xref: /freebsd/sys/dev/mlx4/mlx4_en/mlx4_en_netdev.c (revision c66ec88fed842fbaad62c30d510644ceb7bd2d71)
1 /*
2  * Copyright (c) 2007, 2014 Mellanox Technologies. All rights reserved.
3  *
4  * This software is available to you under a choice of one of two
5  * licenses.  You may choose to be licensed under the terms of the GNU
6  * General Public License (GPL) Version 2, available from the file
7  * COPYING in the main directory of this source tree, or the
8  * OpenIB.org BSD license below:
9  *
10  *     Redistribution and use in source and binary forms, with or
11  *     without modification, are permitted provided that the following
12  *     conditions are met:
13  *
14  *      - Redistributions of source code must retain the above
15  *        copyright notice, this list of conditions and the following
16  *        disclaimer.
17  *
18  *      - Redistributions in binary form must reproduce the above
19  *        copyright notice, this list of conditions and the following
20  *        disclaimer in the documentation and/or other materials
21  *        provided with the distribution.
22  *
23  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
24  * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
25  * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
26  * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
27  * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
28  * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
29  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
30  * SOFTWARE.
31  *
32  */
33 
34 #include <linux/etherdevice.h>
35 #include <linux/delay.h>
36 #include <linux/slab.h>
37 #include <linux/compat.h>
38 #ifdef CONFIG_NET_RX_BUSY_POLL
39 #include <net/busy_poll.h>
40 #endif
41 
42 #include <linux/list.h>
43 #include <linux/if_ether.h>
44 
45 #include <dev/mlx4/driver.h>
46 #include <dev/mlx4/device.h>
47 #include <dev/mlx4/cmd.h>
48 #include <dev/mlx4/cq.h>
49 
50 #include <sys/eventhandler.h>
51 #include <sys/sockio.h>
52 #include <sys/sysctl.h>
53 
54 #include "en.h"
55 #include "en_port.h"
56 
57 DEBUGNET_DEFINE(mlx4_en);
58 
59 static void mlx4_en_sysctl_stat(struct mlx4_en_priv *priv);
60 static void mlx4_en_sysctl_conf(struct mlx4_en_priv *priv);
61 
62 #ifdef CONFIG_NET_RX_BUSY_POLL
63 /* must be called with local_bh_disable()d */
64 static int mlx4_en_low_latency_recv(struct napi_struct *napi)
65 {
66 	struct mlx4_en_cq *cq = container_of(napi, struct mlx4_en_cq, napi);
67 	struct net_device *dev = cq->dev;
68 	struct mlx4_en_priv *priv = netdev_priv(dev);
69 	struct mlx4_en_rx_ring *rx_ring = priv->rx_ring[cq->ring];
70 	int done;
71 
72 	if (!priv->port_up)
73 		return LL_FLUSH_FAILED;
74 
75 	if (!mlx4_en_cq_lock_poll(cq))
76 		return LL_FLUSH_BUSY;
77 
78 	done = mlx4_en_process_rx_cq(dev, cq, 4);
79 #ifdef LL_EXTENDED_STATS
80 	if (likely(done))
81 		rx_ring->cleaned += done;
82 	else
83 		rx_ring->misses++;
84 #endif
85 
86 	mlx4_en_cq_unlock_poll(cq);
87 
88 	return done;
89 }
90 #endif	/* CONFIG_NET_RX_BUSY_POLL */
91 
92 #ifdef CONFIG_RFS_ACCEL
93 
94 struct mlx4_en_filter {
95 	struct list_head next;
96 	struct work_struct work;
97 
98 	u8     ip_proto;
99 	__be32 src_ip;
100 	__be32 dst_ip;
101 	__be16 src_port;
102 	__be16 dst_port;
103 
104 	int rxq_index;
105 	struct mlx4_en_priv *priv;
106 	u32 flow_id;			/* RFS infrastructure id */
107 	int id;				/* mlx4_en driver id */
108 	u64 reg_id;			/* Flow steering API id */
109 	u8 activated;			/* Used to prevent expiry before filter
110 					 * is attached
111 					 */
112 	struct hlist_node filter_chain;
113 };
114 
115 static void mlx4_en_filter_rfs_expire(struct mlx4_en_priv *priv);
116 
117 static enum mlx4_net_trans_rule_id mlx4_ip_proto_to_trans_rule_id(u8 ip_proto)
118 {
119 	switch (ip_proto) {
120 	case IPPROTO_UDP:
121 		return MLX4_NET_TRANS_RULE_ID_UDP;
122 	case IPPROTO_TCP:
123 		return MLX4_NET_TRANS_RULE_ID_TCP;
124 	default:
125 		return MLX4_NET_TRANS_RULE_NUM;
126 	}
127 };
128 
129 static void mlx4_en_filter_work(struct work_struct *work)
130 {
131 	struct mlx4_en_filter *filter = container_of(work,
132 						     struct mlx4_en_filter,
133 						     work);
134 	struct mlx4_en_priv *priv = filter->priv;
135 	struct mlx4_spec_list spec_tcp_udp = {
136 		.id = mlx4_ip_proto_to_trans_rule_id(filter->ip_proto),
137 		{
138 			.tcp_udp = {
139 				.dst_port = filter->dst_port,
140 				.dst_port_msk = (__force __be16)-1,
141 				.src_port = filter->src_port,
142 				.src_port_msk = (__force __be16)-1,
143 			},
144 		},
145 	};
146 	struct mlx4_spec_list spec_ip = {
147 		.id = MLX4_NET_TRANS_RULE_ID_IPV4,
148 		{
149 			.ipv4 = {
150 				.dst_ip = filter->dst_ip,
151 				.dst_ip_msk = (__force __be32)-1,
152 				.src_ip = filter->src_ip,
153 				.src_ip_msk = (__force __be32)-1,
154 			},
155 		},
156 	};
157 	struct mlx4_spec_list spec_eth = {
158 		.id = MLX4_NET_TRANS_RULE_ID_ETH,
159 	};
160 	struct mlx4_net_trans_rule rule = {
161 		.list = LIST_HEAD_INIT(rule.list),
162 		.queue_mode = MLX4_NET_TRANS_Q_LIFO,
163 		.exclusive = 1,
164 		.allow_loopback = 1,
165 		.promisc_mode = MLX4_FS_REGULAR,
166 		.port = priv->port,
167 		.priority = MLX4_DOMAIN_RFS,
168 	};
169 	int rc;
170 	__be64 mac_mask = cpu_to_be64(MLX4_MAC_MASK << 16);
171 
172 	if (spec_tcp_udp.id >= MLX4_NET_TRANS_RULE_NUM) {
173 		en_warn(priv, "RFS: ignoring unsupported ip protocol (%d)\n",
174 			filter->ip_proto);
175 		goto ignore;
176 	}
177 	list_add_tail(&spec_eth.list, &rule.list);
178 	list_add_tail(&spec_ip.list, &rule.list);
179 	list_add_tail(&spec_tcp_udp.list, &rule.list);
180 
181 	rule.qpn = priv->rss_map.qps[filter->rxq_index].qpn;
182 	memcpy(spec_eth.eth.dst_mac, priv->dev->dev_addr, ETH_ALEN);
183 	memcpy(spec_eth.eth.dst_mac_msk, &mac_mask, ETH_ALEN);
184 
185 	filter->activated = 0;
186 
187 	if (filter->reg_id) {
188 		rc = mlx4_flow_detach(priv->mdev->dev, filter->reg_id);
189 		if (rc && rc != -ENOENT)
190 			en_err(priv, "Error detaching flow. rc = %d\n", rc);
191 	}
192 
193 	rc = mlx4_flow_attach(priv->mdev->dev, &rule, &filter->reg_id);
194 	if (rc)
195 		en_err(priv, "Error attaching flow. err = %d\n", rc);
196 
197 ignore:
198 	mlx4_en_filter_rfs_expire(priv);
199 
200 	filter->activated = 1;
201 }
202 
203 static inline struct hlist_head *
204 filter_hash_bucket(struct mlx4_en_priv *priv, __be32 src_ip, __be32 dst_ip,
205 		   __be16 src_port, __be16 dst_port)
206 {
207 	unsigned long l;
208 	int bucket_idx;
209 
210 	l = (__force unsigned long)src_port |
211 	    ((__force unsigned long)dst_port << 2);
212 	l ^= (__force unsigned long)(src_ip ^ dst_ip);
213 
214 	bucket_idx = hash_long(l, MLX4_EN_FILTER_HASH_SHIFT);
215 
216 	return &priv->filter_hash[bucket_idx];
217 }
218 
219 static struct mlx4_en_filter *
220 mlx4_en_filter_alloc(struct mlx4_en_priv *priv, int rxq_index, __be32 src_ip,
221 		     __be32 dst_ip, u8 ip_proto, __be16 src_port,
222 		     __be16 dst_port, u32 flow_id)
223 {
224 	struct mlx4_en_filter *filter = NULL;
225 
226 	filter = kzalloc(sizeof(struct mlx4_en_filter), GFP_ATOMIC);
227 	if (!filter)
228 		return NULL;
229 
230 	filter->priv = priv;
231 	filter->rxq_index = rxq_index;
232 	INIT_WORK(&filter->work, mlx4_en_filter_work);
233 
234 	filter->src_ip = src_ip;
235 	filter->dst_ip = dst_ip;
236 	filter->ip_proto = ip_proto;
237 	filter->src_port = src_port;
238 	filter->dst_port = dst_port;
239 
240 	filter->flow_id = flow_id;
241 
242 	filter->id = priv->last_filter_id++ % RPS_NO_FILTER;
243 
244 	list_add_tail(&filter->next, &priv->filters);
245 	hlist_add_head(&filter->filter_chain,
246 		       filter_hash_bucket(priv, src_ip, dst_ip, src_port,
247 					  dst_port));
248 
249 	return filter;
250 }
251 
252 static void mlx4_en_filter_free(struct mlx4_en_filter *filter)
253 {
254 	struct mlx4_en_priv *priv = filter->priv;
255 	int rc;
256 
257 	list_del(&filter->next);
258 
259 	rc = mlx4_flow_detach(priv->mdev->dev, filter->reg_id);
260 	if (rc && rc != -ENOENT)
261 		en_err(priv, "Error detaching flow. rc = %d\n", rc);
262 
263 	kfree(filter);
264 }
265 
266 static inline struct mlx4_en_filter *
267 mlx4_en_filter_find(struct mlx4_en_priv *priv, __be32 src_ip, __be32 dst_ip,
268 		    u8 ip_proto, __be16 src_port, __be16 dst_port)
269 {
270 	struct mlx4_en_filter *filter;
271 	struct mlx4_en_filter *ret = NULL;
272 
273 	hlist_for_each_entry(filter,
274 			     filter_hash_bucket(priv, src_ip, dst_ip,
275 						src_port, dst_port),
276 			     filter_chain) {
277 		if (filter->src_ip == src_ip &&
278 		    filter->dst_ip == dst_ip &&
279 		    filter->ip_proto == ip_proto &&
280 		    filter->src_port == src_port &&
281 		    filter->dst_port == dst_port) {
282 			ret = filter;
283 			break;
284 		}
285 	}
286 
287 	return ret;
288 }
289 
290 static int
291 mlx4_en_filter_rfs(struct net_device *net_dev, const struct sk_buff *skb,
292 		   u16 rxq_index, u32 flow_id)
293 {
294 	struct mlx4_en_priv *priv = netdev_priv(net_dev);
295 	struct mlx4_en_filter *filter;
296 	const struct iphdr *ip;
297 	const __be16 *ports;
298 	u8 ip_proto;
299 	__be32 src_ip;
300 	__be32 dst_ip;
301 	__be16 src_port;
302 	__be16 dst_port;
303 	int nhoff = skb_network_offset(skb);
304 	int ret = 0;
305 
306 	if (skb->protocol != htons(ETH_P_IP))
307 		return -EPROTONOSUPPORT;
308 
309 	ip = (const struct iphdr *)(skb->data + nhoff);
310 	if (ip_is_fragment(ip))
311 		return -EPROTONOSUPPORT;
312 
313 	if ((ip->protocol != IPPROTO_TCP) && (ip->protocol != IPPROTO_UDP))
314 		return -EPROTONOSUPPORT;
315 	ports = (const __be16 *)(skb->data + nhoff + 4 * ip->ihl);
316 
317 	ip_proto = ip->protocol;
318 	src_ip = ip->saddr;
319 	dst_ip = ip->daddr;
320 	src_port = ports[0];
321 	dst_port = ports[1];
322 
323 	spin_lock_bh(&priv->filters_lock);
324 	filter = mlx4_en_filter_find(priv, src_ip, dst_ip, ip_proto,
325 				     src_port, dst_port);
326 	if (filter) {
327 		if (filter->rxq_index == rxq_index)
328 			goto out;
329 
330 		filter->rxq_index = rxq_index;
331 	} else {
332 		filter = mlx4_en_filter_alloc(priv, rxq_index,
333 					      src_ip, dst_ip, ip_proto,
334 					      src_port, dst_port, flow_id);
335 		if (!filter) {
336 			ret = -ENOMEM;
337 			goto err;
338 		}
339 	}
340 
341 	queue_work(priv->mdev->workqueue, &filter->work);
342 
343 out:
344 	ret = filter->id;
345 err:
346 	spin_unlock_bh(&priv->filters_lock);
347 
348 	return ret;
349 }
350 
351 void mlx4_en_cleanup_filters(struct mlx4_en_priv *priv)
352 {
353 	struct mlx4_en_filter *filter, *tmp;
354 	LIST_HEAD(del_list);
355 
356 	spin_lock_bh(&priv->filters_lock);
357 	list_for_each_entry_safe(filter, tmp, &priv->filters, next) {
358 		list_move(&filter->next, &del_list);
359 		hlist_del(&filter->filter_chain);
360 	}
361 	spin_unlock_bh(&priv->filters_lock);
362 
363 	list_for_each_entry_safe(filter, tmp, &del_list, next) {
364 		cancel_work_sync(&filter->work);
365 		mlx4_en_filter_free(filter);
366 	}
367 }
368 
369 static void mlx4_en_filter_rfs_expire(struct mlx4_en_priv *priv)
370 {
371 	struct mlx4_en_filter *filter = NULL, *tmp, *last_filter = NULL;
372 	LIST_HEAD(del_list);
373 	int i = 0;
374 
375 	spin_lock_bh(&priv->filters_lock);
376 	list_for_each_entry_safe(filter, tmp, &priv->filters, next) {
377 		if (i > MLX4_EN_FILTER_EXPIRY_QUOTA)
378 			break;
379 
380 		if (filter->activated &&
381 		    !work_pending(&filter->work) &&
382 		    rps_may_expire_flow(priv->dev,
383 					filter->rxq_index, filter->flow_id,
384 					filter->id)) {
385 			list_move(&filter->next, &del_list);
386 			hlist_del(&filter->filter_chain);
387 		} else
388 			last_filter = filter;
389 
390 		i++;
391 	}
392 
393 	if (last_filter && (&last_filter->next != priv->filters.next))
394 		list_move(&priv->filters, &last_filter->next);
395 
396 	spin_unlock_bh(&priv->filters_lock);
397 
398 	list_for_each_entry_safe(filter, tmp, &del_list, next)
399 		mlx4_en_filter_free(filter);
400 }
401 #endif
402 
403 static void mlx4_en_vlan_rx_add_vid(void *arg, struct net_device *dev, u16 vid)
404 {
405 	struct mlx4_en_priv *priv = netdev_priv(dev);
406 	struct mlx4_en_dev *mdev = priv->mdev;
407 	int err;
408 	int idx;
409 
410 	if (arg != priv)
411 		return;
412 
413 	en_dbg(HW, priv, "adding VLAN:%d\n", vid);
414 
415 	set_bit(vid, priv->active_vlans);
416 
417 	/* Add VID to port VLAN filter */
418 	mutex_lock(&mdev->state_lock);
419 	if (mdev->device_up && priv->port_up) {
420 		err = mlx4_SET_VLAN_FLTR(mdev->dev, priv);
421 		if (err)
422 			en_err(priv, "Failed configuring VLAN filter\n");
423 	}
424 	if (mlx4_register_vlan(mdev->dev, priv->port, vid, &idx))
425 		en_dbg(HW, priv, "failed adding vlan %d\n", vid);
426 	mutex_unlock(&mdev->state_lock);
427 
428 }
429 
430 static void mlx4_en_vlan_rx_kill_vid(void *arg, struct net_device *dev, u16 vid)
431 {
432 	struct mlx4_en_priv *priv = netdev_priv(dev);
433 	struct mlx4_en_dev *mdev = priv->mdev;
434 	int err;
435 
436 	if (arg != priv)
437 		return;
438 
439 	en_dbg(HW, priv, "Killing VID:%d\n", vid);
440 
441 	clear_bit(vid, priv->active_vlans);
442 
443 	/* Remove VID from port VLAN filter */
444 	mutex_lock(&mdev->state_lock);
445 	mlx4_unregister_vlan(mdev->dev, priv->port, vid);
446 
447 	if (mdev->device_up && priv->port_up) {
448 		err = mlx4_SET_VLAN_FLTR(mdev->dev, priv);
449 		if (err)
450 			en_err(priv, "Failed configuring VLAN filter\n");
451 	}
452 	mutex_unlock(&mdev->state_lock);
453 
454 }
455 
456 static int mlx4_en_tunnel_steer_add(struct mlx4_en_priv *priv, unsigned char *addr,
457 				    int qpn, u64 *reg_id)
458 {
459 	int err;
460 
461 	if (priv->mdev->dev->caps.tunnel_offload_mode != MLX4_TUNNEL_OFFLOAD_MODE_VXLAN ||
462 	    priv->mdev->dev->caps.dmfs_high_steer_mode == MLX4_STEERING_DMFS_A0_STATIC)
463 		return 0; /* do nothing */
464 
465 	err = mlx4_tunnel_steer_add(priv->mdev->dev, addr, priv->port, qpn,
466 				    MLX4_DOMAIN_NIC, reg_id);
467 	if (err) {
468 		en_err(priv, "failed to add vxlan steering rule, err %d\n", err);
469 		return err;
470 	}
471 	en_dbg(DRV, priv, "added vxlan steering rule, mac %pM reg_id %llx\n", addr, (long long)*reg_id);
472 	return 0;
473 }
474 
475 static int mlx4_en_uc_steer_add(struct mlx4_en_priv *priv,
476 				unsigned char *mac, int *qpn, u64 *reg_id)
477 {
478 	struct mlx4_en_dev *mdev = priv->mdev;
479 	struct mlx4_dev *dev = mdev->dev;
480 	int err;
481 
482 	switch (dev->caps.steering_mode) {
483 	case MLX4_STEERING_MODE_B0: {
484 		struct mlx4_qp qp;
485 		u8 gid[16] = {0};
486 
487 		qp.qpn = *qpn;
488 		memcpy(&gid[10], mac, ETH_ALEN);
489 		gid[5] = priv->port;
490 
491 		err = mlx4_unicast_attach(dev, &qp, gid, 0, MLX4_PROT_ETH);
492 		break;
493 	}
494 	case MLX4_STEERING_MODE_DEVICE_MANAGED: {
495 		struct mlx4_spec_list spec_eth = { {NULL} };
496 		__be64 mac_mask = cpu_to_be64(MLX4_MAC_MASK << 16);
497 
498 		struct mlx4_net_trans_rule rule = {
499 			.queue_mode = MLX4_NET_TRANS_Q_FIFO,
500 			.exclusive = 0,
501 			.allow_loopback = 1,
502 			.promisc_mode = MLX4_FS_REGULAR,
503 			.priority = MLX4_DOMAIN_NIC,
504 		};
505 
506 		rule.port = priv->port;
507 		rule.qpn = *qpn;
508 		INIT_LIST_HEAD(&rule.list);
509 
510 		spec_eth.id = MLX4_NET_TRANS_RULE_ID_ETH;
511 		memcpy(spec_eth.eth.dst_mac, mac, ETH_ALEN);
512 		memcpy(spec_eth.eth.dst_mac_msk, &mac_mask, ETH_ALEN);
513 		list_add_tail(&spec_eth.list, &rule.list);
514 
515 		err = mlx4_flow_attach(dev, &rule, reg_id);
516 		break;
517 	}
518 	default:
519 		return -EINVAL;
520 	}
521 	if (err)
522 		en_warn(priv, "Failed Attaching Unicast\n");
523 
524 	return err;
525 }
526 
527 static void mlx4_en_uc_steer_release(struct mlx4_en_priv *priv,
528 				     unsigned char *mac, int qpn, u64 reg_id)
529 {
530 	struct mlx4_en_dev *mdev = priv->mdev;
531 	struct mlx4_dev *dev = mdev->dev;
532 
533 	switch (dev->caps.steering_mode) {
534 	case MLX4_STEERING_MODE_B0: {
535 		struct mlx4_qp qp;
536 		u8 gid[16] = {0};
537 
538 		qp.qpn = qpn;
539 		memcpy(&gid[10], mac, ETH_ALEN);
540 		gid[5] = priv->port;
541 
542 		mlx4_unicast_detach(dev, &qp, gid, MLX4_PROT_ETH);
543 		break;
544 	}
545 	case MLX4_STEERING_MODE_DEVICE_MANAGED: {
546 		mlx4_flow_detach(dev, reg_id);
547 		break;
548 	}
549 	default:
550 		en_err(priv, "Invalid steering mode.\n");
551 	}
552 }
553 
554 static int mlx4_en_get_qp(struct mlx4_en_priv *priv)
555 {
556 	struct mlx4_en_dev *mdev = priv->mdev;
557 	struct mlx4_dev *dev = mdev->dev;
558 	int index = 0;
559 	int err = 0;
560 	int *qpn = &priv->base_qpn;
561 	u64 mac = mlx4_mac_to_u64(IF_LLADDR(priv->dev));
562 
563 	en_dbg(DRV, priv, "Registering MAC: %pM for adding\n",
564 	       IF_LLADDR(priv->dev));
565 	index = mlx4_register_mac(dev, priv->port, mac);
566 	if (index < 0) {
567 		err = index;
568 		en_err(priv, "Failed adding MAC: %pM\n",
569 		       IF_LLADDR(priv->dev));
570 		return err;
571 	}
572 
573 	if (dev->caps.steering_mode == MLX4_STEERING_MODE_A0) {
574 		int base_qpn = mlx4_get_base_qpn(dev, priv->port);
575 		*qpn = base_qpn + index;
576 		return 0;
577 	}
578 
579 	err = mlx4_qp_reserve_range(dev, 1, 1, qpn, MLX4_RESERVE_A0_QP);
580 	en_dbg(DRV, priv, "Reserved qp %d\n", *qpn);
581 	if (err) {
582 		en_err(priv, "Failed to reserve qp for mac registration\n");
583 		mlx4_unregister_mac(dev, priv->port, mac);
584 		return err;
585 	}
586 
587 	return 0;
588 }
589 
590 static void mlx4_en_put_qp(struct mlx4_en_priv *priv)
591 {
592 	struct mlx4_en_dev *mdev = priv->mdev;
593 	struct mlx4_dev *dev = mdev->dev;
594 	int qpn = priv->base_qpn;
595 
596 	if (dev->caps.steering_mode == MLX4_STEERING_MODE_A0) {
597 		u64 mac = mlx4_mac_to_u64(IF_LLADDR(priv->dev));
598 		en_dbg(DRV, priv, "Registering MAC: %pM for deleting\n",
599 		       IF_LLADDR(priv->dev));
600 		mlx4_unregister_mac(dev, priv->port, mac);
601 	} else {
602 		en_dbg(DRV, priv, "Releasing qp: port %d, qpn %d\n",
603 		       priv->port, qpn);
604 		mlx4_qp_release_range(dev, qpn, 1);
605 		priv->flags &= ~MLX4_EN_FLAG_FORCE_PROMISC;
606 	}
607 }
608 
609 static void mlx4_en_clear_uclist(struct net_device *dev)
610 {
611 	struct mlx4_en_priv *priv = netdev_priv(dev);
612 	struct mlx4_en_addr_list *tmp, *uc_to_del;
613 
614 	list_for_each_entry_safe(uc_to_del, tmp, &priv->uc_list, list) {
615 		list_del(&uc_to_del->list);
616 		kfree(uc_to_del);
617 	}
618 }
619 
620 static u_int mlx4_copy_addr(void *arg, struct sockaddr_dl *sdl, u_int cnt)
621 {
622 	struct mlx4_en_priv *priv = arg;
623 	struct mlx4_en_addr_list *tmp;
624 
625 	if (sdl->sdl_alen != ETHER_ADDR_LEN)	/* XXXGL: can that happen? */
626 		return (0);
627 	tmp = kzalloc(sizeof(struct mlx4_en_addr_list), GFP_ATOMIC);
628 	if (tmp == NULL) {
629 		en_err(priv, "Failed to allocate address list\n");
630 		return (0);
631 	}
632 	memcpy(tmp->addr, LLADDR(sdl), ETH_ALEN);
633 	list_add_tail(&tmp->list, &priv->uc_list);
634 
635 	return (1);
636 }
637 
638 static void mlx4_en_cache_uclist(struct net_device *dev)
639 {
640 	struct mlx4_en_priv *priv = netdev_priv(dev);
641 
642 	mlx4_en_clear_uclist(dev);
643 	if_foreach_lladdr(dev, mlx4_copy_addr, priv);
644 }
645 
646 static void mlx4_en_clear_mclist(struct net_device *dev)
647 {
648 	struct mlx4_en_priv *priv = netdev_priv(dev);
649 	struct mlx4_en_addr_list *tmp, *mc_to_del;
650 
651 	list_for_each_entry_safe(mc_to_del, tmp, &priv->mc_list, list) {
652 		list_del(&mc_to_del->list);
653 		kfree(mc_to_del);
654 	}
655 }
656 
657 static u_int mlx4_copy_maddr(void *arg, struct sockaddr_dl *sdl, u_int count)
658 {
659 	struct mlx4_en_priv *priv = arg;
660 	struct mlx4_en_addr_list *tmp;
661 
662 	if (sdl->sdl_alen != ETHER_ADDR_LEN)	/* XXXGL: can that happen? */
663 		return (0);
664 	tmp = kzalloc(sizeof(struct mlx4_en_addr_list), GFP_ATOMIC);
665 	if (tmp == NULL) {
666 		en_err(priv, "Failed to allocate address list\n");
667 		return (0);
668 	}
669 	memcpy(tmp->addr, LLADDR(sdl), ETH_ALEN);
670 	list_add_tail(&tmp->list, &priv->mc_list);
671 	return (1);
672 }
673 
674 static void mlx4_en_cache_mclist(struct net_device *dev)
675 {
676 	struct mlx4_en_priv *priv = netdev_priv(dev);
677 
678 	mlx4_en_clear_mclist(dev);
679 	if_foreach_llmaddr(dev, mlx4_copy_maddr, priv);
680 }
681 
682 static void update_addr_list_flags(struct mlx4_en_priv *priv,
683 				struct list_head *dst,
684 				struct list_head *src)
685 {
686 	struct mlx4_en_addr_list *dst_tmp, *src_tmp, *new_mc;
687 	bool found;
688 
689 	/* Find all the entries that should be removed from dst,
690 	 * These are the entries that are not found in src
691 	 */
692 	list_for_each_entry(dst_tmp, dst, list) {
693 		found = false;
694 		list_for_each_entry(src_tmp, src, list) {
695 			if (!memcmp(dst_tmp->addr, src_tmp->addr, ETH_ALEN)) {
696 				found = true;
697 				break;
698 			}
699 		}
700 		if (!found)
701 			dst_tmp->action = MLX4_ADDR_LIST_REM;
702 	}
703 
704 	/* Add entries that exist in src but not in dst
705 	 * mark them as need to add
706 	 */
707 	list_for_each_entry(src_tmp, src, list) {
708 		found = false;
709 		list_for_each_entry(dst_tmp, dst, list) {
710 			if (!memcmp(dst_tmp->addr, src_tmp->addr, ETH_ALEN)) {
711 				dst_tmp->action = MLX4_ADDR_LIST_NONE;
712 				found = true;
713 				break;
714 			}
715 		}
716 		if (!found) {
717 			new_mc = kmalloc(sizeof(struct mlx4_en_addr_list),
718 					 GFP_KERNEL);
719 			if (!new_mc) {
720 				en_err(priv, "Failed to allocate current multicast list\n");
721 				return;
722 			}
723 			memcpy(new_mc, src_tmp,
724 			       sizeof(struct mlx4_en_addr_list));
725 			new_mc->action = MLX4_ADDR_LIST_ADD;
726 			list_add_tail(&new_mc->list, dst);
727 		}
728 	}
729 }
730 
731 static void mlx4_en_set_rx_mode(struct net_device *dev)
732 {
733 	struct mlx4_en_priv *priv = netdev_priv(dev);
734 
735 	if (!priv->port_up)
736 		return;
737 
738 	queue_work(priv->mdev->workqueue, &priv->rx_mode_task);
739 }
740 
741 static void mlx4_en_set_promisc_mode(struct mlx4_en_priv *priv,
742 				     struct mlx4_en_dev *mdev)
743 {
744 	int err = 0;
745 
746 	if (!(priv->flags & MLX4_EN_FLAG_PROMISC)) {
747 		priv->flags |= MLX4_EN_FLAG_PROMISC;
748 
749 		/* Enable promiscouos mode */
750 		switch (mdev->dev->caps.steering_mode) {
751 		case MLX4_STEERING_MODE_DEVICE_MANAGED:
752 			err = mlx4_flow_steer_promisc_add(mdev->dev,
753 							  priv->port,
754 							  priv->base_qpn,
755 							  MLX4_FS_ALL_DEFAULT);
756 			if (err)
757 				en_err(priv, "Failed enabling promiscuous mode\n");
758 			priv->flags |= MLX4_EN_FLAG_MC_PROMISC;
759 			break;
760 
761 		case MLX4_STEERING_MODE_B0:
762 			err = mlx4_unicast_promisc_add(mdev->dev,
763 						       priv->base_qpn,
764 						       priv->port);
765 			if (err)
766 				en_err(priv, "Failed enabling unicast promiscuous mode\n");
767 
768 			/* Add the default qp number as multicast
769 			 * promisc
770 			 */
771 			if (!(priv->flags & MLX4_EN_FLAG_MC_PROMISC)) {
772 				err = mlx4_multicast_promisc_add(mdev->dev,
773 								 priv->base_qpn,
774 								 priv->port);
775 				if (err)
776 					en_err(priv, "Failed enabling multicast promiscuous mode\n");
777 				priv->flags |= MLX4_EN_FLAG_MC_PROMISC;
778 			}
779 			break;
780 
781 		case MLX4_STEERING_MODE_A0:
782 			err = mlx4_SET_PORT_qpn_calc(mdev->dev,
783 						     priv->port,
784 						     priv->base_qpn,
785 						     1);
786 			if (err)
787 				en_err(priv, "Failed enabling promiscuous mode\n");
788 			break;
789 		}
790 
791 		/* Disable port multicast filter (unconditionally) */
792 		err = mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, 0,
793 					  0, MLX4_MCAST_DISABLE);
794 		if (err)
795 			en_err(priv, "Failed disabling multicast filter\n");
796 	}
797 }
798 
799 static void mlx4_en_clear_promisc_mode(struct mlx4_en_priv *priv,
800 				       struct mlx4_en_dev *mdev)
801 {
802 	int err = 0;
803 
804 	priv->flags &= ~MLX4_EN_FLAG_PROMISC;
805 
806 	/* Disable promiscouos mode */
807 	switch (mdev->dev->caps.steering_mode) {
808 	case MLX4_STEERING_MODE_DEVICE_MANAGED:
809 		err = mlx4_flow_steer_promisc_remove(mdev->dev,
810 						     priv->port,
811 						     MLX4_FS_ALL_DEFAULT);
812 		if (err)
813 			en_err(priv, "Failed disabling promiscuous mode\n");
814 		priv->flags &= ~MLX4_EN_FLAG_MC_PROMISC;
815 		break;
816 
817 	case MLX4_STEERING_MODE_B0:
818 		err = mlx4_unicast_promisc_remove(mdev->dev,
819 						  priv->base_qpn,
820 						  priv->port);
821 		if (err)
822 			en_err(priv, "Failed disabling unicast promiscuous mode\n");
823 		/* Disable Multicast promisc */
824 		if (priv->flags & MLX4_EN_FLAG_MC_PROMISC) {
825 			err = mlx4_multicast_promisc_remove(mdev->dev,
826 							    priv->base_qpn,
827 							    priv->port);
828 			if (err)
829 				en_err(priv, "Failed disabling multicast promiscuous mode\n");
830 			priv->flags &= ~MLX4_EN_FLAG_MC_PROMISC;
831 		}
832 		break;
833 
834 	case MLX4_STEERING_MODE_A0:
835 		err = mlx4_SET_PORT_qpn_calc(mdev->dev,
836 					     priv->port,
837 					     priv->base_qpn, 0);
838 		if (err)
839 			en_err(priv, "Failed disabling promiscuous mode\n");
840 		break;
841 	}
842 }
843 
844 static void mlx4_en_do_multicast(struct mlx4_en_priv *priv,
845 				 struct net_device *dev,
846 				 struct mlx4_en_dev *mdev)
847 {
848 	struct mlx4_en_addr_list *addr_list, *tmp;
849 	u8 mc_list[16] = {0};
850 	int err = 0;
851 	u64 mcast_addr = 0;
852 
853 	/*
854 	 * Enable/disable the multicast filter according to
855 	 * IFF_ALLMULTI and IFF_PROMISC:
856 	 */
857 	if (dev->if_flags & (IFF_ALLMULTI | IFF_PROMISC)) {
858 		err = mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, 0,
859 					  0, MLX4_MCAST_DISABLE);
860 		if (err)
861 			en_err(priv, "Failed disabling multicast filter\n");
862 
863 		/* Add the default qp number as multicast promisc */
864 		if (!(priv->flags & MLX4_EN_FLAG_MC_PROMISC)) {
865 			switch (mdev->dev->caps.steering_mode) {
866 			case MLX4_STEERING_MODE_DEVICE_MANAGED:
867 				err = mlx4_flow_steer_promisc_add(mdev->dev,
868 								  priv->port,
869 								  priv->base_qpn,
870 								  MLX4_FS_MC_DEFAULT);
871 				break;
872 
873 			case MLX4_STEERING_MODE_B0:
874 				err = mlx4_multicast_promisc_add(mdev->dev,
875 								 priv->base_qpn,
876 								 priv->port);
877 				break;
878 
879 			case MLX4_STEERING_MODE_A0:
880 				break;
881 			}
882 			if (err)
883 				en_err(priv, "Failed entering multicast promisc mode\n");
884 			priv->flags |= MLX4_EN_FLAG_MC_PROMISC;
885 		}
886 	} else {
887 		/* Disable Multicast promisc */
888 		if (priv->flags & MLX4_EN_FLAG_MC_PROMISC) {
889 			switch (mdev->dev->caps.steering_mode) {
890 			case MLX4_STEERING_MODE_DEVICE_MANAGED:
891 				err = mlx4_flow_steer_promisc_remove(mdev->dev,
892 								     priv->port,
893 								     MLX4_FS_MC_DEFAULT);
894 				break;
895 
896 			case MLX4_STEERING_MODE_B0:
897 				err = mlx4_multicast_promisc_remove(mdev->dev,
898 								    priv->base_qpn,
899 								    priv->port);
900 				break;
901 
902 			case MLX4_STEERING_MODE_A0:
903 				break;
904 			}
905 			if (err)
906 				en_err(priv, "Failed disabling multicast promiscuous mode\n");
907 			priv->flags &= ~MLX4_EN_FLAG_MC_PROMISC;
908 		}
909 
910 		err = mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, 0,
911 					  0, MLX4_MCAST_DISABLE);
912 		if (err)
913 			en_err(priv, "Failed disabling multicast filter\n");
914 
915 		/* Flush mcast filter and init it with broadcast address */
916 		mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, ETH_BCAST,
917 				    1, MLX4_MCAST_CONFIG);
918 
919 		/* Update multicast list - we cache all addresses so they won't
920 		 * change while HW is updated holding the command semaphor */
921 		mlx4_en_cache_mclist(dev);
922 		list_for_each_entry(addr_list, &priv->mc_list, list) {
923 			mcast_addr = mlx4_mac_to_u64(addr_list->addr);
924 			mlx4_SET_MCAST_FLTR(mdev->dev, priv->port,
925 					mcast_addr, 0, MLX4_MCAST_CONFIG);
926 		}
927 		err = mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, 0,
928 					  0, MLX4_MCAST_ENABLE);
929 		if (err)
930 			en_err(priv, "Failed enabling multicast filter\n");
931 
932 		update_addr_list_flags(priv, &priv->curr_mc_list, &priv->mc_list);
933 
934 		list_for_each_entry_safe(addr_list, tmp, &priv->curr_mc_list, list) {
935 			if (addr_list->action == MLX4_ADDR_LIST_REM) {
936 				/* detach this address and delete from list */
937 				memcpy(&mc_list[10], addr_list->addr, ETH_ALEN);
938 				mc_list[5] = priv->port;
939 				err = mlx4_multicast_detach(mdev->dev,
940 							    &priv->rss_map.indir_qp,
941 							    mc_list,
942 							    MLX4_PROT_ETH,
943 							    addr_list->reg_id);
944 				if (err)
945 					en_err(priv, "Fail to detach multicast address\n");
946 
947 				if (addr_list->tunnel_reg_id) {
948 					err = mlx4_flow_detach(priv->mdev->dev, addr_list->tunnel_reg_id);
949 					if (err)
950 						en_err(priv, "Failed to detach multicast address\n");
951 				}
952 
953 				/* remove from list */
954 				list_del(&addr_list->list);
955 				kfree(addr_list);
956 			} else if (addr_list->action == MLX4_ADDR_LIST_ADD) {
957 				/* attach the address */
958 				memcpy(&mc_list[10], addr_list->addr, ETH_ALEN);
959 				/* needed for B0 steering support */
960 				mc_list[5] = priv->port;
961 				err = mlx4_multicast_attach(mdev->dev,
962 							    &priv->rss_map.indir_qp,
963 							    mc_list,
964 							    priv->port, 0,
965 							    MLX4_PROT_ETH,
966 							    &addr_list->reg_id);
967 				if (err)
968 					en_err(priv, "Fail to attach multicast address\n");
969 
970 				err = mlx4_en_tunnel_steer_add(priv, &mc_list[10], priv->base_qpn,
971 							       &addr_list->tunnel_reg_id);
972 				if (err)
973 					en_err(priv, "Failed to attach multicast address\n");
974 			}
975 		}
976 	}
977 }
978 
979 static void mlx4_en_do_unicast(struct mlx4_en_priv *priv,
980 			       struct net_device *dev,
981 			       struct mlx4_en_dev *mdev)
982 {
983 	struct mlx4_en_addr_list *addr_list, *tmp;
984 	int err;
985 
986 	/* Update unicast list */
987 	mlx4_en_cache_uclist(dev);
988 
989 	update_addr_list_flags(priv, &priv->curr_uc_list, &priv->uc_list);
990 
991 	list_for_each_entry_safe(addr_list, tmp, &priv->curr_uc_list, list) {
992 		if (addr_list->action == MLX4_ADDR_LIST_REM) {
993 			mlx4_en_uc_steer_release(priv, addr_list->addr,
994 						 priv->rss_map.indir_qp.qpn,
995 						 addr_list->reg_id);
996 			/* remove from list */
997 			list_del(&addr_list->list);
998 			kfree(addr_list);
999 		} else if (addr_list->action == MLX4_ADDR_LIST_ADD) {
1000 			err = mlx4_en_uc_steer_add(priv, addr_list->addr,
1001 						   &priv->rss_map.indir_qp.qpn,
1002 						   &addr_list->reg_id);
1003 			if (err)
1004 				en_err(priv, "Fail to add unicast address\n");
1005 		}
1006 	}
1007 }
1008 
1009 static void mlx4_en_do_set_rx_mode(struct work_struct *work)
1010 {
1011 	struct mlx4_en_priv *priv = container_of(work, struct mlx4_en_priv,
1012 						 rx_mode_task);
1013 	struct mlx4_en_dev *mdev = priv->mdev;
1014 	struct net_device *dev = priv->dev;
1015 
1016 	mutex_lock(&mdev->state_lock);
1017 	if (!mdev->device_up) {
1018 		en_dbg(HW, priv, "Card is not up, ignoring rx mode change.\n");
1019 		goto out;
1020 	}
1021 	if (!priv->port_up) {
1022 		en_dbg(HW, priv, "Port is down, ignoring rx mode change.\n");
1023 		goto out;
1024 	}
1025 	if (!mlx4_en_QUERY_PORT(mdev, priv->port)) {
1026 		if (priv->port_state.link_state) {
1027 			priv->last_link_state = MLX4_DEV_EVENT_PORT_UP;
1028 			/* update netif baudrate */
1029 			priv->dev->if_baudrate =
1030 			    IF_Mbps(priv->port_state.link_speed);
1031 			/* Important note: the following call for if_link_state_change
1032 			 * is needed for interface up scenario (start port, link state
1033 			 * change) */
1034 			if_link_state_change(priv->dev, LINK_STATE_UP);
1035 			en_dbg(HW, priv, "Link Up\n");
1036 		}
1037 	}
1038 
1039 	/* Set unicast rules */
1040 	mlx4_en_do_unicast(priv, dev, mdev);
1041 
1042 	/* Promsicuous mode: disable all filters */
1043 	if ((dev->if_flags & IFF_PROMISC) ||
1044 	    (priv->flags & MLX4_EN_FLAG_FORCE_PROMISC)) {
1045 		mlx4_en_set_promisc_mode(priv, mdev);
1046 	} else if (priv->flags & MLX4_EN_FLAG_PROMISC) {
1047 		/* Not in promiscuous mode */
1048 		mlx4_en_clear_promisc_mode(priv, mdev);
1049 	}
1050 
1051 	/* Set multicast rules */
1052 	mlx4_en_do_multicast(priv, dev, mdev);
1053 out:
1054 	mutex_unlock(&mdev->state_lock);
1055 }
1056 
1057 static void mlx4_en_watchdog_timeout(void *arg)
1058 {
1059         struct mlx4_en_priv *priv = arg;
1060         struct mlx4_en_dev *mdev = priv->mdev;
1061 
1062         en_dbg(DRV, priv, "Scheduling watchdog\n");
1063         queue_work(mdev->workqueue, &priv->watchdog_task);
1064         if (priv->port_up)
1065                 callout_reset(&priv->watchdog_timer, MLX4_EN_WATCHDOG_TIMEOUT,
1066                                 mlx4_en_watchdog_timeout, priv);
1067 }
1068 
1069 
1070 
1071 static void mlx4_en_set_default_moderation(struct mlx4_en_priv *priv)
1072 {
1073 	struct mlx4_en_cq *cq;
1074 	int i;
1075 
1076 	/* If we haven't received a specific coalescing setting
1077 	 * (module param), we set the moderation parameters as follows:
1078 	 * - moder_cnt is set to the number of mtu sized packets to
1079 	 *   satisfy our coalescing target.
1080 	 * - moder_time is set to a fixed value.
1081 	 */
1082 	priv->rx_frames = MLX4_EN_RX_COAL_TARGET;
1083 	priv->rx_usecs = MLX4_EN_RX_COAL_TIME;
1084 	priv->tx_frames = MLX4_EN_TX_COAL_PKTS;
1085 	priv->tx_usecs = MLX4_EN_TX_COAL_TIME;
1086 	en_dbg(INTR, priv, "Default coalesing params for mtu: %u - "
1087 	       "rx_frames:%d rx_usecs:%d\n",
1088 	       (unsigned)priv->dev->if_mtu, priv->rx_frames, priv->rx_usecs);
1089 
1090 	/* Setup cq moderation params */
1091 	for (i = 0; i < priv->rx_ring_num; i++) {
1092 		cq = priv->rx_cq[i];
1093 		cq->moder_cnt = priv->rx_frames;
1094 		cq->moder_time = priv->rx_usecs;
1095 		priv->last_moder_time[i] = MLX4_EN_AUTO_CONF;
1096 		priv->last_moder_packets[i] = 0;
1097 		priv->last_moder_bytes[i] = 0;
1098 	}
1099 
1100 	for (i = 0; i < priv->tx_ring_num; i++) {
1101 		cq = priv->tx_cq[i];
1102 		cq->moder_cnt = priv->tx_frames;
1103 		cq->moder_time = priv->tx_usecs;
1104 	}
1105 
1106 	/* Reset auto-moderation params */
1107 	priv->pkt_rate_low = MLX4_EN_RX_RATE_LOW;
1108 	priv->rx_usecs_low = MLX4_EN_RX_COAL_TIME_LOW;
1109 	priv->pkt_rate_high = MLX4_EN_RX_RATE_HIGH;
1110 	priv->rx_usecs_high = MLX4_EN_RX_COAL_TIME_HIGH;
1111 	priv->sample_interval = MLX4_EN_SAMPLE_INTERVAL;
1112 	priv->adaptive_rx_coal = 1;
1113 	priv->last_moder_jiffies = 0;
1114 	priv->last_moder_tx_packets = 0;
1115 }
1116 
1117 static void mlx4_en_auto_moderation(struct mlx4_en_priv *priv)
1118 {
1119 	unsigned long period = (unsigned long) (jiffies - priv->last_moder_jiffies);
1120 	struct mlx4_en_cq *cq;
1121 	unsigned long packets;
1122 	unsigned long rate;
1123 	unsigned long avg_pkt_size;
1124 	unsigned long rx_packets;
1125 	unsigned long rx_bytes;
1126 	unsigned long rx_pkt_diff;
1127 	int moder_time;
1128 	int ring, err;
1129 
1130 	if (!priv->adaptive_rx_coal || period < priv->sample_interval * HZ)
1131 		return;
1132 
1133 	for (ring = 0; ring < priv->rx_ring_num; ring++) {
1134                 spin_lock(&priv->stats_lock);
1135 		rx_packets = priv->rx_ring[ring]->packets;
1136 		rx_bytes = priv->rx_ring[ring]->bytes;
1137 		spin_unlock(&priv->stats_lock);
1138 
1139 		rx_pkt_diff = ((unsigned long) (rx_packets -
1140 				priv->last_moder_packets[ring]));
1141 		packets = rx_pkt_diff;
1142 		rate = packets * HZ / period;
1143 		avg_pkt_size = packets ? ((unsigned long) (rx_bytes -
1144 				priv->last_moder_bytes[ring])) / packets : 0;
1145 
1146 		/* Apply auto-moderation only when packet rate
1147 		 * exceeds a rate that it matters */
1148 		if (rate > (MLX4_EN_RX_RATE_THRESH / priv->rx_ring_num) &&
1149 		    avg_pkt_size > MLX4_EN_AVG_PKT_SMALL) {
1150 			if (rate < priv->pkt_rate_low)
1151 				moder_time = priv->rx_usecs_low;
1152 			else if (rate > priv->pkt_rate_high)
1153 				moder_time = priv->rx_usecs_high;
1154 			else
1155 				moder_time = (rate - priv->pkt_rate_low) *
1156 					(priv->rx_usecs_high - priv->rx_usecs_low) /
1157 					(priv->pkt_rate_high - priv->pkt_rate_low) +
1158 					priv->rx_usecs_low;
1159 		} else {
1160 			moder_time = priv->rx_usecs_low;
1161 		}
1162 
1163 		if (moder_time != priv->last_moder_time[ring]) {
1164 			priv->last_moder_time[ring] = moder_time;
1165 			cq = priv->rx_cq[ring];
1166 			cq->moder_time = moder_time;
1167 			cq->moder_cnt = priv->rx_frames;
1168 			err = mlx4_en_set_cq_moder(priv, cq);
1169 			if (err)
1170 				en_err(priv, "Failed modifying moderation for cq:%d\n",
1171 				       ring);
1172 		}
1173 		priv->last_moder_packets[ring] = rx_packets;
1174 		priv->last_moder_bytes[ring] = rx_bytes;
1175 	}
1176 
1177 	priv->last_moder_jiffies = jiffies;
1178 }
1179 
1180 static void mlx4_en_do_get_stats(struct work_struct *work)
1181 {
1182 	struct delayed_work *delay = to_delayed_work(work);
1183 	struct mlx4_en_priv *priv = container_of(delay, struct mlx4_en_priv,
1184 						 stats_task);
1185 	struct mlx4_en_dev *mdev = priv->mdev;
1186 	int err;
1187 
1188 	mutex_lock(&mdev->state_lock);
1189 	if (mdev->device_up) {
1190 		if (priv->port_up) {
1191 			if (mlx4_is_slave(mdev->dev))
1192 				err = mlx4_en_get_vport_stats(mdev, priv->port);
1193 			else
1194 				err = mlx4_en_DUMP_ETH_STATS(mdev, priv->port, 0);
1195 			if (err)
1196 				en_dbg(HW, priv, "Could not update stats\n");
1197 
1198 			mlx4_en_auto_moderation(priv);
1199 		}
1200 
1201 		queue_delayed_work(mdev->workqueue, &priv->stats_task, STATS_DELAY);
1202 	}
1203 	mutex_unlock(&mdev->state_lock);
1204 }
1205 
1206 /* mlx4_en_service_task - Run service task for tasks that needed to be done
1207  * periodically
1208  */
1209 static void mlx4_en_service_task(struct work_struct *work)
1210 {
1211 	struct delayed_work *delay = to_delayed_work(work);
1212 	struct mlx4_en_priv *priv = container_of(delay, struct mlx4_en_priv,
1213 						 service_task);
1214 	struct mlx4_en_dev *mdev = priv->mdev;
1215 
1216 	mutex_lock(&mdev->state_lock);
1217 	if (mdev->device_up) {
1218 		queue_delayed_work(mdev->workqueue, &priv->service_task,
1219 				   SERVICE_TASK_DELAY);
1220 	}
1221 	mutex_unlock(&mdev->state_lock);
1222 }
1223 
1224 static void mlx4_en_linkstate(struct work_struct *work)
1225 {
1226 	struct mlx4_en_priv *priv = container_of(work, struct mlx4_en_priv,
1227 						 linkstate_task);
1228 	struct mlx4_en_dev *mdev = priv->mdev;
1229 	int linkstate = priv->link_state;
1230 
1231 	mutex_lock(&mdev->state_lock);
1232 	/* If observable port state changed set carrier state and
1233 	 * report to system log */
1234 	if (priv->last_link_state != linkstate) {
1235 		if (linkstate == MLX4_DEV_EVENT_PORT_DOWN) {
1236 			en_info(priv, "Link Down\n");
1237 			if_link_state_change(priv->dev, LINK_STATE_DOWN);
1238 			/* update netif baudrate */
1239 			priv->dev->if_baudrate = 0;
1240 
1241 		/* make sure the port is up before notifying the OS.
1242 		 * This is tricky since we get here on INIT_PORT and
1243 		 * in such case we can't tell the OS the port is up.
1244 		 * To solve this there is a call to if_link_state_change
1245 		 * in set_rx_mode.
1246 		 * */
1247 		} else if (priv->port_up && (linkstate == MLX4_DEV_EVENT_PORT_UP)){
1248 			if (mlx4_en_QUERY_PORT(priv->mdev, priv->port))
1249 				en_info(priv, "Query port failed\n");
1250 			priv->dev->if_baudrate =
1251 			    IF_Mbps(priv->port_state.link_speed);
1252 			en_info(priv, "Link Up\n");
1253 			if_link_state_change(priv->dev, LINK_STATE_UP);
1254 		}
1255 	}
1256 	priv->last_link_state = linkstate;
1257 	mutex_unlock(&mdev->state_lock);
1258 }
1259 
1260 
1261 int mlx4_en_start_port(struct net_device *dev)
1262 {
1263 	struct mlx4_en_priv *priv = netdev_priv(dev);
1264 	struct mlx4_en_dev *mdev = priv->mdev;
1265 	struct mlx4_en_cq *cq;
1266 	struct mlx4_en_tx_ring *tx_ring;
1267 	int rx_index = 0;
1268 	int tx_index = 0;
1269 	int err = 0;
1270 	int i;
1271 	int j;
1272 	u8 mc_list[16] = {0};
1273 
1274 
1275 	if (priv->port_up) {
1276 		en_dbg(DRV, priv, "start port called while port already up\n");
1277 		return 0;
1278 	}
1279 
1280 	INIT_LIST_HEAD(&priv->mc_list);
1281 	INIT_LIST_HEAD(&priv->uc_list);
1282 	INIT_LIST_HEAD(&priv->curr_mc_list);
1283 	INIT_LIST_HEAD(&priv->curr_uc_list);
1284 	INIT_LIST_HEAD(&priv->ethtool_list);
1285 
1286 	/* Calculate Rx buf size */
1287 	dev->if_mtu = min(dev->if_mtu, priv->max_mtu);
1288         mlx4_en_calc_rx_buf(dev);
1289 	en_dbg(DRV, priv, "Rx buf size:%d\n", priv->rx_mb_size);
1290 
1291 	/* Configure rx cq's and rings */
1292 	err = mlx4_en_activate_rx_rings(priv);
1293 	if (err) {
1294 		en_err(priv, "Failed to activate RX rings\n");
1295 		return err;
1296 	}
1297 	for (i = 0; i < priv->rx_ring_num; i++) {
1298 		cq = priv->rx_cq[i];
1299 
1300 		mlx4_en_cq_init_lock(cq);
1301 		err = mlx4_en_activate_cq(priv, cq, i);
1302 		if (err) {
1303 			en_err(priv, "Failed activating Rx CQ\n");
1304 			goto cq_err;
1305 		}
1306 		for (j = 0; j < cq->size; j++)
1307 			cq->buf[j].owner_sr_opcode = MLX4_CQE_OWNER_MASK;
1308 		err = mlx4_en_set_cq_moder(priv, cq);
1309 		if (err) {
1310 			en_err(priv, "Failed setting cq moderation parameters");
1311 			mlx4_en_deactivate_cq(priv, cq);
1312 			goto cq_err;
1313 		}
1314 		mlx4_en_arm_cq(priv, cq);
1315 		priv->rx_ring[i]->cqn = cq->mcq.cqn;
1316 		++rx_index;
1317 	}
1318 
1319 	/* Set qp number */
1320 	en_dbg(DRV, priv, "Getting qp number for port %d\n", priv->port);
1321 	err = mlx4_en_get_qp(priv);
1322 	if (err) {
1323 		en_err(priv, "Failed getting eth qp\n");
1324 		goto cq_err;
1325 	}
1326 	mdev->mac_removed[priv->port] = 0;
1327 
1328 	priv->counter_index =
1329 			mlx4_get_default_counter_index(mdev->dev, priv->port);
1330 
1331 	err = mlx4_en_config_rss_steer(priv);
1332 	if (err) {
1333 		en_err(priv, "Failed configuring rss steering\n");
1334 		goto mac_err;
1335 	}
1336 
1337 	err = mlx4_en_create_drop_qp(priv);
1338 	if (err)
1339 		goto rss_err;
1340 
1341 	/* Configure tx cq's and rings */
1342 	for (i = 0; i < priv->tx_ring_num; i++) {
1343 		/* Configure cq */
1344 		cq = priv->tx_cq[i];
1345 		err = mlx4_en_activate_cq(priv, cq, i);
1346 		if (err) {
1347 			en_err(priv, "Failed activating Tx CQ\n");
1348 			goto tx_err;
1349 		}
1350 		err = mlx4_en_set_cq_moder(priv, cq);
1351 		if (err) {
1352 			en_err(priv, "Failed setting cq moderation parameters");
1353 			mlx4_en_deactivate_cq(priv, cq);
1354 			goto tx_err;
1355 		}
1356 		en_dbg(DRV, priv, "Resetting index of collapsed CQ:%d to -1\n", i);
1357 		cq->buf->wqe_index = cpu_to_be16(0xffff);
1358 
1359 		/* Configure ring */
1360 		tx_ring = priv->tx_ring[i];
1361 
1362 		err = mlx4_en_activate_tx_ring(priv, tx_ring, cq->mcq.cqn,
1363 					       i / priv->num_tx_rings_p_up);
1364 		if (err) {
1365 			en_err(priv, "Failed activating Tx ring %d\n", i);
1366 			mlx4_en_deactivate_cq(priv, cq);
1367 			goto tx_err;
1368 		}
1369 
1370 		/* Arm CQ for TX completions */
1371 		mlx4_en_arm_cq(priv, cq);
1372 
1373 		/* Set initial ownership of all Tx TXBBs to SW (1) */
1374 		for (j = 0; j < tx_ring->buf_size; j += STAMP_STRIDE)
1375 			*((u32 *) (tx_ring->buf + j)) = INIT_OWNER_BIT;
1376 		++tx_index;
1377 	}
1378 
1379 	/* Configure port */
1380 	err = mlx4_SET_PORT_general(mdev->dev, priv->port,
1381 				    priv->rx_mb_size,
1382 				    priv->prof->tx_pause,
1383 				    priv->prof->tx_ppp,
1384 				    priv->prof->rx_pause,
1385 				    priv->prof->rx_ppp);
1386 	if (err) {
1387 		en_err(priv, "Failed setting port general configurations for port %d, with error %d\n",
1388 		       priv->port, err);
1389 		goto tx_err;
1390 	}
1391 	/* Set default qp number */
1392 	err = mlx4_SET_PORT_qpn_calc(mdev->dev, priv->port, priv->base_qpn, 0);
1393 	if (err) {
1394 		en_err(priv, "Failed setting default qp numbers\n");
1395 		goto tx_err;
1396 	}
1397 
1398 	/* Init port */
1399 	en_dbg(HW, priv, "Initializing port\n");
1400 	err = mlx4_INIT_PORT(mdev->dev, priv->port);
1401 	if (err) {
1402 		en_err(priv, "Failed Initializing port\n");
1403 		goto tx_err;
1404 	}
1405 
1406 	/* Attach rx QP to bradcast address */
1407 	memset(&mc_list[10], 0xff, ETH_ALEN);
1408 	mc_list[5] = priv->port; /* needed for B0 steering support */
1409 	if (mlx4_multicast_attach(mdev->dev, &priv->rss_map.indir_qp, mc_list,
1410 				  priv->port, 0, MLX4_PROT_ETH,
1411 				  &priv->broadcast_id))
1412 		mlx4_warn(mdev, "Failed Attaching Broadcast\n");
1413 
1414 	/* Must redo promiscuous mode setup. */
1415 	priv->flags &= ~(MLX4_EN_FLAG_PROMISC | MLX4_EN_FLAG_MC_PROMISC);
1416 
1417 	/* Schedule multicast task to populate multicast list */
1418 	queue_work(mdev->workqueue, &priv->rx_mode_task);
1419 
1420 	priv->port_up = true;
1421 
1422         /* Enable the queues. */
1423         dev->if_drv_flags &= ~IFF_DRV_OACTIVE;
1424         dev->if_drv_flags |= IFF_DRV_RUNNING;
1425 #ifdef CONFIG_DEBUG_FS
1426 	mlx4_en_create_debug_files(priv);
1427 #endif
1428         callout_reset(&priv->watchdog_timer, MLX4_EN_WATCHDOG_TIMEOUT,
1429                     mlx4_en_watchdog_timeout, priv);
1430 
1431 
1432 	return 0;
1433 
1434 tx_err:
1435 	while (tx_index--) {
1436 		mlx4_en_deactivate_tx_ring(priv, priv->tx_ring[tx_index]);
1437 		mlx4_en_deactivate_cq(priv, priv->tx_cq[tx_index]);
1438 	}
1439 	mlx4_en_destroy_drop_qp(priv);
1440 rss_err:
1441 	mlx4_en_release_rss_steer(priv);
1442 mac_err:
1443 	mlx4_en_put_qp(priv);
1444 cq_err:
1445 	while (rx_index--)
1446 		mlx4_en_deactivate_cq(priv, priv->rx_cq[rx_index]);
1447 	for (i = 0; i < priv->rx_ring_num; i++)
1448 		mlx4_en_deactivate_rx_ring(priv, priv->rx_ring[i]);
1449 
1450 	return err; /* need to close devices */
1451 }
1452 
1453 
1454 void mlx4_en_stop_port(struct net_device *dev)
1455 {
1456 	struct mlx4_en_priv *priv = netdev_priv(dev);
1457 	struct mlx4_en_dev *mdev = priv->mdev;
1458 	struct mlx4_en_addr_list *addr_list, *tmp;
1459 	int i;
1460 	u8 mc_list[16] = {0};
1461 
1462 	if (!priv->port_up) {
1463 		en_dbg(DRV, priv, "stop port called while port already down\n");
1464 		return;
1465 	}
1466 
1467 #ifdef CONFIG_DEBUG_FS
1468 	mlx4_en_delete_debug_files(priv);
1469 #endif
1470 
1471 	/* close port*/
1472 	mlx4_CLOSE_PORT(mdev->dev, priv->port);
1473 
1474 	/* Set port as not active */
1475 	priv->port_up = false;
1476 	priv->counter_index = MLX4_SINK_COUNTER_INDEX(mdev->dev);
1477 
1478 	/* Promsicuous mode */
1479 	if (mdev->dev->caps.steering_mode ==
1480 	    MLX4_STEERING_MODE_DEVICE_MANAGED) {
1481 		priv->flags &= ~(MLX4_EN_FLAG_PROMISC |
1482 				 MLX4_EN_FLAG_MC_PROMISC);
1483 		mlx4_flow_steer_promisc_remove(mdev->dev,
1484 					       priv->port,
1485 					       MLX4_FS_ALL_DEFAULT);
1486 		mlx4_flow_steer_promisc_remove(mdev->dev,
1487 					       priv->port,
1488 					       MLX4_FS_MC_DEFAULT);
1489 	} else if (priv->flags & MLX4_EN_FLAG_PROMISC) {
1490 		priv->flags &= ~MLX4_EN_FLAG_PROMISC;
1491 
1492 		/* Disable promiscouos mode */
1493 		mlx4_unicast_promisc_remove(mdev->dev, priv->base_qpn,
1494 					    priv->port);
1495 
1496 		/* Disable Multicast promisc */
1497 		if (priv->flags & MLX4_EN_FLAG_MC_PROMISC) {
1498 			mlx4_multicast_promisc_remove(mdev->dev, priv->base_qpn,
1499 						      priv->port);
1500 			priv->flags &= ~MLX4_EN_FLAG_MC_PROMISC;
1501 		}
1502 	}
1503 
1504 	/* Detach All unicasts */
1505 	list_for_each_entry(addr_list, &priv->curr_uc_list, list) {
1506 		mlx4_en_uc_steer_release(priv, addr_list->addr,
1507 					 priv->rss_map.indir_qp.qpn,
1508 					 addr_list->reg_id);
1509 	}
1510 	mlx4_en_clear_uclist(dev);
1511 	list_for_each_entry_safe(addr_list, tmp, &priv->curr_uc_list, list) {
1512 		list_del(&addr_list->list);
1513 		kfree(addr_list);
1514 	}
1515 
1516 	/* Detach All multicasts */
1517 	memset(&mc_list[10], 0xff, ETH_ALEN);
1518 	mc_list[5] = priv->port; /* needed for B0 steering support */
1519 	mlx4_multicast_detach(mdev->dev, &priv->rss_map.indir_qp, mc_list,
1520 			      MLX4_PROT_ETH, priv->broadcast_id);
1521 	list_for_each_entry(addr_list, &priv->curr_mc_list, list) {
1522 		memcpy(&mc_list[10], addr_list->addr, ETH_ALEN);
1523 		mc_list[5] = priv->port;
1524 		mlx4_multicast_detach(mdev->dev, &priv->rss_map.indir_qp,
1525 				      mc_list, MLX4_PROT_ETH, addr_list->reg_id);
1526 	}
1527 	mlx4_en_clear_mclist(dev);
1528 	list_for_each_entry_safe(addr_list, tmp, &priv->curr_mc_list, list) {
1529 		list_del(&addr_list->list);
1530 		kfree(addr_list);
1531 	}
1532 
1533 	/* Flush multicast filter */
1534 	mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, 0, 1, MLX4_MCAST_CONFIG);
1535 	mlx4_en_destroy_drop_qp(priv);
1536 
1537 	/* Free TX Rings */
1538 	for (i = 0; i < priv->tx_ring_num; i++) {
1539 		mlx4_en_deactivate_tx_ring(priv, priv->tx_ring[i]);
1540 		mlx4_en_deactivate_cq(priv, priv->tx_cq[i]);
1541 	}
1542 	msleep(10);
1543 
1544 	for (i = 0; i < priv->tx_ring_num; i++)
1545 		mlx4_en_free_tx_buf(dev, priv->tx_ring[i]);
1546 
1547 	/* Free RSS qps */
1548 	mlx4_en_release_rss_steer(priv);
1549 
1550 	/* Unregister Mac address for the port */
1551 	mlx4_en_put_qp(priv);
1552 	mdev->mac_removed[priv->port] = 1;
1553 
1554 	/* Free RX Rings */
1555 	for (i = 0; i < priv->rx_ring_num; i++) {
1556 		struct mlx4_en_cq *cq = priv->rx_cq[i];
1557 		mlx4_en_deactivate_rx_ring(priv, priv->rx_ring[i]);
1558 		mlx4_en_deactivate_cq(priv, cq);
1559 	}
1560 
1561         callout_stop(&priv->watchdog_timer);
1562 
1563         dev->if_drv_flags &= ~(IFF_DRV_RUNNING | IFF_DRV_OACTIVE);
1564 }
1565 
1566 static void mlx4_en_restart(struct work_struct *work)
1567 {
1568 	struct mlx4_en_priv *priv = container_of(work, struct mlx4_en_priv,
1569 						 watchdog_task);
1570 	struct mlx4_en_dev *mdev = priv->mdev;
1571 	struct net_device *dev = priv->dev;
1572 	struct mlx4_en_tx_ring *ring;
1573 	int i;
1574 
1575 
1576 	if (priv->blocked == 0 || priv->port_up == 0)
1577 		return;
1578 	for (i = 0; i < priv->tx_ring_num; i++) {
1579 		int watchdog_time;
1580 
1581 		ring = priv->tx_ring[i];
1582 		watchdog_time = READ_ONCE(ring->watchdog_time);
1583 		if (watchdog_time != 0 &&
1584 		    time_after(ticks, ring->watchdog_time))
1585 			goto reset;
1586 	}
1587 	return;
1588 
1589 reset:
1590 	priv->port_stats.tx_timeout++;
1591 	en_dbg(DRV, priv, "Watchdog task called for port %d\n", priv->port);
1592 
1593 	mutex_lock(&mdev->state_lock);
1594 	if (priv->port_up) {
1595 		mlx4_en_stop_port(dev);
1596                 //for (i = 0; i < priv->tx_ring_num; i++)
1597                 //        netdev_tx_reset_queue(priv->tx_ring[i]->tx_queue);
1598 		if (mlx4_en_start_port(dev))
1599 			en_err(priv, "Failed restarting port %d\n", priv->port);
1600 	}
1601 	mutex_unlock(&mdev->state_lock);
1602 }
1603 
1604 static void mlx4_en_clear_stats(struct net_device *dev)
1605 {
1606 	struct mlx4_en_priv *priv = netdev_priv(dev);
1607 	struct mlx4_en_dev *mdev = priv->mdev;
1608 	int i;
1609 
1610 	if (!mlx4_is_slave(mdev->dev))
1611 		if (mlx4_en_DUMP_ETH_STATS(mdev, priv->port, 1))
1612 			en_dbg(HW, priv, "Failed dumping statistics\n");
1613 
1614 	memset(&priv->pstats, 0, sizeof(priv->pstats));
1615 	memset(&priv->pkstats, 0, sizeof(priv->pkstats));
1616 	memset(&priv->port_stats, 0, sizeof(priv->port_stats));
1617 	memset(&priv->vport_stats, 0, sizeof(priv->vport_stats));
1618 
1619 	for (i = 0; i < priv->tx_ring_num; i++) {
1620 		priv->tx_ring[i]->bytes = 0;
1621 		priv->tx_ring[i]->packets = 0;
1622 		priv->tx_ring[i]->tx_csum = 0;
1623 		priv->tx_ring[i]->oversized_packets = 0;
1624 	}
1625 	for (i = 0; i < priv->rx_ring_num; i++) {
1626 		priv->rx_ring[i]->bytes = 0;
1627 		priv->rx_ring[i]->packets = 0;
1628 		priv->rx_ring[i]->csum_ok = 0;
1629 		priv->rx_ring[i]->csum_none = 0;
1630 	}
1631 }
1632 
1633 static void mlx4_en_open(void* arg)
1634 {
1635 
1636         struct mlx4_en_priv *priv;
1637         struct mlx4_en_dev *mdev;
1638         struct net_device *dev;
1639         int err = 0;
1640 
1641         priv = arg;
1642         mdev = priv->mdev;
1643         dev = priv->dev;
1644 
1645 
1646 	mutex_lock(&mdev->state_lock);
1647 
1648 	if (!mdev->device_up) {
1649 		en_err(priv, "Cannot open - device down/disabled\n");
1650 		goto out;
1651 	}
1652 
1653 	/* Reset HW statistics and SW counters */
1654 	mlx4_en_clear_stats(dev);
1655 
1656 	err = mlx4_en_start_port(dev);
1657 	if (err)
1658 		en_err(priv, "Failed starting port:%d\n", priv->port);
1659 
1660 out:
1661 	mutex_unlock(&mdev->state_lock);
1662 	return;
1663 }
1664 
1665 void mlx4_en_free_resources(struct mlx4_en_priv *priv)
1666 {
1667 	int i;
1668 
1669 #ifdef CONFIG_RFS_ACCEL
1670 	if (priv->dev->rx_cpu_rmap) {
1671 		free_irq_cpu_rmap(priv->dev->rx_cpu_rmap);
1672 		priv->dev->rx_cpu_rmap = NULL;
1673 	}
1674 #endif
1675 
1676 	for (i = 0; i < priv->tx_ring_num; i++) {
1677 		if (priv->tx_ring && priv->tx_ring[i])
1678 			mlx4_en_destroy_tx_ring(priv, &priv->tx_ring[i]);
1679 		if (priv->tx_cq && priv->tx_cq[i])
1680 			mlx4_en_destroy_cq(priv, &priv->tx_cq[i]);
1681 	}
1682 
1683 	for (i = 0; i < priv->rx_ring_num; i++) {
1684 		if (priv->rx_ring[i])
1685 			mlx4_en_destroy_rx_ring(priv, &priv->rx_ring[i],
1686 				priv->prof->rx_ring_size);
1687 		if (priv->rx_cq[i])
1688 			mlx4_en_destroy_cq(priv, &priv->rx_cq[i]);
1689 	}
1690 
1691 	if (priv->stat_sysctl != NULL)
1692 		sysctl_ctx_free(&priv->stat_ctx);
1693 }
1694 
1695 int mlx4_en_alloc_resources(struct mlx4_en_priv *priv)
1696 {
1697 	struct mlx4_en_port_profile *prof = priv->prof;
1698 	int i;
1699 	int node = 0;
1700 
1701 	/* Create rx Rings */
1702 	for (i = 0; i < priv->rx_ring_num; i++) {
1703 		if (mlx4_en_create_cq(priv, &priv->rx_cq[i],
1704 				      prof->rx_ring_size, i, RX, node))
1705 			goto err;
1706 
1707 		if (mlx4_en_create_rx_ring(priv, &priv->rx_ring[i],
1708 					   prof->rx_ring_size, node))
1709 			goto err;
1710 	}
1711 
1712 	/* Create tx Rings */
1713 	for (i = 0; i < priv->tx_ring_num; i++) {
1714 		if (mlx4_en_create_cq(priv, &priv->tx_cq[i],
1715 				      prof->tx_ring_size, i, TX, node))
1716 			goto err;
1717 
1718 		if (mlx4_en_create_tx_ring(priv, &priv->tx_ring[i],
1719 					   prof->tx_ring_size, TXBB_SIZE, node, i))
1720 			goto err;
1721 	}
1722 
1723 #ifdef CONFIG_RFS_ACCEL
1724 	priv->dev->rx_cpu_rmap = alloc_irq_cpu_rmap(priv->rx_ring_num);
1725 	if (!priv->dev->rx_cpu_rmap)
1726 		goto err;
1727 #endif
1728         /* Re-create stat sysctls in case the number of rings changed. */
1729 	mlx4_en_sysctl_stat(priv);
1730 	return 0;
1731 
1732 err:
1733 	en_err(priv, "Failed to allocate NIC resources\n");
1734 	for (i = 0; i < priv->rx_ring_num; i++) {
1735 		if (priv->rx_ring[i])
1736 			mlx4_en_destroy_rx_ring(priv, &priv->rx_ring[i],
1737 						prof->rx_ring_size);
1738 		if (priv->rx_cq[i])
1739 			mlx4_en_destroy_cq(priv, &priv->rx_cq[i]);
1740 	}
1741 	for (i = 0; i < priv->tx_ring_num; i++) {
1742 		if (priv->tx_ring[i])
1743 			mlx4_en_destroy_tx_ring(priv, &priv->tx_ring[i]);
1744 		if (priv->tx_cq[i])
1745 			mlx4_en_destroy_cq(priv, &priv->tx_cq[i]);
1746 	}
1747 	priv->port_up = false;
1748 	return -ENOMEM;
1749 }
1750 
1751 struct en_port_attribute {
1752 	struct attribute attr;
1753 	ssize_t (*show)(struct en_port *, struct en_port_attribute *, char *buf);
1754 	ssize_t (*store)(struct en_port *, struct en_port_attribute *, char *buf, size_t count);
1755 };
1756 
1757 #define PORT_ATTR_RO(_name) \
1758 struct en_port_attribute en_port_attr_##_name = __ATTR_RO(_name)
1759 
1760 #define EN_PORT_ATTR(_name, _mode, _show, _store) \
1761 struct en_port_attribute en_port_attr_##_name = __ATTR(_name, _mode, _show, _store)
1762 
1763 void mlx4_en_destroy_netdev(struct net_device *dev)
1764 {
1765 	struct mlx4_en_priv *priv = netdev_priv(dev);
1766 	struct mlx4_en_dev *mdev = priv->mdev;
1767 
1768 	en_dbg(DRV, priv, "Destroying netdev on port:%d\n", priv->port);
1769 
1770 	/* don't allow more IOCTLs */
1771 	priv->gone = 1;
1772 
1773 	/* XXX wait a bit to allow IOCTL handlers to complete */
1774 	pause("W", hz);
1775 
1776 	if (priv->vlan_attach != NULL)
1777 		EVENTHANDLER_DEREGISTER(vlan_config, priv->vlan_attach);
1778 	if (priv->vlan_detach != NULL)
1779 		EVENTHANDLER_DEREGISTER(vlan_unconfig, priv->vlan_detach);
1780 
1781 	mutex_lock(&mdev->state_lock);
1782 	mlx4_en_stop_port(dev);
1783 	mutex_unlock(&mdev->state_lock);
1784 
1785 	/* Unregister device - this will close the port if it was up */
1786 	if (priv->registered)
1787 		ether_ifdetach(dev);
1788 
1789 	if (priv->allocated)
1790 		mlx4_free_hwq_res(mdev->dev, &priv->res, MLX4_EN_PAGE_SIZE);
1791 
1792 	cancel_delayed_work(&priv->stats_task);
1793 	cancel_delayed_work(&priv->service_task);
1794 	/* flush any pending task for this netdev */
1795 	flush_workqueue(mdev->workqueue);
1796         callout_drain(&priv->watchdog_timer);
1797 
1798 	/* Detach the netdev so tasks would not attempt to access it */
1799 	mutex_lock(&mdev->state_lock);
1800 	mdev->pndev[priv->port] = NULL;
1801 	mutex_unlock(&mdev->state_lock);
1802 
1803 
1804 	mlx4_en_free_resources(priv);
1805 
1806 	/* freeing the sysctl conf cannot be called from within mlx4_en_free_resources */
1807 	if (priv->conf_sysctl != NULL)
1808 		sysctl_ctx_free(&priv->conf_ctx);
1809 
1810 	kfree(priv->tx_ring);
1811 	kfree(priv->tx_cq);
1812 
1813         kfree(priv);
1814         if_free(dev);
1815 
1816 }
1817 
1818 static int mlx4_en_change_mtu(struct net_device *dev, int new_mtu)
1819 {
1820 	struct mlx4_en_priv *priv = netdev_priv(dev);
1821 	struct mlx4_en_dev *mdev = priv->mdev;
1822 	int err = 0;
1823 
1824 	en_dbg(DRV, priv, "Change MTU called - current:%u new:%u\n",
1825 	       (unsigned)dev->if_mtu, (unsigned)new_mtu);
1826 
1827 	if ((new_mtu < MLX4_EN_MIN_MTU) || (new_mtu > priv->max_mtu)) {
1828 		en_err(priv, "Bad MTU size:%d, max %u.\n", new_mtu,
1829 		    priv->max_mtu);
1830 		return -EPERM;
1831 	}
1832 	mutex_lock(&mdev->state_lock);
1833 	dev->if_mtu = new_mtu;
1834 	if (dev->if_drv_flags & IFF_DRV_RUNNING) {
1835 		if (!mdev->device_up) {
1836 			/* NIC is probably restarting - let watchdog task reset
1837 			 *                          * the port */
1838 			en_dbg(DRV, priv, "Change MTU called with card down!?\n");
1839 		} else {
1840 			mlx4_en_stop_port(dev);
1841 			err = mlx4_en_start_port(dev);
1842 			if (err) {
1843 				en_err(priv, "Failed restarting port:%d\n",
1844 						priv->port);
1845 				queue_work(mdev->workqueue, &priv->watchdog_task);
1846 			}
1847 		}
1848 	}
1849 	mutex_unlock(&mdev->state_lock);
1850 	return 0;
1851 }
1852 
1853 static int mlx4_en_calc_media(struct mlx4_en_priv *priv)
1854 {
1855 	int trans_type;
1856 	int active;
1857 
1858 	active = IFM_ETHER;
1859 	if (priv->last_link_state == MLX4_DEV_EVENT_PORT_DOWN)
1860 		return (active);
1861 	active |= IFM_FDX;
1862 	trans_type = priv->port_state.transceiver;
1863 	/* XXX I don't know all of the transceiver values. */
1864 	switch (priv->port_state.link_speed) {
1865 	case 100:
1866 		active |= IFM_100_T;
1867 		break;
1868 	case 1000:
1869 		active |= IFM_1000_T;
1870 		break;
1871 	case 10000:
1872 		if (trans_type > 0 && trans_type <= 0xC)
1873 			active |= IFM_10G_SR;
1874 		else if (trans_type == 0x80 || trans_type == 0)
1875 			active |= IFM_10G_CX4;
1876 		break;
1877 	case 40000:
1878 		active |= IFM_40G_CR4;
1879 		break;
1880 	}
1881 	if (priv->prof->tx_pause)
1882 		active |= IFM_ETH_TXPAUSE;
1883 	if (priv->prof->rx_pause)
1884 		active |= IFM_ETH_RXPAUSE;
1885 
1886 	return (active);
1887 }
1888 
1889 static void mlx4_en_media_status(struct ifnet *dev, struct ifmediareq *ifmr)
1890 {
1891 	struct mlx4_en_priv *priv;
1892 
1893 	priv = dev->if_softc;
1894 	ifmr->ifm_status = IFM_AVALID;
1895 	if (priv->last_link_state != MLX4_DEV_EVENT_PORT_DOWN)
1896 		ifmr->ifm_status |= IFM_ACTIVE;
1897 	ifmr->ifm_active = mlx4_en_calc_media(priv);
1898 
1899 	return;
1900 }
1901 
1902 static int mlx4_en_media_change(struct ifnet *dev)
1903 {
1904 	struct mlx4_en_priv *priv;
1905         struct ifmedia *ifm;
1906 	int rxpause;
1907 	int txpause;
1908 	int error;
1909 
1910 	priv = dev->if_softc;
1911 	ifm = &priv->media;
1912 	rxpause = txpause = 0;
1913 	error = 0;
1914 
1915 	if (IFM_TYPE(ifm->ifm_media) != IFM_ETHER)
1916 		return (EINVAL);
1917         switch (IFM_SUBTYPE(ifm->ifm_media)) {
1918         case IFM_AUTO:
1919 		break;
1920 	case IFM_10G_SR:
1921 	case IFM_10G_CX4:
1922 	case IFM_1000_T:
1923 	case IFM_40G_CR4:
1924 		if ((IFM_SUBTYPE(ifm->ifm_media)
1925 			== IFM_SUBTYPE(mlx4_en_calc_media(priv)))
1926 			&& (ifm->ifm_media & IFM_FDX))
1927 			break;
1928 		/* Fallthrough */
1929 	default:
1930                 printf("%s: Only auto media type\n", if_name(dev));
1931                 return (EINVAL);
1932 	}
1933 	/* Allow user to set/clear pause */
1934 	if (IFM_OPTIONS(ifm->ifm_media) & IFM_ETH_RXPAUSE)
1935 		rxpause = 1;
1936 	if (IFM_OPTIONS(ifm->ifm_media) & IFM_ETH_TXPAUSE)
1937 		txpause = 1;
1938 	if (priv->prof->tx_pause != txpause || priv->prof->rx_pause != rxpause) {
1939 		priv->prof->tx_pause = txpause;
1940 		priv->prof->rx_pause = rxpause;
1941 		error = -mlx4_SET_PORT_general(priv->mdev->dev, priv->port,
1942 		     priv->rx_mb_size + ETHER_CRC_LEN, priv->prof->tx_pause,
1943 		     priv->prof->tx_ppp, priv->prof->rx_pause,
1944 		     priv->prof->rx_ppp);
1945 	}
1946 	return (error);
1947 }
1948 
1949 static int mlx4_en_ioctl(struct ifnet *dev, u_long command, caddr_t data)
1950 {
1951 	struct mlx4_en_priv *priv;
1952 	struct mlx4_en_dev *mdev;
1953 	struct ifreq *ifr;
1954 	int error;
1955 	int mask;
1956 	struct ifrsskey *ifrk;
1957 	const u32 *key;
1958 	struct ifrsshash *ifrh;
1959 	u8 rss_mask;
1960 
1961 	error = 0;
1962 	mask = 0;
1963 	priv = dev->if_softc;
1964 
1965 	/* check if detaching */
1966 	if (priv == NULL || priv->gone != 0)
1967 		return (ENXIO);
1968 
1969 	mdev = priv->mdev;
1970 	ifr = (struct ifreq *) data;
1971 
1972 	switch (command) {
1973 	case SIOCSIFMTU:
1974 		error = -mlx4_en_change_mtu(dev, ifr->ifr_mtu);
1975 		break;
1976 	case SIOCSIFFLAGS:
1977 		if (dev->if_flags & IFF_UP) {
1978 			if ((dev->if_drv_flags & IFF_DRV_RUNNING) == 0) {
1979 				mutex_lock(&mdev->state_lock);
1980 				mlx4_en_start_port(dev);
1981 				mutex_unlock(&mdev->state_lock);
1982 			} else {
1983 				mlx4_en_set_rx_mode(dev);
1984 			}
1985 		} else {
1986 			mutex_lock(&mdev->state_lock);
1987 			if (dev->if_drv_flags & IFF_DRV_RUNNING) {
1988 				mlx4_en_stop_port(dev);
1989 				if_link_state_change(dev, LINK_STATE_DOWN);
1990 			}
1991 			mutex_unlock(&mdev->state_lock);
1992 		}
1993 		break;
1994 	case SIOCADDMULTI:
1995 	case SIOCDELMULTI:
1996 		mlx4_en_set_rx_mode(dev);
1997 		break;
1998 	case SIOCSIFMEDIA:
1999 	case SIOCGIFMEDIA:
2000 		error = ifmedia_ioctl(dev, ifr, &priv->media, command);
2001 		break;
2002 	case SIOCSIFCAP:
2003 		mutex_lock(&mdev->state_lock);
2004 		mask = ifr->ifr_reqcap ^ dev->if_capenable;
2005 		if (mask & IFCAP_TXCSUM) {
2006 			dev->if_capenable ^= IFCAP_TXCSUM;
2007 			dev->if_hwassist ^= (CSUM_TCP | CSUM_UDP | CSUM_IP);
2008 
2009 			if (IFCAP_TSO4 & dev->if_capenable &&
2010 			    !(IFCAP_TXCSUM & dev->if_capenable)) {
2011 				mask &= ~IFCAP_TSO4;
2012 				dev->if_capenable &= ~IFCAP_TSO4;
2013 				dev->if_hwassist &= ~CSUM_IP_TSO;
2014 				if_printf(dev,
2015 				    "tso4 disabled due to -txcsum.\n");
2016 			}
2017 		}
2018 		if (mask & IFCAP_TXCSUM_IPV6) {
2019 			dev->if_capenable ^= IFCAP_TXCSUM_IPV6;
2020 			dev->if_hwassist ^= (CSUM_UDP_IPV6 | CSUM_TCP_IPV6);
2021 
2022 			if (IFCAP_TSO6 & dev->if_capenable &&
2023 			    !(IFCAP_TXCSUM_IPV6 & dev->if_capenable)) {
2024 				mask &= ~IFCAP_TSO6;
2025 				dev->if_capenable &= ~IFCAP_TSO6;
2026 				dev->if_hwassist &= ~CSUM_IP6_TSO;
2027 				if_printf(dev,
2028 				    "tso6 disabled due to -txcsum6.\n");
2029 			}
2030 		}
2031 		if (mask & IFCAP_RXCSUM)
2032 			dev->if_capenable ^= IFCAP_RXCSUM;
2033 		if (mask & IFCAP_RXCSUM_IPV6)
2034 			dev->if_capenable ^= IFCAP_RXCSUM_IPV6;
2035 
2036 		if (mask & IFCAP_TSO4) {
2037 			if (!(IFCAP_TSO4 & dev->if_capenable) &&
2038 			    !(IFCAP_TXCSUM & dev->if_capenable)) {
2039 				if_printf(dev, "enable txcsum first.\n");
2040 				error = EAGAIN;
2041 				goto out;
2042 			}
2043 			dev->if_capenable ^= IFCAP_TSO4;
2044 			dev->if_hwassist ^= CSUM_IP_TSO;
2045 		}
2046 		if (mask & IFCAP_TSO6) {
2047 			if (!(IFCAP_TSO6 & dev->if_capenable) &&
2048 			    !(IFCAP_TXCSUM_IPV6 & dev->if_capenable)) {
2049 				if_printf(dev, "enable txcsum6 first.\n");
2050 				error = EAGAIN;
2051 				goto out;
2052 			}
2053 			dev->if_capenable ^= IFCAP_TSO6;
2054 			dev->if_hwassist ^= CSUM_IP6_TSO;
2055 		}
2056 		if (mask & IFCAP_LRO)
2057 			dev->if_capenable ^= IFCAP_LRO;
2058 		if (mask & IFCAP_VLAN_HWTAGGING)
2059 			dev->if_capenable ^= IFCAP_VLAN_HWTAGGING;
2060 		if (mask & IFCAP_VLAN_HWFILTER)
2061 			dev->if_capenable ^= IFCAP_VLAN_HWFILTER;
2062 		if (mask & IFCAP_WOL_MAGIC)
2063 			dev->if_capenable ^= IFCAP_WOL_MAGIC;
2064 		if (dev->if_drv_flags & IFF_DRV_RUNNING)
2065 			mlx4_en_start_port(dev);
2066 out:
2067 		mutex_unlock(&mdev->state_lock);
2068 		VLAN_CAPABILITIES(dev);
2069 		break;
2070 #if __FreeBSD_version >= 1100036
2071 	case SIOCGI2C: {
2072 		struct ifi2creq i2c;
2073 
2074 		error = copyin(ifr_data_get_ptr(ifr), &i2c, sizeof(i2c));
2075 		if (error)
2076 			break;
2077 		if (i2c.len > sizeof(i2c.data)) {
2078 			error = EINVAL;
2079 			break;
2080 		}
2081 		/*
2082 		 * Note that we ignore i2c.addr here. The driver hardcodes
2083 		 * the address to 0x50, while standard expects it to be 0xA0.
2084 		 */
2085 		error = mlx4_get_module_info(mdev->dev, priv->port,
2086 		    i2c.offset, i2c.len, i2c.data);
2087 		if (error < 0) {
2088 			error = -error;
2089 			break;
2090 		}
2091 		error = copyout(&i2c, ifr_data_get_ptr(ifr), sizeof(i2c));
2092 		break;
2093 	}
2094 #endif
2095 	case SIOCGIFRSSKEY:
2096 		ifrk = (struct ifrsskey *)data;
2097 		ifrk->ifrk_func = RSS_FUNC_TOEPLITZ;
2098 		mutex_lock(&mdev->state_lock);
2099 		key = mlx4_en_get_rss_key(priv, &ifrk->ifrk_keylen);
2100 		if (ifrk->ifrk_keylen > RSS_KEYLEN)
2101 			error = EINVAL;
2102 		else
2103 			memcpy(ifrk->ifrk_key, key, ifrk->ifrk_keylen);
2104 		mutex_unlock(&mdev->state_lock);
2105 		break;
2106 
2107 	case SIOCGIFRSSHASH:
2108 		mutex_lock(&mdev->state_lock);
2109 		rss_mask = mlx4_en_get_rss_mask(priv);
2110 		mutex_unlock(&mdev->state_lock);
2111 		ifrh = (struct ifrsshash *)data;
2112 		ifrh->ifrh_func = RSS_FUNC_TOEPLITZ;
2113 		ifrh->ifrh_types = 0;
2114 		if (rss_mask & MLX4_RSS_IPV4)
2115 			ifrh->ifrh_types |= RSS_TYPE_IPV4;
2116 		if (rss_mask & MLX4_RSS_TCP_IPV4)
2117 			ifrh->ifrh_types |= RSS_TYPE_TCP_IPV4;
2118 		if (rss_mask & MLX4_RSS_IPV6)
2119 			ifrh->ifrh_types |= RSS_TYPE_IPV6;
2120 		if (rss_mask & MLX4_RSS_TCP_IPV6)
2121 			ifrh->ifrh_types |= RSS_TYPE_TCP_IPV6;
2122 		if (rss_mask & MLX4_RSS_UDP_IPV4)
2123 			ifrh->ifrh_types |= RSS_TYPE_UDP_IPV4;
2124 		if (rss_mask & MLX4_RSS_UDP_IPV6)
2125 			ifrh->ifrh_types |= RSS_TYPE_UDP_IPV6;
2126 		break;
2127 
2128 	default:
2129 		error = ether_ioctl(dev, command, data);
2130 		break;
2131 	}
2132 
2133 	return (error);
2134 }
2135 
2136 
2137 int mlx4_en_init_netdev(struct mlx4_en_dev *mdev, int port,
2138 			struct mlx4_en_port_profile *prof)
2139 {
2140 	struct net_device *dev;
2141 	struct mlx4_en_priv *priv;
2142 	uint8_t dev_addr[ETHER_ADDR_LEN];
2143 	int err;
2144 	int i;
2145 
2146 	priv = kzalloc(sizeof(*priv), GFP_KERNEL);
2147 	dev = priv->dev = if_alloc(IFT_ETHER);
2148 	if (dev == NULL) {
2149 		en_err(priv, "Net device allocation failed\n");
2150 		kfree(priv);
2151 		return -ENOMEM;
2152 	}
2153 	dev->if_softc = priv;
2154 	if_initname(dev, "mlxen", (device_get_unit(
2155 	    mdev->pdev->dev.bsddev) * MLX4_MAX_PORTS) + port - 1);
2156 	dev->if_mtu = ETHERMTU;
2157 	dev->if_init = mlx4_en_open;
2158 	dev->if_flags = IFF_BROADCAST | IFF_SIMPLEX | IFF_MULTICAST;
2159 	dev->if_ioctl = mlx4_en_ioctl;
2160 	dev->if_transmit = mlx4_en_transmit;
2161 	dev->if_qflush = mlx4_en_qflush;
2162 	dev->if_snd.ifq_maxlen = prof->tx_ring_size;
2163 
2164 	/*
2165 	 * Initialize driver private data
2166 	 */
2167 	priv->counter_index = 0xff;
2168 	spin_lock_init(&priv->stats_lock);
2169 	INIT_WORK(&priv->rx_mode_task, mlx4_en_do_set_rx_mode);
2170 	INIT_WORK(&priv->watchdog_task, mlx4_en_restart);
2171 	INIT_WORK(&priv->linkstate_task, mlx4_en_linkstate);
2172 	INIT_DELAYED_WORK(&priv->stats_task, mlx4_en_do_get_stats);
2173 	INIT_DELAYED_WORK(&priv->service_task, mlx4_en_service_task);
2174 	callout_init(&priv->watchdog_timer, 1);
2175 #ifdef CONFIG_RFS_ACCEL
2176 	INIT_LIST_HEAD(&priv->filters);
2177 	spin_lock_init(&priv->filters_lock);
2178 #endif
2179 
2180 	priv->msg_enable = MLX4_EN_MSG_LEVEL;
2181 	priv->dev = dev;
2182 	priv->mdev = mdev;
2183 	priv->ddev = &mdev->pdev->dev;
2184 	priv->prof = prof;
2185 	priv->port = port;
2186 	priv->port_up = false;
2187 	priv->flags = prof->flags;
2188 
2189 	priv->num_tx_rings_p_up = mdev->profile.num_tx_rings_p_up;
2190 	priv->tx_ring_num = prof->tx_ring_num;
2191 	priv->tx_ring = kcalloc(MAX_TX_RINGS,
2192 				sizeof(struct mlx4_en_tx_ring *), GFP_KERNEL);
2193 	if (!priv->tx_ring) {
2194 		err = -ENOMEM;
2195 		goto out;
2196 	}
2197 	priv->tx_cq = kcalloc(sizeof(struct mlx4_en_cq *), MAX_TX_RINGS,
2198 			GFP_KERNEL);
2199 	if (!priv->tx_cq) {
2200 		err = -ENOMEM;
2201 		goto out;
2202 	}
2203 
2204 	priv->rx_ring_num = prof->rx_ring_num;
2205 	priv->cqe_factor = (mdev->dev->caps.cqe_size == 64) ? 1 : 0;
2206 	priv->mac_index = -1;
2207 	priv->last_ifq_jiffies = 0;
2208 	priv->if_counters_rx_errors = 0;
2209 	priv->if_counters_rx_no_buffer = 0;
2210 #ifdef CONFIG_MLX4_EN_DCB
2211 	if (!mlx4_is_slave(priv->mdev->dev)) {
2212 		priv->dcbx_cap = DCB_CAP_DCBX_HOST;
2213 		priv->flags |= MLX4_EN_FLAG_DCB_ENABLED;
2214 		if (mdev->dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_ETS_CFG) {
2215 			dev->dcbnl_ops = &mlx4_en_dcbnl_ops;
2216 		} else {
2217 			en_info(priv, "QoS disabled - no HW support\n");
2218 			dev->dcbnl_ops = &mlx4_en_dcbnl_pfc_ops;
2219 		}
2220 	}
2221 #endif
2222 
2223 	/* Query for default mac and max mtu */
2224 	priv->max_mtu = mdev->dev->caps.eth_mtu_cap[priv->port];
2225         priv->mac = mdev->dev->caps.def_mac[priv->port];
2226         if (ILLEGAL_MAC(priv->mac)) {
2227 #if BITS_PER_LONG == 64
2228                 en_err(priv, "Port: %d, invalid mac burned: 0x%lx, quiting\n",
2229                                 priv->port, priv->mac);
2230 #elif BITS_PER_LONG == 32
2231                 en_err(priv, "Port: %d, invalid mac burned: 0x%llx, quiting\n",
2232                                 priv->port, priv->mac);
2233 #endif
2234                 err = -EINVAL;
2235                 goto out;
2236         }
2237 
2238 	mlx4_en_sysctl_conf(priv);
2239 
2240 	err = mlx4_en_alloc_resources(priv);
2241 	if (err)
2242 		goto out;
2243 
2244 	/* Allocate page for receive rings */
2245 	err = mlx4_alloc_hwq_res(mdev->dev, &priv->res,
2246 				MLX4_EN_PAGE_SIZE, MLX4_EN_PAGE_SIZE);
2247 	if (err) {
2248 		en_err(priv, "Failed to allocate page for rx qps\n");
2249 		goto out;
2250 	}
2251 	priv->allocated = 1;
2252 
2253 	/*
2254 	 * Set driver features
2255 	 */
2256 	dev->if_capabilities |= IFCAP_HWCSUM | IFCAP_HWCSUM_IPV6;
2257 	dev->if_capabilities |= IFCAP_VLAN_MTU | IFCAP_VLAN_HWTAGGING;
2258 	dev->if_capabilities |= IFCAP_VLAN_HWCSUM | IFCAP_VLAN_HWFILTER;
2259 	dev->if_capabilities |= IFCAP_LINKSTATE | IFCAP_JUMBO_MTU;
2260 	dev->if_capabilities |= IFCAP_LRO;
2261 	dev->if_capabilities |= IFCAP_HWSTATS;
2262 
2263 	if (mdev->LSO_support)
2264 		dev->if_capabilities |= IFCAP_TSO4 | IFCAP_TSO6 | IFCAP_VLAN_HWTSO;
2265 
2266 #if __FreeBSD_version >= 1100000
2267 	/* set TSO limits so that we don't have to drop TX packets */
2268 	dev->if_hw_tsomax = MLX4_EN_TX_MAX_PAYLOAD_SIZE - (ETHER_HDR_LEN + ETHER_VLAN_ENCAP_LEN) /* hdr */;
2269 	dev->if_hw_tsomaxsegcount = MLX4_EN_TX_MAX_MBUF_FRAGS - 1 /* hdr */;
2270 	dev->if_hw_tsomaxsegsize = MLX4_EN_TX_MAX_MBUF_SIZE;
2271 #endif
2272 
2273 	dev->if_capenable = dev->if_capabilities;
2274 
2275 	dev->if_hwassist = 0;
2276 	if (dev->if_capenable & (IFCAP_TSO4 | IFCAP_TSO6))
2277 		dev->if_hwassist |= CSUM_TSO;
2278 	if (dev->if_capenable & IFCAP_TXCSUM)
2279 		dev->if_hwassist |= (CSUM_TCP | CSUM_UDP | CSUM_IP);
2280 	if (dev->if_capenable & IFCAP_TXCSUM_IPV6)
2281 		dev->if_hwassist |= (CSUM_UDP_IPV6 | CSUM_TCP_IPV6);
2282 
2283 
2284         /* Register for VLAN events */
2285 	priv->vlan_attach = EVENTHANDLER_REGISTER(vlan_config,
2286             mlx4_en_vlan_rx_add_vid, priv, EVENTHANDLER_PRI_FIRST);
2287 	priv->vlan_detach = EVENTHANDLER_REGISTER(vlan_unconfig,
2288             mlx4_en_vlan_rx_kill_vid, priv, EVENTHANDLER_PRI_FIRST);
2289 
2290 	mdev->pndev[priv->port] = dev;
2291 
2292 	priv->last_link_state = MLX4_DEV_EVENT_PORT_DOWN;
2293         mlx4_en_set_default_moderation(priv);
2294 
2295 	/* Set default MAC */
2296 	for (i = 0; i < ETHER_ADDR_LEN; i++)
2297 		dev_addr[ETHER_ADDR_LEN - 1 - i] = (u8) (priv->mac >> (8 * i));
2298 
2299 
2300 	ether_ifattach(dev, dev_addr);
2301 	if_link_state_change(dev, LINK_STATE_DOWN);
2302 	ifmedia_init(&priv->media, IFM_IMASK | IFM_ETH_FMASK,
2303 	    mlx4_en_media_change, mlx4_en_media_status);
2304 	ifmedia_add(&priv->media, IFM_ETHER | IFM_FDX | IFM_1000_T, 0, NULL);
2305 	ifmedia_add(&priv->media, IFM_ETHER | IFM_FDX | IFM_10G_SR, 0, NULL);
2306 	ifmedia_add(&priv->media, IFM_ETHER | IFM_FDX | IFM_10G_CX4, 0, NULL);
2307 	ifmedia_add(&priv->media, IFM_ETHER | IFM_FDX | IFM_40G_CR4, 0, NULL);
2308 	ifmedia_add(&priv->media, IFM_ETHER | IFM_AUTO, 0, NULL);
2309 	ifmedia_set(&priv->media, IFM_ETHER | IFM_AUTO);
2310 
2311 	DEBUGNET_SET(dev, mlx4_en);
2312 
2313 	en_warn(priv, "Using %d TX rings\n", prof->tx_ring_num);
2314 	en_warn(priv, "Using %d RX rings\n", prof->rx_ring_num);
2315 
2316 	priv->registered = 1;
2317 
2318         en_warn(priv, "Using %d TX rings\n", prof->tx_ring_num);
2319         en_warn(priv, "Using %d RX rings\n", prof->rx_ring_num);
2320 
2321 
2322 	priv->rx_mb_size = dev->if_mtu + ETH_HLEN + VLAN_HLEN + ETH_FCS_LEN;
2323 	err = mlx4_SET_PORT_general(mdev->dev, priv->port,
2324 				    priv->rx_mb_size,
2325 				    prof->tx_pause, prof->tx_ppp,
2326 				    prof->rx_pause, prof->rx_ppp);
2327 	if (err) {
2328 		en_err(priv, "Failed setting port general configurations "
2329 		       "for port %d, with error %d\n", priv->port, err);
2330 		goto out;
2331 	}
2332 
2333 	/* Init port */
2334 	en_warn(priv, "Initializing port\n");
2335 	err = mlx4_INIT_PORT(mdev->dev, priv->port);
2336 	if (err) {
2337 		en_err(priv, "Failed Initializing port\n");
2338 		goto out;
2339 	}
2340 
2341 	queue_delayed_work(mdev->workqueue, &priv->stats_task, STATS_DELAY);
2342 
2343         if (mdev->dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_TS)
2344                 queue_delayed_work(mdev->workqueue, &priv->service_task, SERVICE_TASK_DELAY);
2345 
2346 	return 0;
2347 
2348 out:
2349 	mlx4_en_destroy_netdev(dev);
2350 	return err;
2351 }
2352 
2353 static int mlx4_en_set_ring_size(struct net_device *dev,
2354     int rx_size, int tx_size)
2355 {
2356         struct mlx4_en_priv *priv = netdev_priv(dev);
2357         struct mlx4_en_dev *mdev = priv->mdev;
2358         int port_up = 0;
2359         int err = 0;
2360 
2361         rx_size = roundup_pow_of_two(rx_size);
2362         rx_size = max_t(u32, rx_size, MLX4_EN_MIN_RX_SIZE);
2363         rx_size = min_t(u32, rx_size, MLX4_EN_MAX_RX_SIZE);
2364         tx_size = roundup_pow_of_two(tx_size);
2365         tx_size = max_t(u32, tx_size, MLX4_EN_MIN_TX_SIZE);
2366         tx_size = min_t(u32, tx_size, MLX4_EN_MAX_TX_SIZE);
2367 
2368         if (rx_size == (priv->port_up ?
2369             priv->rx_ring[0]->actual_size : priv->rx_ring[0]->size) &&
2370             tx_size == priv->tx_ring[0]->size)
2371                 return 0;
2372         mutex_lock(&mdev->state_lock);
2373         if (priv->port_up) {
2374                 port_up = 1;
2375                 mlx4_en_stop_port(dev);
2376         }
2377         mlx4_en_free_resources(priv);
2378         priv->prof->tx_ring_size = tx_size;
2379         priv->prof->rx_ring_size = rx_size;
2380         err = mlx4_en_alloc_resources(priv);
2381         if (err) {
2382                 en_err(priv, "Failed reallocating port resources\n");
2383                 goto out;
2384         }
2385         if (port_up) {
2386                 err = mlx4_en_start_port(dev);
2387                 if (err)
2388                         en_err(priv, "Failed starting port\n");
2389         }
2390 out:
2391         mutex_unlock(&mdev->state_lock);
2392         return err;
2393 }
2394 static int mlx4_en_set_rx_ring_size(SYSCTL_HANDLER_ARGS)
2395 {
2396         struct mlx4_en_priv *priv;
2397         int size;
2398         int error;
2399 
2400         priv = arg1;
2401         size = priv->prof->rx_ring_size;
2402         error = sysctl_handle_int(oidp, &size, 0, req);
2403         if (error || !req->newptr)
2404                 return (error);
2405         error = -mlx4_en_set_ring_size(priv->dev, size,
2406             priv->prof->tx_ring_size);
2407         return (error);
2408 }
2409 
2410 static int mlx4_en_set_tx_ring_size(SYSCTL_HANDLER_ARGS)
2411 {
2412         struct mlx4_en_priv *priv;
2413         int size;
2414         int error;
2415 
2416         priv = arg1;
2417         size = priv->prof->tx_ring_size;
2418         error = sysctl_handle_int(oidp, &size, 0, req);
2419         if (error || !req->newptr)
2420                 return (error);
2421         error = -mlx4_en_set_ring_size(priv->dev, priv->prof->rx_ring_size,
2422             size);
2423 
2424         return (error);
2425 }
2426 
2427 static int mlx4_en_get_module_info(struct net_device *dev,
2428 				   struct ethtool_modinfo *modinfo)
2429 {
2430 	struct mlx4_en_priv *priv = netdev_priv(dev);
2431 	struct mlx4_en_dev *mdev = priv->mdev;
2432 	int ret;
2433 	u8 data[4];
2434 
2435 	/* Read first 2 bytes to get Module & REV ID */
2436 	ret = mlx4_get_module_info(mdev->dev, priv->port,
2437 				   0/*offset*/, 2/*size*/, data);
2438 
2439 	if (ret < 2) {
2440 		en_err(priv, "Failed to read eeprom module first two bytes, error: 0x%x\n", -ret);
2441 		return -EIO;
2442 	}
2443 
2444 	switch (data[0] /* identifier */) {
2445 	case MLX4_MODULE_ID_QSFP:
2446 		modinfo->type = ETH_MODULE_SFF_8436;
2447 		modinfo->eeprom_len = ETH_MODULE_SFF_8436_LEN;
2448 		break;
2449 	case MLX4_MODULE_ID_QSFP_PLUS:
2450 		if (data[1] >= 0x3) { /* revision id */
2451 			modinfo->type = ETH_MODULE_SFF_8636;
2452 			modinfo->eeprom_len = ETH_MODULE_SFF_8636_LEN;
2453 		} else {
2454 			modinfo->type = ETH_MODULE_SFF_8436;
2455 			modinfo->eeprom_len = ETH_MODULE_SFF_8436_LEN;
2456 		}
2457 		break;
2458 	case MLX4_MODULE_ID_QSFP28:
2459 		modinfo->type = ETH_MODULE_SFF_8636;
2460 		modinfo->eeprom_len = ETH_MODULE_SFF_8636_LEN;
2461 		break;
2462 	case MLX4_MODULE_ID_SFP:
2463 		modinfo->type = ETH_MODULE_SFF_8472;
2464 		modinfo->eeprom_len = ETH_MODULE_SFF_8472_LEN;
2465 		break;
2466 	default:
2467 		en_err(priv, "mlx4_en_get_module_info :  Not recognized cable type\n");
2468 		return -EINVAL;
2469 	}
2470 
2471 	return 0;
2472 }
2473 
2474 static int mlx4_en_get_module_eeprom(struct net_device *dev,
2475 				     struct ethtool_eeprom *ee,
2476 				     u8 *data)
2477 {
2478 	struct mlx4_en_priv *priv = netdev_priv(dev);
2479 	struct mlx4_en_dev *mdev = priv->mdev;
2480 	int offset = ee->offset;
2481 	int i = 0, ret;
2482 
2483 	if (ee->len == 0)
2484 		return -EINVAL;
2485 
2486 	memset(data, 0, ee->len);
2487 
2488 	while (i < ee->len) {
2489 		en_dbg(DRV, priv,
2490 		       "mlx4_get_module_info i(%d) offset(%d) len(%d)\n",
2491 		       i, offset, ee->len - i);
2492 
2493 		ret = mlx4_get_module_info(mdev->dev, priv->port,
2494 					   offset, ee->len - i, data + i);
2495 
2496 		if (!ret) /* Done reading */
2497 			return 0;
2498 
2499 		if (ret < 0) {
2500 			en_err(priv,
2501 			       "mlx4_get_module_info i(%d) offset(%d) bytes_to_read(%d) - FAILED (0x%x)\n",
2502 			       i, offset, ee->len - i, ret);
2503 			return -1;
2504 		}
2505 
2506 		i += ret;
2507 		offset += ret;
2508 	}
2509 	return 0;
2510 }
2511 
2512 static void mlx4_en_print_eeprom(u8 *data, __u32 len)
2513 {
2514 	int		i;
2515 	int		j = 0;
2516 	int		row = 0;
2517 	const int	NUM_OF_BYTES = 16;
2518 
2519 	printf("\nOffset\t\tValues\n");
2520 	printf("------\t\t------\n");
2521 	while(row < len){
2522 		printf("0x%04x\t\t",row);
2523 		for(i=0; i < NUM_OF_BYTES; i++){
2524 			printf("%02x ", data[j]);
2525 			row++;
2526 			j++;
2527 		}
2528 		printf("\n");
2529 	}
2530 }
2531 
2532 /* Read cable EEPROM module information by first inspecting the first
2533  * two bytes to get the length and then read the rest of the information.
2534  * The information is printed to dmesg. */
2535 static int mlx4_en_read_eeprom(SYSCTL_HANDLER_ARGS)
2536 {
2537 
2538 	u8*		data;
2539 	int		error;
2540 	int		result = 0;
2541 	struct		mlx4_en_priv *priv;
2542 	struct		net_device *dev;
2543 	struct		ethtool_modinfo modinfo;
2544 	struct		ethtool_eeprom ee;
2545 
2546 	error = sysctl_handle_int(oidp, &result, 0, req);
2547 	if (error || !req->newptr)
2548 		return (error);
2549 
2550 	if (result == 1) {
2551 		priv = arg1;
2552 		dev = priv->dev;
2553 		data = kmalloc(PAGE_SIZE, GFP_KERNEL);
2554 
2555 		error = mlx4_en_get_module_info(dev, &modinfo);
2556 		if (error) {
2557 			en_err(priv,
2558 			       "mlx4_en_get_module_info returned with error - FAILED (0x%x)\n",
2559 			       -error);
2560 			goto out;
2561 		}
2562 
2563 		ee.len = modinfo.eeprom_len;
2564 		ee.offset = 0;
2565 
2566 		error = mlx4_en_get_module_eeprom(dev, &ee, data);
2567 		if (error) {
2568 			en_err(priv,
2569 			       "mlx4_en_get_module_eeprom returned with error - FAILED (0x%x)\n",
2570 			       -error);
2571 			/* Continue printing partial information in case of an error */
2572 		}
2573 
2574 		/* EEPROM information will be printed in dmesg */
2575 		mlx4_en_print_eeprom(data, ee.len);
2576 out:
2577 		kfree(data);
2578 	}
2579 	/* Return zero to prevent sysctl failure. */
2580 	return (0);
2581 }
2582 
2583 static int mlx4_en_set_tx_ppp(SYSCTL_HANDLER_ARGS)
2584 {
2585         struct mlx4_en_priv *priv;
2586         int ppp;
2587         int error;
2588 
2589         priv = arg1;
2590         ppp = priv->prof->tx_ppp;
2591         error = sysctl_handle_int(oidp, &ppp, 0, req);
2592         if (error || !req->newptr)
2593                 return (error);
2594         if (ppp > 0xff || ppp < 0)
2595                 return (-EINVAL);
2596         priv->prof->tx_ppp = ppp;
2597         error = -mlx4_SET_PORT_general(priv->mdev->dev, priv->port,
2598                                        priv->rx_mb_size + ETHER_CRC_LEN,
2599                                        priv->prof->tx_pause,
2600                                        priv->prof->tx_ppp,
2601                                        priv->prof->rx_pause,
2602                                        priv->prof->rx_ppp);
2603 
2604         return (error);
2605 }
2606 
2607 static int mlx4_en_set_rx_ppp(SYSCTL_HANDLER_ARGS)
2608 {
2609         struct mlx4_en_priv *priv;
2610         struct mlx4_en_dev *mdev;
2611         int ppp;
2612         int error;
2613         int port_up;
2614 
2615         port_up = 0;
2616         priv = arg1;
2617         mdev = priv->mdev;
2618         ppp = priv->prof->rx_ppp;
2619         error = sysctl_handle_int(oidp, &ppp, 0, req);
2620         if (error || !req->newptr)
2621                 return (error);
2622         if (ppp > 0xff || ppp < 0)
2623                 return (-EINVAL);
2624         /* See if we have to change the number of tx queues. */
2625         if (!ppp != !priv->prof->rx_ppp) {
2626                 mutex_lock(&mdev->state_lock);
2627                 if (priv->port_up) {
2628                         port_up = 1;
2629                         mlx4_en_stop_port(priv->dev);
2630                 }
2631                 mlx4_en_free_resources(priv);
2632                 priv->prof->rx_ppp = ppp;
2633                 error = -mlx4_en_alloc_resources(priv);
2634                 if (error)
2635                         en_err(priv, "Failed reallocating port resources\n");
2636                 if (error == 0 && port_up) {
2637                         error = -mlx4_en_start_port(priv->dev);
2638                         if (error)
2639                                 en_err(priv, "Failed starting port\n");
2640                 }
2641                 mutex_unlock(&mdev->state_lock);
2642                 return (error);
2643 
2644         }
2645         priv->prof->rx_ppp = ppp;
2646         error = -mlx4_SET_PORT_general(priv->mdev->dev, priv->port,
2647                                        priv->rx_mb_size + ETHER_CRC_LEN,
2648                                        priv->prof->tx_pause,
2649                                        priv->prof->tx_ppp,
2650                                        priv->prof->rx_pause,
2651                                        priv->prof->rx_ppp);
2652 
2653         return (error);
2654 }
2655 
2656 static void mlx4_en_sysctl_conf(struct mlx4_en_priv *priv)
2657 {
2658         struct net_device *dev;
2659         struct sysctl_ctx_list *ctx;
2660         struct sysctl_oid *node;
2661         struct sysctl_oid_list *node_list;
2662         struct sysctl_oid *coal;
2663         struct sysctl_oid_list *coal_list;
2664 	const char *pnameunit;
2665         dev = priv->dev;
2666         ctx = &priv->conf_ctx;
2667 	pnameunit = device_get_nameunit(priv->mdev->pdev->dev.bsddev);
2668 
2669         sysctl_ctx_init(ctx);
2670         priv->conf_sysctl = SYSCTL_ADD_NODE(ctx, SYSCTL_STATIC_CHILDREN(_hw),
2671             OID_AUTO, dev->if_xname, CTLFLAG_RD | CTLFLAG_MPSAFE, 0,
2672 	    "mlx4 10gig ethernet");
2673         node = SYSCTL_ADD_NODE(ctx, SYSCTL_CHILDREN(priv->conf_sysctl), OID_AUTO,
2674             "conf", CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, "Configuration");
2675         node_list = SYSCTL_CHILDREN(node);
2676 
2677         SYSCTL_ADD_UINT(ctx, node_list, OID_AUTO, "msg_enable",
2678             CTLFLAG_RW, &priv->msg_enable, 0,
2679             "Driver message enable bitfield");
2680         SYSCTL_ADD_UINT(ctx, node_list, OID_AUTO, "rx_rings",
2681             CTLFLAG_RD, &priv->rx_ring_num, 0,
2682             "Number of receive rings");
2683         SYSCTL_ADD_UINT(ctx, node_list, OID_AUTO, "tx_rings",
2684             CTLFLAG_RD, &priv->tx_ring_num, 0,
2685             "Number of transmit rings");
2686         SYSCTL_ADD_PROC(ctx, node_list, OID_AUTO, "rx_size",
2687             CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, priv, 0,
2688             mlx4_en_set_rx_ring_size, "I", "Receive ring size");
2689         SYSCTL_ADD_PROC(ctx, node_list, OID_AUTO, "tx_size",
2690             CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, priv, 0,
2691             mlx4_en_set_tx_ring_size, "I", "Transmit ring size");
2692         SYSCTL_ADD_PROC(ctx, node_list, OID_AUTO, "tx_ppp",
2693             CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, priv, 0,
2694             mlx4_en_set_tx_ppp, "I", "TX Per-priority pause");
2695         SYSCTL_ADD_PROC(ctx, node_list, OID_AUTO, "rx_ppp",
2696             CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, priv, 0,
2697             mlx4_en_set_rx_ppp, "I", "RX Per-priority pause");
2698         SYSCTL_ADD_UINT(ctx, node_list, OID_AUTO, "port_num",
2699             CTLFLAG_RD, &priv->port, 0,
2700             "Port Number");
2701         SYSCTL_ADD_STRING(ctx, node_list, OID_AUTO, "device_name",
2702 	    CTLFLAG_RD, __DECONST(void *, pnameunit), 0,
2703 	    "PCI device name");
2704         /* Add coalescer configuration. */
2705         coal = SYSCTL_ADD_NODE(ctx, node_list, OID_AUTO,
2706             "coalesce", CTLFLAG_RD | CTLFLAG_MPSAFE, NULL,
2707 	    "Interrupt coalesce configuration");
2708         coal_list = SYSCTL_CHILDREN(coal);
2709         SYSCTL_ADD_UINT(ctx, coal_list, OID_AUTO, "pkt_rate_low",
2710             CTLFLAG_RW, &priv->pkt_rate_low, 0,
2711             "Packets per-second for minimum delay");
2712         SYSCTL_ADD_UINT(ctx, coal_list, OID_AUTO, "rx_usecs_low",
2713             CTLFLAG_RW, &priv->rx_usecs_low, 0,
2714             "Minimum RX delay in micro-seconds");
2715         SYSCTL_ADD_UINT(ctx, coal_list, OID_AUTO, "pkt_rate_high",
2716             CTLFLAG_RW, &priv->pkt_rate_high, 0,
2717             "Packets per-second for maximum delay");
2718         SYSCTL_ADD_UINT(ctx, coal_list, OID_AUTO, "rx_usecs_high",
2719             CTLFLAG_RW, &priv->rx_usecs_high, 0,
2720             "Maximum RX delay in micro-seconds");
2721         SYSCTL_ADD_UINT(ctx, coal_list, OID_AUTO, "sample_interval",
2722             CTLFLAG_RW, &priv->sample_interval, 0,
2723             "adaptive frequency in units of HZ ticks");
2724         SYSCTL_ADD_UINT(ctx, coal_list, OID_AUTO, "adaptive_rx_coal",
2725             CTLFLAG_RW, &priv->adaptive_rx_coal, 0,
2726             "Enable adaptive rx coalescing");
2727 	/* EEPROM support */
2728 	SYSCTL_ADD_PROC(ctx, node_list, OID_AUTO, "eeprom_info",
2729 	    CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, priv, 0,
2730 	    mlx4_en_read_eeprom, "I", "EEPROM information");
2731 }
2732 
2733 static void mlx4_en_sysctl_stat(struct mlx4_en_priv *priv)
2734 {
2735 	struct sysctl_ctx_list *ctx;
2736 	struct sysctl_oid_list *node_list;
2737 	struct sysctl_oid *ring_node;
2738 	struct sysctl_oid_list *ring_list;
2739 	struct mlx4_en_tx_ring *tx_ring;
2740 	struct mlx4_en_rx_ring *rx_ring;
2741 	char namebuf[128];
2742 	int i;
2743 
2744 	ctx = &priv->stat_ctx;
2745 	sysctl_ctx_init(ctx);
2746 	priv->stat_sysctl = SYSCTL_ADD_NODE(ctx, SYSCTL_CHILDREN(priv->conf_sysctl), OID_AUTO,
2747 	    "stat", CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, "Statistics");
2748 	node_list = SYSCTL_CHILDREN(priv->stat_sysctl);
2749 
2750 #ifdef MLX4_EN_PERF_STAT
2751 	SYSCTL_ADD_UINT(ctx, node_list, OID_AUTO, "tx_poll", CTLFLAG_RD,
2752 	    &priv->pstats.tx_poll, "TX Poll calls");
2753 	SYSCTL_ADD_QUAD(ctx, node_list, OID_AUTO, "tx_pktsz_avg", CTLFLAG_RD,
2754 	    &priv->pstats.tx_pktsz_avg, "TX average packet size");
2755 	SYSCTL_ADD_UINT(ctx, node_list, OID_AUTO, "inflight_avg", CTLFLAG_RD,
2756 	    &priv->pstats.inflight_avg, "TX average packets in-flight");
2757 	SYSCTL_ADD_UINT(ctx, node_list, OID_AUTO, "tx_coal_avg", CTLFLAG_RD,
2758 	    &priv->pstats.tx_coal_avg, "TX average coalesced completions");
2759 	SYSCTL_ADD_UINT(ctx, node_list, OID_AUTO, "rx_coal_avg", CTLFLAG_RD,
2760 	    &priv->pstats.rx_coal_avg, "RX average coalesced completions");
2761 #endif
2762 
2763 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "tso_packets", CTLFLAG_RD,
2764 	    &priv->port_stats.tso_packets, 0, "TSO packets sent");
2765 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "queue_stopped", CTLFLAG_RD,
2766 	    &priv->port_stats.queue_stopped, 0, "Queue full");
2767 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "wake_queue", CTLFLAG_RD,
2768 	    &priv->port_stats.wake_queue, 0, "Queue resumed after full");
2769 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "tx_timeout", CTLFLAG_RD,
2770 	    &priv->port_stats.tx_timeout, 0, "Transmit timeouts");
2771 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "tx_oversized_packets", CTLFLAG_RD,
2772 	    &priv->port_stats.oversized_packets, 0, "TX oversized packets, m_defrag failed");
2773 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_alloc_failed", CTLFLAG_RD,
2774 	    &priv->port_stats.rx_alloc_failed, 0, "RX failed to allocate mbuf");
2775 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_chksum_good", CTLFLAG_RD,
2776             &priv->port_stats.rx_chksum_good, 0, "RX checksum offload success");
2777 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_chksum_none", CTLFLAG_RD,
2778 	    &priv->port_stats.rx_chksum_none, 0, "RX without checksum offload");
2779 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "tx_chksum_offload",
2780 	    CTLFLAG_RD, &priv->port_stats.tx_chksum_offload, 0,
2781 	    "TX checksum offloads");
2782 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "defrag_attempts",
2783 	    CTLFLAG_RD, &priv->port_stats.defrag_attempts, 0,
2784 	    "Oversized chains defragged");
2785 
2786 	/* Could strdup the names and add in a loop.  This is simpler. */
2787 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_bytes", CTLFLAG_RD,
2788 	    &priv->pkstats.rx_bytes, 0, "RX Bytes");
2789 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_packets", CTLFLAG_RD,
2790 	    &priv->pkstats.rx_packets, 0, "RX packets");
2791 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_multicast_packets", CTLFLAG_RD,
2792 	    &priv->pkstats.rx_multicast_packets, 0, "RX Multicast Packets");
2793 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_broadcast_packets", CTLFLAG_RD,
2794 	    &priv->pkstats.rx_broadcast_packets, 0, "RX Broadcast Packets");
2795 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_errors", CTLFLAG_RD,
2796 	    &priv->pkstats.rx_errors, 0, "RX Errors");
2797 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_dropped", CTLFLAG_RD,
2798 	    &priv->pkstats.rx_dropped, 0, "RX Dropped");
2799 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_length_errors", CTLFLAG_RD,
2800 	    &priv->pkstats.rx_length_errors, 0, "RX Length Errors");
2801 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_over_errors", CTLFLAG_RD,
2802 	    &priv->pkstats.rx_over_errors, 0, "RX Over Errors");
2803 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_crc_errors", CTLFLAG_RD,
2804 	    &priv->pkstats.rx_crc_errors, 0, "RX CRC Errors");
2805 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_jabbers", CTLFLAG_RD,
2806 	    &priv->pkstats.rx_jabbers, 0, "RX Jabbers");
2807 
2808 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_in_range_length_error", CTLFLAG_RD,
2809 	    &priv->pkstats.rx_in_range_length_error, 0, "RX IN_Range Length Error");
2810 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_out_range_length_error",
2811 	    CTLFLAG_RD, &priv->pkstats.rx_out_range_length_error, 0,
2812 	    "RX Out Range Length Error");
2813 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_lt_64_bytes_packets", CTLFLAG_RD,
2814 	    &priv->pkstats.rx_lt_64_bytes_packets, 0, "RX Lt 64 Bytes Packets");
2815 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_127_bytes_packets", CTLFLAG_RD,
2816 	    &priv->pkstats.rx_127_bytes_packets, 0, "RX 127 bytes Packets");
2817 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_255_bytes_packets", CTLFLAG_RD,
2818 	    &priv->pkstats.rx_255_bytes_packets, 0, "RX 255 bytes Packets");
2819 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_511_bytes_packets", CTLFLAG_RD,
2820 	    &priv->pkstats.rx_511_bytes_packets, 0, "RX 511 bytes Packets");
2821 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_1023_bytes_packets", CTLFLAG_RD,
2822 	    &priv->pkstats.rx_1023_bytes_packets, 0, "RX 1023 bytes Packets");
2823 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_1518_bytes_packets", CTLFLAG_RD,
2824 	    &priv->pkstats.rx_1518_bytes_packets, 0, "RX 1518 bytes Packets");
2825 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_1522_bytes_packets", CTLFLAG_RD,
2826 	    &priv->pkstats.rx_1522_bytes_packets, 0, "RX 1522 bytes Packets");
2827 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_1548_bytes_packets", CTLFLAG_RD,
2828 	    &priv->pkstats.rx_1548_bytes_packets, 0, "RX 1548 bytes Packets");
2829 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_gt_1548_bytes_packets", CTLFLAG_RD,
2830 	    &priv->pkstats.rx_gt_1548_bytes_packets, 0,
2831 	    "RX Greater Then 1548 bytes Packets");
2832 
2833 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "tx_packets", CTLFLAG_RD,
2834 	    &priv->pkstats.tx_packets, 0, "TX packets");
2835 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "tx_bytes", CTLFLAG_RD,
2836 	    &priv->pkstats.tx_bytes, 0, "TX Bytes");
2837 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "tx_multicast_packets", CTLFLAG_RD,
2838 	    &priv->pkstats.tx_multicast_packets, 0, "TX Multicast Packets");
2839 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "tx_broadcast_packets", CTLFLAG_RD,
2840 	    &priv->pkstats.tx_broadcast_packets, 0, "TX Broadcast Packets");
2841 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "tx_errors", CTLFLAG_RD,
2842 	    &priv->pkstats.tx_errors, 0, "TX Errors");
2843 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "tx_dropped", CTLFLAG_RD,
2844 	    &priv->pkstats.tx_dropped, 0, "TX Dropped");
2845 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "tx_lt_64_bytes_packets", CTLFLAG_RD,
2846 	    &priv->pkstats.tx_lt_64_bytes_packets, 0, "TX Less Then 64 Bytes Packets");
2847 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "tx_127_bytes_packets", CTLFLAG_RD,
2848 	    &priv->pkstats.tx_127_bytes_packets, 0, "TX 127 Bytes Packets");
2849 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "tx_255_bytes_packets", CTLFLAG_RD,
2850 	    &priv->pkstats.tx_255_bytes_packets, 0, "TX 255 Bytes Packets");
2851 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "tx_511_bytes_packets", CTLFLAG_RD,
2852 	    &priv->pkstats.tx_511_bytes_packets, 0, "TX 511 Bytes Packets");
2853 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "tx_1023_bytes_packets", CTLFLAG_RD,
2854 	    &priv->pkstats.tx_1023_bytes_packets, 0, "TX 1023 Bytes Packets");
2855 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "tx_1518_bytes_packets", CTLFLAG_RD,
2856 	    &priv->pkstats.tx_1518_bytes_packets, 0, "TX 1518 Bytes Packets");
2857 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "tx_1522_bytes_packets", CTLFLAG_RD,
2858 	    &priv->pkstats.tx_1522_bytes_packets, 0, "TX 1522 Bytes Packets");
2859 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "tx_1548_bytes_packets", CTLFLAG_RD,
2860 	    &priv->pkstats.tx_1548_bytes_packets, 0, "TX 1548 Bytes Packets");
2861 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "tx_gt_1548_bytes_packets", CTLFLAG_RD,
2862 	    &priv->pkstats.tx_gt_1548_bytes_packets, 0,
2863 	    "TX Greater Then 1548 Bytes Packets");
2864 
2865 	for (i = 0; i < priv->tx_ring_num; i++) {
2866 		tx_ring = priv->tx_ring[i];
2867 		snprintf(namebuf, sizeof(namebuf), "tx_ring%d", i);
2868 		ring_node = SYSCTL_ADD_NODE(ctx, node_list, OID_AUTO, namebuf,
2869 		    CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, "TX Ring");
2870 		ring_list = SYSCTL_CHILDREN(ring_node);
2871 		SYSCTL_ADD_U64(ctx, ring_list, OID_AUTO, "packets",
2872 		    CTLFLAG_RD, &tx_ring->packets, 0, "TX packets");
2873 		SYSCTL_ADD_U64(ctx, ring_list, OID_AUTO, "bytes",
2874 		    CTLFLAG_RD, &tx_ring->bytes, 0, "TX bytes");
2875 		SYSCTL_ADD_U64(ctx, ring_list, OID_AUTO, "tso_packets",
2876 		    CTLFLAG_RD, &tx_ring->tso_packets, 0, "TSO packets");
2877 		SYSCTL_ADD_U64(ctx, ring_list, OID_AUTO, "defrag_attempts",
2878 		    CTLFLAG_RD, &tx_ring->defrag_attempts, 0,
2879 		    "Oversized chains defragged");
2880 	}
2881 
2882 	for (i = 0; i < priv->rx_ring_num; i++) {
2883 		rx_ring = priv->rx_ring[i];
2884 		snprintf(namebuf, sizeof(namebuf), "rx_ring%d", i);
2885 		ring_node = SYSCTL_ADD_NODE(ctx, node_list, OID_AUTO, namebuf,
2886 		    CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, "RX Ring");
2887 		ring_list = SYSCTL_CHILDREN(ring_node);
2888 		SYSCTL_ADD_U64(ctx, ring_list, OID_AUTO, "packets",
2889 		    CTLFLAG_RD, &rx_ring->packets, 0, "RX packets");
2890 		SYSCTL_ADD_U64(ctx, ring_list, OID_AUTO, "bytes",
2891 		    CTLFLAG_RD, &rx_ring->bytes, 0, "RX bytes");
2892 		SYSCTL_ADD_U64(ctx, ring_list, OID_AUTO, "error",
2893 		    CTLFLAG_RD, &rx_ring->errors, 0, "RX soft errors");
2894 	}
2895 }
2896 
2897 #ifdef DEBUGNET
2898 static void
2899 mlx4_en_debugnet_init(struct ifnet *dev, int *nrxr, int *ncl, int *clsize)
2900 {
2901 	struct mlx4_en_priv *priv;
2902 
2903 	priv = if_getsoftc(dev);
2904 	mutex_lock(&priv->mdev->state_lock);
2905 	*nrxr = priv->rx_ring_num;
2906 	*ncl = DEBUGNET_MAX_IN_FLIGHT;
2907 	*clsize = priv->rx_mb_size;
2908 	mutex_unlock(&priv->mdev->state_lock);
2909 }
2910 
2911 static void
2912 mlx4_en_debugnet_event(struct ifnet *dev, enum debugnet_ev event)
2913 {
2914 }
2915 
2916 static int
2917 mlx4_en_debugnet_transmit(struct ifnet *dev, struct mbuf *m)
2918 {
2919 	struct mlx4_en_priv *priv;
2920 	int err;
2921 
2922 	priv = if_getsoftc(dev);
2923 	if ((if_getdrvflags(dev) & (IFF_DRV_RUNNING | IFF_DRV_OACTIVE)) !=
2924 	    IFF_DRV_RUNNING || !priv->link_state)
2925 		return (ENOENT);
2926 
2927 	err = mlx4_en_xmit(priv, 0, &m);
2928 	if (err != 0 && m != NULL)
2929 		m_freem(m);
2930 	return (err);
2931 }
2932 
2933 static int
2934 mlx4_en_debugnet_poll(struct ifnet *dev, int count)
2935 {
2936 	struct mlx4_en_priv *priv;
2937 
2938 	priv = if_getsoftc(dev);
2939 	if ((if_getdrvflags(dev) & IFF_DRV_RUNNING) == 0 || !priv->link_state)
2940 		return (ENOENT);
2941 
2942 	mlx4_poll_interrupts(priv->mdev->dev);
2943 
2944 	return (0);
2945 }
2946 #endif /* DEBUGNET */
2947