en_netdev.c revision 280018
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 55static void mlx4_en_sysctl_stat(struct mlx4_en_priv *priv); 56static void mlx4_en_sysctl_conf(struct mlx4_en_priv *priv); 57static int mlx4_en_unit; 58 59#ifdef CONFIG_NET_RX_BUSY_POLL 60/* must be called with local_bh_disable()d */ 61static 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 91struct 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 112static void mlx4_en_filter_rfs_expire(struct mlx4_en_priv *priv); 113 114static 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 126static 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 194ignore: 195 mlx4_en_filter_rfs_expire(priv); 196 197 filter->activated = 1; 198} 199 200static inline struct hlist_head * 201filter_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 216static struct mlx4_en_filter * 217mlx4_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 249static 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 263static inline struct mlx4_en_filter * 264mlx4_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 288static int 289mlx4_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 341out: 342 ret = filter->id; 343err: 344 spin_unlock_bh(&priv->filters_lock); 345 346 return ret; 347} 348 349void 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 368static 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 402static 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 429static 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 455static 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 507static 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 534static 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, ®_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 585alloc_err: 586 mlx4_en_uc_steer_release(priv, IF_LLADDR(priv->dev), *qpn, reg_id); 587 588steer_err: 589 mlx4_qp_release_range(dev, *qpn, 1); 590 591qp_err: 592 mlx4_unregister_mac(dev, priv->port, mac); 593 return err; 594} 595 596static 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 636static 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 647static 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 TAILQ_FOREACH(ifma, &dev->if_multiaddrs, ifma_link) { 654 if (ifma->ifma_addr->sa_family != AF_LINK) 655 continue; 656 if (((struct sockaddr_dl *)ifma->ifma_addr)->sdl_alen != 657 ETHER_ADDR_LEN) 658 continue; 659 /* Make sure the list didn't grow. */ 660 tmp = kzalloc(sizeof(struct mlx4_en_mc_list), GFP_ATOMIC); 661 memcpy(tmp->addr, 662 LLADDR((struct sockaddr_dl *)ifma->ifma_addr), ETH_ALEN); 663 list_add_tail(&tmp->list, &priv->mc_list); 664 } 665} 666 667static void update_mclist_flags(struct mlx4_en_priv *priv, 668 struct list_head *dst, 669 struct list_head *src) 670{ 671 struct mlx4_en_mc_list *dst_tmp, *src_tmp, *new_mc; 672 bool found; 673 674 /* Find all the entries that should be removed from dst, 675 * These are the entries that are not found in src 676 */ 677 list_for_each_entry(dst_tmp, dst, list) { 678 found = false; 679 list_for_each_entry(src_tmp, src, list) { 680 if (!memcmp(dst_tmp->addr, src_tmp->addr, ETH_ALEN)) { 681 found = true; 682 break; 683 } 684 } 685 if (!found) 686 dst_tmp->action = MCLIST_REM; 687 } 688 689 /* Add entries that exist in src but not in dst 690 * mark them as need to add 691 */ 692 list_for_each_entry(src_tmp, src, list) { 693 found = false; 694 list_for_each_entry(dst_tmp, dst, list) { 695 if (!memcmp(dst_tmp->addr, src_tmp->addr, ETH_ALEN)) { 696 dst_tmp->action = MCLIST_NONE; 697 found = true; 698 break; 699 } 700 } 701 if (!found) { 702 new_mc = kmalloc(sizeof(struct mlx4_en_mc_list), 703 GFP_KERNEL); 704 if (!new_mc) { 705 en_err(priv, "Failed to allocate current multicast list\n"); 706 return; 707 } 708 memcpy(new_mc, src_tmp, 709 sizeof(struct mlx4_en_mc_list)); 710 new_mc->action = MCLIST_ADD; 711 list_add_tail(&new_mc->list, dst); 712 } 713 } 714} 715 716static void mlx4_en_set_rx_mode(struct net_device *dev) 717{ 718 struct mlx4_en_priv *priv = netdev_priv(dev); 719 720 if (!priv->port_up) 721 return; 722 723 queue_work(priv->mdev->workqueue, &priv->rx_mode_task); 724} 725 726static void mlx4_en_set_promisc_mode(struct mlx4_en_priv *priv, 727 struct mlx4_en_dev *mdev) 728{ 729 int err = 0; 730 if (!(priv->flags & MLX4_EN_FLAG_PROMISC)) { 731 priv->flags |= MLX4_EN_FLAG_PROMISC; 732 733 /* Enable promiscouos mode */ 734 switch (mdev->dev->caps.steering_mode) { 735 case MLX4_STEERING_MODE_DEVICE_MANAGED: 736 err = mlx4_flow_steer_promisc_add(mdev->dev, 737 priv->port, 738 priv->base_qpn, 739 MLX4_FS_ALL_DEFAULT); 740 if (err) 741 en_err(priv, "Failed enabling promiscuous mode\n"); 742 priv->flags |= MLX4_EN_FLAG_MC_PROMISC; 743 break; 744 745 case MLX4_STEERING_MODE_B0: 746 err = mlx4_unicast_promisc_add(mdev->dev, 747 priv->base_qpn, 748 priv->port); 749 if (err) 750 en_err(priv, "Failed enabling unicast promiscuous mode\n"); 751 752 /* Add the default qp number as multicast 753 * promisc 754 */ 755 if (!(priv->flags & MLX4_EN_FLAG_MC_PROMISC)) { 756 err = mlx4_multicast_promisc_add(mdev->dev, 757 priv->base_qpn, 758 priv->port); 759 if (err) 760 en_err(priv, "Failed enabling multicast promiscuous mode\n"); 761 priv->flags |= MLX4_EN_FLAG_MC_PROMISC; 762 } 763 break; 764 765 case MLX4_STEERING_MODE_A0: 766 err = mlx4_SET_PORT_qpn_calc(mdev->dev, 767 priv->port, 768 priv->base_qpn, 769 1); 770 if (err) 771 en_err(priv, "Failed enabling promiscuous mode\n"); 772 break; 773 } 774 775 /* Disable port multicast filter (unconditionally) */ 776 err = mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, 0, 777 0, MLX4_MCAST_DISABLE); 778 if (err) 779 en_err(priv, "Failed disabling multicast filter\n"); 780 } 781} 782 783static void mlx4_en_clear_promisc_mode(struct mlx4_en_priv *priv, 784 struct mlx4_en_dev *mdev) 785{ 786 int err = 0; 787 788 priv->flags &= ~MLX4_EN_FLAG_PROMISC; 789 790 /* Disable promiscouos mode */ 791 switch (mdev->dev->caps.steering_mode) { 792 case MLX4_STEERING_MODE_DEVICE_MANAGED: 793 err = mlx4_flow_steer_promisc_remove(mdev->dev, 794 priv->port, 795 MLX4_FS_ALL_DEFAULT); 796 if (err) 797 en_err(priv, "Failed disabling promiscuous mode\n"); 798 priv->flags &= ~MLX4_EN_FLAG_MC_PROMISC; 799 break; 800 801 case MLX4_STEERING_MODE_B0: 802 err = mlx4_unicast_promisc_remove(mdev->dev, 803 priv->base_qpn, 804 priv->port); 805 if (err) 806 en_err(priv, "Failed disabling unicast promiscuous mode\n"); 807 /* Disable Multicast promisc */ 808 if (priv->flags & MLX4_EN_FLAG_MC_PROMISC) { 809 err = mlx4_multicast_promisc_remove(mdev->dev, 810 priv->base_qpn, 811 priv->port); 812 if (err) 813 en_err(priv, "Failed disabling multicast promiscuous mode\n"); 814 priv->flags &= ~MLX4_EN_FLAG_MC_PROMISC; 815 } 816 break; 817 818 case MLX4_STEERING_MODE_A0: 819 err = mlx4_SET_PORT_qpn_calc(mdev->dev, 820 priv->port, 821 priv->base_qpn, 0); 822 if (err) 823 en_err(priv, "Failed disabling promiscuous mode\n"); 824 break; 825 } 826} 827 828static void mlx4_en_do_multicast(struct mlx4_en_priv *priv, 829 struct net_device *dev, 830 struct mlx4_en_dev *mdev) 831{ 832 struct mlx4_en_mc_list *mclist, *tmp; 833 u8 mc_list[16] = {0}; 834 int err = 0; 835 u64 mcast_addr = 0; 836 837 838 /* Enable/disable the multicast filter according to IFF_ALLMULTI */ 839 if (dev->if_flags & IFF_ALLMULTI) { 840 err = mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, 0, 841 0, MLX4_MCAST_DISABLE); 842 if (err) 843 en_err(priv, "Failed disabling multicast filter\n"); 844 845 /* Add the default qp number as multicast promisc */ 846 if (!(priv->flags & MLX4_EN_FLAG_MC_PROMISC)) { 847 switch (mdev->dev->caps.steering_mode) { 848 case MLX4_STEERING_MODE_DEVICE_MANAGED: 849 err = mlx4_flow_steer_promisc_add(mdev->dev, 850 priv->port, 851 priv->base_qpn, 852 MLX4_FS_MC_DEFAULT); 853 break; 854 855 case MLX4_STEERING_MODE_B0: 856 err = mlx4_multicast_promisc_add(mdev->dev, 857 priv->base_qpn, 858 priv->port); 859 break; 860 861 case MLX4_STEERING_MODE_A0: 862 break; 863 } 864 if (err) 865 en_err(priv, "Failed entering multicast promisc mode\n"); 866 priv->flags |= MLX4_EN_FLAG_MC_PROMISC; 867 } 868 } else { 869 /* Disable Multicast promisc */ 870 if (priv->flags & MLX4_EN_FLAG_MC_PROMISC) { 871 switch (mdev->dev->caps.steering_mode) { 872 case MLX4_STEERING_MODE_DEVICE_MANAGED: 873 err = mlx4_flow_steer_promisc_remove(mdev->dev, 874 priv->port, 875 MLX4_FS_MC_DEFAULT); 876 break; 877 878 case MLX4_STEERING_MODE_B0: 879 err = mlx4_multicast_promisc_remove(mdev->dev, 880 priv->base_qpn, 881 priv->port); 882 break; 883 884 case MLX4_STEERING_MODE_A0: 885 break; 886 } 887 if (err) 888 en_err(priv, "Failed disabling multicast promiscuous mode\n"); 889 priv->flags &= ~MLX4_EN_FLAG_MC_PROMISC; 890 } 891 892 err = mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, 0, 893 0, MLX4_MCAST_DISABLE); 894 if (err) 895 en_err(priv, "Failed disabling multicast filter\n"); 896 897 /* Flush mcast filter and init it with broadcast address */ 898 mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, ETH_BCAST, 899 1, MLX4_MCAST_CONFIG); 900 901 /* Update multicast list - we cache all addresses so they won't 902 * change while HW is updated holding the command semaphor */ 903 mlx4_en_cache_mclist(dev); 904 list_for_each_entry(mclist, &priv->mc_list, list) { 905 mcast_addr = mlx4_mac_to_u64(mclist->addr); 906 mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, 907 mcast_addr, 0, MLX4_MCAST_CONFIG); 908 } 909 err = mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, 0, 910 0, MLX4_MCAST_ENABLE); 911 if (err) 912 en_err(priv, "Failed enabling multicast filter\n"); 913 914 update_mclist_flags(priv, &priv->curr_list, &priv->mc_list); 915 list_for_each_entry_safe(mclist, tmp, &priv->curr_list, list) { 916 if (mclist->action == MCLIST_REM) { 917 /* detach this address and delete from list */ 918 memcpy(&mc_list[10], mclist->addr, ETH_ALEN); 919 mc_list[5] = priv->port; 920 err = mlx4_multicast_detach(mdev->dev, 921 &priv->rss_map.indir_qp, 922 mc_list, 923 MLX4_PROT_ETH, 924 mclist->reg_id); 925 if (err) 926 en_err(priv, "Fail to detach multicast address\n"); 927 928 /* remove from list */ 929 list_del(&mclist->list); 930 kfree(mclist); 931 } else if (mclist->action == MCLIST_ADD) { 932 /* attach the address */ 933 memcpy(&mc_list[10], mclist->addr, ETH_ALEN); 934 /* needed for B0 steering support */ 935 mc_list[5] = priv->port; 936 err = mlx4_multicast_attach(mdev->dev, 937 &priv->rss_map.indir_qp, 938 mc_list, 939 priv->port, 0, 940 MLX4_PROT_ETH, 941 &mclist->reg_id); 942 if (err) 943 en_err(priv, "Fail to attach multicast address\n"); 944 945 } 946 } 947 } 948} 949 950static void mlx4_en_do_set_rx_mode(struct work_struct *work) 951{ 952 struct mlx4_en_priv *priv = container_of(work, struct mlx4_en_priv, 953 rx_mode_task); 954 struct mlx4_en_dev *mdev = priv->mdev; 955 struct net_device *dev = priv->dev; 956 957 958 mutex_lock(&mdev->state_lock); 959 if (!mdev->device_up) { 960 en_dbg(HW, priv, "Card is not up, ignoring rx mode change.\n"); 961 goto out; 962 } 963 if (!priv->port_up) { 964 en_dbg(HW, priv, "Port is down, ignoring rx mode change.\n"); 965 goto out; 966 } 967 if (!mlx4_en_QUERY_PORT(mdev, priv->port)) { 968 if (priv->port_state.link_state) { 969 priv->last_link_state = MLX4_DEV_EVENT_PORT_UP; 970 /* Important note: the following call for if_link_state_change 971 * is needed for interface up scenario (start port, link state 972 * change) */ 973 /* update netif baudrate */ 974 priv->dev->if_baudrate = 975 IF_Mbps(priv->port_state.link_speed); 976 if_link_state_change(priv->dev, LINK_STATE_UP); 977 en_dbg(HW, priv, "Link Up\n"); 978 } 979 } 980 981 /* Promsicuous mode: disable all filters */ 982 if ((dev->if_flags & IFF_PROMISC) || 983 (priv->flags & MLX4_EN_FLAG_FORCE_PROMISC)) { 984 mlx4_en_set_promisc_mode(priv, mdev); 985 goto out; 986 } 987 988 /* Not in promiscuous mode */ 989 if (priv->flags & MLX4_EN_FLAG_PROMISC) 990 mlx4_en_clear_promisc_mode(priv, mdev); 991 992 mlx4_en_do_multicast(priv, dev, mdev); 993out: 994 mutex_unlock(&mdev->state_lock); 995} 996 997#ifdef CONFIG_NET_POLL_CONTROLLER 998static void mlx4_en_netpoll(struct net_device *dev) 999{ 1000 struct mlx4_en_priv *priv = netdev_priv(dev); 1001 struct mlx4_en_cq *cq; 1002 unsigned long flags; 1003 int i; 1004 1005 for (i = 0; i < priv->rx_ring_num; i++) { 1006 cq = priv->rx_cq[i]; 1007 spin_lock_irqsave(&cq->lock, flags); 1008 napi_synchronize(&cq->napi); 1009 mlx4_en_process_rx_cq(dev, cq, 0); 1010 spin_unlock_irqrestore(&cq->lock, flags); 1011 } 1012} 1013#endif 1014 1015static void mlx4_en_watchdog_timeout(void *arg) 1016{ 1017 struct mlx4_en_priv *priv = arg; 1018 struct mlx4_en_dev *mdev = priv->mdev; 1019 1020 en_dbg(DRV, priv, "Scheduling watchdog\n"); 1021 queue_work(mdev->workqueue, &priv->watchdog_task); 1022 if (priv->port_up) 1023 callout_reset(&priv->watchdog_timer, MLX4_EN_WATCHDOG_TIMEOUT, 1024 mlx4_en_watchdog_timeout, priv); 1025} 1026 1027 1028 1029static void mlx4_en_set_default_moderation(struct mlx4_en_priv *priv) 1030{ 1031 struct mlx4_en_cq *cq; 1032 int i; 1033 1034 /* If we haven't received a specific coalescing setting 1035 * (module param), we set the moderation parameters as follows: 1036 * - moder_cnt is set to the number of mtu sized packets to 1037 * satisfy our coelsing target. 1038 * - moder_time is set to a fixed value. 1039 */ 1040 priv->rx_frames = MLX4_EN_RX_COAL_TARGET / priv->dev->if_mtu + 1; 1041 priv->rx_usecs = MLX4_EN_RX_COAL_TIME; 1042 priv->tx_frames = MLX4_EN_TX_COAL_PKTS; 1043 priv->tx_usecs = MLX4_EN_TX_COAL_TIME; 1044 en_dbg(INTR, priv, "Default coalesing params for mtu: %u - " 1045 "rx_frames:%d rx_usecs:%d\n", 1046 (unsigned)priv->dev->if_mtu, priv->rx_frames, priv->rx_usecs); 1047 1048 /* Setup cq moderation params */ 1049 for (i = 0; i < priv->rx_ring_num; i++) { 1050 cq = priv->rx_cq[i]; 1051 cq->moder_cnt = priv->rx_frames; 1052 cq->moder_time = priv->rx_usecs; 1053 priv->last_moder_time[i] = MLX4_EN_AUTO_CONF; 1054 priv->last_moder_packets[i] = 0; 1055 priv->last_moder_bytes[i] = 0; 1056 } 1057 1058 for (i = 0; i < priv->tx_ring_num; i++) { 1059 cq = priv->tx_cq[i]; 1060 cq->moder_cnt = priv->tx_frames; 1061 cq->moder_time = priv->tx_usecs; 1062 } 1063 1064 /* Reset auto-moderation params */ 1065 priv->pkt_rate_low = MLX4_EN_RX_RATE_LOW; 1066 priv->rx_usecs_low = MLX4_EN_RX_COAL_TIME_LOW; 1067 priv->pkt_rate_high = MLX4_EN_RX_RATE_HIGH; 1068 priv->rx_usecs_high = MLX4_EN_RX_COAL_TIME_HIGH; 1069 priv->sample_interval = MLX4_EN_SAMPLE_INTERVAL; 1070 priv->adaptive_rx_coal = 1; 1071 priv->last_moder_jiffies = 0; 1072 priv->last_moder_tx_packets = 0; 1073} 1074 1075static void mlx4_en_auto_moderation(struct mlx4_en_priv *priv) 1076{ 1077 unsigned long period = (unsigned long) (jiffies - priv->last_moder_jiffies); 1078 struct mlx4_en_cq *cq; 1079 unsigned long packets; 1080 unsigned long rate; 1081 unsigned long avg_pkt_size; 1082 unsigned long rx_packets; 1083 unsigned long rx_bytes; 1084 unsigned long rx_pkt_diff; 1085 int moder_time; 1086 int ring, err; 1087 1088 if (!priv->adaptive_rx_coal || period < priv->sample_interval * HZ) 1089 return; 1090 1091 for (ring = 0; ring < priv->rx_ring_num; ring++) { 1092 spin_lock(&priv->stats_lock); 1093 rx_packets = priv->rx_ring[ring]->packets; 1094 rx_bytes = priv->rx_ring[ring]->bytes; 1095 spin_unlock(&priv->stats_lock); 1096 1097 rx_pkt_diff = ((unsigned long) (rx_packets - 1098 priv->last_moder_packets[ring])); 1099 packets = rx_pkt_diff; 1100 rate = packets * HZ / period; 1101 avg_pkt_size = packets ? ((unsigned long) (rx_bytes - 1102 priv->last_moder_bytes[ring])) / packets : 0; 1103 1104 /* Apply auto-moderation only when packet rate 1105 * exceeds a rate that it matters */ 1106 if (rate > (MLX4_EN_RX_RATE_THRESH / priv->rx_ring_num) && 1107 avg_pkt_size > MLX4_EN_AVG_PKT_SMALL) { 1108 if (rate < priv->pkt_rate_low) 1109 moder_time = priv->rx_usecs_low; 1110 else if (rate > priv->pkt_rate_high) 1111 moder_time = priv->rx_usecs_high; 1112 else 1113 moder_time = (rate - priv->pkt_rate_low) * 1114 (priv->rx_usecs_high - priv->rx_usecs_low) / 1115 (priv->pkt_rate_high - priv->pkt_rate_low) + 1116 priv->rx_usecs_low; 1117 } else { 1118 moder_time = priv->rx_usecs_low; 1119 } 1120 1121 if (moder_time != priv->last_moder_time[ring]) { 1122 priv->last_moder_time[ring] = moder_time; 1123 cq = priv->rx_cq[ring]; 1124 cq->moder_time = moder_time; 1125 err = mlx4_en_set_cq_moder(priv, cq); 1126 if (err) 1127 en_err(priv, "Failed modifying moderation for cq:%d\n", 1128 ring); 1129 } 1130 priv->last_moder_packets[ring] = rx_packets; 1131 priv->last_moder_bytes[ring] = rx_bytes; 1132 } 1133 1134 priv->last_moder_jiffies = jiffies; 1135} 1136 1137static void mlx4_en_do_get_stats(struct work_struct *work) 1138{ 1139 struct delayed_work *delay = to_delayed_work(work); 1140 struct mlx4_en_priv *priv = container_of(delay, struct mlx4_en_priv, 1141 stats_task); 1142 struct mlx4_en_dev *mdev = priv->mdev; 1143 int err; 1144 1145 mutex_lock(&mdev->state_lock); 1146 if (mdev->device_up) { 1147 if (priv->port_up) { 1148 err = mlx4_en_DUMP_ETH_STATS(mdev, priv->port, 0); 1149 if (err) 1150 en_dbg(HW, priv, "Could not update stats\n"); 1151 1152 mlx4_en_auto_moderation(priv); 1153 } 1154 1155 queue_delayed_work(mdev->workqueue, &priv->stats_task, STATS_DELAY); 1156 } 1157 mutex_unlock(&mdev->state_lock); 1158} 1159 1160/* mlx4_en_service_task - Run service task for tasks that needed to be done 1161 * periodically 1162 */ 1163static void mlx4_en_service_task(struct work_struct *work) 1164{ 1165 struct delayed_work *delay = to_delayed_work(work); 1166 struct mlx4_en_priv *priv = container_of(delay, struct mlx4_en_priv, 1167 service_task); 1168 struct mlx4_en_dev *mdev = priv->mdev; 1169 1170 mutex_lock(&mdev->state_lock); 1171 if (mdev->device_up) { 1172 queue_delayed_work(mdev->workqueue, &priv->service_task, 1173 SERVICE_TASK_DELAY); 1174 } 1175 mutex_unlock(&mdev->state_lock); 1176} 1177 1178static void mlx4_en_linkstate(struct work_struct *work) 1179{ 1180 struct mlx4_en_priv *priv = container_of(work, struct mlx4_en_priv, 1181 linkstate_task); 1182 struct mlx4_en_dev *mdev = priv->mdev; 1183 int linkstate = priv->link_state; 1184 1185 mutex_lock(&mdev->state_lock); 1186 /* If observable port state changed set carrier state and 1187 * report to system log */ 1188 if (priv->last_link_state != linkstate) { 1189 if (linkstate == MLX4_DEV_EVENT_PORT_DOWN) { 1190 en_info(priv, "Link Down\n"); 1191 if_link_state_change(priv->dev, LINK_STATE_DOWN); 1192 /* update netif baudrate */ 1193 priv->dev->if_baudrate = 0; 1194 1195 /* make sure the port is up before notifying the OS. 1196 * This is tricky since we get here on INIT_PORT and 1197 * in such case we can't tell the OS the port is up. 1198 * To solve this there is a call to if_link_state_change 1199 * in set_rx_mode. 1200 * */ 1201 } else if (priv->port_up && (linkstate == MLX4_DEV_EVENT_PORT_UP)){ 1202 if (mlx4_en_QUERY_PORT(priv->mdev, priv->port)) 1203 en_info(priv, "Query port failed\n"); 1204 priv->dev->if_baudrate = 1205 IF_Mbps(priv->port_state.link_speed); 1206 en_info(priv, "Link Up\n"); 1207 if_link_state_change(priv->dev, LINK_STATE_UP); 1208 } 1209 } 1210 priv->last_link_state = linkstate; 1211 mutex_unlock(&mdev->state_lock); 1212} 1213 1214 1215int mlx4_en_start_port(struct net_device *dev) 1216{ 1217 struct mlx4_en_priv *priv = netdev_priv(dev); 1218 struct mlx4_en_dev *mdev = priv->mdev; 1219 struct mlx4_en_cq *cq; 1220 struct mlx4_en_tx_ring *tx_ring; 1221 int rx_index = 0; 1222 int tx_index = 0; 1223 int err = 0; 1224 int i; 1225 int j; 1226 u8 mc_list[16] = {0}; 1227 1228 1229 if (priv->port_up) { 1230 en_dbg(DRV, priv, "start port called while port already up\n"); 1231 return 0; 1232 } 1233 1234 INIT_LIST_HEAD(&priv->mc_list); 1235 INIT_LIST_HEAD(&priv->curr_list); 1236 INIT_LIST_HEAD(&priv->ethtool_list); 1237 1238 /* Calculate Rx buf size */ 1239 dev->if_mtu = min(dev->if_mtu, priv->max_mtu); 1240 mlx4_en_calc_rx_buf(dev); 1241 priv->rx_alloc_size = max_t(int, 2 * roundup_pow_of_two(priv->rx_mb_size), 1242 PAGE_SIZE); 1243 priv->rx_alloc_order = get_order(priv->rx_alloc_size); 1244 priv->rx_buf_size = roundup_pow_of_two(priv->rx_mb_size); 1245 priv->log_rx_info = ROUNDUP_LOG2(sizeof(struct mlx4_en_rx_buf)); 1246 en_dbg(DRV, priv, "Rx buf size:%d\n", priv->rx_mb_size); 1247 1248 /* Configure rx cq's and rings */ 1249 err = mlx4_en_activate_rx_rings(priv); 1250 if (err) { 1251 en_err(priv, "Failed to activate RX rings\n"); 1252 return err; 1253 } 1254 for (i = 0; i < priv->rx_ring_num; i++) { 1255 cq = priv->rx_cq[i]; 1256 1257 mlx4_en_cq_init_lock(cq); 1258 err = mlx4_en_activate_cq(priv, cq, i); 1259 if (err) { 1260 en_err(priv, "Failed activating Rx CQ\n"); 1261 goto cq_err; 1262 } 1263 for (j = 0; j < cq->size; j++) 1264 cq->buf[j].owner_sr_opcode = MLX4_CQE_OWNER_MASK; 1265 err = mlx4_en_set_cq_moder(priv, cq); 1266 if (err) { 1267 en_err(priv, "Failed setting cq moderation parameters"); 1268 mlx4_en_deactivate_cq(priv, cq); 1269 goto cq_err; 1270 } 1271 mlx4_en_arm_cq(priv, cq); 1272 priv->rx_ring[i]->cqn = cq->mcq.cqn; 1273 ++rx_index; 1274 } 1275 1276 /* Set qp number */ 1277 en_dbg(DRV, priv, "Getting qp number for port %d\n", priv->port); 1278 err = mlx4_en_get_qp(priv); 1279 if (err) { 1280 en_err(priv, "Failed getting eth qp\n"); 1281 goto cq_err; 1282 } 1283 mdev->mac_removed[priv->port] = 0; 1284 1285 /* gets default allocated counter index from func cap */ 1286 /* or sink counter index if no resources */ 1287 priv->counter_index = mdev->dev->caps.def_counter_index[priv->port - 1]; 1288 1289 en_dbg(DRV, priv, "%s: default counter index %d for port %d\n", 1290 __func__, priv->counter_index, priv->port); 1291 1292 err = mlx4_en_config_rss_steer(priv); 1293 if (err) { 1294 en_err(priv, "Failed configuring rss steering\n"); 1295 goto mac_err; 1296 } 1297 1298 err = mlx4_en_create_drop_qp(priv); 1299 if (err) 1300 goto rss_err; 1301 1302 /* Configure tx cq's and rings */ 1303 for (i = 0; i < priv->tx_ring_num; i++) { 1304 /* Configure cq */ 1305 cq = priv->tx_cq[i]; 1306 err = mlx4_en_activate_cq(priv, cq, i); 1307 if (err) { 1308 en_err(priv, "Failed activating Tx CQ\n"); 1309 goto tx_err; 1310 } 1311 err = mlx4_en_set_cq_moder(priv, cq); 1312 if (err) { 1313 en_err(priv, "Failed setting cq moderation parameters"); 1314 mlx4_en_deactivate_cq(priv, cq); 1315 goto tx_err; 1316 } 1317 en_dbg(DRV, priv, "Resetting index of collapsed CQ:%d to -1\n", i); 1318 cq->buf->wqe_index = cpu_to_be16(0xffff); 1319 1320 /* Configure ring */ 1321 tx_ring = priv->tx_ring[i]; 1322 1323 err = mlx4_en_activate_tx_ring(priv, tx_ring, cq->mcq.cqn, 1324 i / priv->num_tx_rings_p_up); 1325 if (err) { 1326 en_err(priv, "Failed activating Tx ring %d\n", i); 1327 mlx4_en_deactivate_cq(priv, cq); 1328 goto tx_err; 1329 } 1330 1331 /* Arm CQ for TX completions */ 1332 mlx4_en_arm_cq(priv, cq); 1333 1334 /* Set initial ownership of all Tx TXBBs to SW (1) */ 1335 for (j = 0; j < tx_ring->buf_size; j += STAMP_STRIDE) 1336 *((u32 *) (tx_ring->buf + j)) = 0xffffffff; 1337 ++tx_index; 1338 } 1339 1340 /* Configure port */ 1341 err = mlx4_SET_PORT_general(mdev->dev, priv->port, 1342 priv->rx_mb_size, 1343 priv->prof->tx_pause, 1344 priv->prof->tx_ppp, 1345 priv->prof->rx_pause, 1346 priv->prof->rx_ppp); 1347 if (err) { 1348 en_err(priv, "Failed setting port general configurations for port %d, with error %d\n", 1349 priv->port, err); 1350 goto tx_err; 1351 } 1352 /* Set default qp number */ 1353 err = mlx4_SET_PORT_qpn_calc(mdev->dev, priv->port, priv->base_qpn, 0); 1354 if (err) { 1355 en_err(priv, "Failed setting default qp numbers\n"); 1356 goto tx_err; 1357 } 1358 1359 /* Init port */ 1360 en_dbg(HW, priv, "Initializing port\n"); 1361 err = mlx4_INIT_PORT(mdev->dev, priv->port); 1362 if (err) { 1363 en_err(priv, "Failed Initializing port\n"); 1364 goto tx_err; 1365 } 1366 1367 /* Attach rx QP to bradcast address */ 1368 memset(&mc_list[10], 0xff, ETH_ALEN); 1369 mc_list[5] = priv->port; /* needed for B0 steering support */ 1370 if (mlx4_multicast_attach(mdev->dev, &priv->rss_map.indir_qp, mc_list, 1371 priv->port, 0, MLX4_PROT_ETH, 1372 &priv->broadcast_id)) 1373 mlx4_warn(mdev, "Failed Attaching Broadcast\n"); 1374 1375 /* Must redo promiscuous mode setup. */ 1376 priv->flags &= ~(MLX4_EN_FLAG_PROMISC | MLX4_EN_FLAG_MC_PROMISC); 1377 1378 /* Schedule multicast task to populate multicast list */ 1379 queue_work(mdev->workqueue, &priv->rx_mode_task); 1380 1381 mlx4_set_stats_bitmap(mdev->dev, priv->stats_bitmap); 1382 1383 priv->port_up = true; 1384 1385 /* Enable the queues. */ 1386 dev->if_drv_flags &= ~IFF_DRV_OACTIVE; 1387 dev->if_drv_flags |= IFF_DRV_RUNNING; 1388#ifdef CONFIG_DEBUG_FS 1389 mlx4_en_create_debug_files(priv); 1390#endif 1391 callout_reset(&priv->watchdog_timer, MLX4_EN_WATCHDOG_TIMEOUT, 1392 mlx4_en_watchdog_timeout, priv); 1393 1394 1395 return 0; 1396 1397tx_err: 1398 while (tx_index--) { 1399 mlx4_en_deactivate_tx_ring(priv, priv->tx_ring[tx_index]); 1400 mlx4_en_deactivate_cq(priv, priv->tx_cq[tx_index]); 1401 } 1402 mlx4_en_destroy_drop_qp(priv); 1403rss_err: 1404 mlx4_en_release_rss_steer(priv); 1405mac_err: 1406 mlx4_en_put_qp(priv); 1407cq_err: 1408 while (rx_index--) 1409 mlx4_en_deactivate_cq(priv, priv->rx_cq[rx_index]); 1410 for (i = 0; i < priv->rx_ring_num; i++) 1411 mlx4_en_deactivate_rx_ring(priv, priv->rx_ring[i]); 1412 1413 return err; /* need to close devices */ 1414} 1415 1416 1417void mlx4_en_stop_port(struct net_device *dev) 1418{ 1419 struct mlx4_en_priv *priv = netdev_priv(dev); 1420 struct mlx4_en_dev *mdev = priv->mdev; 1421 struct mlx4_en_mc_list *mclist, *tmp; 1422 int i; 1423 u8 mc_list[16] = {0}; 1424 1425 if (!priv->port_up) { 1426 en_dbg(DRV, priv, "stop port called while port already down\n"); 1427 return; 1428 } 1429 1430#ifdef CONFIG_DEBUG_FS 1431 mlx4_en_delete_debug_files(priv); 1432#endif 1433 1434 /* close port*/ 1435 mlx4_CLOSE_PORT(mdev->dev, priv->port); 1436 1437 /* Set port as not active */ 1438 priv->port_up = false; 1439 if (priv->counter_index != 0xff) { 1440 mlx4_counter_free(mdev->dev, priv->port, priv->counter_index); 1441 priv->counter_index = 0xff; 1442 } 1443 1444 /* Promsicuous mode */ 1445 if (mdev->dev->caps.steering_mode == 1446 MLX4_STEERING_MODE_DEVICE_MANAGED) { 1447 priv->flags &= ~(MLX4_EN_FLAG_PROMISC | 1448 MLX4_EN_FLAG_MC_PROMISC); 1449 mlx4_flow_steer_promisc_remove(mdev->dev, 1450 priv->port, 1451 MLX4_FS_ALL_DEFAULT); 1452 mlx4_flow_steer_promisc_remove(mdev->dev, 1453 priv->port, 1454 MLX4_FS_MC_DEFAULT); 1455 } else if (priv->flags & MLX4_EN_FLAG_PROMISC) { 1456 priv->flags &= ~MLX4_EN_FLAG_PROMISC; 1457 1458 /* Disable promiscouos mode */ 1459 mlx4_unicast_promisc_remove(mdev->dev, priv->base_qpn, 1460 priv->port); 1461 1462 /* Disable Multicast promisc */ 1463 if (priv->flags & MLX4_EN_FLAG_MC_PROMISC) { 1464 mlx4_multicast_promisc_remove(mdev->dev, priv->base_qpn, 1465 priv->port); 1466 priv->flags &= ~MLX4_EN_FLAG_MC_PROMISC; 1467 } 1468 } 1469 1470 /* Detach All multicasts */ 1471 memset(&mc_list[10], 0xff, ETH_ALEN); 1472 mc_list[5] = priv->port; /* needed for B0 steering support */ 1473 mlx4_multicast_detach(mdev->dev, &priv->rss_map.indir_qp, mc_list, 1474 MLX4_PROT_ETH, priv->broadcast_id); 1475 list_for_each_entry(mclist, &priv->curr_list, list) { 1476 memcpy(&mc_list[10], mclist->addr, ETH_ALEN); 1477 mc_list[5] = priv->port; 1478 mlx4_multicast_detach(mdev->dev, &priv->rss_map.indir_qp, 1479 mc_list, MLX4_PROT_ETH, mclist->reg_id); 1480 } 1481 mlx4_en_clear_list(dev); 1482 list_for_each_entry_safe(mclist, tmp, &priv->curr_list, list) { 1483 list_del(&mclist->list); 1484 kfree(mclist); 1485 } 1486 1487 /* Flush multicast filter */ 1488 mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, 0, 1, MLX4_MCAST_CONFIG); 1489 mlx4_en_destroy_drop_qp(priv); 1490 1491 /* Free TX Rings */ 1492 for (i = 0; i < priv->tx_ring_num; i++) { 1493 mlx4_en_deactivate_tx_ring(priv, priv->tx_ring[i]); 1494 mlx4_en_deactivate_cq(priv, priv->tx_cq[i]); 1495 } 1496 msleep(10); 1497 1498 for (i = 0; i < priv->tx_ring_num; i++) 1499 mlx4_en_free_tx_buf(dev, priv->tx_ring[i]); 1500 1501 /* Free RSS qps */ 1502 mlx4_en_release_rss_steer(priv); 1503 1504 /* Unregister Mac address for the port */ 1505 mlx4_en_put_qp(priv); 1506 mdev->mac_removed[priv->port] = 1; 1507 1508 /* Free RX Rings */ 1509 for (i = 0; i < priv->rx_ring_num; i++) { 1510 struct mlx4_en_cq *cq = priv->rx_cq[i]; 1511 mlx4_en_deactivate_rx_ring(priv, priv->rx_ring[i]); 1512 mlx4_en_deactivate_cq(priv, cq); 1513 } 1514 1515 callout_stop(&priv->watchdog_timer); 1516 1517 dev->if_drv_flags &= ~(IFF_DRV_RUNNING | IFF_DRV_OACTIVE); 1518} 1519 1520static void mlx4_en_restart(struct work_struct *work) 1521{ 1522 struct mlx4_en_priv *priv = container_of(work, struct mlx4_en_priv, 1523 watchdog_task); 1524 struct mlx4_en_dev *mdev = priv->mdev; 1525 struct net_device *dev = priv->dev; 1526 struct mlx4_en_tx_ring *ring; 1527 int i; 1528 1529 1530 if (priv->blocked == 0 || priv->port_up == 0) 1531 return; 1532 for (i = 0; i < priv->tx_ring_num; i++) { 1533 ring = priv->tx_ring[i]; 1534 if (ring->blocked && 1535 ring->watchdog_time + MLX4_EN_WATCHDOG_TIMEOUT < ticks) 1536 goto reset; 1537 } 1538 return; 1539 1540reset: 1541 priv->port_stats.tx_timeout++; 1542 en_dbg(DRV, priv, "Watchdog task called for port %d\n", priv->port); 1543 1544 mutex_lock(&mdev->state_lock); 1545 if (priv->port_up) { 1546 mlx4_en_stop_port(dev); 1547 //for (i = 0; i < priv->tx_ring_num; i++) 1548 // netdev_tx_reset_queue(priv->tx_ring[i]->tx_queue); 1549 if (mlx4_en_start_port(dev)) 1550 en_err(priv, "Failed restarting port %d\n", priv->port); 1551 } 1552 mutex_unlock(&mdev->state_lock); 1553} 1554 1555static void mlx4_en_clear_stats(struct net_device *dev) 1556{ 1557 struct mlx4_en_priv *priv = netdev_priv(dev); 1558 struct mlx4_en_dev *mdev = priv->mdev; 1559 int i; 1560 1561 if (!mlx4_is_slave(mdev->dev)) 1562 if (mlx4_en_DUMP_ETH_STATS(mdev, priv->port, 1)) 1563 en_dbg(HW, priv, "Failed dumping statistics\n"); 1564 1565 memset(&priv->pstats, 0, sizeof(priv->pstats)); 1566 memset(&priv->pkstats, 0, sizeof(priv->pkstats)); 1567 memset(&priv->port_stats, 0, sizeof(priv->port_stats)); 1568 memset(&priv->vport_stats, 0, sizeof(priv->vport_stats)); 1569 1570 for (i = 0; i < priv->tx_ring_num; i++) { 1571 priv->tx_ring[i]->bytes = 0; 1572 priv->tx_ring[i]->packets = 0; 1573 priv->tx_ring[i]->tx_csum = 0; 1574 } 1575 for (i = 0; i < priv->rx_ring_num; i++) { 1576 priv->rx_ring[i]->bytes = 0; 1577 priv->rx_ring[i]->packets = 0; 1578 priv->rx_ring[i]->csum_ok = 0; 1579 priv->rx_ring[i]->csum_none = 0; 1580 } 1581} 1582 1583static void mlx4_en_open(void* arg) 1584{ 1585 1586 struct mlx4_en_priv *priv; 1587 struct mlx4_en_dev *mdev; 1588 struct net_device *dev; 1589 int err = 0; 1590 1591 priv = arg; 1592 mdev = priv->mdev; 1593 dev = priv->dev; 1594 1595 1596 mutex_lock(&mdev->state_lock); 1597 1598 if (!mdev->device_up) { 1599 en_err(priv, "Cannot open - device down/disabled\n"); 1600 goto out; 1601 } 1602 1603 /* Reset HW statistics and SW counters */ 1604 mlx4_en_clear_stats(dev); 1605 1606 err = mlx4_en_start_port(dev); 1607 if (err) 1608 en_err(priv, "Failed starting port:%d\n", priv->port); 1609 1610out: 1611 mutex_unlock(&mdev->state_lock); 1612 return; 1613} 1614 1615void mlx4_en_free_resources(struct mlx4_en_priv *priv) 1616{ 1617 int i; 1618 1619#ifdef CONFIG_RFS_ACCEL 1620 if (priv->dev->rx_cpu_rmap) { 1621 free_irq_cpu_rmap(priv->dev->rx_cpu_rmap); 1622 priv->dev->rx_cpu_rmap = NULL; 1623 } 1624#endif 1625 1626 for (i = 0; i < priv->tx_ring_num; i++) { 1627 if (priv->tx_ring && priv->tx_ring[i]) 1628 mlx4_en_destroy_tx_ring(priv, &priv->tx_ring[i]); 1629 if (priv->tx_cq && priv->tx_cq[i]) 1630 mlx4_en_destroy_cq(priv, &priv->tx_cq[i]); 1631 } 1632 1633 for (i = 0; i < priv->rx_ring_num; i++) { 1634 if (priv->rx_ring[i]) 1635 mlx4_en_destroy_rx_ring(priv, &priv->rx_ring[i], 1636 priv->prof->rx_ring_size, priv->stride); 1637 if (priv->rx_cq[i]) 1638 mlx4_en_destroy_cq(priv, &priv->rx_cq[i]); 1639 } 1640 1641 if (priv->sysctl) 1642 sysctl_ctx_free(&priv->stat_ctx); 1643 1644 1645} 1646 1647int mlx4_en_alloc_resources(struct mlx4_en_priv *priv) 1648{ 1649 struct mlx4_en_port_profile *prof = priv->prof; 1650 int i; 1651 int node = 0; 1652 1653 /* Create rx Rings */ 1654 for (i = 0; i < priv->rx_ring_num; i++) { 1655 if (mlx4_en_create_cq(priv, &priv->rx_cq[i], 1656 prof->rx_ring_size, i, RX, node)) 1657 goto err; 1658 1659 if (mlx4_en_create_rx_ring(priv, &priv->rx_ring[i], 1660 prof->rx_ring_size, node)) 1661 goto err; 1662 } 1663 1664 /* Create tx Rings */ 1665 for (i = 0; i < priv->tx_ring_num; i++) { 1666 if (mlx4_en_create_cq(priv, &priv->tx_cq[i], 1667 prof->tx_ring_size, i, TX, node)) 1668 goto err; 1669 1670 if (mlx4_en_create_tx_ring(priv, &priv->tx_ring[i], 1671 prof->tx_ring_size, TXBB_SIZE, node, i)) 1672 goto err; 1673 } 1674 1675#ifdef CONFIG_RFS_ACCEL 1676 priv->dev->rx_cpu_rmap = alloc_irq_cpu_rmap(priv->rx_ring_num); 1677 if (!priv->dev->rx_cpu_rmap) 1678 goto err; 1679#endif 1680 /* Re-create stat sysctls in case the number of rings changed. */ 1681 mlx4_en_sysctl_stat(priv); 1682 return 0; 1683 1684err: 1685 en_err(priv, "Failed to allocate NIC resources\n"); 1686 for (i = 0; i < priv->rx_ring_num; i++) { 1687 if (priv->rx_ring[i]) 1688 mlx4_en_destroy_rx_ring(priv, &priv->rx_ring[i], 1689 prof->rx_ring_size, 1690 priv->stride); 1691 if (priv->rx_cq[i]) 1692 mlx4_en_destroy_cq(priv, &priv->rx_cq[i]); 1693 } 1694 for (i = 0; i < priv->tx_ring_num; i++) { 1695 if (priv->tx_ring[i]) 1696 mlx4_en_destroy_tx_ring(priv, &priv->tx_ring[i]); 1697 if (priv->tx_cq[i]) 1698 mlx4_en_destroy_cq(priv, &priv->tx_cq[i]); 1699 } 1700 priv->port_up = false; 1701 return -ENOMEM; 1702} 1703 1704struct en_port_attribute { 1705 struct attribute attr; 1706 ssize_t (*show)(struct en_port *, struct en_port_attribute *, char *buf); 1707 ssize_t (*store)(struct en_port *, struct en_port_attribute *, char *buf, size_t count); 1708}; 1709 1710#define PORT_ATTR_RO(_name) \ 1711struct en_port_attribute en_port_attr_##_name = __ATTR_RO(_name) 1712 1713#define EN_PORT_ATTR(_name, _mode, _show, _store) \ 1714struct en_port_attribute en_port_attr_##_name = __ATTR(_name, _mode, _show, _store) 1715 1716void mlx4_en_destroy_netdev(struct net_device *dev) 1717{ 1718 struct mlx4_en_priv *priv = netdev_priv(dev); 1719 struct mlx4_en_dev *mdev = priv->mdev; 1720 1721 en_dbg(DRV, priv, "Destroying netdev on port:%d\n", priv->port); 1722 1723 if (priv->vlan_attach != NULL) 1724 EVENTHANDLER_DEREGISTER(vlan_config, priv->vlan_attach); 1725 if (priv->vlan_detach != NULL) 1726 EVENTHANDLER_DEREGISTER(vlan_unconfig, priv->vlan_detach); 1727 1728 /* Unregister device - this will close the port if it was up */ 1729 if (priv->registered) 1730 ether_ifdetach(dev); 1731 1732 if (priv->allocated) 1733 mlx4_free_hwq_res(mdev->dev, &priv->res, MLX4_EN_PAGE_SIZE); 1734 1735 mutex_lock(&mdev->state_lock); 1736 mlx4_en_stop_port(dev); 1737 mutex_unlock(&mdev->state_lock); 1738 1739 1740 cancel_delayed_work(&priv->stats_task); 1741 cancel_delayed_work(&priv->service_task); 1742 /* flush any pending task for this netdev */ 1743 flush_workqueue(mdev->workqueue); 1744 callout_drain(&priv->watchdog_timer); 1745 1746 /* Detach the netdev so tasks would not attempt to access it */ 1747 mutex_lock(&mdev->state_lock); 1748 mdev->pndev[priv->port] = NULL; 1749 mutex_unlock(&mdev->state_lock); 1750 1751 1752 mlx4_en_free_resources(priv); 1753 1754 /* freeing the sysctl conf cannot be called from within mlx4_en_free_resources */ 1755 if (priv->sysctl) 1756 sysctl_ctx_free(&priv->conf_ctx); 1757 1758 kfree(priv->tx_ring); 1759 kfree(priv->tx_cq); 1760 1761 kfree(priv); 1762 if_free(dev); 1763 1764} 1765 1766static int mlx4_en_change_mtu(struct net_device *dev, int new_mtu) 1767{ 1768 struct mlx4_en_priv *priv = netdev_priv(dev); 1769 struct mlx4_en_dev *mdev = priv->mdev; 1770 int err = 0; 1771 1772 en_dbg(DRV, priv, "Change MTU called - current:%u new:%u\n", 1773 (unsigned)dev->if_mtu, (unsigned)new_mtu); 1774 1775 if ((new_mtu < MLX4_EN_MIN_MTU) || (new_mtu > priv->max_mtu)) { 1776 en_err(priv, "Bad MTU size:%d.\n", new_mtu); 1777 return -EPERM; 1778 } 1779 mutex_lock(&mdev->state_lock); 1780 dev->if_mtu = new_mtu; 1781 if (dev->if_drv_flags & IFF_DRV_RUNNING) { 1782 if (!mdev->device_up) { 1783 /* NIC is probably restarting - let watchdog task reset 1784 * * the port */ 1785 en_dbg(DRV, priv, "Change MTU called with card down!?\n"); 1786 } else { 1787 mlx4_en_stop_port(dev); 1788 err = mlx4_en_start_port(dev); 1789 if (err) { 1790 en_err(priv, "Failed restarting port:%d\n", 1791 priv->port); 1792 queue_work(mdev->workqueue, &priv->watchdog_task); 1793 } 1794 } 1795 } 1796 mutex_unlock(&mdev->state_lock); 1797 return 0; 1798} 1799 1800static int mlx4_en_calc_media(struct mlx4_en_priv *priv) 1801{ 1802 int trans_type; 1803 int active; 1804 1805 active = IFM_ETHER; 1806 if (priv->last_link_state == MLX4_DEV_EVENT_PORT_DOWN) 1807 return (active); 1808 /* 1809 * [ShaharK] mlx4_en_QUERY_PORT sleeps and cannot be called under a 1810 * non-sleepable lock. 1811 * I moved it to the periodic mlx4_en_do_get_stats. 1812 if (mlx4_en_QUERY_PORT(priv->mdev, priv->port)) 1813 return (active); 1814 */ 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 1840static 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 1853static 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 1900static 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_HWCSUM) 1948 dev->if_capenable ^= IFCAP_HWCSUM; 1949 if (mask & IFCAP_TSO4) 1950 dev->if_capenable ^= IFCAP_TSO4; 1951 if (mask & IFCAP_TSO6) 1952 dev->if_capenable ^= IFCAP_TSO6; 1953 if (mask & IFCAP_LRO) 1954 dev->if_capenable ^= IFCAP_LRO; 1955 if (mask & IFCAP_VLAN_HWTAGGING) 1956 dev->if_capenable ^= IFCAP_VLAN_HWTAGGING; 1957 if (mask & IFCAP_VLAN_HWFILTER) 1958 dev->if_capenable ^= IFCAP_VLAN_HWFILTER; 1959 if (mask & IFCAP_WOL_MAGIC) 1960 dev->if_capenable ^= IFCAP_WOL_MAGIC; 1961 if (dev->if_drv_flags & IFF_DRV_RUNNING) 1962 mlx4_en_start_port(dev); 1963 mutex_unlock(&mdev->state_lock); 1964 VLAN_CAPABILITIES(dev); 1965 break; 1966 default: 1967 error = ether_ioctl(dev, command, data); 1968 break; 1969 } 1970 1971 return (error); 1972} 1973 1974 1975int mlx4_en_init_netdev(struct mlx4_en_dev *mdev, int port, 1976 struct mlx4_en_port_profile *prof) 1977{ 1978 struct net_device *dev; 1979 struct mlx4_en_priv *priv; 1980 uint8_t dev_addr[ETHER_ADDR_LEN]; 1981 int err; 1982 int i; 1983 1984 priv = kzalloc(sizeof(*priv), GFP_KERNEL); 1985 dev = priv->dev = if_alloc(IFT_ETHER); 1986 if (dev == NULL) { 1987 en_err(priv, "Net device allocation failed\n"); 1988 kfree(priv); 1989 return -ENOMEM; 1990 } 1991 dev->if_softc = priv; 1992 if_initname(dev, "mlxen", atomic_fetchadd_int(&mlx4_en_unit, 1)); 1993 dev->if_mtu = ETHERMTU; 1994 dev->if_init = mlx4_en_open; 1995 dev->if_flags = IFF_BROADCAST | IFF_SIMPLEX | IFF_MULTICAST; 1996 dev->if_ioctl = mlx4_en_ioctl; 1997 dev->if_transmit = mlx4_en_transmit; 1998 dev->if_qflush = mlx4_en_qflush; 1999 dev->if_snd.ifq_maxlen = prof->tx_ring_size; 2000 2001 /* 2002 * Initialize driver private data 2003 */ 2004 priv->counter_index = 0xff; 2005 spin_lock_init(&priv->stats_lock); 2006 INIT_WORK(&priv->rx_mode_task, mlx4_en_do_set_rx_mode); 2007 INIT_WORK(&priv->watchdog_task, mlx4_en_restart); 2008 INIT_WORK(&priv->linkstate_task, mlx4_en_linkstate); 2009 INIT_DELAYED_WORK(&priv->stats_task, mlx4_en_do_get_stats); 2010 INIT_DELAYED_WORK(&priv->service_task, mlx4_en_service_task); 2011 callout_init(&priv->watchdog_timer, 1); 2012#ifdef CONFIG_RFS_ACCEL 2013 INIT_LIST_HEAD(&priv->filters); 2014 spin_lock_init(&priv->filters_lock); 2015#endif 2016 2017 priv->msg_enable = MLX4_EN_MSG_LEVEL; 2018 priv->dev = dev; 2019 priv->mdev = mdev; 2020 priv->ddev = &mdev->pdev->dev; 2021 priv->prof = prof; 2022 priv->port = port; 2023 priv->port_up = false; 2024 priv->flags = prof->flags; 2025 priv->ctrl_flags = cpu_to_be32(MLX4_WQE_CTRL_CQ_UPDATE | 2026 MLX4_WQE_CTRL_SOLICITED); 2027 2028 priv->num_tx_rings_p_up = mdev->profile.num_tx_rings_p_up; 2029 priv->tx_ring_num = prof->tx_ring_num; 2030 priv->tx_ring = kcalloc(MAX_TX_RINGS, 2031 sizeof(struct mlx4_en_tx_ring *), GFP_KERNEL); 2032 if (!priv->tx_ring) { 2033 err = -ENOMEM; 2034 goto out; 2035 } 2036 priv->tx_cq = kcalloc(sizeof(struct mlx4_en_cq *), MAX_TX_RINGS, 2037 GFP_KERNEL); 2038 if (!priv->tx_cq) { 2039 err = -ENOMEM; 2040 goto out; 2041 } 2042 2043 priv->rx_ring_num = prof->rx_ring_num; 2044 priv->cqe_factor = (mdev->dev->caps.cqe_size == 64) ? 1 : 0; 2045 priv->mac_index = -1; 2046 priv->last_ifq_jiffies = 0; 2047 priv->if_counters_rx_errors = 0; 2048 priv->if_counters_rx_no_buffer = 0; 2049#ifdef CONFIG_MLX4_EN_DCB 2050 if (!mlx4_is_slave(priv->mdev->dev)) { 2051 priv->dcbx_cap = DCB_CAP_DCBX_HOST; 2052 priv->flags |= MLX4_EN_FLAG_DCB_ENABLED; 2053 if (mdev->dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_ETS_CFG) { 2054 dev->dcbnl_ops = &mlx4_en_dcbnl_ops; 2055 } else { 2056 en_info(priv, "QoS disabled - no HW support\n"); 2057 dev->dcbnl_ops = &mlx4_en_dcbnl_pfc_ops; 2058 } 2059 } 2060#endif 2061 2062 for (i = 0; i < MLX4_EN_MAC_HASH_SIZE; ++i) 2063 INIT_HLIST_HEAD(&priv->mac_hash[i]); 2064 2065 2066 /* Query for default mac and max mtu */ 2067 priv->max_mtu = mdev->dev->caps.eth_mtu_cap[priv->port]; 2068 priv->mac = mdev->dev->caps.def_mac[priv->port]; 2069 if (ILLEGAL_MAC(priv->mac)) { 2070#if BITS_PER_LONG == 64 2071 en_err(priv, "Port: %d, invalid mac burned: 0x%lx, quiting\n", 2072 priv->port, priv->mac); 2073#elif BITS_PER_LONG == 32 2074 en_err(priv, "Port: %d, invalid mac burned: 0x%llx, quiting\n", 2075 priv->port, priv->mac); 2076#endif 2077 err = -EINVAL; 2078 goto out; 2079 } 2080 2081 2082 2083 priv->stride = roundup_pow_of_two(sizeof(struct mlx4_en_rx_desc) + 2084 DS_SIZE); 2085 2086 mlx4_en_sysctl_conf(priv); 2087 2088 err = mlx4_en_alloc_resources(priv); 2089 if (err) 2090 goto out; 2091 2092 /* Allocate page for receive rings */ 2093 err = mlx4_alloc_hwq_res(mdev->dev, &priv->res, 2094 MLX4_EN_PAGE_SIZE, MLX4_EN_PAGE_SIZE); 2095 if (err) { 2096 en_err(priv, "Failed to allocate page for rx qps\n"); 2097 goto out; 2098 } 2099 priv->allocated = 1; 2100 2101 /* 2102 * Set driver features 2103 */ 2104 dev->if_capabilities |= IFCAP_RXCSUM | IFCAP_TXCSUM; 2105 dev->if_capabilities |= IFCAP_VLAN_MTU | IFCAP_VLAN_HWTAGGING; 2106 dev->if_capabilities |= IFCAP_VLAN_HWCSUM | IFCAP_VLAN_HWFILTER; 2107 dev->if_capabilities |= IFCAP_LINKSTATE | IFCAP_JUMBO_MTU; 2108 dev->if_capabilities |= IFCAP_LRO; 2109 2110 if (mdev->LSO_support) 2111 dev->if_capabilities |= IFCAP_TSO4 | IFCAP_TSO6 | IFCAP_VLAN_HWTSO; 2112 2113 /* set TSO limits so that we don't have to drop TX packets */ 2114 dev->if_hw_tsomax = 65536 - (ETHER_HDR_LEN + ETHER_VLAN_ENCAP_LEN); 2115 dev->if_hw_tsomaxsegcount = 16; 2116 dev->if_hw_tsomaxsegsize = 65536; /* XXX can do up to 4GByte */ 2117 2118 dev->if_capenable = dev->if_capabilities; 2119 2120 dev->if_hwassist = 0; 2121 if (dev->if_capenable & (IFCAP_TSO4 | IFCAP_TSO6)) 2122 dev->if_hwassist |= CSUM_TSO; 2123 if (dev->if_capenable & IFCAP_TXCSUM) 2124 dev->if_hwassist |= (CSUM_TCP | CSUM_UDP | CSUM_IP); 2125 2126 2127 /* Register for VLAN events */ 2128 priv->vlan_attach = EVENTHANDLER_REGISTER(vlan_config, 2129 mlx4_en_vlan_rx_add_vid, priv, EVENTHANDLER_PRI_FIRST); 2130 priv->vlan_detach = EVENTHANDLER_REGISTER(vlan_unconfig, 2131 mlx4_en_vlan_rx_kill_vid, priv, EVENTHANDLER_PRI_FIRST); 2132 2133 mdev->pndev[priv->port] = dev; 2134 2135 priv->last_link_state = MLX4_DEV_EVENT_PORT_DOWN; 2136 mlx4_en_set_default_moderation(priv); 2137 2138 /* Set default MAC */ 2139 for (i = 0; i < ETHER_ADDR_LEN; i++) 2140 dev_addr[ETHER_ADDR_LEN - 1 - i] = (u8) (priv->mac >> (8 * i)); 2141 2142 2143 ether_ifattach(dev, dev_addr); 2144 if_link_state_change(dev, LINK_STATE_DOWN); 2145 ifmedia_init(&priv->media, IFM_IMASK | IFM_ETH_FMASK, 2146 mlx4_en_media_change, mlx4_en_media_status); 2147 ifmedia_add(&priv->media, IFM_ETHER | IFM_FDX | IFM_1000_T, 0, NULL); 2148 ifmedia_add(&priv->media, IFM_ETHER | IFM_FDX | IFM_10G_SR, 0, NULL); 2149 ifmedia_add(&priv->media, IFM_ETHER | IFM_FDX | IFM_10G_CX4, 0, NULL); 2150 ifmedia_add(&priv->media, IFM_ETHER | IFM_FDX | IFM_40G_CR4, 0, NULL); 2151 ifmedia_add(&priv->media, IFM_ETHER | IFM_AUTO, 0, NULL); 2152 ifmedia_set(&priv->media, IFM_ETHER | IFM_AUTO); 2153 2154 en_warn(priv, "Using %d TX rings\n", prof->tx_ring_num); 2155 en_warn(priv, "Using %d RX rings\n", prof->rx_ring_num); 2156 2157 priv->registered = 1; 2158 2159 en_warn(priv, "Using %d TX rings\n", prof->tx_ring_num); 2160 en_warn(priv, "Using %d RX rings\n", prof->rx_ring_num); 2161 2162 2163 priv->rx_mb_size = dev->if_mtu + ETH_HLEN + VLAN_HLEN + ETH_FCS_LEN; 2164 err = mlx4_SET_PORT_general(mdev->dev, priv->port, 2165 priv->rx_mb_size, 2166 prof->tx_pause, prof->tx_ppp, 2167 prof->rx_pause, prof->rx_ppp); 2168 if (err) { 2169 en_err(priv, "Failed setting port general configurations " 2170 "for port %d, with error %d\n", priv->port, err); 2171 goto out; 2172 } 2173 2174 /* Init port */ 2175 en_warn(priv, "Initializing port\n"); 2176 err = mlx4_INIT_PORT(mdev->dev, priv->port); 2177 if (err) { 2178 en_err(priv, "Failed Initializing port\n"); 2179 goto out; 2180 } 2181 2182 queue_delayed_work(mdev->workqueue, &priv->stats_task, STATS_DELAY); 2183 2184 if (mdev->dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_TS) 2185 queue_delayed_work(mdev->workqueue, &priv->service_task, SERVICE_TASK_DELAY); 2186 2187 2188 2189 return 0; 2190 2191out: 2192 mlx4_en_destroy_netdev(dev); 2193 return err; 2194} 2195 2196static int mlx4_en_set_ring_size(struct net_device *dev, 2197 int rx_size, int tx_size) 2198{ 2199 struct mlx4_en_priv *priv = netdev_priv(dev); 2200 struct mlx4_en_dev *mdev = priv->mdev; 2201 int port_up = 0; 2202 int err = 0; 2203 2204 rx_size = roundup_pow_of_two(rx_size); 2205 rx_size = max_t(u32, rx_size, MLX4_EN_MIN_RX_SIZE); 2206 rx_size = min_t(u32, rx_size, MLX4_EN_MAX_RX_SIZE); 2207 tx_size = roundup_pow_of_two(tx_size); 2208 tx_size = max_t(u32, tx_size, MLX4_EN_MIN_TX_SIZE); 2209 tx_size = min_t(u32, tx_size, MLX4_EN_MAX_TX_SIZE); 2210 2211 if (rx_size == (priv->port_up ? 2212 priv->rx_ring[0]->actual_size : priv->rx_ring[0]->size) && 2213 tx_size == priv->tx_ring[0]->size) 2214 return 0; 2215 mutex_lock(&mdev->state_lock); 2216 if (priv->port_up) { 2217 port_up = 1; 2218 mlx4_en_stop_port(dev); 2219 } 2220 mlx4_en_free_resources(priv); 2221 priv->prof->tx_ring_size = tx_size; 2222 priv->prof->rx_ring_size = rx_size; 2223 err = mlx4_en_alloc_resources(priv); 2224 if (err) { 2225 en_err(priv, "Failed reallocating port resources\n"); 2226 goto out; 2227 } 2228 if (port_up) { 2229 err = mlx4_en_start_port(dev); 2230 if (err) 2231 en_err(priv, "Failed starting port\n"); 2232 } 2233out: 2234 mutex_unlock(&mdev->state_lock); 2235 return err; 2236} 2237static int mlx4_en_set_rx_ring_size(SYSCTL_HANDLER_ARGS) 2238{ 2239 struct mlx4_en_priv *priv; 2240 int size; 2241 int error; 2242 2243 priv = arg1; 2244 size = priv->prof->rx_ring_size; 2245 error = sysctl_handle_int(oidp, &size, 0, req); 2246 if (error || !req->newptr) 2247 return (error); 2248 error = -mlx4_en_set_ring_size(priv->dev, size, 2249 priv->prof->tx_ring_size); 2250 return (error); 2251} 2252 2253static int mlx4_en_set_tx_ring_size(SYSCTL_HANDLER_ARGS) 2254{ 2255 struct mlx4_en_priv *priv; 2256 int size; 2257 int error; 2258 2259 priv = arg1; 2260 size = priv->prof->tx_ring_size; 2261 error = sysctl_handle_int(oidp, &size, 0, req); 2262 if (error || !req->newptr) 2263 return (error); 2264 error = -mlx4_en_set_ring_size(priv->dev, priv->prof->rx_ring_size, 2265 size); 2266 2267 return (error); 2268} 2269 2270static int mlx4_en_set_tx_ppp(SYSCTL_HANDLER_ARGS) 2271{ 2272 struct mlx4_en_priv *priv; 2273 int ppp; 2274 int error; 2275 2276 priv = arg1; 2277 ppp = priv->prof->tx_ppp; 2278 error = sysctl_handle_int(oidp, &ppp, 0, req); 2279 if (error || !req->newptr) 2280 return (error); 2281 if (ppp > 0xff || ppp < 0) 2282 return (-EINVAL); 2283 priv->prof->tx_ppp = ppp; 2284 error = -mlx4_SET_PORT_general(priv->mdev->dev, priv->port, 2285 priv->rx_mb_size + ETHER_CRC_LEN, 2286 priv->prof->tx_pause, 2287 priv->prof->tx_ppp, 2288 priv->prof->rx_pause, 2289 priv->prof->rx_ppp); 2290 2291 return (error); 2292} 2293 2294static int mlx4_en_set_rx_ppp(SYSCTL_HANDLER_ARGS) 2295{ 2296 struct mlx4_en_priv *priv; 2297 struct mlx4_en_dev *mdev; 2298 int ppp; 2299 int error; 2300 int port_up; 2301 2302 port_up = 0; 2303 priv = arg1; 2304 mdev = priv->mdev; 2305 ppp = priv->prof->rx_ppp; 2306 error = sysctl_handle_int(oidp, &ppp, 0, req); 2307 if (error || !req->newptr) 2308 return (error); 2309 if (ppp > 0xff || ppp < 0) 2310 return (-EINVAL); 2311 /* See if we have to change the number of tx queues. */ 2312 if (!ppp != !priv->prof->rx_ppp) { 2313 mutex_lock(&mdev->state_lock); 2314 if (priv->port_up) { 2315 port_up = 1; 2316 mlx4_en_stop_port(priv->dev); 2317 } 2318 mlx4_en_free_resources(priv); 2319 priv->prof->rx_ppp = ppp; 2320 error = -mlx4_en_alloc_resources(priv); 2321 if (error) 2322 en_err(priv, "Failed reallocating port resources\n"); 2323 if (error == 0 && port_up) { 2324 error = -mlx4_en_start_port(priv->dev); 2325 if (error) 2326 en_err(priv, "Failed starting port\n"); 2327 } 2328 mutex_unlock(&mdev->state_lock); 2329 return (error); 2330 2331 } 2332 priv->prof->rx_ppp = ppp; 2333 error = -mlx4_SET_PORT_general(priv->mdev->dev, priv->port, 2334 priv->rx_mb_size + ETHER_CRC_LEN, 2335 priv->prof->tx_pause, 2336 priv->prof->tx_ppp, 2337 priv->prof->rx_pause, 2338 priv->prof->rx_ppp); 2339 2340 return (error); 2341} 2342 2343static void mlx4_en_sysctl_conf(struct mlx4_en_priv *priv) 2344{ 2345 struct net_device *dev; 2346 struct sysctl_ctx_list *ctx; 2347 struct sysctl_oid *node; 2348 struct sysctl_oid_list *node_list; 2349 struct sysctl_oid *coal; 2350 struct sysctl_oid_list *coal_list; 2351 const char *pnameunit; 2352 2353 dev = priv->dev; 2354 ctx = &priv->conf_ctx; 2355 pnameunit = device_get_nameunit(priv->mdev->pdev->dev.bsddev); 2356 2357 sysctl_ctx_init(ctx); 2358 priv->sysctl = SYSCTL_ADD_NODE(ctx, SYSCTL_STATIC_CHILDREN(_hw), 2359 OID_AUTO, dev->if_xname, CTLFLAG_RD, 0, "mlx4 10gig ethernet"); 2360 node = SYSCTL_ADD_NODE(ctx, SYSCTL_CHILDREN(priv->sysctl), OID_AUTO, 2361 "conf", CTLFLAG_RD, NULL, "Configuration"); 2362 node_list = SYSCTL_CHILDREN(node); 2363 2364 SYSCTL_ADD_UINT(ctx, node_list, OID_AUTO, "msg_enable", 2365 CTLFLAG_RW, &priv->msg_enable, 0, 2366 "Driver message enable bitfield"); 2367 SYSCTL_ADD_UINT(ctx, node_list, OID_AUTO, "rx_rings", 2368 CTLFLAG_RD, &priv->rx_ring_num, 0, 2369 "Number of receive rings"); 2370 SYSCTL_ADD_UINT(ctx, node_list, OID_AUTO, "tx_rings", 2371 CTLFLAG_RD, &priv->tx_ring_num, 0, 2372 "Number of transmit rings"); 2373 SYSCTL_ADD_PROC(ctx, node_list, OID_AUTO, "rx_size", 2374 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, priv, 0, 2375 mlx4_en_set_rx_ring_size, "I", "Receive ring size"); 2376 SYSCTL_ADD_PROC(ctx, node_list, OID_AUTO, "tx_size", 2377 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, priv, 0, 2378 mlx4_en_set_tx_ring_size, "I", "Transmit ring size"); 2379 SYSCTL_ADD_PROC(ctx, node_list, OID_AUTO, "tx_ppp", 2380 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, priv, 0, 2381 mlx4_en_set_tx_ppp, "I", "TX Per-priority pause"); 2382 SYSCTL_ADD_PROC(ctx, node_list, OID_AUTO, "rx_ppp", 2383 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, priv, 0, 2384 mlx4_en_set_rx_ppp, "I", "RX Per-priority pause"); 2385 SYSCTL_ADD_UINT(ctx, node_list, OID_AUTO, "port_num", 2386 CTLFLAG_RD, &priv->port, 0, 2387 "Port Number"); 2388 SYSCTL_ADD_STRING(ctx, node_list, OID_AUTO, "device_name", 2389 CTLFLAG_RD, __DECONST(void *, pnameunit), 0, 2390 "PCI device name"); 2391 2392 /* Add coalescer configuration. */ 2393 coal = SYSCTL_ADD_NODE(ctx, node_list, OID_AUTO, 2394 "coalesce", CTLFLAG_RD, NULL, "Interrupt coalesce configuration"); 2395 coal_list = SYSCTL_CHILDREN(node); 2396 SYSCTL_ADD_UINT(ctx, coal_list, OID_AUTO, "pkt_rate_low", 2397 CTLFLAG_RW, &priv->pkt_rate_low, 0, 2398 "Packets per-second for minimum delay"); 2399 SYSCTL_ADD_UINT(ctx, coal_list, OID_AUTO, "rx_usecs_low", 2400 CTLFLAG_RW, &priv->rx_usecs_low, 0, 2401 "Minimum RX delay in micro-seconds"); 2402 SYSCTL_ADD_UINT(ctx, coal_list, OID_AUTO, "pkt_rate_high", 2403 CTLFLAG_RW, &priv->pkt_rate_high, 0, 2404 "Packets per-second for maximum delay"); 2405 SYSCTL_ADD_UINT(ctx, coal_list, OID_AUTO, "rx_usecs_high", 2406 CTLFLAG_RW, &priv->rx_usecs_high, 0, 2407 "Maximum RX delay in micro-seconds"); 2408 SYSCTL_ADD_UINT(ctx, coal_list, OID_AUTO, "sample_interval", 2409 CTLFLAG_RW, &priv->sample_interval, 0, 2410 "adaptive frequency in units of HZ ticks"); 2411 SYSCTL_ADD_UINT(ctx, coal_list, OID_AUTO, "adaptive_rx_coal", 2412 CTLFLAG_RW, &priv->adaptive_rx_coal, 0, 2413 "Enable adaptive rx coalescing"); 2414} 2415 2416static void mlx4_en_sysctl_stat(struct mlx4_en_priv *priv) 2417{ 2418 struct net_device *dev; 2419 struct sysctl_ctx_list *ctx; 2420 struct sysctl_oid *node; 2421 struct sysctl_oid_list *node_list; 2422 struct sysctl_oid *ring_node; 2423 struct sysctl_oid_list *ring_list; 2424 struct mlx4_en_tx_ring *tx_ring; 2425 struct mlx4_en_rx_ring *rx_ring; 2426 char namebuf[128]; 2427 int i; 2428 2429 dev = priv->dev; 2430 2431 ctx = &priv->stat_ctx; 2432 sysctl_ctx_init(ctx); 2433 node = SYSCTL_ADD_NODE(ctx, SYSCTL_CHILDREN(priv->sysctl), OID_AUTO, 2434 "stat", CTLFLAG_RD, NULL, "Statistics"); 2435 node_list = SYSCTL_CHILDREN(node); 2436 2437#ifdef MLX4_EN_PERF_STAT 2438 SYSCTL_ADD_UINT(ctx, node_list, OID_AUTO, "tx_poll", CTLFLAG_RD, 2439 &priv->pstats.tx_poll, "TX Poll calls"); 2440 SYSCTL_ADD_QUAD(ctx, node_list, OID_AUTO, "tx_pktsz_avg", CTLFLAG_RD, 2441 &priv->pstats.tx_pktsz_avg, "TX average packet size"); 2442 SYSCTL_ADD_UINT(ctx, node_list, OID_AUTO, "inflight_avg", CTLFLAG_RD, 2443 &priv->pstats.inflight_avg, "TX average packets in-flight"); 2444 SYSCTL_ADD_UINT(ctx, node_list, OID_AUTO, "tx_coal_avg", CTLFLAG_RD, 2445 &priv->pstats.tx_coal_avg, "TX average coalesced completions"); 2446 SYSCTL_ADD_UINT(ctx, node_list, OID_AUTO, "rx_coal_avg", CTLFLAG_RD, 2447 &priv->pstats.rx_coal_avg, "RX average coalesced completions"); 2448#endif 2449 2450 SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "tso_packets", CTLFLAG_RD, 2451 &priv->port_stats.tso_packets, "TSO packets sent"); 2452 SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "queue_stopped", CTLFLAG_RD, 2453 &priv->port_stats.queue_stopped, "Queue full"); 2454 SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "wake_queue", CTLFLAG_RD, 2455 &priv->port_stats.wake_queue, "Queue resumed after full"); 2456 SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "tx_timeout", CTLFLAG_RD, 2457 &priv->port_stats.tx_timeout, "Transmit timeouts"); 2458 SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "rx_alloc_failed", CTLFLAG_RD, 2459 &priv->port_stats.rx_alloc_failed, "RX failed to allocate mbuf"); 2460 SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "rx_chksum_good", CTLFLAG_RD, 2461 &priv->port_stats.rx_chksum_good, "RX checksum offload success"); 2462 SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "rx_chksum_none", CTLFLAG_RD, 2463 &priv->port_stats.rx_chksum_none, "RX without checksum offload"); 2464 SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "tx_chksum_offload", 2465 CTLFLAG_RD, &priv->port_stats.tx_chksum_offload, 2466 "TX checksum offloads"); 2467 2468 /* Could strdup the names and add in a loop. This is simpler. */ 2469 SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "rx_bytes", CTLFLAG_RD, 2470 &priv->pkstats.rx_bytes, "RX Bytes"); 2471 SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "rx_packets", CTLFLAG_RD, 2472 &priv->pkstats.rx_packets, "RX packets"); 2473 SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "rx_multicast_packets", CTLFLAG_RD, 2474 &priv->pkstats.rx_multicast_packets, "RX Multicast Packets"); 2475 SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "rx_broadcast_packets", CTLFLAG_RD, 2476 &priv->pkstats.rx_broadcast_packets, "RX Broadcast Packets"); 2477 SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "rx_errors", CTLFLAG_RD, 2478 &priv->pkstats.rx_errors, "RX Errors"); 2479 SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "rx_dropped", CTLFLAG_RD, 2480 &priv->pkstats.rx_dropped, "RX Dropped"); 2481 SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "rx_length_errors", CTLFLAG_RD, 2482 &priv->pkstats.rx_length_errors, "RX Length Errors"); 2483 SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "rx_over_errors", CTLFLAG_RD, 2484 &priv->pkstats.rx_over_errors, "RX Over Errors"); 2485 SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "rx_crc_errors", CTLFLAG_RD, 2486 &priv->pkstats.rx_crc_errors, "RX CRC Errors"); 2487 SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "rx_jabbers", CTLFLAG_RD, 2488 &priv->pkstats.rx_jabbers, "RX Jabbers"); 2489 2490 2491 SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "rx_in_range_length_error", CTLFLAG_RD, 2492 &priv->pkstats.rx_in_range_length_error, "RX IN_Range Length Error"); 2493 SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "rx_out_range_length_error", 2494 CTLFLAG_RD, &priv->pkstats.rx_out_range_length_error, 2495 "RX Out Range Length Error"); 2496 SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "rx_lt_64_bytes_packets", CTLFLAG_RD, 2497 &priv->pkstats.rx_lt_64_bytes_packets, "RX Lt 64 Bytes Packets"); 2498 SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "rx_127_bytes_packets", CTLFLAG_RD, 2499 &priv->pkstats.rx_127_bytes_packets, "RX 127 bytes Packets"); 2500 SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "rx_255_bytes_packets", CTLFLAG_RD, 2501 &priv->pkstats.rx_255_bytes_packets, "RX 255 bytes Packets"); 2502 SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "rx_511_bytes_packets", CTLFLAG_RD, 2503 &priv->pkstats.rx_511_bytes_packets, "RX 511 bytes Packets"); 2504 SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "rx_1023_bytes_packets", CTLFLAG_RD, 2505 &priv->pkstats.rx_1023_bytes_packets, "RX 1023 bytes Packets"); 2506 SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "rx_1518_bytes_packets", CTLFLAG_RD, 2507 &priv->pkstats.rx_1518_bytes_packets, "RX 1518 bytes Packets"); 2508 SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "rx_1522_bytes_packets", CTLFLAG_RD, 2509 &priv->pkstats.rx_1522_bytes_packets, "RX 1522 bytes Packets"); 2510 SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "rx_1548_bytes_packets", CTLFLAG_RD, 2511 &priv->pkstats.rx_1548_bytes_packets, "RX 1548 bytes Packets"); 2512 SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "rx_gt_1548_bytes_packets", CTLFLAG_RD, 2513 &priv->pkstats.rx_gt_1548_bytes_packets, 2514 "RX Greater Then 1548 bytes Packets"); 2515 2516struct mlx4_en_pkt_stats { 2517 unsigned long tx_packets; 2518 unsigned long tx_bytes; 2519 unsigned long tx_multicast_packets; 2520 unsigned long tx_broadcast_packets; 2521 unsigned long tx_errors; 2522 unsigned long tx_dropped; 2523 unsigned long tx_lt_64_bytes_packets; 2524 unsigned long tx_127_bytes_packets; 2525 unsigned long tx_255_bytes_packets; 2526 unsigned long tx_511_bytes_packets; 2527 unsigned long tx_1023_bytes_packets; 2528 unsigned long tx_1518_bytes_packets; 2529 unsigned long tx_1522_bytes_packets; 2530 unsigned long tx_1548_bytes_packets; 2531 unsigned long tx_gt_1548_bytes_packets; 2532 unsigned long rx_prio[NUM_PRIORITIES][NUM_PRIORITY_STATS]; 2533 unsigned long tx_prio[NUM_PRIORITIES][NUM_PRIORITY_STATS]; 2534#define NUM_PKT_STATS 72 2535}; 2536 2537 2538 SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "tx_packets", CTLFLAG_RD, 2539 &priv->pkstats.tx_packets, "TX packets"); 2540 SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "tx_bytes", CTLFLAG_RD, 2541 &priv->pkstats.tx_packets, "TX Bytes"); 2542 SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "tx_multicast_packets", CTLFLAG_RD, 2543 &priv->pkstats.tx_multicast_packets, "TX Multicast Packets"); 2544 SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "tx_broadcast_packets", CTLFLAG_RD, 2545 &priv->pkstats.tx_broadcast_packets, "TX Broadcast Packets"); 2546 SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "tx_errors", CTLFLAG_RD, 2547 &priv->pkstats.tx_errors, "TX Errors"); 2548 SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "tx_dropped", CTLFLAG_RD, 2549 &priv->pkstats.tx_dropped, "TX Dropped"); 2550 SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "tx_lt_64_bytes_packets", CTLFLAG_RD, 2551 &priv->pkstats.tx_lt_64_bytes_packets, "TX Less Then 64 Bytes Packets"); 2552 SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "tx_127_bytes_packets", CTLFLAG_RD, 2553 &priv->pkstats.tx_127_bytes_packets, "TX 127 Bytes Packets"); 2554 SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "tx_255_bytes_packets", CTLFLAG_RD, 2555 &priv->pkstats.tx_255_bytes_packets, "TX 255 Bytes Packets"); 2556 SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "tx_511_bytes_packets", CTLFLAG_RD, 2557 &priv->pkstats.tx_511_bytes_packets, "TX 511 Bytes Packets"); 2558 SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "tx_1023_bytes_packets", CTLFLAG_RD, 2559 &priv->pkstats.tx_1023_bytes_packets, "TX 1023 Bytes Packets"); 2560 SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "tx_1518_bytes_packets", CTLFLAG_RD, 2561 &priv->pkstats.tx_1518_bytes_packets, "TX 1518 Bytes Packets"); 2562 SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "tx_1522_bytes_packets", CTLFLAG_RD, 2563 &priv->pkstats.tx_1522_bytes_packets, "TX 1522 Bytes Packets"); 2564 SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "tx_1548_bytes_packets", CTLFLAG_RD, 2565 &priv->pkstats.tx_1548_bytes_packets, "TX 1548 Bytes Packets"); 2566 SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "tx_gt_1548_bytes_packets", CTLFLAG_RD, 2567 &priv->pkstats.tx_gt_1548_bytes_packets, 2568 "TX Greater Then 1548 Bytes Packets"); 2569 2570 2571 2572 for (i = 0; i < priv->tx_ring_num; i++) { 2573 tx_ring = priv->tx_ring[i]; 2574 snprintf(namebuf, sizeof(namebuf), "tx_ring%d", i); 2575 ring_node = SYSCTL_ADD_NODE(ctx, node_list, OID_AUTO, namebuf, 2576 CTLFLAG_RD, NULL, "TX Ring"); 2577 ring_list = SYSCTL_CHILDREN(ring_node); 2578 SYSCTL_ADD_ULONG(ctx, ring_list, OID_AUTO, "packets", 2579 CTLFLAG_RD, &tx_ring->packets, "TX packets"); 2580 SYSCTL_ADD_ULONG(ctx, ring_list, OID_AUTO, "bytes", 2581 CTLFLAG_RD, &tx_ring->bytes, "TX bytes"); 2582 2583 } 2584 for (i = 0; i < priv->rx_ring_num; i++) { 2585 rx_ring = priv->rx_ring[i]; 2586 snprintf(namebuf, sizeof(namebuf), "rx_ring%d", i); 2587 ring_node = SYSCTL_ADD_NODE(ctx, node_list, OID_AUTO, namebuf, 2588 CTLFLAG_RD, NULL, "RX Ring"); 2589 ring_list = SYSCTL_CHILDREN(ring_node); 2590 SYSCTL_ADD_ULONG(ctx, ring_list, OID_AUTO, "packets", 2591 CTLFLAG_RD, &rx_ring->packets, "RX packets"); 2592 SYSCTL_ADD_ULONG(ctx, ring_list, OID_AUTO, "bytes", 2593 CTLFLAG_RD, &rx_ring->bytes, "RX bytes"); 2594 SYSCTL_ADD_ULONG(ctx, ring_list, OID_AUTO, "error", 2595 CTLFLAG_RD, &rx_ring->errors, "RX soft errors"); 2596 } 2597} 2598