEN);
1374         mc_list[5] = priv->port; /* needed for B0 steering support */
1375         if (mlx4_multicast_attach(mdev->dev, &priv->rss_map.indir_qp, mc_list,
1376                                   priv->port, 0, MLX4_PROT_ETH,
1377                                   &priv->broadcast_id))
1378                 mlx4_warn(mdev, "Failed Attaching Broadcast\n");
1379
1380         /* Must redo promiscuous mode setup. */
1381         priv->flags &= ~(MLX4_EN_FLAG_PROMISC | MLX4_EN_FLAG_MC_PROMISC);
1382
1383         /* Schedule multicast task to populate multicast list */
1384         queue_work(mdev->workqueue, &priv->rx_mode_task);
1385
1386         mlx4_set_stats_bitmap(mdev->dev, priv->stats_bitmap);
1387
1388         priv->port_up = true;
1389
1390         /* Enable the queues. */
1391         dev->if_drv_flags &= ~IFF_DRV_OACTIVE;
1392         dev->if_drv_flags |= IFF_DRV_RUNNING;
1393 #ifdef CONFIG_DEBUG_FS
1394         mlx4_en_create_debug_files(priv);
1395 #endif
1396         callout_reset(&priv->watchdog_timer, MLX4_EN_WATCHDOG_TIMEOUT,
1397                     mlx4_en_watchdog_timeout, priv);
1398
1399
1400         return 0;
1401
1402 tx_err:
1403         while (tx_index--) {
1404                 mlx4_en_deactivate_tx_ring(priv, priv->tx_ring[tx_index]);
1405                 mlx4_en_deactivate_cq(priv, priv->tx_cq[tx_index]);
1406         }
1407         mlx4_en_destroy_drop_qp(priv);
1408 rss_err:
1409         mlx4_en_release_rss_steer(priv);
1410 mac_err:
1411         mlx4_en_put_qp(priv);
1412 cq_err:
1413         while (rx_index--)
1414                 mlx4_en_deactivate_cq(priv, priv->rx_cq[rx_index]);
1415         for (i = 0; i < priv->rx_ring_num; i++)
1416                 mlx4_en_deactivate_rx_ring(priv, priv->rx_ring[i]);
1417
1418         return err; /* need to close devices */
1419 }
1420
1421
1422 void mlx4_en_stop_port(struct net_device *dev)
1423 {
1424         struct mlx4_en_priv *priv = netdev_priv(dev);
1425         struct mlx4_en_dev *mdev = priv->mdev;
1426         struct mlx4_en_mc_list *mclist, *tmp;
1427         int i;
1428         u8 mc_list[16] = {0};
1429
1430         if (!priv->port_up) {
1431                 en_dbg(DRV, priv, "stop port called while port already down\n");
1432                 return;
1433         }
1434
1435 #ifdef CONFIG_DEBUG_FS
1436         mlx4_en_delete_debug_files(priv);
1437 #endif
1438
1439         /* close port*/
1440         mlx4_CLOSE_PORT(mdev->dev, priv->port);
1441
1442         /* Set port as not active */
1443         priv->port_up = false;
1444         if (priv->counter_index != 0xff) {
1445                 mlx4_counter_free(mdev->dev, priv->port, priv->counter_index);
1446                 priv->counter_index = 0xff;
1447         }
1448
1449         /* Promsicuous mode */
1450         if (mdev->dev->caps.steering_mode ==
1451             MLX4_STEERING_MODE_DEVICE_MANAGED) {
1452                 priv->flags &= ~(MLX4_EN_FLAG_PROMISC |
1453                                  MLX4_EN_FLAG_MC_PROMISC);
1454                 mlx4_flow_steer_promisc_remove(mdev->dev,
1455                                                priv->port,
1456                                                MLX4_FS_ALL_DEFAULT);
1457                 mlx4_flow_steer_promisc_remove(mdev->dev,
1458                                                priv->port,
1459                                                MLX4_FS_MC_DEFAULT);
1460         } else if (priv->flags & MLX4_EN_FLAG_PROMISC) {
1461                 priv->flags &= ~MLX4_EN_FLAG_PROMISC;
1462
1463                 /* Disable promiscouos mode */
1464                 mlx4_unicast_promisc_remove(mdev->dev, priv->base_qpn,
1465                                             priv->port);
1466
1467                 /* Disable Multicast promisc */
1468                 if (priv->flags & MLX4_EN_FLAG_MC_PROMISC) {
1469                         mlx4_multicast_promisc_remove(mdev->dev, priv->base_qpn,
1470                                                       priv->port);
1471                         priv->flags &= ~MLX4_EN_FLAG_MC_PROMISC;
1472                 }
1473         }
1474
1475         /* Detach All multicasts */
1476         memset(&mc_list[10], 0xff, ETH_ALEN);
1477         mc_list[5] = priv->port; /* needed for B0 steering support */
1478         mlx4_multicast_detach(mdev->dev, &priv->rss_map.indir_qp, mc_list,
1479                               MLX4_PROT_ETH, priv->broadcast_id);
1480         list_for_each_entry(mclist, &priv->curr_list, list) {
1481                 memcpy(&mc_list[10], mclist->addr, ETH_ALEN);
1482                 mc_list[5] = priv->port;
1483                 mlx4_multicast_detach(mdev->dev, &priv->rss_map.indir_qp,
1484                                       mc_list, MLX4_PROT_ETH, mclist->reg_id);
1485         }
1486         mlx4_en_clear_list(dev);
1487         list_for_each_entry_safe(mclist, tmp, &priv->curr_list, list) {
1488                 list_del(&mclist->list);
1489                 kfree(mclist);
1490         }
1491
1492         /* Flush multicast filter */
1493         mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, 0, 1, MLX4_MCAST_CONFIG);
1494         mlx4_en_destroy_drop_qp(priv);
1495
1496         /* Free TX Rings */
1497         for (i = 0; i < priv->tx_ring_num; i++) {
1498                 mlx4_en_deactivate_tx_ring(priv, priv->tx_ring[i]);
1499                 mlx4_en_deactivate_cq(priv, priv->tx_cq[i]);
1500         }
1501         msleep(10);
1502
1503         for (i = 0; i < priv->tx_ring_num; i++)
1504                 mlx4_en_free_tx_buf(dev, priv->tx_ring[i]);
1505
1506         /* Free RSS qps */
1507         mlx4_en_release_rss_steer(priv);
1508
1509         /* Unregister Mac address for the port */
1510         mlx4_en_put_qp(priv);
1511         mdev->mac_removed[priv->port] = 1;
1512
1513         /* Free RX Rings */
1514         for (i = 0; i < priv->rx_ring_num; i++) {
1515                 struct mlx4_en_cq *cq = priv->rx_cq[i];
1516                 mlx4_en_deactivate_rx_ring(priv, priv->rx_ring[i]);
1517                 mlx4_en_deactivate_cq(priv, cq);
1518         }
1519
1520         callout_stop(&priv->watchdog_timer);
1521
1522         dev->if_drv_flags &= ~(IFF_DRV_RUNNING | IFF_DRV_OACTIVE);
1523 }
1524
1525 static void mlx4_en_restart(struct work_struct *work)
1526 {
1527         struct mlx4_en_priv *priv = container_of(work, struct mlx4_en_priv,
1528                                                  watchdog_task);
1529         struct mlx4_en_dev *mdev = priv->mdev;
1530         struct net_device *dev = priv->dev;
1531         struct mlx4_en_tx_ring *ring;
1532         int i;
1533
1534
1535         if (priv->blocked == 0 || priv->port_up == 0)
1536                 return;
1537         for (i = 0; i < priv->tx_ring_num; i++) {
1538                 ring = priv->tx_ring[i];
1539                 if (ring->blocked &&
1540                                 ring->watchdog_time + MLX4_EN_WATCHDOG_TIMEOUT < ticks)
1541                         goto reset;
1542         }
1543         return;
1544
1545 reset:
1546         priv->port_stats.tx_timeout++;
1547         en_dbg(DRV, priv, "Watchdog task called for port %d\n", priv->port);
1548
1549         mutex_lock(&mdev->state_lock);
1550         if (priv->port_up) {
1551                 mlx4_en_stop_port(dev);
1552                 //for (i = 0; i < priv->tx_ring_num; i++)
1553                 //        netdev_tx_reset_queue(priv->tx_ring[i]->tx_queue);
1554                 if (mlx4_en_start_port(dev))
1555                         en_err(priv, "Failed restarting port %d\n", priv->port);
1556         }
1557         mutex_unlock(&mdev->state_lock);
1558 }
1559
1560 static void mlx4_en_clear_stats(struct net_device *dev)
1561 {
1562         struct mlx4_en_priv *priv = netdev_priv(dev);
1563         struct mlx4_en_dev *mdev = priv->mdev;
1564         int i;
1565
1566         if (!mlx4_is_slave(mdev->dev))
1567                 if (mlx4_en_DUMP_ETH_STATS(mdev, priv->port, 1))
1568                         en_dbg(HW, priv, "Failed dumping statistics\n");
1569
1570         memset(&priv->pstats, 0, sizeof(priv->pstats));
1571         memset(&priv->pkstats, 0, sizeof(priv->pkstats));
1572         memset(&priv->port_stats, 0, sizeof(priv->port_stats));
1573         memset(&priv->vport_stats, 0, sizeof(priv->vport_stats));
1574
1575         for (i = 0; i < priv->tx_ring_num; i++) {
1576                 priv->tx_ring[i]->bytes = 0;
1577                 priv->tx_ring[i]->packets = 0;
1578                 priv->tx_ring[i]->tx_csum = 0;
1579                 priv->tx_ring[i]->oversized_packets = 0;
1580         }
1581         for (i = 0; i < priv->rx_ring_num; i++) {
1582                 priv->rx_ring[i]->bytes = 0;
1583                 priv->rx_ring[i]->packets = 0;
1584                 priv->rx_ring[i]->csum_ok = 0;
1585                 priv->rx_ring[i]->csum_none = 0;
1586         }
1587 }
1588
1589 static void mlx4_en_open(void* arg)
1590 {
1591
1592         struct mlx4_en_priv *priv;
1593         struct mlx4_en_dev *mdev;
1594         struct net_device *dev;
1595         int err = 0;
1596
1597         priv = arg;
1598         mdev = priv->mdev;
1599         dev = priv->dev;
1600
1601
1602         mutex_lock(&mdev->state_lock);
1603
1604         if (!mdev->device_up) {
1605                 en_err(priv, "Cannot open - device down/disabled\n");
1606                 goto out;
1607         }
1608
1609         /* Reset HW statistics and SW counters */
1610         mlx4_en_clear_stats(dev);
1611
1612         err = mlx4_en_start_port(dev);
1613         if (err)
1614                 en_err(priv, "Failed starting port:%d\n", priv->port);
1615
1616 out:
1617         mutex_unlock(&mdev->state_lock);
1618         return;
1619 }
1620
1621 void mlx4_en_free_resources(struct mlx4_en_priv *priv)
1622 {
1623         int i;
1624
1625 #ifdef CONFIG_RFS_ACCEL
1626         if (priv->dev->rx_cpu_rmap) {
1627                 free_irq_cpu_rmap(priv->dev->rx_cpu_rmap);
1628                 priv->dev->rx_cpu_rmap = NULL;
1629         }
1630 #endif
1631
1632         for (i = 0; i < priv->tx_ring_num; i++) {
1633                 if (priv->tx_ring && priv->tx_ring[i])
1634                         mlx4_en_destroy_tx_ring(priv, &priv->tx_ring[i]);
1635                 if (priv->tx_cq && priv->tx_cq[i])
1636                         mlx4_en_destroy_cq(priv, &priv->tx_cq[i]);
1637         }
1638
1639         for (i = 0; i < priv->rx_ring_num; i++) {
1640                 if (priv->rx_ring[i])
1641                         mlx4_en_destroy_rx_ring(priv, &priv->rx_ring[i],
1642                                 priv->prof->rx_ring_size, priv->stride);
1643                 if (priv->rx_cq[i])
1644                         mlx4_en_destroy_cq(priv, &priv->rx_cq[i]);
1645         }
1646
1647         if (priv->sysctl)
1648                 sysctl_ctx_free(&priv->stat_ctx);
1649 }
1650
1651 int mlx4_en_alloc_resources(struct mlx4_en_priv *priv)
1652 {
1653         struct mlx4_en_port_profile *prof = priv->prof;
1654         int i;
1655         int node = 0;
1656
1657         /* Create rx Rings */
1658         for (i = 0; i < priv->rx_ring_num; i++) {
1659                 if (mlx4_en_create_cq(priv, &priv->rx_cq[i],
1660                                       prof->rx_ring_size, i, RX, node))
1661                         goto err;
1662
1663                 if (mlx4_en_create_rx_ring(priv, &priv->rx_ring[i],
1664                                            prof->rx_ring_size, node))
1665                         goto err;
1666         }
1667
1668         /* Create tx Rings */
1669         for (i = 0; i < priv->tx_ring_num; i++) {
1670                 if (mlx4_en_create_cq(priv, &priv->tx_cq[i],
1671                                       prof->tx_ring_size, i, TX, node))
1672                         goto err;
1673
1674                 if (mlx4_en_create_tx_ring(priv, &priv->tx_ring[i],
1675                                            prof->tx_ring_size, TXBB_SIZE, node, i))
1676                         goto err;
1677         }
1678
1679 #ifdef CONFIG_RFS_ACCEL
1680         priv->dev->rx_cpu_rmap = alloc_irq_cpu_rmap(priv->rx_ring_num);
1681         if (!priv->dev->rx_cpu_rmap)
1682                 goto err;
1683 #endif
1684         /* Re-create stat sysctls in case the number of rings changed. */
1685         mlx4_en_sysctl_stat(priv);
1686         return 0;
1687
1688 err:
1689         en_err(priv, "Failed to allocate NIC resources\n");
1690         for (i = 0; i < priv->rx_ring_num; i++) {
1691                 if (priv->rx_ring[i])
1692                         mlx4_en_destroy_rx_ring(priv, &priv->rx_ring[i],
1693                                                 prof->rx_ring_size,
1694                                                 priv->stride);
1695                 if (priv->rx_cq[i])
1696                         mlx4_en_destroy_cq(priv, &priv->rx_cq[i]);
1697         }
1698         for (i = 0; i < priv->tx_ring_num; i++) {
1699                 if (priv->tx_ring[i])
1700                         mlx4_en_destroy_tx_ring(priv, &priv->tx_ring[i]);
1701                 if (priv->tx_cq[i])
1702                         mlx4_en_destroy_cq(priv, &priv->tx_cq[i]);
1703         }
1704         priv->port_up = false;
1705         return -ENOMEM;
1706 }
1707
1708 struct en_port_attribute {
1709         struct attribute attr;
1710         ssize_t (*show)(struct en_port *, struct en_port_attribute *, char *buf);
1711         ssize_t (*store)(struct en_port *, struct en_port_attribute *, char *buf, size_t count);
1712 };
1713
1714 #define PORT_ATTR_RO(_name) \
1715 struct en_port_attribute en_port_attr_##_name = __ATTR_RO(_name)
1716
1717 #define EN_PORT_ATTR(_name, _mode, _show, _store) \
1718 struct en_port_attribute en_port_attr_##_name = __ATTR(_name, _mode, _show, _store)
1719
1720 void mlx4_en_destroy_netdev(struct net_device *dev)
1721 {
1722         struct mlx4_en_priv *priv = netdev_priv(dev);
1723         struct mlx4_en_dev *mdev = priv->mdev;
1724
1725         en_dbg(DRV, priv, "Destroying netdev on port:%d\n", priv->port);
1726
1727         if (priv->vlan_attach != NULL)
1728                 EVENTHANDLER_DEREGISTER(vlan_config, priv->vlan_attach);
1729         if (priv->vlan_detach != NULL)
1730                 EVENTHANDLER_DEREGISTER(vlan_unconfig, priv->vlan_detach);
1731
1732         /* Unregister device - this will close the port if it was up */
1733         if (priv->registered) {
1734                 mutex_lock(&mdev->state_lock);
1735                 ether_ifdetach(dev);
1736                 mutex_unlock(&mdev->state_lock);
1737         }
1738
1739         if (priv->allocated)
1740                 mlx4_free_hwq_res(mdev->dev, &priv->res, MLX4_EN_PAGE_SIZE);
1741
1742         mutex_lock(&mdev->state_lock);
1743         mlx4_en_stop_port(dev);
1744         mutex_unlock(&mdev->state_lock);
1745
1746
1747         cancel_delayed_work(&priv->stats_task);
1748         cancel_delayed_work(&priv->service_task);
1749         /* flush any pending task for this netdev */
1750         flush_workqueue(mdev->workqueue);
1751         callout_drain(&priv->watchdog_timer);
1752
1753         /* Detach the netdev so tasks would not attempt to access it */
1754         mutex_lock(&mdev->state_lock);
1755         mdev->pndev[priv->port] = NULL;
1756         mutex_unlock(&mdev->state_lock);
1757
1758
1759         mlx4_en_free_resources(priv);
1760
1761         /* freeing the sysctl conf cannot be called from within mlx4_en_free_resources */
1762         if (priv->sysctl)
1763                 sysctl_ctx_free(&priv->conf_ctx);
1764
1765         kfree(priv->tx_ring);
1766         kfree(priv->tx_cq);
1767
1768         kfree(priv);
1769         if_free(dev);
1770
1771 }
1772
1773 static int mlx4_en_change_mtu(struct net_device *dev, int new_mtu)
1774 {
1775         struct mlx4_en_priv *priv = netdev_priv(dev);
1776         struct mlx4_en_dev *mdev = priv->mdev;
1777         int err = 0;
1778
1779         en_dbg(DRV, priv, "Change MTU called - current:%u new:%u\n",
1780                (unsigned)dev->if_mtu, (unsigned)new_mtu);
1781
1782         if ((new_mtu < MLX4_EN_MIN_MTU) || (new_mtu > priv->max_mtu)) {
1783                 en_err(priv, "Bad MTU size:%d.\n", new_mtu);
1784                 return -EPERM;
1785         }
1786         mutex_lock(&mdev->state_lock);
1787         dev->if_mtu = new_mtu;
1788         if (dev->if_drv_flags & IFF_DRV_RUNNING) {
1789                 if (!mdev->device_up) {
1790                         /* NIC is probably restarting - let watchdog task reset
1791                          *                          * the port */
1792                         en_dbg(DRV, priv, "Change MTU called with card down!?\n");
1793                 } else {
1794                         mlx4_en_stop_port(dev);
1795                         err = mlx4_en_start_port(dev);
1796                         if (err) {
1797                                 en_err(priv, "Failed restarting port:%d\n",
1798                                                 priv->port);
1799                                 queue_work(mdev->workqueue, &priv->watchdog_task);
1800                         }
1801                 }
1802         }
1803         mutex_unlock(&mdev->state_lock);
1804         return 0;
1805 }
1806
1807 static int mlx4_en_calc_media(struct mlx4_en_priv *priv)
1808 {
1809         int trans_type;
1810         int active;
1811
1812         active = IFM_ETHER;
1813         if (priv->last_link_state == MLX4_DEV_EVENT_PORT_DOWN)
1814                 return (active);
1815         active |= IFM_FDX;
1816         trans_type = priv->port_state.transciver;
1817         /* XXX I don't know all of the transceiver values. */
1818         switch (priv->port_state.link_speed) {
1819         case 1000:
1820                 active |= IFM_1000_T;
1821                 break;
1822         case 10000:
1823                 if (trans_type > 0 && trans_type <= 0xC)
1824                         active |= IFM_10G_SR;
1825                 else if (trans_type == 0x80 || trans_type == 0)
1826                         active |= IFM_10G_CX4;
1827                 break;
1828         case 40000:
1829                 active |= IFM_40G_CR4;
1830                 break;
1831         }
1832         if (priv->prof->tx_pause)
1833                 active |= IFM_ETH_TXPAUSE;
1834         if (priv->prof->rx_pause)
1835                 active |= IFM_ETH_RXPAUSE;
1836
1837         return (active);
1838 }
1839
1840 static void mlx4_en_media_status(struct ifnet *dev, struct ifmediareq *ifmr)
1841 {
1842         struct mlx4_en_priv *priv;
1843
1844         priv = dev->if_softc;
1845         ifmr->ifm_status = IFM_AVALID;
1846         if (priv->last_link_state != MLX4_DEV_EVENT_PORT_DOWN)
1847                 ifmr->ifm_status |= IFM_ACTIVE;
1848         ifmr->ifm_active = mlx4_en_calc_media(priv);
1849
1850         return;
1851 }
1852
1853 static int mlx4_en_media_change(struct ifnet *dev)
1854 {
1855         struct mlx4_en_priv *priv;
1856         struct ifmedia *ifm;
1857         int rxpause;
1858         int txpause;
1859         int error;
1860
1861         priv = dev->if_softc;
1862         ifm = &priv->media;
1863         rxpause = txpause = 0;
1864         error = 0;
1865
1866         if (IFM_TYPE(ifm->ifm_media) != IFM_ETHER)
1867                 return (EINVAL);
1868         switch (IFM_SUBTYPE(ifm->ifm_media)) {
1869         case IFM_AUTO:
1870                 break;
1871         case IFM_10G_SR:
1872         case IFM_10G_CX4:
1873         case IFM_1000_T:
1874         case IFM_40G_CR4:
1875                 if ((IFM_SUBTYPE(ifm->ifm_media)
1876                         == IFM_SUBTYPE(mlx4_en_calc_media(priv)))
1877                         && (ifm->ifm_media & IFM_FDX))
1878                         break;
1879                 /* Fallthrough */
1880         default:
1881                 printf("%s: Only auto media type\n", if_name(dev));
1882                 return (EINVAL);
1883         }
1884         /* Allow user to set/clear pause */
1885         if (IFM_OPTIONS(ifm->ifm_media) & IFM_ETH_RXPAUSE)
1886                 rxpause = 1;
1887         if (IFM_OPTIONS(ifm->ifm_media) & IFM_ETH_TXPAUSE)
1888                 txpause = 1;
1889         if (priv->prof->tx_pause != txpause || priv->prof->rx_pause != rxpause) {
1890                 priv->prof->tx_pause = txpause;
1891                 priv->prof->rx_pause = rxpause;
1892                 error = -mlx4_SET_PORT_general(priv->mdev->dev, priv->port,
1893                      priv->rx_mb_size + ETHER_CRC_LEN, priv->prof->tx_pause,
1894                      priv->prof->tx_ppp, priv->prof->rx_pause,
1895                      priv->prof->rx_ppp);
1896         }
1897         return (error);
1898 }
1899
1900 static int mlx4_en_ioctl(struct ifnet *dev, u_long command, caddr_t data)
1901 {
1902         struct mlx4_en_priv *priv;
1903         struct mlx4_en_dev *mdev;
1904         struct ifreq *ifr;
1905         int error;
1906         int mask;
1907
1908         error = 0;
1909         mask = 0;
1910         priv = dev->if_softc;
1911         mdev = priv->mdev;
1912         ifr = (struct ifreq *) data;
1913         switch (command) {
1914
1915         case SIOCSIFMTU:
1916                 error = -mlx4_en_change_mtu(dev, ifr->ifr_mtu);
1917                 break;
1918         case SIOCSIFFLAGS:
1919                 if (dev->if_flags & IFF_UP) {
1920                         if ((dev->if_drv_flags & IFF_DRV_RUNNING) == 0) {
1921                                 mutex_lock(&mdev->state_lock);
1922                                 mlx4_en_start_port(dev);
1923                                 mutex_unlock(&mdev->state_lock);
1924                         } else {
1925                                 mlx4_en_set_rx_mode(dev);
1926                         }
1927                 } else {
1928                         mutex_lock(&mdev->state_lock);
1929                         if (dev->if_drv_flags & IFF_DRV_RUNNING) {
1930                                 mlx4_en_stop_port(dev);
1931                                 if_link_state_change(dev, LINK_STATE_DOWN);
1932                         }
1933                         mutex_unlock(&mdev->state_lock);
1934                 }
1935                 break;
1936         case SIOCADDMULTI:
1937         case SIOCDELMULTI:
1938                 mlx4_en_set_rx_mode(dev);
1939                 break;
1940         case SIOCSIFMEDIA:
1941         case SIOCGIFMEDIA:
1942                 error = ifmedia_ioctl(dev, ifr, &priv->media, command);
1943                 break;
1944         case SIOCSIFCAP:
1945                 mutex_lock(&mdev->state_lock);
1946                 mask = ifr->ifr_reqcap ^ dev->if_capenable;
1947                 if (mask & IFCAP_TXCSUM) {
1948                         dev->if_capenable ^= IFCAP_TXCSUM;
1949                         dev->if_hwassist ^= (CSUM_TCP | CSUM_UDP | CSUM_IP);
1950
1951                         if (IFCAP_TSO4 & dev->if_capenable &&
1952                             !(IFCAP_TXCSUM & dev->if_capenable)) {
1953                                 dev->if_capenable &= ~IFCAP_TSO4;
1954                                 dev->if_hwassist &= ~CSUM_TSO;
1955                                 if_printf(dev,
1956                                     "tso4 disabled due to -txcsum.\n");
1957                         }
1958                 }
1959                 if (mask & IFCAP_TXCSUM_IPV6) {
1960                         dev->if_capenable ^= IFCAP_TXCSUM_IPV6;
1961                         dev->if_hwassist ^= (CSUM_UDP_IPV6 | CSUM_TCP_IPV6);
1962
1963                         if (IFCAP_TSO6 & dev->if_capenable &&
1964                             !(IFCAP_TXCSUM_IPV6 & dev->if_capenable)) {
1965                                 dev->if_capenable &= ~IFCAP_TSO6;
1966                                 dev->if_hwassist &= ~CSUM_TSO;
1967                                 if_printf(dev,
1968                                     "tso6 disabled due to -txcsum6.\n");
1969                         }
1970                 }
1971                 if (mask & IFCAP_RXCSUM)
1972                         dev->if_capenable ^= IFCAP_RXCSUM;
1973                 if (mask & IFCAP_RXCSUM_IPV6)
1974                         dev->if_capenable ^= IFCAP_RXCSUM_IPV6;
1975
1976                 if (mask & IFCAP_TSO4) {
1977                         if (!(IFCAP_TSO4 & dev->if_capenable) &&
1978                             !(IFCAP_TXCSUM & dev->if_capenable)) {
1979                                 if_printf(dev, "enable txcsum first.\n");
1980                                 error = EAGAIN;
1981                                 goto out;
1982                         }
1983                         dev->if_capenable ^= IFCAP_TSO4;
1984                 }
1985                 if (mask & IFCAP_TSO6) {
1986                         if (!(IFCAP_TSO6 & dev->if_capenable) &&
1987                             !(IFCAP_TXCSUM_IPV6 & dev->if_capenable)) {
1988                                 if_printf(dev, "enable txcsum6 first.\n");
1989                                 error = EAGAIN;
1990                                 goto out;
1991                         }
1992                         dev->if_capenable ^= IFCAP_TSO6;
1993                 }
1994                 if (dev->if_capenable & (IFCAP_TSO4 | IFCAP_TSO6))
1995                         dev->if_hwassist |= CSUM_TSO;
1996                 else
1997                         dev->if_hwassist &= ~CSUM_TSO;
1998                 if (mask & IFCAP_LRO)
1999                         dev->if_capenable ^= IFCAP_LRO;
2000                 if (mask & IFCAP_VLAN_HWTAGGING)
2001                         dev->if_capenable ^= IFCAP_VLAN_HWTAGGING;
2002                 if (mask & IFCAP_VLAN_HWFILTER)
2003                         dev->if_capenable ^= IFCAP_VLAN_HWFILTER;
2004                 if (mask & IFCAP_WOL_MAGIC)
2005                         dev->if_capenable ^= IFCAP_WOL_MAGIC;
2006                 if (dev->if_drv_flags & IFF_DRV_RUNNING)
2007                         mlx4_en_start_port(dev);
2008 out:
2009                 mutex_unlock(&mdev->state_lock);
2010                 VLAN_CAPABILITIES(dev);
2011                 break;
2012 #if __FreeBSD_version >= 1100036
2013         case SIOCGI2C: {
2014                 struct ifi2creq i2c;
2015
2016                 error = copyin(ifr->ifr_data, &i2c, sizeof(i2c));
2017                 if (error)
2018                         break;
2019                 if (i2c.len > sizeof(i2c.data)) {
2020                         error = EINVAL;
2021                         break;
2022                 }
2023                 /*
2024                  * Note that we ignore i2c.addr here. The driver hardcodes
2025                  * the address to 0x50, while standard expects it to be 0xA0.
2026                  */
2027                 error = mlx4_get_module_info(mdev->dev, priv->port,
2028                     i2c.offset, i2c.len, i2c.data);
2029                 if (error < 0) {
2030                         error = -error;
2031                         break;
2032                 }
2033                 error = copyout(&i2c, ifr->ifr_data, sizeof(i2c));
2034                 break;
2035         }
2036 #endif
2037         default:
2038                 error = ether_ioctl(dev, command, data);
2039                 break;
2040         }
2041
2042         return (error);
2043 }
2044
2045
2046 int mlx4_en_init_netdev(struct mlx4_en_dev *mdev, int port,
2047                         struct mlx4_en_port_profile *prof)
2048 {
2049         struct net_device *dev;
2050         struct mlx4_en_priv *priv;
2051         uint8_t dev_addr[ETHER_ADDR_LEN];
2052         int err;
2053         int i;
2054
2055         priv = kzalloc(sizeof(*priv), GFP_KERNEL);
2056         dev = priv->dev = if_alloc(IFT_ETHER);
2057         if (dev == NULL) {
2058                 en_err(priv, "Net device allocation failed\n");
2059                 kfree(priv);
2060                 return -ENOMEM;
2061         }
2062         dev->if_softc = priv;
2063         if_initname(dev, "mlxen", atomic_fetchadd_int(&mlx4_en_unit, 1));
2064         dev->if_mtu = ETHERMTU;
2065         dev->if_init = mlx4_en_open;
2066         dev->if_flags = IFF_BROADCAST | IFF_SIMPLEX | IFF_MULTICAST;
2067         dev->if_ioctl = mlx4_en_ioctl;
2068         dev->if_transmit = mlx4_en_transmit;
2069         dev->if_qflush = mlx4_en_qflush;
2070         dev->if_snd.ifq_maxlen = prof->tx_ring_size;
2071
2072         /*
2073          * Initialize driver private data
2074          */
2075         priv->counter_index = 0xff;
2076         spin_lock_init(&priv->stats_lock);
2077         INIT_WORK(&priv->rx_mode_task, mlx4_en_do_set_rx_mode);
2078         INIT_WORK(&priv->watchdog_task, mlx4_en_restart);
2079         INIT_WORK(&priv->linkstate_task, mlx4_en_linkstate);
2080         INIT_DELAYED_WORK(&priv->stats_task, mlx4_en_do_get_stats);
2081         INIT_DELAYED_WORK(&priv->service_task, mlx4_en_service_task);
2082         callout_init(&priv->watchdog_timer, 1);
2083 #ifdef CONFIG_RFS_ACCEL
2084         INIT_LIST_HEAD(&priv->filters);
2085         spin_lock_init(&priv->filters_lock);
2086 #endif
2087
2088         priv->msg_enable = MLX4_EN_MSG_LEVEL;
2089         priv->dev = dev;
2090         priv->mdev = mdev;
2091         priv->ddev = &mdev->pdev->dev;
2092         priv->prof = prof;
2093         priv->port = port;
2094         priv->port_up = false;
2095         priv->flags = prof->flags;
2096
2097         priv->num_tx_rings_p_up = mdev->profile.num_tx_rings_p_up;
2098         priv->tx_ring_num = prof->tx_ring_num;
2099         priv->tx_ring = kcalloc(MAX_TX_RINGS,
2100                                 sizeof(struct mlx4_en_tx_ring *), GFP_KERNEL);
2101         if (!priv->tx_ring) {
2102                 err = -ENOMEM;
2103                 goto out;
2104         }
2105         priv->tx_cq = kcalloc(sizeof(struct mlx4_en_cq *), MAX_TX_RINGS,
2106                         GFP_KERNEL);
2107         if (!priv->tx_cq) {
2108                 err = -ENOMEM;
2109                 goto out;
2110         }
2111
2112         priv->rx_ring_num = prof->rx_ring_num;
2113         priv->cqe_factor = (mdev->dev->caps.cqe_size == 64) ? 1 : 0;
2114         priv->mac_index = -1;
2115         priv->last_ifq_jiffies = 0;
2116         priv->if_counters_rx_errors = 0;
2117         priv->if_counters_rx_no_buffer = 0;
2118 #ifdef CONFIG_MLX4_EN_DCB
2119         if (!mlx4_is_slave(priv->mdev->dev)) {
2120                 priv->dcbx_cap = DCB_CAP_DCBX_HOST;
2121                 priv->flags |= MLX4_EN_FLAG_DCB_ENABLED;
2122                 if (mdev->dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_ETS_CFG) {
2123                         dev->dcbnl_ops = &mlx4_en_dcbnl_ops;
2124                 } else {
2125                         en_info(priv, "QoS disabled - no HW support\n");
2126                         dev->dcbnl_ops = &mlx4_en_dcbnl_pfc_ops;
2127                 }
2128         }
2129 #endif
2130
2131         for (i = 0; i < MLX4_EN_MAC_HASH_SIZE; ++i)
2132                 INIT_HLIST_HEAD(&priv->mac_hash[i]);
2133
2134         /* Query for default mac and max mtu */
2135         priv->max_mtu = mdev->dev->caps.eth_mtu_cap[priv->port];
2136         priv->mac = mdev->dev->caps.def_mac[priv->port];
2137         if (ILLEGAL_MAC(priv->mac)) {
2138 #if BITS_PER_LONG == 64
2139                 en_err(priv, "Port: %d, invalid mac burned: 0x%lx, quiting\n",
2140                                 priv->port, priv->mac);
2141 #elif BITS_PER_LONG == 32
2142                 en_err(priv, "Port: %d, invalid mac burned: 0x%llx, quiting\n",
2143                                 priv->port, priv->mac);
2144 #endif
2145                 err = -EINVAL;
2146                 goto out;
2147         }
2148
2149         priv->stride = roundup_pow_of_two(sizeof(struct mlx4_en_rx_desc) +
2150                                           DS_SIZE);
2151
2152         mlx4_en_sysctl_conf(priv);
2153
2154         err = mlx4_en_alloc_resources(priv);
2155         if (err)
2156                 goto out;
2157
2158         /* Allocate page for receive rings */
2159         err = mlx4_alloc_hwq_res(mdev->dev, &priv->res,
2160                                 MLX4_EN_PAGE_SIZE, MLX4_EN_PAGE_SIZE);
2161         if (err) {
2162                 en_err(priv, "Failed to allocate page for rx qps\n");
2163                 goto out;
2164         }
2165         priv->allocated = 1;
2166
2167         /*
2168          * Set driver features
2169          */
2170         dev->if_capabilities |= IFCAP_HWCSUM | IFCAP_HWCSUM_IPV6;
2171         dev->if_capabilities |= IFCAP_VLAN_MTU | IFCAP_VLAN_HWTAGGING;
2172         dev->if_capabilities |= IFCAP_VLAN_HWCSUM | IFCAP_VLAN_HWFILTER;
2173         dev->if_capabilities |= IFCAP_LINKSTATE | IFCAP_JUMBO_MTU;
2174         dev->if_capabilities |= IFCAP_LRO;
2175
2176         if (mdev->LSO_support)
2177                 dev->if_capabilities |= IFCAP_TSO4 | IFCAP_TSO6 | IFCAP_VLAN_HWTSO;
2178
2179         /* set TSO limits so that we don't have to drop TX packets */
2180         dev->if_hw_tsomax = MLX4_EN_TX_MAX_PAYLOAD_SIZE - (ETHER_HDR_LEN + ETHER_VLAN_ENCAP_LEN) /* hdr */;
2181         dev->if_hw_tsomaxsegcount = MLX4_EN_TX_MAX_MBUF_FRAGS - 1 /* hdr */;
2182         dev->if_hw_tsomaxsegsize = MLX4_EN_TX_MAX_MBUF_SIZE;
2183
2184         dev->if_capenable = dev->if_capabilities;
2185
2186         dev->if_hwassist = 0;
2187         if (dev->if_capenable & (IFCAP_TSO4 | IFCAP_TSO6))
2188                 dev->if_hwassist |= CSUM_TSO;
2189         if (dev->if_capenable & IFCAP_TXCSUM)
2190                 dev->if_hwassist |= (CSUM_TCP | CSUM_UDP | CSUM_IP);
2191         if (dev->if_capenable & IFCAP_TXCSUM_IPV6)
2192                 dev->if_hwassist |= (CSUM_UDP_IPV6 | CSUM_TCP_IPV6);
2193
2194
2195         /* Register for VLAN events */
2196         priv->vlan_attach = EVENTHANDLER_REGISTER(vlan_config,
2197             mlx4_en_vlan_rx_add_vid, priv, EVENTHANDLER_PRI_FIRST);
2198         priv->vlan_detach = EVENTHANDLER_REGISTER(vlan_unconfig,
2199             mlx4_en_vlan_rx_kill_vid, priv, EVENTHANDLER_PRI_FIRST);
2200
2201         mdev->pndev[priv->port] = dev;
2202
2203         priv->last_link_state = MLX4_DEV_EVENT_PORT_DOWN;
2204         mlx4_en_set_default_moderation(priv);
2205
2206         /* Set default MAC */
2207         for (i = 0; i < ETHER_ADDR_LEN; i++)
2208                 dev_addr[ETHER_ADDR_LEN - 1 - i] = (u8) (priv->mac >> (8 * i));
2209
2210
2211         ether_ifattach(dev, dev_addr);
2212         if_link_state_change(dev, LINK_STATE_DOWN);
2213         ifmedia_init(&priv->media, IFM_IMASK | IFM_ETH_FMASK,
2214             mlx4_en_media_change, mlx4_en_media_status);
2215         ifmedia_add(&priv->media, IFM_ETHER | IFM_FDX | IFM_1000_T, 0, NULL);
2216         ifmedia_add(&priv->media, IFM_ETHER | IFM_FDX | IFM_10G_SR, 0, NULL);
2217         ifmedia_add(&priv->media, IFM_ETHER | IFM_FDX | IFM_10G_CX4, 0, NULL);
2218         ifmedia_add(&priv->media, IFM_ETHER | IFM_FDX | IFM_40G_CR4, 0, NULL);
2219         ifmedia_add(&priv->media, IFM_ETHER | IFM_AUTO, 0, NULL);
2220         ifmedia_set(&priv->media, IFM_ETHER | IFM_AUTO);
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         priv->registered = 1;
2226
2227         en_warn(priv, "Using %d TX rings\n", prof->tx_ring_num);
2228         en_warn(priv, "Using %d RX rings\n", prof->rx_ring_num);
2229
2230
2231         priv->rx_mb_size = dev->if_mtu + ETH_HLEN + VLAN_HLEN + ETH_FCS_LEN;
2232         err = mlx4_SET_PORT_general(mdev->dev, priv->port,
2233                                     priv->rx_mb_size,
2234                                     prof->tx_pause, prof->tx_ppp,
2235                                     prof->rx_pause, prof->rx_ppp);
2236         if (err) {
2237                 en_err(priv, "Failed setting port general configurations "
2238                        "for port %d, with error %d\n", priv->port, err);
2239                 goto out;
2240         }
2241
2242         /* Init port */
2243         en_warn(priv, "Initializing port\n");
2244         err = mlx4_INIT_PORT(mdev->dev, priv->port);
2245         if (err) {
2246                 en_err(priv, "Failed Initializing port\n");
2247                 goto out;
2248         }
2249
2250         queue_delayed_work(mdev->workqueue, &priv->stats_task, STATS_DELAY);
2251
2252         if (mdev->dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_TS)
2253                 queue_delayed_work(mdev->workqueue, &priv->service_task, SERVICE_TASK_DELAY);
2254
2255         return 0;
2256
2257 out:
2258         mlx4_en_destroy_netdev(dev);
2259         return err;
2260 }
2261
2262 static int mlx4_en_set_ring_size(struct net_device *dev,
2263     int rx_size, int tx_size)
2264 {
2265         struct mlx4_en_priv *priv = netdev_priv(dev);
2266         struct mlx4_en_dev *mdev = priv->mdev;
2267         int port_up = 0;
2268         int err = 0;
2269
2270         rx_size = roundup_pow_of_two(rx_size);
2271         rx_size = max_t(u32, rx_size, MLX4_EN_MIN_RX_SIZE);
2272         rx_size = min_t(u32, rx_size, MLX4_EN_MAX_RX_SIZE);
2273         tx_size = roundup_pow_of_two(tx_size);
2274         tx_size = max_t(u32, tx_size, MLX4_EN_MIN_TX_SIZE);
2275         tx_size = min_t(u32, tx_size, MLX4_EN_MAX_TX_SIZE);
2276
2277         if (rx_size == (priv->port_up ?
2278             priv->rx_ring[0]->actual_size : priv->rx_ring[0]->size) &&
2279             tx_size == priv->tx_ring[0]->size)
2280                 return 0;
2281         mutex_lock(&mdev->state_lock);
2282         if (priv->port_up) {
2283                 port_up = 1;
2284                 mlx4_en_stop_port(dev);
2285         }
2286         mlx4_en_free_resources(priv);
2287         priv->prof->tx_ring_size = tx_size;
2288         priv->prof->rx_ring_size = rx_size;
2289         err = mlx4_en_alloc_resources(priv);
2290         if (err) {
2291                 en_err(priv, "Failed reallocating port resources\n");
2292                 goto out;
2293         }
2294         if (port_up) {
2295                 err = mlx4_en_start_port(dev);
2296                 if (err)
2297                         en_err(priv, "Failed starting port\n");
2298         }
2299 out:
2300         mutex_unlock(&mdev->state_lock);
2301         return err;
2302 }
2303 static int mlx4_en_set_rx_ring_size(SYSCTL_HANDLER_ARGS)
2304 {
2305         struct mlx4_en_priv *priv;
2306         int size;
2307         int error;
2308
2309         priv = arg1;
2310         size = priv->prof->rx_ring_size;
2311         error = sysctl_handle_int(oidp, &size, 0, req);
2312         if (error || !req->newptr)
2313                 return (error);
2314         error = -mlx4_en_set_ring_size(priv->dev, size,
2315             priv->prof->tx_ring_size);
2316         return (error);
2317 }
2318
2319 static int mlx4_en_set_tx_ring_size(SYSCTL_HANDLER_ARGS)
2320 {
2321         struct mlx4_en_priv *priv;
2322         int size;
2323         int error;
2324
2325         priv = arg1;
2326         size = priv->prof->tx_ring_size;
2327         error = sysctl_handle_int(oidp, &size, 0, req);
2328         if (error || !req->newptr)
2329                 return (error);
2330         error = -mlx4_en_set_ring_size(priv->dev, priv->prof->rx_ring_size,
2331             size);
2332
2333         return (error);
2334 }
2335
2336 static int mlx4_en_get_module_info(struct net_device *dev,
2337                                    struct ethtool_modinfo *modinfo)
2338 {
2339         struct mlx4_en_priv *priv = netdev_priv(dev);
2340         struct mlx4_en_dev *mdev = priv->mdev;
2341         int ret;
2342         u8 data[4];
2343
2344         /* Read first 2 bytes to get Module & REV ID */
2345         ret = mlx4_get_module_info(mdev->dev, priv->port,
2346                                    0/*offset*/, 2/*size*/, data);
2347
2348         if (ret < 2) {
2349                 en_err(priv, "Failed to read eeprom module first two bytes, error: 0x%x\n", -ret);
2350                 return -EIO;
2351         }
2352
2353         switch (data[0] /* identifier */) {
2354         case MLX4_MODULE_ID_QSFP:
2355                 modinfo->type = ETH_MODULE_SFF_8436;
2356                 modinfo->eeprom_len = ETH_MODULE_SFF_8436_LEN;
2357                 break;
2358         case MLX4_MODULE_ID_QSFP_PLUS:
2359                 if (data[1] >= 0x3) { /* revision id */
2360                         modinfo->type = ETH_MODULE_SFF_8636;
2361                         modinfo->eeprom_len = ETH_MODULE_SFF_8636_LEN;
2362                 } else {
2363                         modinfo->type = ETH_MODULE_SFF_8436;
2364                         modinfo->eeprom_len = ETH_MODULE_SFF_8436_LEN;
2365                 }
2366                 break;
2367         case MLX4_MODULE_ID_QSFP28:
2368                 modinfo->type = ETH_MODULE_SFF_8636;
2369                 modinfo->eeprom_len = ETH_MODULE_SFF_8636_LEN;
2370                 break;
2371         case MLX4_MODULE_ID_SFP:
2372                 modinfo->type = ETH_MODULE_SFF_8472;
2373                 modinfo->eeprom_len = ETH_MODULE_SFF_8472_LEN;
2374                 break;
2375         default:
2376                 en_err(priv, "mlx4_en_get_module_info :  Not recognized cable type\n");
2377                 return -EINVAL;
2378         }
2379
2380         return 0;
2381 }
2382
2383 static int mlx4_en_get_module_eeprom(struct net_device *dev,
2384                                      struct ethtool_eeprom *ee,
2385                                      u8 *data)
2386 {
2387         struct mlx4_en_priv *priv = netdev_priv(dev);
2388         struct mlx4_en_dev *mdev = priv->mdev;
2389         int offset = ee->offset;
2390         int i = 0, ret;
2391
2392         if (ee->len == 0)
2393                 return -EINVAL;
2394
2395         memset(data, 0, ee->len);
2396
2397         while (i < ee->len) {
2398                 en_dbg(DRV, priv,
2399                        "mlx4_get_module_info i(%d) offset(%d) len(%d)\n",
2400                        i, offset, ee->len - i);
2401
2402                 ret = mlx4_get_module_info(mdev->dev, priv->port,
2403                                            offset, ee->len - i, data + i);
2404
2405                 if (!ret) /* Done reading */
2406                         return 0;
2407
2408                 if (ret < 0) {
2409                         en_err(priv,
2410                                "mlx4_get_module_info i(%d) offset(%d) bytes_to_read(%d) - FAILED (0x%x)\n",
2411                                i, offset, ee->len - i, ret);
2412                         return -1;
2413                 }
2414
2415                 i += ret;
2416                 offset += ret;
2417         }
2418         return 0;
2419 }
2420
2421 static void mlx4_en_print_eeprom(u8 *data, __u32 len)
2422 {
2423         int             i;
2424         int             j = 0;
2425         int             row = 0;
2426         const int       NUM_OF_BYTES = 16;
2427
2428         printf("\nOffset\t\tValues\n");
2429         printf("------\t\t------\n");
2430         while(row < len){
2431                 printf("0x%04x\t\t",row);
2432                 for(i=0; i < NUM_OF_BYTES; i++){
2433                         printf("%02x ", data[j]);
2434                         row++;
2435                         j++;
2436                 }
2437                 printf("\n");
2438         }
2439 }
2440
2441 /* Read cable EEPROM module information by first inspecting the first
2442  * two bytes to get the length and then read the rest of the information.
2443  * The information is printed to dmesg. */
2444 static int mlx4_en_read_eeprom(SYSCTL_HANDLER_ARGS)
2445 {
2446
2447         u8*             data;
2448         int             error;
2449         int             result = 0;
2450         struct          mlx4_en_priv *priv;
2451         struct          net_device *dev;
2452         struct          ethtool_modinfo modinfo;
2453         struct          ethtool_eeprom ee;
2454
2455         error = sysctl_handle_int(oidp, &result, 0, req);
2456         if (error || !req->newptr)
2457                 return (error);
2458
2459         if (result == 1) {
2460                 priv = arg1;
2461                 dev = priv->dev;
2462                 data = kmalloc(PAGE_SIZE, GFP_KERNEL);
2463
2464                 error = mlx4_en_get_module_info(dev, &modinfo);
2465                 if (error) {
2466                         en_err(priv,
2467                                "mlx4_en_get_module_info returned with error - FAILED (0x%x)\n",
2468                                -error);
2469                         goto out;
2470                 }
2471
2472                 ee.len = modinfo.eeprom_len;
2473                 ee.offset = 0;
2474
2475                 error = mlx4_en_get_module_eeprom(dev, &ee, data);
2476                 if (error) {
2477                         en_err(priv,
2478                                "mlx4_en_get_module_eeprom returned with error - FAILED (0x%x)\n",
2479                                -error);
2480                         /* Continue printing partial information in case of an error */
2481                 }
2482
2483                 /* EEPROM information will be printed in dmesg */
2484                 mlx4_en_print_eeprom(data, ee.len);
2485 out:
2486                 kfree(data);
2487         }
2488         /* Return zero to prevent sysctl failure. */
2489         return (0);
2490 }
2491
2492 static int mlx4_en_set_tx_ppp(SYSCTL_HANDLER_ARGS)
2493 {
2494         struct mlx4_en_priv *priv;
2495         int ppp;
2496         int error;
2497
2498         priv = arg1;
2499         ppp = priv->prof->tx_ppp;
2500         error = sysctl_handle_int(oidp, &ppp, 0, req);
2501         if (error || !req->newptr)
2502                 return (error);
2503         if (ppp > 0xff || ppp < 0)
2504                 return (-EINVAL);
2505         priv->prof->tx_ppp = ppp;
2506         error = -mlx4_SET_PORT_general(priv->mdev->dev, priv->port,
2507                                        priv->rx_mb_size + ETHER_CRC_LEN,
2508                                        priv->prof->tx_pause,
2509                                        priv->prof->tx_ppp,
2510                                        priv->prof->rx_pause,
2511                                        priv->prof->rx_ppp);
2512
2513         return (error);
2514 }
2515
2516 static int mlx4_en_set_rx_ppp(SYSCTL_HANDLER_ARGS)
2517 {
2518         struct mlx4_en_priv *priv;
2519         struct mlx4_en_dev *mdev;
2520         int ppp;
2521         int error;
2522         int port_up;
2523
2524         port_up = 0;
2525         priv = arg1;
2526         mdev = priv->mdev;
2527         ppp = priv->prof->rx_ppp;
2528         error = sysctl_handle_int(oidp, &ppp, 0, req);
2529         if (error || !req->newptr)
2530                 return (error);
2531         if (ppp > 0xff || ppp < 0)
2532                 return (-EINVAL);
2533         /* See if we have to change the number of tx queues. */
2534         if (!ppp != !priv->prof->rx_ppp) {
2535                 mutex_lock(&mdev->state_lock);
2536                 if (priv->port_up) {
2537                         port_up = 1;
2538                         mlx4_en_stop_port(priv->dev);
2539                 }
2540                 mlx4_en_free_resources(priv);
2541                 priv->prof->rx_ppp = ppp;
2542                 error = -mlx4_en_alloc_resources(priv);
2543                 if (error)
2544                         en_err(priv, "Failed reallocating port resources\n");
2545                 if (error == 0 && port_up) {
2546                         error = -mlx4_en_start_port(priv->dev);
2547                         if (error)
2548                                 en_err(priv, "Failed starting port\n");
2549                 }
2550                 mutex_unlock(&mdev->state_lock);
2551                 return (error);
2552
2553         }
2554         priv->prof->rx_ppp = ppp;
2555         error = -mlx4_SET_PORT_general(priv->mdev->dev, priv->port,
2556                                        priv->rx_mb_size + ETHER_CRC_LEN,
2557                                        priv->prof->tx_pause,
2558                                        priv->prof->tx_ppp,
2559                                        priv->prof->rx_pause,
2560                                        priv->prof->rx_ppp);
2561
2562         return (error);
2563 }
2564
2565 static void mlx4_en_sysctl_conf(struct mlx4_en_priv *priv)
2566 {
2567         struct net_device *dev;
2568         struct sysctl_ctx_list *ctx;
2569         struct sysctl_oid *node;
2570         struct sysctl_oid_list *node_list;
2571         struct sysctl_oid *coal;
2572         struct sysctl_oid_list *coal_list;
2573         const char *pnameunit;
2574
2575         dev = priv->dev;
2576         ctx = &priv->conf_ctx;
2577         pnameunit = device_get_nameunit(priv->mdev->pdev->dev.bsddev);
2578
2579         sysctl_ctx_init(ctx);
2580         priv->sysctl = SYSCTL_ADD_NODE(ctx, SYSCTL_STATIC_CHILDREN(_hw),
2581             OID_AUTO, dev->if_xname, CTLFLAG_RD, 0, "mlx4 10gig ethernet");
2582         node = SYSCTL_ADD_NODE(ctx, SYSCTL_CHILDREN(priv->sysctl), OID_AUTO,
2583             "conf", CTLFLAG_RD, NULL, "Configuration");
2584         node_list = SYSCTL_CHILDREN(node);
2585
2586         SYSCTL_ADD_UINT(ctx, node_list, OID_AUTO, "msg_enable",
2587             CTLFLAG_RW, &priv->msg_enable, 0,
2588             "Driver message enable bitfield");
2589         SYSCTL_ADD_UINT(ctx, node_list, OID_AUTO, "rx_rings",
2590             CTLFLAG_RD, &priv->rx_ring_num, 0,
2591             "Number of receive rings");
2592         SYSCTL_ADD_UINT(ctx, node_list, OID_AUTO, "tx_rings",
2593             CTLFLAG_RD, &priv->tx_ring_num, 0,
2594             "Number of transmit rings");
2595         SYSCTL_ADD_PROC(ctx, node_list, OID_AUTO, "rx_size",
2596             CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, priv, 0,
2597             mlx4_en_set_rx_ring_size, "I", "Receive ring size");
2598         SYSCTL_ADD_PROC(ctx, node_list, OID_AUTO, "tx_size",
2599             CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, priv, 0,
2600             mlx4_en_set_tx_ring_size, "I", "Transmit ring size");
2601         SYSCTL_ADD_PROC(ctx, node_list, OID_AUTO, "tx_ppp",
2602             CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, priv, 0,
2603             mlx4_en_set_tx_ppp, "I", "TX Per-priority pause");
2604         SYSCTL_ADD_PROC(ctx, node_list, OID_AUTO, "rx_ppp",
2605             CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, priv, 0,
2606             mlx4_en_set_rx_ppp, "I", "RX Per-priority pause");
2607         SYSCTL_ADD_UINT(ctx, node_list, OID_AUTO, "port_num",
2608             CTLFLAG_RD, &priv->port, 0,
2609             "Port Number");
2610         SYSCTL_ADD_STRING(ctx, node_list, OID_AUTO, "device_name",
2611             CTLFLAG_RD, __DECONST(void *, pnameunit), 0,
2612             "PCI device name");
2613
2614         /* Add coalescer configuration. */
2615         coal = SYSCTL_ADD_NODE(ctx, node_list, OID_AUTO,
2616             "coalesce", CTLFLAG_RD, NULL, "Interrupt coalesce configuration");
2617         coal_list = SYSCTL_CHILDREN(coal);
2618         SYSCTL_ADD_UINT(ctx, coal_list, OID_AUTO, "pkt_rate_low",
2619             CTLFLAG_RW, &priv->pkt_rate_low, 0,
2620             "Packets per-second for minimum delay");
2621         SYSCTL_ADD_UINT(ctx, coal_list, OID_AUTO, "rx_usecs_low",
2622             CTLFLAG_RW, &priv->rx_usecs_low, 0,
2623             "Minimum RX delay in micro-seconds");
2624         SYSCTL_ADD_UINT(ctx, coal_list, OID_AUTO, "pkt_rate_high",
2625             CTLFLAG_RW, &priv->pkt_rate_high, 0,
2626             "Packets per-second for maximum delay");
2627         SYSCTL_ADD_UINT(ctx, coal_list, OID_AUTO, "rx_usecs_high",
2628             CTLFLAG_RW, &priv->rx_usecs_high, 0,
2629             "Maximum RX delay in micro-seconds");
2630         SYSCTL_ADD_UINT(ctx, coal_list, OID_AUTO, "sample_interval",
2631             CTLFLAG_RW, &priv->sample_interval, 0,
2632             "adaptive frequency in units of HZ ticks");
2633         SYSCTL_ADD_UINT(ctx, coal_list, OID_AUTO, "adaptive_rx_coal",
2634             CTLFLAG_RW, &priv->adaptive_rx_coal, 0,
2635             "Enable adaptive rx coalescing");
2636         /* EEPROM support */
2637         SYSCTL_ADD_PROC(ctx, node_list, OID_AUTO, "eeprom_info",
2638             CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, priv, 0,
2639             mlx4_en_read_eeprom, "I", "EEPROM information");
2640 }
2641
2642 static void mlx4_en_sysctl_stat(struct mlx4_en_priv *priv)
2643 {
2644         struct sysctl_ctx_list *ctx;
2645         struct sysctl_oid *node;
2646         struct sysctl_oid_list *node_list;
2647         struct sysctl_oid *ring_node;
2648         struct sysctl_oid_list *ring_list;
2649         struct mlx4_en_tx_ring *tx_ring;
2650         struct mlx4_en_rx_ring *rx_ring;
2651         char namebuf[128];
2652         int i;
2653
2654         ctx = &priv->stat_ctx;
2655         sysctl_ctx_init(ctx);
2656         node = SYSCTL_ADD_NODE(ctx, SYSCTL_CHILDREN(priv->sysctl), OID_AUTO,
2657             "stat", CTLFLAG_RD, NULL, "Statistics");
2658         node_list = SYSCTL_CHILDREN(node);
2659
2660 #ifdef MLX4_EN_PERF_STAT
2661         SYSCTL_ADD_UINT(ctx, node_list, OID_AUTO, "tx_poll", CTLFLAG_RD,
2662             &priv->pstats.tx_poll, "TX Poll calls");
2663         SYSCTL_ADD_QUAD(ctx, node_list, OID_AUTO, "tx_pktsz_avg", CTLFLAG_RD,
2664             &priv->pstats.tx_pktsz_avg, "TX average packet size");
2665         SYSCTL_ADD_UINT(ctx, node_list, OID_AUTO, "inflight_avg", CTLFLAG_RD,
2666             &priv->pstats.inflight_avg, "TX average packets in-flight");
2667         SYSCTL_ADD_UINT(ctx, node_list, OID_AUTO, "tx_coal_avg", CTLFLAG_RD,
2668             &priv->pstats.tx_coal_avg, "TX average coalesced completions");
2669         SYSCTL_ADD_UINT(ctx, node_list, OID_AUTO, "rx_coal_avg", CTLFLAG_RD,
2670             &priv->pstats.rx_coal_avg, "RX average coalesced completions");
2671 #endif
2672
2673         SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "tso_packets", CTLFLAG_RD,
2674             &priv->port_stats.tso_packets, "TSO packets sent");
2675         SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "queue_stopped", CTLFLAG_RD,
2676             &priv->port_stats.queue_stopped, "Queue full");
2677         SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "wake_queue", CTLFLAG_RD,
2678             &priv->port_stats.wake_queue, "Queue resumed after full");
2679         SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "tx_timeout", CTLFLAG_RD,
2680             &priv->port_stats.tx_timeout, "Transmit timeouts");
2681         SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "tx_oversized_packets", CTLFLAG_RD,
2682             &priv->port_stats.oversized_packets, "TX oversized packets, m_defrag failed");
2683         SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "rx_alloc_failed", CTLFLAG_RD,
2684             &priv->port_stats.rx_alloc_failed, "RX failed to allocate mbuf");
2685         SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "rx_chksum_good", CTLFLAG_RD,
2686             &priv->port_stats.rx_chksum_good, "RX checksum offload success");
2687         SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "rx_chksum_none", CTLFLAG_RD,
2688             &priv->port_stats.rx_chksum_none, "RX without checksum offload");
2689         SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "tx_chksum_offload",
2690             CTLFLAG_RD, &priv->port_stats.tx_chksum_offload,
2691             "TX checksum offloads");
2692
2693         /* Could strdup the names and add in a loop.  This is simpler. */
2694         SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "rx_bytes", CTLFLAG_RD,
2695             &priv->pkstats.rx_bytes, "RX Bytes");
2696         SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "rx_packets", CTLFLAG_RD,
2697             &priv->pkstats.rx_packets, "RX packets");
2698         SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "rx_multicast_packets", CTLFLAG_RD,
2699             &priv->pkstats.rx_multicast_packets, "RX Multicast Packets");
2700         SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "rx_broadcast_packets", CTLFLAG_RD,
2701             &priv->pkstats.rx_broadcast_packets, "RX Broadcast Packets");
2702         SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "rx_errors", CTLFLAG_RD,
2703             &priv->pkstats.rx_errors, "RX Errors");
2704         SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "rx_dropped", CTLFLAG_RD,
2705             &priv->pkstats.rx_dropped, "RX Dropped");
2706         SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "rx_length_errors", CTLFLAG_RD,
2707             &priv->pkstats.rx_length_errors, "RX Length Errors");
2708         SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "rx_over_errors", CTLFLAG_RD,
2709             &priv->pkstats.rx_over_errors, "RX Over Errors");
2710         SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "rx_crc_errors", CTLFLAG_RD,
2711             &priv->pkstats.rx_crc_errors, "RX CRC Errors");
2712         SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "rx_jabbers", CTLFLAG_RD,
2713             &priv->pkstats.rx_jabbers, "RX Jabbers");
2714
2715
2716         SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "rx_in_range_length_error", CTLFLAG_RD,
2717             &priv->pkstats.rx_in_range_length_error, "RX IN_Range Length Error");
2718         SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "rx_out_range_length_error",
2719                 CTLFLAG_RD, &priv->pkstats.rx_out_range_length_error,
2720                 "RX Out Range Length Error");
2721         SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "rx_lt_64_bytes_packets", CTLFLAG_RD,
2722             &priv->pkstats.rx_lt_64_bytes_packets, "RX Lt 64 Bytes Packets");
2723         SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "rx_127_bytes_packets", CTLFLAG_RD,
2724             &priv->pkstats.rx_127_bytes_packets, "RX 127 bytes Packets");
2725         SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "rx_255_bytes_packets", CTLFLAG_RD,
2726             &priv->pkstats.rx_255_bytes_packets, "RX 255 bytes Packets");
2727         SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "rx_511_bytes_packets", CTLFLAG_RD,
2728             &priv->pkstats.rx_511_bytes_packets, "RX 511 bytes Packets");
2729         SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "rx_1023_bytes_packets", CTLFLAG_RD,
2730             &priv->pkstats.rx_1023_bytes_packets, "RX 1023 bytes Packets");
2731         SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "rx_1518_bytes_packets", CTLFLAG_RD,
2732             &priv->pkstats.rx_1518_bytes_packets, "RX 1518 bytes Packets");
2733         SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "rx_1522_bytes_packets", CTLFLAG_RD,
2734             &priv->pkstats.rx_1522_bytes_packets, "RX 1522 bytes Packets");
2735         SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "rx_1548_bytes_packets", CTLFLAG_RD,
2736             &priv->pkstats.rx_1548_bytes_packets, "RX 1548 bytes Packets");
2737         SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "rx_gt_1548_bytes_packets", CTLFLAG_RD,
2738             &priv->pkstats.rx_gt_1548_bytes_packets,
2739             "RX Greater Then 1548 bytes Packets");
2740
2741 struct mlx4_en_pkt_stats {
2742         unsigned long tx_packets;
2743         unsigned long tx_bytes;
2744         unsigned long tx_multicast_packets;
2745         unsigned long tx_broadcast_packets;
2746         unsigned long tx_errors;
2747         unsigned long tx_dropped;
2748         unsigned long tx_lt_64_bytes_packets;
2749         unsigned long tx_127_bytes_packets;
2750         unsigned long tx_255_bytes_packets;
2751         unsigned long tx_511_bytes_packets;
2752         unsigned long tx_1023_bytes_packets;
2753         unsigned long tx_1518_bytes_packets;
2754         unsigned long tx_1522_bytes_packets;
2755         unsigned long tx_1548_bytes_packets;
2756         unsigned long tx_gt_1548_bytes_packets;
2757         unsigned long rx_prio[NUM_PRIORITIES][NUM_PRIORITY_STATS];
2758         unsigned long tx_prio[NUM_PRIORITIES][NUM_PRIORITY_STATS];
2759 #define NUM_PKT_STATS           72
2760 };
2761
2762
2763         SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "tx_packets", CTLFLAG_RD,
2764             &priv->pkstats.tx_packets, "TX packets");
2765         SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "tx_bytes", CTLFLAG_RD,
2766             &priv->pkstats.tx_bytes, "TX Bytes");
2767         SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "tx_multicast_packets", CTLFLAG_RD,
2768             &priv->pkstats.tx_multicast_packets, "TX Multicast Packets");
2769         SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "tx_broadcast_packets", CTLFLAG_RD,
2770             &priv->pkstats.tx_broadcast_packets, "TX Broadcast Packets");
2771         SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "tx_errors", CTLFLAG_RD,
2772             &priv->pkstats.tx_errors, "TX Errors");
2773         SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "tx_dropped", CTLFLAG_RD,
2774             &priv->pkstats.tx_dropped, "TX Dropped");
2775         SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "tx_lt_64_bytes_packets", CTLFLAG_RD,
2776             &priv->pkstats.tx_lt_64_bytes_packets, "TX Less Then 64 Bytes Packets");
2777         SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "tx_127_bytes_packets", CTLFLAG_RD,
2778             &priv->pkstats.tx_127_bytes_packets, "TX 127 Bytes Packets");
2779         SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "tx_255_bytes_packets", CTLFLAG_RD,
2780             &priv->pkstats.tx_255_bytes_packets, "TX 255 Bytes Packets");
2781         SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "tx_511_bytes_packets", CTLFLAG_RD,
2782             &priv->pkstats.tx_511_bytes_packets, "TX 511 Bytes Packets");
2783         SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "tx_1023_bytes_packets", CTLFLAG_RD,
2784             &priv->pkstats.tx_1023_bytes_packets, "TX 1023 Bytes Packets");
2785         SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "tx_1518_bytes_packets", CTLFLAG_RD,
2786             &priv->pkstats.tx_1518_bytes_packets, "TX 1518 Bytes Packets");
2787         SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "tx_1522_bytes_packets", CTLFLAG_RD,
2788             &priv->pkstats.tx_1522_bytes_packets, "TX 1522 Bytes Packets");
2789         SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "tx_1548_bytes_packets", CTLFLAG_RD,
2790             &priv->pkstats.tx_1548_bytes_packets, "TX 1548 Bytes Packets");
2791         SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "tx_gt_1548_bytes_packets", CTLFLAG_RD,
2792             &priv->pkstats.tx_gt_1548_bytes_packets,
2793             "TX Greater Then 1548 Bytes Packets");
2794
2795
2796
2797         for (i = 0; i < priv->tx_ring_num; i++) {
2798                 tx_ring = priv->tx_ring[i];
2799                 snprintf(namebuf, sizeof(namebuf), "tx_ring%d", i);
2800                 ring_node = SYSCTL_ADD_NODE(ctx, node_list, OID_AUTO, namebuf,
2801                     CTLFLAG_RD, NULL, "TX Ring");
2802                 ring_list = SYSCTL_CHILDREN(ring_node);
2803                 SYSCTL_ADD_ULONG(ctx, ring_list, OID_AUTO, "packets",
2804                     CTLFLAG_RD, &tx_ring->packets, "TX packets");
2805                 SYSCTL_ADD_ULONG(ctx, ring_list, OID_AUTO, "bytes",
2806                     CTLFLAG_RD, &tx_ring->bytes, "TX bytes");
2807         }
2808
2809         for (i = 0; i < priv->rx_ring_num; i++) {
2810                 rx_ring = priv->rx_ring[i];
2811                 snprintf(namebuf, sizeof(namebuf), "rx_ring%d", i);
2812                 ring_node = SYSCTL_ADD_NODE(ctx, node_list, OID_AUTO, namebuf,
2813                     CTLFLAG_RD, NULL, "RX Ring");
2814                 ring_list = SYSCTL_CHILDREN(ring_node);
2815                 SYSCTL_ADD_ULONG(ctx, ring_list, OID_AUTO, "packets",
2816                     CTLFLAG_RD, &rx_ring->packets, "RX packets");
2817                 SYSCTL_ADD_ULONG(ctx, ring_list, OID_AUTO, "bytes",
2818                     CTLFLAG_RD, &rx_ring->bytes, "RX bytes");
2819                 SYSCTL_ADD_ULONG(ctx, ring_list, OID_AUTO, "error",
2820                     CTLFLAG_RD, &rx_ring->errors, "RX soft errors");
2821         }
2822 }