1 /*
2  * Copyright (c) 2007 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/tcp.h>
36 #include <linux/if_vlan.h>
37 #include <linux/delay.h>
38 #include <linux/slab.h>
39 #include <linux/hash.h>
40 #include <net/ip.h>
41 #include <net/busy_poll.h>
42 #include <net/vxlan.h>
43 
44 #include <linux/mlx4/driver.h>
45 #include <linux/mlx4/device.h>
46 #include <linux/mlx4/cmd.h>
47 #include <linux/mlx4/cq.h>
48 
49 #include "mlx4_en.h"
50 #include "en_port.h"
51 
52 int mlx4_en_setup_tc(struct net_device *dev, u8 up)
53 {
54 	struct mlx4_en_priv *priv = netdev_priv(dev);
55 	int i;
56 	unsigned int offset = 0;
57 
58 	if (up && up != MLX4_EN_NUM_UP)
59 		return -EINVAL;
60 
61 	netdev_set_num_tc(dev, up);
62 
63 	/* Partition Tx queues evenly amongst UP's */
64 	for (i = 0; i < up; i++) {
65 		netdev_set_tc_queue(dev, i, priv->num_tx_rings_p_up, offset);
66 		offset += priv->num_tx_rings_p_up;
67 	}
68 
69 	return 0;
70 }
71 
72 #ifdef CONFIG_NET_RX_BUSY_POLL
73 /* must be called with local_bh_disable()d */
74 static int mlx4_en_low_latency_recv(struct napi_struct *napi)
75 {
76 	struct mlx4_en_cq *cq = container_of(napi, struct mlx4_en_cq, napi);
77 	struct net_device *dev = cq->dev;
78 	struct mlx4_en_priv *priv = netdev_priv(dev);
79 	struct mlx4_en_rx_ring *rx_ring = priv->rx_ring[cq->ring];
80 	int done;
81 
82 	if (!priv->port_up)
83 		return LL_FLUSH_FAILED;
84 
85 	if (!mlx4_en_cq_lock_poll(cq))
86 		return LL_FLUSH_BUSY;
87 
88 	done = mlx4_en_process_rx_cq(dev, cq, 4);
89 	if (likely(done))
90 		rx_ring->cleaned += done;
91 	else
92 		rx_ring->misses++;
93 
94 	mlx4_en_cq_unlock_poll(cq);
95 
96 	return done;
97 }
98 #endif	/* CONFIG_NET_RX_BUSY_POLL */
99 
100 #ifdef CONFIG_RFS_ACCEL
101 
102 struct mlx4_en_filter {
103 	struct list_head next;
104 	struct work_struct work;
105 
106 	u8     ip_proto;
107 	__be32 src_ip;
108 	__be32 dst_ip;
109 	__be16 src_port;
110 	__be16 dst_port;
111 
112 	int rxq_index;
113 	struct mlx4_en_priv *priv;
114 	u32 flow_id;			/* RFS infrastructure id */
115 	int id;				/* mlx4_en driver id */
116 	u64 reg_id;			/* Flow steering API id */
117 	u8 activated;			/* Used to prevent expiry before filter
118 					 * is attached
119 					 */
120 	struct hlist_node filter_chain;
121 };
122 
123 static void mlx4_en_filter_rfs_expire(struct mlx4_en_priv *priv);
124 
125 static enum mlx4_net_trans_rule_id mlx4_ip_proto_to_trans_rule_id(u8 ip_proto)
126 {
127 	switch (ip_proto) {
128 	case IPPROTO_UDP:
129 		return MLX4_NET_TRANS_RULE_ID_UDP;
130 	case IPPROTO_TCP:
131 		return MLX4_NET_TRANS_RULE_ID_TCP;
132 	default:
133 		return MLX4_NET_TRANS_RULE_NUM;
134 	}
135 };
136 
137 static void mlx4_en_filter_work(struct work_struct *work)
138 {
139 	struct mlx4_en_filter *filter = container_of(work,
140 						     struct mlx4_en_filter,
141 						     work);
142 	struct mlx4_en_priv *priv = filter->priv;
143 	struct mlx4_spec_list spec_tcp_udp = {
144 		.id = mlx4_ip_proto_to_trans_rule_id(filter->ip_proto),
145 		{
146 			.tcp_udp = {
147 				.dst_port = filter->dst_port,
148 				.dst_port_msk = (__force __be16)-1,
149 				.src_port = filter->src_port,
150 				.src_port_msk = (__force __be16)-1,
151 			},
152 		},
153 	};
154 	struct mlx4_spec_list spec_ip = {
155 		.id = MLX4_NET_TRANS_RULE_ID_IPV4,
156 		{
157 			.ipv4 = {
158 				.dst_ip = filter->dst_ip,
159 				.dst_ip_msk = (__force __be32)-1,
160 				.src_ip = filter->src_ip,
161 				.src_ip_msk = (__force __be32)-1,
162 			},
163 		},
164 	};
165 	struct mlx4_spec_list spec_eth = {
166 		.id = MLX4_NET_TRANS_RULE_ID_ETH,
167 	};
168 	struct mlx4_net_trans_rule rule = {
169 		.list = LIST_HEAD_INIT(rule.list),
170 		.queue_mode = MLX4_NET_TRANS_Q_LIFO,
171 		.exclusive = 1,
172 		.allow_loopback = 1,
173 		.promisc_mode = MLX4_FS_REGULAR,
174 		.port = priv->port,
175 		.priority = MLX4_DOMAIN_RFS,
176 	};
177 	int rc;
178 	__be64 mac_mask = cpu_to_be64(MLX4_MAC_MASK << 16);
179 
180 	if (spec_tcp_udp.id >= MLX4_NET_TRANS_RULE_NUM) {
181 		en_warn(priv, "RFS: ignoring unsupported ip protocol (%d)\n",
182 			filter->ip_proto);
183 		goto ignore;
184 	}
185 	list_add_tail(&spec_eth.list, &rule.list);
186 	list_add_tail(&spec_ip.list, &rule.list);
187 	list_add_tail(&spec_tcp_udp.list, &rule.list);
188 
189 	rule.qpn = priv->rss_map.qps[filter->rxq_index].qpn;
190 	memcpy(spec_eth.eth.dst_mac, priv->dev->dev_addr, ETH_ALEN);
191 	memcpy(spec_eth.eth.dst_mac_msk, &mac_mask, ETH_ALEN);
192 
193 	filter->activated = 0;
194 
195 	if (filter->reg_id) {
196 		rc = mlx4_flow_detach(priv->mdev->dev, filter->reg_id);
197 		if (rc && rc != -ENOENT)
198 			en_err(priv, "Error detaching flow. rc = %d\n", rc);
199 	}
200 
201 	rc = mlx4_flow_attach(priv->mdev->dev, &rule, &filter->reg_id);
202 	if (rc)
203 		en_err(priv, "Error attaching flow. err = %d\n", rc);
204 
205 ignore:
206 	mlx4_en_filter_rfs_expire(priv);
207 
208 	filter->activated = 1;
209 }
210 
211 static inline struct hlist_head *
212 filter_hash_bucket(struct mlx4_en_priv *priv, __be32 src_ip, __be32 dst_ip,
213 		   __be16 src_port, __be16 dst_port)
214 {
215 	unsigned long l;
216 	int bucket_idx;
217 
218 	l = (__force unsigned long)src_port |
219 	    ((__force unsigned long)dst_port << 2);
220 	l ^= (__force unsigned long)(src_ip ^ dst_ip);
221 
222 	bucket_idx = hash_long(l, MLX4_EN_FILTER_HASH_SHIFT);
223 
224 	return &priv->filter_hash[bucket_idx];
225 }
226 
227 static struct mlx4_en_filter *
228 mlx4_en_filter_alloc(struct mlx4_en_priv *priv, int rxq_index, __be32 src_ip,
229 		     __be32 dst_ip, u8 ip_proto, __be16 src_port,
230 		     __be16 dst_port, u32 flow_id)
231 {
232 	struct mlx4_en_filter *filter = NULL;
233 
234 	filter = kzalloc(sizeof(struct mlx4_en_filter), GFP_ATOMIC);
235 	if (!filter)
236 		return NULL;
237 
238 	filter->priv = priv;
239 	filter->rxq_index = rxq_index;
240 	INIT_WORK(&filter->work, mlx4_en_filter_work);
241 
242 	filter->src_ip = src_ip;
243 	filter->dst_ip = dst_ip;
244 	filter->ip_proto = ip_proto;
245 	filter->src_port = src_port;
246 	filter->dst_port = dst_port;
247 
248 	filter->flow_id = flow_id;
249 
250 	filter->id = priv->last_filter_id++ % RPS_NO_FILTER;
251 
252 	list_add_tail(&filter->next, &priv->filters);
253 	hlist_add_head(&filter->filter_chain,
254 		       filter_hash_bucket(priv, src_ip, dst_ip, src_port,
255 					  dst_port));
256 
257 	return filter;
258 }
259 
260 static void mlx4_en_filter_free(struct mlx4_en_filter *filter)
261 {
262 	struct mlx4_en_priv *priv = filter->priv;
263 	int rc;
264 
265 	list_del(&filter->next);
266 
267 	rc = mlx4_flow_detach(priv->mdev->dev, filter->reg_id);
268 	if (rc && rc != -ENOENT)
269 		en_err(priv, "Error detaching flow. rc = %d\n", rc);
270 
271 	kfree(filter);
272 }
273 
274 static inline struct mlx4_en_filter *
275 mlx4_en_filter_find(struct mlx4_en_priv *priv, __be32 src_ip, __be32 dst_ip,
276 		    u8 ip_proto, __be16 src_port, __be16 dst_port)
277 {
278 	struct mlx4_en_filter *filter;
279 	struct mlx4_en_filter *ret = NULL;
280 
281 	hlist_for_each_entry(filter,
282 			     filter_hash_bucket(priv, src_ip, dst_ip,
283 						src_port, dst_port),
284 			     filter_chain) {
285 		if (filter->src_ip == src_ip &&
286 		    filter->dst_ip == dst_ip &&
287 		    filter->ip_proto == ip_proto &&
288 		    filter->src_port == src_port &&
289 		    filter->dst_port == dst_port) {
290 			ret = filter;
291 			break;
292 		}
293 	}
294 
295 	return ret;
296 }
297 
298 static int
299 mlx4_en_filter_rfs(struct net_device *net_dev, const struct sk_buff *skb,
300 		   u16 rxq_index, u32 flow_id)
301 {
302 	struct mlx4_en_priv *priv = netdev_priv(net_dev);
303 	struct mlx4_en_filter *filter;
304 	const struct iphdr *ip;
305 	const __be16 *ports;
306 	u8 ip_proto;
307 	__be32 src_ip;
308 	__be32 dst_ip;
309 	__be16 src_port;
310 	__be16 dst_port;
311 	int nhoff = skb_network_offset(skb);
312 	int ret = 0;
313 
314 	if (skb->protocol != htons(ETH_P_IP))
315 		return -EPROTONOSUPPORT;
316 
317 	ip = (const struct iphdr *)(skb->data + nhoff);
318 	if (ip_is_fragment(ip))
319 		return -EPROTONOSUPPORT;
320 
321 	if ((ip->protocol != IPPROTO_TCP) && (ip->protocol != IPPROTO_UDP))
322 		return -EPROTONOSUPPORT;
323 	ports = (const __be16 *)(skb->data + nhoff + 4 * ip->ihl);
324 
325 	ip_proto = ip->protocol;
326 	src_ip = ip->saddr;
327 	dst_ip = ip->daddr;
328 	src_port = ports[0];
329 	dst_port = ports[1];
330 
331 	spin_lock_bh(&priv->filters_lock);
332 	filter = mlx4_en_filter_find(priv, src_ip, dst_ip, ip_proto,
333 				     src_port, dst_port);
334 	if (filter) {
335 		if (filter->rxq_index == rxq_index)
336 			goto out;
337 
338 		filter->rxq_index = rxq_index;
339 	} else {
340 		filter = mlx4_en_filter_alloc(priv, rxq_index,
341 					      src_ip, dst_ip, ip_proto,
342 					      src_port, dst_port, flow_id);
343 		if (!filter) {
344 			ret = -ENOMEM;
345 			goto err;
346 		}
347 	}
348 
349 	queue_work(priv->mdev->workqueue, &filter->work);
350 
351 out:
352 	ret = filter->id;
353 err:
354 	spin_unlock_bh(&priv->filters_lock);
355 
356 	return ret;
357 }
358 
359 void mlx4_en_cleanup_filters(struct mlx4_en_priv *priv)
360 {
361 	struct mlx4_en_filter *filter, *tmp;
362 	LIST_HEAD(del_list);
363 
364 	spin_lock_bh(&priv->filters_lock);
365 	list_for_each_entry_safe(filter, tmp, &priv->filters, next) {
366 		list_move(&filter->next, &del_list);
367 		hlist_del(&filter->filter_chain);
368 	}
369 	spin_unlock_bh(&priv->filters_lock);
370 
371 	list_for_each_entry_safe(filter, tmp, &del_list, next) {
372 		cancel_work_sync(&filter->work);
373 		mlx4_en_filter_free(filter);
374 	}
375 }
376 
377 static void mlx4_en_filter_rfs_expire(struct mlx4_en_priv *priv)
378 {
379 	struct mlx4_en_filter *filter = NULL, *tmp, *last_filter = NULL;
380 	LIST_HEAD(del_list);
381 	int i = 0;
382 
383 	spin_lock_bh(&priv->filters_lock);
384 	list_for_each_entry_safe(filter, tmp, &priv->filters, next) {
385 		if (i > MLX4_EN_FILTER_EXPIRY_QUOTA)
386 			break;
387 
388 		if (filter->activated &&
389 		    !work_pending(&filter->work) &&
390 		    rps_may_expire_flow(priv->dev,
391 					filter->rxq_index, filter->flow_id,
392 					filter->id)) {
393 			list_move(&filter->next, &del_list);
394 			hlist_del(&filter->filter_chain);
395 		} else
396 			last_filter = filter;
397 
398 		i++;
399 	}
400 
401 	if (last_filter && (&last_filter->next != priv->filters.next))
402 		list_move(&priv->filters, &last_filter->next);
403 
404 	spin_unlock_bh(&priv->filters_lock);
405 
406 	list_for_each_entry_safe(filter, tmp, &del_list, next)
407 		mlx4_en_filter_free(filter);
408 }
409 #endif
410 
411 static int mlx4_en_vlan_rx_add_vid(struct net_device *dev,
412 				   __be16 proto, u16 vid)
413 {
414 	struct mlx4_en_priv *priv = netdev_priv(dev);
415 	struct mlx4_en_dev *mdev = priv->mdev;
416 	int err;
417 	int idx;
418 
419 	en_dbg(HW, priv, "adding VLAN:%d\n", vid);
420 
421 	set_bit(vid, priv->active_vlans);
422 
423 	/* Add VID to port VLAN filter */
424 	mutex_lock(&mdev->state_lock);
425 	if (mdev->device_up && priv->port_up) {
426 		err = mlx4_SET_VLAN_FLTR(mdev->dev, priv);
427 		if (err)
428 			en_err(priv, "Failed configuring VLAN filter\n");
429 	}
430 	if (mlx4_register_vlan(mdev->dev, priv->port, vid, &idx))
431 		en_dbg(HW, priv, "failed adding vlan %d\n", vid);
432 	mutex_unlock(&mdev->state_lock);
433 
434 	return 0;
435 }
436 
437 static int mlx4_en_vlan_rx_kill_vid(struct net_device *dev,
438 				    __be16 proto, u16 vid)
439 {
440 	struct mlx4_en_priv *priv = netdev_priv(dev);
441 	struct mlx4_en_dev *mdev = priv->mdev;
442 	int err;
443 
444 	en_dbg(HW, priv, "Killing VID:%d\n", vid);
445 
446 	clear_bit(vid, priv->active_vlans);
447 
448 	/* Remove VID from port VLAN filter */
449 	mutex_lock(&mdev->state_lock);
450 	mlx4_unregister_vlan(mdev->dev, priv->port, vid);
451 
452 	if (mdev->device_up && priv->port_up) {
453 		err = mlx4_SET_VLAN_FLTR(mdev->dev, priv);
454 		if (err)
455 			en_err(priv, "Failed configuring VLAN filter\n");
456 	}
457 	mutex_unlock(&mdev->state_lock);
458 
459 	return 0;
460 }
461 
462 static void mlx4_en_u64_to_mac(unsigned char dst_mac[ETH_ALEN + 2], u64 src_mac)
463 {
464 	int i;
465 	for (i = ETH_ALEN - 1; i >= 0; --i) {
466 		dst_mac[i] = src_mac & 0xff;
467 		src_mac >>= 8;
468 	}
469 	memset(&dst_mac[ETH_ALEN], 0, 2);
470 }
471 
472 
473 static int mlx4_en_tunnel_steer_add(struct mlx4_en_priv *priv, unsigned char *addr,
474 				    int qpn, u64 *reg_id)
475 {
476 	int err;
477 
478 	if (priv->mdev->dev->caps.tunnel_offload_mode != MLX4_TUNNEL_OFFLOAD_MODE_VXLAN ||
479 	    priv->mdev->dev->caps.dmfs_high_steer_mode == MLX4_STEERING_DMFS_A0_STATIC)
480 		return 0; /* do nothing */
481 
482 	err = mlx4_tunnel_steer_add(priv->mdev->dev, addr, priv->port, qpn,
483 				    MLX4_DOMAIN_NIC, reg_id);
484 	if (err) {
485 		en_err(priv, "failed to add vxlan steering rule, err %d\n", err);
486 		return err;
487 	}
488 	en_dbg(DRV, priv, "added vxlan steering rule, mac %pM reg_id %llx\n", addr, *reg_id);
489 	return 0;
490 }
491 
492 
493 static int mlx4_en_uc_steer_add(struct mlx4_en_priv *priv,
494 				unsigned char *mac, int *qpn, u64 *reg_id)
495 {
496 	struct mlx4_en_dev *mdev = priv->mdev;
497 	struct mlx4_dev *dev = mdev->dev;
498 	int err;
499 
500 	switch (dev->caps.steering_mode) {
501 	case MLX4_STEERING_MODE_B0: {
502 		struct mlx4_qp qp;
503 		u8 gid[16] = {0};
504 
505 		qp.qpn = *qpn;
506 		memcpy(&gid[10], mac, ETH_ALEN);
507 		gid[5] = priv->port;
508 
509 		err = mlx4_unicast_attach(dev, &qp, gid, 0, MLX4_PROT_ETH);
510 		break;
511 	}
512 	case MLX4_STEERING_MODE_DEVICE_MANAGED: {
513 		struct mlx4_spec_list spec_eth = { {NULL} };
514 		__be64 mac_mask = cpu_to_be64(MLX4_MAC_MASK << 16);
515 
516 		struct mlx4_net_trans_rule rule = {
517 			.queue_mode = MLX4_NET_TRANS_Q_FIFO,
518 			.exclusive = 0,
519 			.allow_loopback = 1,
520 			.promisc_mode = MLX4_FS_REGULAR,
521 			.priority = MLX4_DOMAIN_NIC,
522 		};
523 
524 		rule.port = priv->port;
525 		rule.qpn = *qpn;
526 		INIT_LIST_HEAD(&rule.list);
527 
528 		spec_eth.id = MLX4_NET_TRANS_RULE_ID_ETH;
529 		memcpy(spec_eth.eth.dst_mac, mac, ETH_ALEN);
530 		memcpy(spec_eth.eth.dst_mac_msk, &mac_mask, ETH_ALEN);
531 		list_add_tail(&spec_eth.list, &rule.list);
532 
533 		err = mlx4_flow_attach(dev, &rule, reg_id);
534 		break;
535 	}
536 	default:
537 		return -EINVAL;
538 	}
539 	if (err)
540 		en_warn(priv, "Failed Attaching Unicast\n");
541 
542 	return err;
543 }
544 
545 static void mlx4_en_uc_steer_release(struct mlx4_en_priv *priv,
546 				     unsigned char *mac, int qpn, u64 reg_id)
547 {
548 	struct mlx4_en_dev *mdev = priv->mdev;
549 	struct mlx4_dev *dev = mdev->dev;
550 
551 	switch (dev->caps.steering_mode) {
552 	case MLX4_STEERING_MODE_B0: {
553 		struct mlx4_qp qp;
554 		u8 gid[16] = {0};
555 
556 		qp.qpn = qpn;
557 		memcpy(&gid[10], mac, ETH_ALEN);
558 		gid[5] = priv->port;
559 
560 		mlx4_unicast_detach(dev, &qp, gid, MLX4_PROT_ETH);
561 		break;
562 	}
563 	case MLX4_STEERING_MODE_DEVICE_MANAGED: {
564 		mlx4_flow_detach(dev, reg_id);
565 		break;
566 	}
567 	default:
568 		en_err(priv, "Invalid steering mode.\n");
569 	}
570 }
571 
572 static int mlx4_en_get_qp(struct mlx4_en_priv *priv)
573 {
574 	struct mlx4_en_dev *mdev = priv->mdev;
575 	struct mlx4_dev *dev = mdev->dev;
576 	int index = 0;
577 	int err = 0;
578 	int *qpn = &priv->base_qpn;
579 	u64 mac = mlx4_mac_to_u64(priv->dev->dev_addr);
580 
581 	en_dbg(DRV, priv, "Registering MAC: %pM for adding\n",
582 	       priv->dev->dev_addr);
583 	index = mlx4_register_mac(dev, priv->port, mac);
584 	if (index < 0) {
585 		err = index;
586 		en_err(priv, "Failed adding MAC: %pM\n",
587 		       priv->dev->dev_addr);
588 		return err;
589 	}
590 
591 	if (dev->caps.steering_mode == MLX4_STEERING_MODE_A0) {
592 		int base_qpn = mlx4_get_base_qpn(dev, priv->port);
593 		*qpn = base_qpn + index;
594 		return 0;
595 	}
596 
597 	err = mlx4_qp_reserve_range(dev, 1, 1, qpn, MLX4_RESERVE_A0_QP);
598 	en_dbg(DRV, priv, "Reserved qp %d\n", *qpn);
599 	if (err) {
600 		en_err(priv, "Failed to reserve qp for mac registration\n");
601 		mlx4_unregister_mac(dev, priv->port, mac);
602 		return err;
603 	}
604 
605 	return 0;
606 }
607 
608 static void mlx4_en_put_qp(struct mlx4_en_priv *priv)
609 {
610 	struct mlx4_en_dev *mdev = priv->mdev;
611 	struct mlx4_dev *dev = mdev->dev;
612 	int qpn = priv->base_qpn;
613 
614 	if (dev->caps.steering_mode == MLX4_STEERING_MODE_A0) {
615 		u64 mac = mlx4_mac_to_u64(priv->dev->dev_addr);
616 		en_dbg(DRV, priv, "Registering MAC: %pM for deleting\n",
617 		       priv->dev->dev_addr);
618 		mlx4_unregister_mac(dev, priv->port, mac);
619 	} else {
620 		en_dbg(DRV, priv, "Releasing qp: port %d, qpn %d\n",
621 		       priv->port, qpn);
622 		mlx4_qp_release_range(dev, qpn, 1);
623 		priv->flags &= ~MLX4_EN_FLAG_FORCE_PROMISC;
624 	}
625 }
626 
627 static int mlx4_en_replace_mac(struct mlx4_en_priv *priv, int qpn,
628 			       unsigned char *new_mac, unsigned char *prev_mac)
629 {
630 	struct mlx4_en_dev *mdev = priv->mdev;
631 	struct mlx4_dev *dev = mdev->dev;
632 	int err = 0;
633 	u64 new_mac_u64 = mlx4_mac_to_u64(new_mac);
634 
635 	if (dev->caps.steering_mode != MLX4_STEERING_MODE_A0) {
636 		struct hlist_head *bucket;
637 		unsigned int mac_hash;
638 		struct mlx4_mac_entry *entry;
639 		struct hlist_node *tmp;
640 		u64 prev_mac_u64 = mlx4_mac_to_u64(prev_mac);
641 
642 		bucket = &priv->mac_hash[prev_mac[MLX4_EN_MAC_HASH_IDX]];
643 		hlist_for_each_entry_safe(entry, tmp, bucket, hlist) {
644 			if (ether_addr_equal_64bits(entry->mac, prev_mac)) {
645 				mlx4_en_uc_steer_release(priv, entry->mac,
646 							 qpn, entry->reg_id);
647 				mlx4_unregister_mac(dev, priv->port,
648 						    prev_mac_u64);
649 				hlist_del_rcu(&entry->hlist);
650 				synchronize_rcu();
651 				memcpy(entry->mac, new_mac, ETH_ALEN);
652 				entry->reg_id = 0;
653 				mac_hash = new_mac[MLX4_EN_MAC_HASH_IDX];
654 				hlist_add_head_rcu(&entry->hlist,
655 						   &priv->mac_hash[mac_hash]);
656 				mlx4_register_mac(dev, priv->port, new_mac_u64);
657 				err = mlx4_en_uc_steer_add(priv, new_mac,
658 							   &qpn,
659 							   &entry->reg_id);
660 				if (err)
661 					return err;
662 				if (priv->tunnel_reg_id) {
663 					mlx4_flow_detach(priv->mdev->dev, priv->tunnel_reg_id);
664 					priv->tunnel_reg_id = 0;
665 				}
666 				err = mlx4_en_tunnel_steer_add(priv, new_mac, qpn,
667 							       &priv->tunnel_reg_id);
668 				return err;
669 			}
670 		}
671 		return -EINVAL;
672 	}
673 
674 	return __mlx4_replace_mac(dev, priv->port, qpn, new_mac_u64);
675 }
676 
677 static int mlx4_en_do_set_mac(struct mlx4_en_priv *priv,
678 			      unsigned char new_mac[ETH_ALEN + 2])
679 {
680 	int err = 0;
681 
682 	if (priv->port_up) {
683 		/* Remove old MAC and insert the new one */
684 		err = mlx4_en_replace_mac(priv, priv->base_qpn,
685 					  new_mac, priv->current_mac);
686 		if (err)
687 			en_err(priv, "Failed changing HW MAC address\n");
688 	} else
689 		en_dbg(HW, priv, "Port is down while registering mac, exiting...\n");
690 
691 	if (!err)
692 		memcpy(priv->current_mac, new_mac, sizeof(priv->current_mac));
693 
694 	return err;
695 }
696 
697 static int mlx4_en_set_mac(struct net_device *dev, void *addr)
698 {
699 	struct mlx4_en_priv *priv = netdev_priv(dev);
700 	struct mlx4_en_dev *mdev = priv->mdev;
701 	struct sockaddr *saddr = addr;
702 	unsigned char new_mac[ETH_ALEN + 2];
703 	int err;
704 
705 	if (!is_valid_ether_addr(saddr->sa_data))
706 		return -EADDRNOTAVAIL;
707 
708 	mutex_lock(&mdev->state_lock);
709 	memcpy(new_mac, saddr->sa_data, ETH_ALEN);
710 	err = mlx4_en_do_set_mac(priv, new_mac);
711 	if (!err)
712 		memcpy(dev->dev_addr, saddr->sa_data, ETH_ALEN);
713 	mutex_unlock(&mdev->state_lock);
714 
715 	return err;
716 }
717 
718 static void mlx4_en_clear_list(struct net_device *dev)
719 {
720 	struct mlx4_en_priv *priv = netdev_priv(dev);
721 	struct mlx4_en_mc_list *tmp, *mc_to_del;
722 
723 	list_for_each_entry_safe(mc_to_del, tmp, &priv->mc_list, list) {
724 		list_del(&mc_to_del->list);
725 		kfree(mc_to_del);
726 	}
727 }
728 
729 static void mlx4_en_cache_mclist(struct net_device *dev)
730 {
731 	struct mlx4_en_priv *priv = netdev_priv(dev);
732 	struct netdev_hw_addr *ha;
733 	struct mlx4_en_mc_list *tmp;
734 
735 	mlx4_en_clear_list(dev);
736 	netdev_for_each_mc_addr(ha, dev) {
737 		tmp = kzalloc(sizeof(struct mlx4_en_mc_list), GFP_ATOMIC);
738 		if (!tmp) {
739 			mlx4_en_clear_list(dev);
740 			return;
741 		}
742 		memcpy(tmp->addr, ha->addr, ETH_ALEN);
743 		list_add_tail(&tmp->list, &priv->mc_list);
744 	}
745 }
746 
747 static void update_mclist_flags(struct mlx4_en_priv *priv,
748 				struct list_head *dst,
749 				struct list_head *src)
750 {
751 	struct mlx4_en_mc_list *dst_tmp, *src_tmp, *new_mc;
752 	bool found;
753 
754 	/* Find all the entries that should be removed from dst,
755 	 * These are the entries that are not found in src
756 	 */
757 	list_for_each_entry(dst_tmp, dst, list) {
758 		found = false;
759 		list_for_each_entry(src_tmp, src, list) {
760 			if (ether_addr_equal(dst_tmp->addr, src_tmp->addr)) {
761 				found = true;
762 				break;
763 			}
764 		}
765 		if (!found)
766 			dst_tmp->action = MCLIST_REM;
767 	}
768 
769 	/* Add entries that exist in src but not in dst
770 	 * mark them as need to add
771 	 */
772 	list_for_each_entry(src_tmp, src, list) {
773 		found = false;
774 		list_for_each_entry(dst_tmp, dst, list) {
775 			if (ether_addr_equal(dst_tmp->addr, src_tmp->addr)) {
776 				dst_tmp->action = MCLIST_NONE;
777 				found = true;
778 				break;
779 			}
780 		}
781 		if (!found) {
782 			new_mc = kmemdup(src_tmp,
783 					 sizeof(struct mlx4_en_mc_list),
784 					 GFP_KERNEL);
785 			if (!new_mc)
786 				return;
787 
788 			new_mc->action = MCLIST_ADD;
789 			list_add_tail(&new_mc->list, dst);
790 		}
791 	}
792 }
793 
794 static void mlx4_en_set_rx_mode(struct net_device *dev)
795 {
796 	struct mlx4_en_priv *priv = netdev_priv(dev);
797 
798 	if (!priv->port_up)
799 		return;
800 
801 	queue_work(priv->mdev->workqueue, &priv->rx_mode_task);
802 }
803 
804 static void mlx4_en_set_promisc_mode(struct mlx4_en_priv *priv,
805 				     struct mlx4_en_dev *mdev)
806 {
807 	int err = 0;
808 
809 	if (!(priv->flags & MLX4_EN_FLAG_PROMISC)) {
810 		if (netif_msg_rx_status(priv))
811 			en_warn(priv, "Entering promiscuous mode\n");
812 		priv->flags |= MLX4_EN_FLAG_PROMISC;
813 
814 		/* Enable promiscouos mode */
815 		switch (mdev->dev->caps.steering_mode) {
816 		case MLX4_STEERING_MODE_DEVICE_MANAGED:
817 			err = mlx4_flow_steer_promisc_add(mdev->dev,
818 							  priv->port,
819 							  priv->base_qpn,
820 							  MLX4_FS_ALL_DEFAULT);
821 			if (err)
822 				en_err(priv, "Failed enabling promiscuous mode\n");
823 			priv->flags |= MLX4_EN_FLAG_MC_PROMISC;
824 			break;
825 
826 		case MLX4_STEERING_MODE_B0:
827 			err = mlx4_unicast_promisc_add(mdev->dev,
828 						       priv->base_qpn,
829 						       priv->port);
830 			if (err)
831 				en_err(priv, "Failed enabling unicast promiscuous mode\n");
832 
833 			/* Add the default qp number as multicast
834 			 * promisc
835 			 */
836 			if (!(priv->flags & MLX4_EN_FLAG_MC_PROMISC)) {
837 				err = mlx4_multicast_promisc_add(mdev->dev,
838 								 priv->base_qpn,
839 								 priv->port);
840 				if (err)
841 					en_err(priv, "Failed enabling multicast promiscuous mode\n");
842 				priv->flags |= MLX4_EN_FLAG_MC_PROMISC;
843 			}
844 			break;
845 
846 		case MLX4_STEERING_MODE_A0:
847 			err = mlx4_SET_PORT_qpn_calc(mdev->dev,
848 						     priv->port,
849 						     priv->base_qpn,
850 						     1);
851 			if (err)
852 				en_err(priv, "Failed enabling promiscuous mode\n");
853 			break;
854 		}
855 
856 		/* Disable port multicast filter (unconditionally) */
857 		err = mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, 0,
858 					  0, MLX4_MCAST_DISABLE);
859 		if (err)
860 			en_err(priv, "Failed disabling multicast filter\n");
861 	}
862 }
863 
864 static void mlx4_en_clear_promisc_mode(struct mlx4_en_priv *priv,
865 				       struct mlx4_en_dev *mdev)
866 {
867 	int err = 0;
868 
869 	if (netif_msg_rx_status(priv))
870 		en_warn(priv, "Leaving promiscuous mode\n");
871 	priv->flags &= ~MLX4_EN_FLAG_PROMISC;
872 
873 	/* Disable promiscouos mode */
874 	switch (mdev->dev->caps.steering_mode) {
875 	case MLX4_STEERING_MODE_DEVICE_MANAGED:
876 		err = mlx4_flow_steer_promisc_remove(mdev->dev,
877 						     priv->port,
878 						     MLX4_FS_ALL_DEFAULT);
879 		if (err)
880 			en_err(priv, "Failed disabling promiscuous mode\n");
881 		priv->flags &= ~MLX4_EN_FLAG_MC_PROMISC;
882 		break;
883 
884 	case MLX4_STEERING_MODE_B0:
885 		err = mlx4_unicast_promisc_remove(mdev->dev,
886 						  priv->base_qpn,
887 						  priv->port);
888 		if (err)
889 			en_err(priv, "Failed disabling unicast promiscuous mode\n");
890 		/* Disable Multicast promisc */
891 		if (priv->flags & MLX4_EN_FLAG_MC_PROMISC) {
892 			err = mlx4_multicast_promisc_remove(mdev->dev,
893 							    priv->base_qpn,
894 							    priv->port);
895 			if (err)
896 				en_err(priv, "Failed disabling multicast promiscuous mode\n");
897 			priv->flags &= ~MLX4_EN_FLAG_MC_PROMISC;
898 		}
899 		break;
900 
901 	case MLX4_STEERING_MODE_A0:
902 		err = mlx4_SET_PORT_qpn_calc(mdev->dev,
903 					     priv->port,
904 					     priv->base_qpn, 0);
905 		if (err)
906 			en_err(priv, "Failed disabling promiscuous mode\n");
907 		break;
908 	}
909 }
910 
911 static void mlx4_en_do_multicast(struct mlx4_en_priv *priv,
912 				 struct net_device *dev,
913 				 struct mlx4_en_dev *mdev)
914 {
915 	struct mlx4_en_mc_list *mclist, *tmp;
916 	u64 mcast_addr = 0;
917 	u8 mc_list[16] = {0};
918 	int err = 0;
919 
920 	/* Enable/disable the multicast filter according to IFF_ALLMULTI */
921 	if (dev->flags & IFF_ALLMULTI) {
922 		err = mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, 0,
923 					  0, MLX4_MCAST_DISABLE);
924 		if (err)
925 			en_err(priv, "Failed disabling multicast filter\n");
926 
927 		/* Add the default qp number as multicast promisc */
928 		if (!(priv->flags & MLX4_EN_FLAG_MC_PROMISC)) {
929 			switch (mdev->dev->caps.steering_mode) {
930 			case MLX4_STEERING_MODE_DEVICE_MANAGED:
931 				err = mlx4_flow_steer_promisc_add(mdev->dev,
932 								  priv->port,
933 								  priv->base_qpn,
934 								  MLX4_FS_MC_DEFAULT);
935 				break;
936 
937 			case MLX4_STEERING_MODE_B0:
938 				err = mlx4_multicast_promisc_add(mdev->dev,
939 								 priv->base_qpn,
940 								 priv->port);
941 				break;
942 
943 			case MLX4_STEERING_MODE_A0:
944 				break;
945 			}
946 			if (err)
947 				en_err(priv, "Failed entering multicast promisc mode\n");
948 			priv->flags |= MLX4_EN_FLAG_MC_PROMISC;
949 		}
950 	} else {
951 		/* Disable Multicast promisc */
952 		if (priv->flags & MLX4_EN_FLAG_MC_PROMISC) {
953 			switch (mdev->dev->caps.steering_mode) {
954 			case MLX4_STEERING_MODE_DEVICE_MANAGED:
955 				err = mlx4_flow_steer_promisc_remove(mdev->dev,
956 								     priv->port,
957 								     MLX4_FS_MC_DEFAULT);
958 				break;
959 
960 			case MLX4_STEERING_MODE_B0:
961 				err = mlx4_multicast_promisc_remove(mdev->dev,
962 								    priv->base_qpn,
963 								    priv->port);
964 				break;
965 
966 			case MLX4_STEERING_MODE_A0:
967 				break;
968 			}
969 			if (err)
970 				en_err(priv, "Failed disabling multicast promiscuous mode\n");
971 			priv->flags &= ~MLX4_EN_FLAG_MC_PROMISC;
972 		}
973 
974 		err = mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, 0,
975 					  0, MLX4_MCAST_DISABLE);
976 		if (err)
977 			en_err(priv, "Failed disabling multicast filter\n");
978 
979 		/* Flush mcast filter and init it with broadcast address */
980 		mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, ETH_BCAST,
981 				    1, MLX4_MCAST_CONFIG);
982 
983 		/* Update multicast list - we cache all addresses so they won't
984 		 * change while HW is updated holding the command semaphor */
985 		netif_addr_lock_bh(dev);
986 		mlx4_en_cache_mclist(dev);
987 		netif_addr_unlock_bh(dev);
988 		list_for_each_entry(mclist, &priv->mc_list, list) {
989 			mcast_addr = mlx4_mac_to_u64(mclist->addr);
990 			mlx4_SET_MCAST_FLTR(mdev->dev, priv->port,
991 					    mcast_addr, 0, MLX4_MCAST_CONFIG);
992 		}
993 		err = mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, 0,
994 					  0, MLX4_MCAST_ENABLE);
995 		if (err)
996 			en_err(priv, "Failed enabling multicast filter\n");
997 
998 		update_mclist_flags(priv, &priv->curr_list, &priv->mc_list);
999 		list_for_each_entry_safe(mclist, tmp, &priv->curr_list, list) {
1000 			if (mclist->action == MCLIST_REM) {
1001 				/* detach this address and delete from list */
1002 				memcpy(&mc_list[10], mclist->addr, ETH_ALEN);
1003 				mc_list[5] = priv->port;
1004 				err = mlx4_multicast_detach(mdev->dev,
1005 							    &priv->rss_map.indir_qp,
1006 							    mc_list,
1007 							    MLX4_PROT_ETH,
1008 							    mclist->reg_id);
1009 				if (err)
1010 					en_err(priv, "Fail to detach multicast address\n");
1011 
1012 				if (mclist->tunnel_reg_id) {
1013 					err = mlx4_flow_detach(priv->mdev->dev, mclist->tunnel_reg_id);
1014 					if (err)
1015 						en_err(priv, "Failed to detach multicast address\n");
1016 				}
1017 
1018 				/* remove from list */
1019 				list_del(&mclist->list);
1020 				kfree(mclist);
1021 			} else if (mclist->action == MCLIST_ADD) {
1022 				/* attach the address */
1023 				memcpy(&mc_list[10], mclist->addr, ETH_ALEN);
1024 				/* needed for B0 steering support */
1025 				mc_list[5] = priv->port;
1026 				err = mlx4_multicast_attach(mdev->dev,
1027 							    &priv->rss_map.indir_qp,
1028 							    mc_list,
1029 							    priv->port, 0,
1030 							    MLX4_PROT_ETH,
1031 							    &mclist->reg_id);
1032 				if (err)
1033 					en_err(priv, "Fail to attach multicast address\n");
1034 
1035 				err = mlx4_en_tunnel_steer_add(priv, &mc_list[10], priv->base_qpn,
1036 							       &mclist->tunnel_reg_id);
1037 				if (err)
1038 					en_err(priv, "Failed to attach multicast address\n");
1039 			}
1040 		}
1041 	}
1042 }
1043 
1044 static void mlx4_en_do_uc_filter(struct mlx4_en_priv *priv,
1045 				 struct net_device *dev,
1046 				 struct mlx4_en_dev *mdev)
1047 {
1048 	struct netdev_hw_addr *ha;
1049 	struct mlx4_mac_entry *entry;
1050 	struct hlist_node *tmp;
1051 	bool found;
1052 	u64 mac;
1053 	int err = 0;
1054 	struct hlist_head *bucket;
1055 	unsigned int i;
1056 	int removed = 0;
1057 	u32 prev_flags;
1058 
1059 	/* Note that we do not need to protect our mac_hash traversal with rcu,
1060 	 * since all modification code is protected by mdev->state_lock
1061 	 */
1062 
1063 	/* find what to remove */
1064 	for (i = 0; i < MLX4_EN_MAC_HASH_SIZE; ++i) {
1065 		bucket = &priv->mac_hash[i];
1066 		hlist_for_each_entry_safe(entry, tmp, bucket, hlist) {
1067 			found = false;
1068 			netdev_for_each_uc_addr(ha, dev) {
1069 				if (ether_addr_equal_64bits(entry->mac,
1070 							    ha->addr)) {
1071 					found = true;
1072 					break;
1073 				}
1074 			}
1075 
1076 			/* MAC address of the port is not in uc list */
1077 			if (ether_addr_equal_64bits(entry->mac,
1078 						    priv->current_mac))
1079 				found = true;
1080 
1081 			if (!found) {
1082 				mac = mlx4_mac_to_u64(entry->mac);
1083 				mlx4_en_uc_steer_release(priv, entry->mac,
1084 							 priv->base_qpn,
1085 							 entry->reg_id);
1086 				mlx4_unregister_mac(mdev->dev, priv->port, mac);
1087 
1088 				hlist_del_rcu(&entry->hlist);
1089 				kfree_rcu(entry, rcu);
1090 				en_dbg(DRV, priv, "Removed MAC %pM on port:%d\n",
1091 				       entry->mac, priv->port);
1092 				++removed;
1093 			}
1094 		}
1095 	}
1096 
1097 	/* if we didn't remove anything, there is no use in trying to add
1098 	 * again once we are in a forced promisc mode state
1099 	 */
1100 	if ((priv->flags & MLX4_EN_FLAG_FORCE_PROMISC) && 0 == removed)
1101 		return;
1102 
1103 	prev_flags = priv->flags;
1104 	priv->flags &= ~MLX4_EN_FLAG_FORCE_PROMISC;
1105 
1106 	/* find what to add */
1107 	netdev_for_each_uc_addr(ha, dev) {
1108 		found = false;
1109 		bucket = &priv->mac_hash[ha->addr[MLX4_EN_MAC_HASH_IDX]];
1110 		hlist_for_each_entry(entry, bucket, hlist) {
1111 			if (ether_addr_equal_64bits(entry->mac, ha->addr)) {
1112 				found = true;
1113 				break;
1114 			}
1115 		}
1116 
1117 		if (!found) {
1118 			entry = kmalloc(sizeof(*entry), GFP_KERNEL);
1119 			if (!entry) {
1120 				en_err(priv, "Failed adding MAC %pM on port:%d (out of memory)\n",
1121 				       ha->addr, priv->port);
1122 				priv->flags |= MLX4_EN_FLAG_FORCE_PROMISC;
1123 				break;
1124 			}
1125 			mac = mlx4_mac_to_u64(ha->addr);
1126 			memcpy(entry->mac, ha->addr, ETH_ALEN);
1127 			err = mlx4_register_mac(mdev->dev, priv->port, mac);
1128 			if (err < 0) {
1129 				en_err(priv, "Failed registering MAC %pM on port %d: %d\n",
1130 				       ha->addr, priv->port, err);
1131 				kfree(entry);
1132 				priv->flags |= MLX4_EN_FLAG_FORCE_PROMISC;
1133 				break;
1134 			}
1135 			err = mlx4_en_uc_steer_add(priv, ha->addr,
1136 						   &priv->base_qpn,
1137 						   &entry->reg_id);
1138 			if (err) {
1139 				en_err(priv, "Failed adding MAC %pM on port %d: %d\n",
1140 				       ha->addr, priv->port, err);
1141 				mlx4_unregister_mac(mdev->dev, priv->port, mac);
1142 				kfree(entry);
1143 				priv->flags |= MLX4_EN_FLAG_FORCE_PROMISC;
1144 				break;
1145 			} else {
1146 				unsigned int mac_hash;
1147 				en_dbg(DRV, priv, "Added MAC %pM on port:%d\n",
1148 				       ha->addr, priv->port);
1149 				mac_hash = ha->addr[MLX4_EN_MAC_HASH_IDX];
1150 				bucket = &priv->mac_hash[mac_hash];
1151 				hlist_add_head_rcu(&entry->hlist, bucket);
1152 			}
1153 		}
1154 	}
1155 
1156 	if (priv->flags & MLX4_EN_FLAG_FORCE_PROMISC) {
1157 		en_warn(priv, "Forcing promiscuous mode on port:%d\n",
1158 			priv->port);
1159 	} else if (prev_flags & MLX4_EN_FLAG_FORCE_PROMISC) {
1160 		en_warn(priv, "Stop forcing promiscuous mode on port:%d\n",
1161 			priv->port);
1162 	}
1163 }
1164 
1165 static void mlx4_en_do_set_rx_mode(struct work_struct *work)
1166 {
1167 	struct mlx4_en_priv *priv = container_of(work, struct mlx4_en_priv,
1168 						 rx_mode_task);
1169 	struct mlx4_en_dev *mdev = priv->mdev;
1170 	struct net_device *dev = priv->dev;
1171 
1172 	mutex_lock(&mdev->state_lock);
1173 	if (!mdev->device_up) {
1174 		en_dbg(HW, priv, "Card is not up, ignoring rx mode change.\n");
1175 		goto out;
1176 	}
1177 	if (!priv->port_up) {
1178 		en_dbg(HW, priv, "Port is down, ignoring rx mode change.\n");
1179 		goto out;
1180 	}
1181 
1182 	if (!netif_carrier_ok(dev)) {
1183 		if (!mlx4_en_QUERY_PORT(mdev, priv->port)) {
1184 			if (priv->port_state.link_state) {
1185 				priv->last_link_state = MLX4_DEV_EVENT_PORT_UP;
1186 				netif_carrier_on(dev);
1187 				en_dbg(LINK, priv, "Link Up\n");
1188 			}
1189 		}
1190 	}
1191 
1192 	if (dev->priv_flags & IFF_UNICAST_FLT)
1193 		mlx4_en_do_uc_filter(priv, dev, mdev);
1194 
1195 	/* Promsicuous mode: disable all filters */
1196 	if ((dev->flags & IFF_PROMISC) ||
1197 	    (priv->flags & MLX4_EN_FLAG_FORCE_PROMISC)) {
1198 		mlx4_en_set_promisc_mode(priv, mdev);
1199 		goto out;
1200 	}
1201 
1202 	/* Not in promiscuous mode */
1203 	if (priv->flags & MLX4_EN_FLAG_PROMISC)
1204 		mlx4_en_clear_promisc_mode(priv, mdev);
1205 
1206 	mlx4_en_do_multicast(priv, dev, mdev);
1207 out:
1208 	mutex_unlock(&mdev->state_lock);
1209 }
1210 
1211 #ifdef CONFIG_NET_POLL_CONTROLLER
1212 static void mlx4_en_netpoll(struct net_device *dev)
1213 {
1214 	struct mlx4_en_priv *priv = netdev_priv(dev);
1215 	struct mlx4_en_cq *cq;
1216 	int i;
1217 
1218 	for (i = 0; i < priv->rx_ring_num; i++) {
1219 		cq = priv->rx_cq[i];
1220 		napi_schedule(&cq->napi);
1221 	}
1222 }
1223 #endif
1224 
1225 static int mlx4_en_set_rss_steer_rules(struct mlx4_en_priv *priv)
1226 {
1227 	u64 reg_id;
1228 	int err = 0;
1229 	int *qpn = &priv->base_qpn;
1230 	struct mlx4_mac_entry *entry;
1231 
1232 	err = mlx4_en_uc_steer_add(priv, priv->dev->dev_addr, qpn, &reg_id);
1233 	if (err)
1234 		return err;
1235 
1236 	err = mlx4_en_tunnel_steer_add(priv, priv->dev->dev_addr, *qpn,
1237 				       &priv->tunnel_reg_id);
1238 	if (err)
1239 		goto tunnel_err;
1240 
1241 	entry = kmalloc(sizeof(*entry), GFP_KERNEL);
1242 	if (!entry) {
1243 		err = -ENOMEM;
1244 		goto alloc_err;
1245 	}
1246 
1247 	memcpy(entry->mac, priv->dev->dev_addr, sizeof(entry->mac));
1248 	memcpy(priv->current_mac, entry->mac, sizeof(priv->current_mac));
1249 	entry->reg_id = reg_id;
1250 	hlist_add_head_rcu(&entry->hlist,
1251 			   &priv->mac_hash[entry->mac[MLX4_EN_MAC_HASH_IDX]]);
1252 
1253 	return 0;
1254 
1255 alloc_err:
1256 	if (priv->tunnel_reg_id)
1257 		mlx4_flow_detach(priv->mdev->dev, priv->tunnel_reg_id);
1258 
1259 tunnel_err:
1260 	mlx4_en_uc_steer_release(priv, priv->dev->dev_addr, *qpn, reg_id);
1261 	return err;
1262 }
1263 
1264 static void mlx4_en_delete_rss_steer_rules(struct mlx4_en_priv *priv)
1265 {
1266 	u64 mac;
1267 	unsigned int i;
1268 	int qpn = priv->base_qpn;
1269 	struct hlist_head *bucket;
1270 	struct hlist_node *tmp;
1271 	struct mlx4_mac_entry *entry;
1272 
1273 	for (i = 0; i < MLX4_EN_MAC_HASH_SIZE; ++i) {
1274 		bucket = &priv->mac_hash[i];
1275 		hlist_for_each_entry_safe(entry, tmp, bucket, hlist) {
1276 			mac = mlx4_mac_to_u64(entry->mac);
1277 			en_dbg(DRV, priv, "Registering MAC:%pM for deleting\n",
1278 			       entry->mac);
1279 			mlx4_en_uc_steer_release(priv, entry->mac,
1280 						 qpn, entry->reg_id);
1281 
1282 			mlx4_unregister_mac(priv->mdev->dev, priv->port, mac);
1283 			hlist_del_rcu(&entry->hlist);
1284 			kfree_rcu(entry, rcu);
1285 		}
1286 	}
1287 
1288 	if (priv->tunnel_reg_id) {
1289 		mlx4_flow_detach(priv->mdev->dev, priv->tunnel_reg_id);
1290 		priv->tunnel_reg_id = 0;
1291 	}
1292 }
1293 
1294 static void mlx4_en_tx_timeout(struct net_device *dev)
1295 {
1296 	struct mlx4_en_priv *priv = netdev_priv(dev);
1297 	struct mlx4_en_dev *mdev = priv->mdev;
1298 	int i;
1299 
1300 	if (netif_msg_timer(priv))
1301 		en_warn(priv, "Tx timeout called on port:%d\n", priv->port);
1302 
1303 	for (i = 0; i < priv->tx_ring_num; i++) {
1304 		if (!netif_tx_queue_stopped(netdev_get_tx_queue(dev, i)))
1305 			continue;
1306 		en_warn(priv, "TX timeout on queue: %d, QP: 0x%x, CQ: 0x%x, Cons: 0x%x, Prod: 0x%x\n",
1307 			i, priv->tx_ring[i]->qpn, priv->tx_ring[i]->cqn,
1308 			priv->tx_ring[i]->cons, priv->tx_ring[i]->prod);
1309 	}
1310 
1311 	priv->port_stats.tx_timeout++;
1312 	en_dbg(DRV, priv, "Scheduling watchdog\n");
1313 	queue_work(mdev->workqueue, &priv->watchdog_task);
1314 }
1315 
1316 
1317 static struct net_device_stats *mlx4_en_get_stats(struct net_device *dev)
1318 {
1319 	struct mlx4_en_priv *priv = netdev_priv(dev);
1320 
1321 	spin_lock_bh(&priv->stats_lock);
1322 	memcpy(&priv->ret_stats, &priv->stats, sizeof(priv->stats));
1323 	spin_unlock_bh(&priv->stats_lock);
1324 
1325 	return &priv->ret_stats;
1326 }
1327 
1328 static void mlx4_en_set_default_moderation(struct mlx4_en_priv *priv)
1329 {
1330 	struct mlx4_en_cq *cq;
1331 	int i;
1332 
1333 	/* If we haven't received a specific coalescing setting
1334 	 * (module param), we set the moderation parameters as follows:
1335 	 * - moder_cnt is set to the number of mtu sized packets to
1336 	 *   satisfy our coalescing target.
1337 	 * - moder_time is set to a fixed value.
1338 	 */
1339 	priv->rx_frames = MLX4_EN_RX_COAL_TARGET;
1340 	priv->rx_usecs = MLX4_EN_RX_COAL_TIME;
1341 	priv->tx_frames = MLX4_EN_TX_COAL_PKTS;
1342 	priv->tx_usecs = MLX4_EN_TX_COAL_TIME;
1343 	en_dbg(INTR, priv, "Default coalesing params for mtu:%d - rx_frames:%d rx_usecs:%d\n",
1344 	       priv->dev->mtu, priv->rx_frames, priv->rx_usecs);
1345 
1346 	/* Setup cq moderation params */
1347 	for (i = 0; i < priv->rx_ring_num; i++) {
1348 		cq = priv->rx_cq[i];
1349 		cq->moder_cnt = priv->rx_frames;
1350 		cq->moder_time = priv->rx_usecs;
1351 		priv->last_moder_time[i] = MLX4_EN_AUTO_CONF;
1352 		priv->last_moder_packets[i] = 0;
1353 		priv->last_moder_bytes[i] = 0;
1354 	}
1355 
1356 	for (i = 0; i < priv->tx_ring_num; i++) {
1357 		cq = priv->tx_cq[i];
1358 		cq->moder_cnt = priv->tx_frames;
1359 		cq->moder_time = priv->tx_usecs;
1360 	}
1361 
1362 	/* Reset auto-moderation params */
1363 	priv->pkt_rate_low = MLX4_EN_RX_RATE_LOW;
1364 	priv->rx_usecs_low = MLX4_EN_RX_COAL_TIME_LOW;
1365 	priv->pkt_rate_high = MLX4_EN_RX_RATE_HIGH;
1366 	priv->rx_usecs_high = MLX4_EN_RX_COAL_TIME_HIGH;
1367 	priv->sample_interval = MLX4_EN_SAMPLE_INTERVAL;
1368 	priv->adaptive_rx_coal = 1;
1369 	priv->last_moder_jiffies = 0;
1370 	priv->last_moder_tx_packets = 0;
1371 }
1372 
1373 static void mlx4_en_auto_moderation(struct mlx4_en_priv *priv)
1374 {
1375 	unsigned long period = (unsigned long) (jiffies - priv->last_moder_jiffies);
1376 	struct mlx4_en_cq *cq;
1377 	unsigned long packets;
1378 	unsigned long rate;
1379 	unsigned long avg_pkt_size;
1380 	unsigned long rx_packets;
1381 	unsigned long rx_bytes;
1382 	unsigned long rx_pkt_diff;
1383 	int moder_time;
1384 	int ring, err;
1385 
1386 	if (!priv->adaptive_rx_coal || period < priv->sample_interval * HZ)
1387 		return;
1388 
1389 	for (ring = 0; ring < priv->rx_ring_num; ring++) {
1390 		spin_lock_bh(&priv->stats_lock);
1391 		rx_packets = priv->rx_ring[ring]->packets;
1392 		rx_bytes = priv->rx_ring[ring]->bytes;
1393 		spin_unlock_bh(&priv->stats_lock);
1394 
1395 		rx_pkt_diff = ((unsigned long) (rx_packets -
1396 				priv->last_moder_packets[ring]));
1397 		packets = rx_pkt_diff;
1398 		rate = packets * HZ / period;
1399 		avg_pkt_size = packets ? ((unsigned long) (rx_bytes -
1400 				priv->last_moder_bytes[ring])) / packets : 0;
1401 
1402 		/* Apply auto-moderation only when packet rate
1403 		 * exceeds a rate that it matters */
1404 		if (rate > (MLX4_EN_RX_RATE_THRESH / priv->rx_ring_num) &&
1405 		    avg_pkt_size > MLX4_EN_AVG_PKT_SMALL) {
1406 			if (rate < priv->pkt_rate_low)
1407 				moder_time = priv->rx_usecs_low;
1408 			else if (rate > priv->pkt_rate_high)
1409 				moder_time = priv->rx_usecs_high;
1410 			else
1411 				moder_time = (rate - priv->pkt_rate_low) *
1412 					(priv->rx_usecs_high - priv->rx_usecs_low) /
1413 					(priv->pkt_rate_high - priv->pkt_rate_low) +
1414 					priv->rx_usecs_low;
1415 		} else {
1416 			moder_time = priv->rx_usecs_low;
1417 		}
1418 
1419 		if (moder_time != priv->last_moder_time[ring]) {
1420 			priv->last_moder_time[ring] = moder_time;
1421 			cq = priv->rx_cq[ring];
1422 			cq->moder_time = moder_time;
1423 			cq->moder_cnt = priv->rx_frames;
1424 			err = mlx4_en_set_cq_moder(priv, cq);
1425 			if (err)
1426 				en_err(priv, "Failed modifying moderation for cq:%d\n",
1427 				       ring);
1428 		}
1429 		priv->last_moder_packets[ring] = rx_packets;
1430 		priv->last_moder_bytes[ring] = rx_bytes;
1431 	}
1432 
1433 	priv->last_moder_jiffies = jiffies;
1434 }
1435 
1436 static void mlx4_en_do_get_stats(struct work_struct *work)
1437 {
1438 	struct delayed_work *delay = to_delayed_work(work);
1439 	struct mlx4_en_priv *priv = container_of(delay, struct mlx4_en_priv,
1440 						 stats_task);
1441 	struct mlx4_en_dev *mdev = priv->mdev;
1442 	int err;
1443 
1444 	mutex_lock(&mdev->state_lock);
1445 	if (mdev->device_up) {
1446 		if (priv->port_up) {
1447 			err = mlx4_en_DUMP_ETH_STATS(mdev, priv->port, 0);
1448 			if (err)
1449 				en_dbg(HW, priv, "Could not update stats\n");
1450 
1451 			mlx4_en_auto_moderation(priv);
1452 		}
1453 
1454 		queue_delayed_work(mdev->workqueue, &priv->stats_task, STATS_DELAY);
1455 	}
1456 	if (mdev->mac_removed[MLX4_MAX_PORTS + 1 - priv->port]) {
1457 		mlx4_en_do_set_mac(priv, priv->current_mac);
1458 		mdev->mac_removed[MLX4_MAX_PORTS + 1 - priv->port] = 0;
1459 	}
1460 	mutex_unlock(&mdev->state_lock);
1461 }
1462 
1463 /* mlx4_en_service_task - Run service task for tasks that needed to be done
1464  * periodically
1465  */
1466 static void mlx4_en_service_task(struct work_struct *work)
1467 {
1468 	struct delayed_work *delay = to_delayed_work(work);
1469 	struct mlx4_en_priv *priv = container_of(delay, struct mlx4_en_priv,
1470 						 service_task);
1471 	struct mlx4_en_dev *mdev = priv->mdev;
1472 
1473 	mutex_lock(&mdev->state_lock);
1474 	if (mdev->device_up) {
1475 		if (mdev->dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_TS)
1476 			mlx4_en_ptp_overflow_check(mdev);
1477 
1478 		mlx4_en_recover_from_oom(priv);
1479 		queue_delayed_work(mdev->workqueue, &priv->service_task,
1480 				   SERVICE_TASK_DELAY);
1481 	}
1482 	mutex_unlock(&mdev->state_lock);
1483 }
1484 
1485 static void mlx4_en_linkstate(struct work_struct *work)
1486 {
1487 	struct mlx4_en_priv *priv = container_of(work, struct mlx4_en_priv,
1488 						 linkstate_task);
1489 	struct mlx4_en_dev *mdev = priv->mdev;
1490 	int linkstate = priv->link_state;
1491 
1492 	mutex_lock(&mdev->state_lock);
1493 	/* If observable port state changed set carrier state and
1494 	 * report to system log */
1495 	if (priv->last_link_state != linkstate) {
1496 		if (linkstate == MLX4_DEV_EVENT_PORT_DOWN) {
1497 			en_info(priv, "Link Down\n");
1498 			netif_carrier_off(priv->dev);
1499 		} else {
1500 			en_info(priv, "Link Up\n");
1501 			netif_carrier_on(priv->dev);
1502 		}
1503 	}
1504 	priv->last_link_state = linkstate;
1505 	mutex_unlock(&mdev->state_lock);
1506 }
1507 
1508 static int mlx4_en_init_affinity_hint(struct mlx4_en_priv *priv, int ring_idx)
1509 {
1510 	struct mlx4_en_rx_ring *ring = priv->rx_ring[ring_idx];
1511 	int numa_node = priv->mdev->dev->numa_node;
1512 
1513 	if (!zalloc_cpumask_var(&ring->affinity_mask, GFP_KERNEL))
1514 		return -ENOMEM;
1515 
1516 	cpumask_set_cpu(cpumask_local_spread(ring_idx, numa_node),
1517 			ring->affinity_mask);
1518 	return 0;
1519 }
1520 
1521 static void mlx4_en_free_affinity_hint(struct mlx4_en_priv *priv, int ring_idx)
1522 {
1523 	free_cpumask_var(priv->rx_ring[ring_idx]->affinity_mask);
1524 }
1525 
1526 int mlx4_en_start_port(struct net_device *dev)
1527 {
1528 	struct mlx4_en_priv *priv = netdev_priv(dev);
1529 	struct mlx4_en_dev *mdev = priv->mdev;
1530 	struct mlx4_en_cq *cq;
1531 	struct mlx4_en_tx_ring *tx_ring;
1532 	int rx_index = 0;
1533 	int tx_index = 0;
1534 	int err = 0;
1535 	int i;
1536 	int j;
1537 	u8 mc_list[16] = {0};
1538 
1539 	if (priv->port_up) {
1540 		en_dbg(DRV, priv, "start port called while port already up\n");
1541 		return 0;
1542 	}
1543 
1544 	INIT_LIST_HEAD(&priv->mc_list);
1545 	INIT_LIST_HEAD(&priv->curr_list);
1546 	INIT_LIST_HEAD(&priv->ethtool_list);
1547 	memset(&priv->ethtool_rules[0], 0,
1548 	       sizeof(struct ethtool_flow_id) * MAX_NUM_OF_FS_RULES);
1549 
1550 	/* Calculate Rx buf size */
1551 	dev->mtu = min(dev->mtu, priv->max_mtu);
1552 	mlx4_en_calc_rx_buf(dev);
1553 	en_dbg(DRV, priv, "Rx buf size:%d\n", priv->rx_skb_size);
1554 
1555 	/* Configure rx cq's and rings */
1556 	err = mlx4_en_activate_rx_rings(priv);
1557 	if (err) {
1558 		en_err(priv, "Failed to activate RX rings\n");
1559 		return err;
1560 	}
1561 	for (i = 0; i < priv->rx_ring_num; i++) {
1562 		cq = priv->rx_cq[i];
1563 
1564 		mlx4_en_cq_init_lock(cq);
1565 
1566 		err = mlx4_en_init_affinity_hint(priv, i);
1567 		if (err) {
1568 			en_err(priv, "Failed preparing IRQ affinity hint\n");
1569 			goto cq_err;
1570 		}
1571 
1572 		err = mlx4_en_activate_cq(priv, cq, i);
1573 		if (err) {
1574 			en_err(priv, "Failed activating Rx CQ\n");
1575 			mlx4_en_free_affinity_hint(priv, i);
1576 			goto cq_err;
1577 		}
1578 
1579 		for (j = 0; j < cq->size; j++) {
1580 			struct mlx4_cqe *cqe = NULL;
1581 
1582 			cqe = mlx4_en_get_cqe(cq->buf, j, priv->cqe_size) +
1583 			      priv->cqe_factor;
1584 			cqe->owner_sr_opcode = MLX4_CQE_OWNER_MASK;
1585 		}
1586 
1587 		err = mlx4_en_set_cq_moder(priv, cq);
1588 		if (err) {
1589 			en_err(priv, "Failed setting cq moderation parameters\n");
1590 			mlx4_en_deactivate_cq(priv, cq);
1591 			mlx4_en_free_affinity_hint(priv, i);
1592 			goto cq_err;
1593 		}
1594 		mlx4_en_arm_cq(priv, cq);
1595 		priv->rx_ring[i]->cqn = cq->mcq.cqn;
1596 		++rx_index;
1597 	}
1598 
1599 	/* Set qp number */
1600 	en_dbg(DRV, priv, "Getting qp number for port %d\n", priv->port);
1601 	err = mlx4_en_get_qp(priv);
1602 	if (err) {
1603 		en_err(priv, "Failed getting eth qp\n");
1604 		goto cq_err;
1605 	}
1606 	mdev->mac_removed[priv->port] = 0;
1607 
1608 	priv->counter_index =
1609 			mlx4_get_default_counter_index(mdev->dev, priv->port);
1610 
1611 	err = mlx4_en_config_rss_steer(priv);
1612 	if (err) {
1613 		en_err(priv, "Failed configuring rss steering\n");
1614 		goto mac_err;
1615 	}
1616 
1617 	err = mlx4_en_create_drop_qp(priv);
1618 	if (err)
1619 		goto rss_err;
1620 
1621 	/* Configure tx cq's and rings */
1622 	for (i = 0; i < priv->tx_ring_num; i++) {
1623 		/* Configure cq */
1624 		cq = priv->tx_cq[i];
1625 		err = mlx4_en_activate_cq(priv, cq, i);
1626 		if (err) {
1627 			en_err(priv, "Failed allocating Tx CQ\n");
1628 			goto tx_err;
1629 		}
1630 		err = mlx4_en_set_cq_moder(priv, cq);
1631 		if (err) {
1632 			en_err(priv, "Failed setting cq moderation parameters\n");
1633 			mlx4_en_deactivate_cq(priv, cq);
1634 			goto tx_err;
1635 		}
1636 		en_dbg(DRV, priv, "Resetting index of collapsed CQ:%d to -1\n", i);
1637 		cq->buf->wqe_index = cpu_to_be16(0xffff);
1638 
1639 		/* Configure ring */
1640 		tx_ring = priv->tx_ring[i];
1641 		err = mlx4_en_activate_tx_ring(priv, tx_ring, cq->mcq.cqn,
1642 			i / priv->num_tx_rings_p_up);
1643 		if (err) {
1644 			en_err(priv, "Failed allocating Tx ring\n");
1645 			mlx4_en_deactivate_cq(priv, cq);
1646 			goto tx_err;
1647 		}
1648 		tx_ring->tx_queue = netdev_get_tx_queue(dev, i);
1649 
1650 		/* Arm CQ for TX completions */
1651 		mlx4_en_arm_cq(priv, cq);
1652 
1653 		/* Set initial ownership of all Tx TXBBs to SW (1) */
1654 		for (j = 0; j < tx_ring->buf_size; j += STAMP_STRIDE)
1655 			*((u32 *) (tx_ring->buf + j)) = 0xffffffff;
1656 		++tx_index;
1657 	}
1658 
1659 	/* Configure port */
1660 	err = mlx4_SET_PORT_general(mdev->dev, priv->port,
1661 				    priv->rx_skb_size + ETH_FCS_LEN,
1662 				    priv->prof->tx_pause,
1663 				    priv->prof->tx_ppp,
1664 				    priv->prof->rx_pause,
1665 				    priv->prof->rx_ppp);
1666 	if (err) {
1667 		en_err(priv, "Failed setting port general configurations for port %d, with error %d\n",
1668 		       priv->port, err);
1669 		goto tx_err;
1670 	}
1671 	/* Set default qp number */
1672 	err = mlx4_SET_PORT_qpn_calc(mdev->dev, priv->port, priv->base_qpn, 0);
1673 	if (err) {
1674 		en_err(priv, "Failed setting default qp numbers\n");
1675 		goto tx_err;
1676 	}
1677 
1678 	if (mdev->dev->caps.tunnel_offload_mode == MLX4_TUNNEL_OFFLOAD_MODE_VXLAN) {
1679 		err = mlx4_SET_PORT_VXLAN(mdev->dev, priv->port, VXLAN_STEER_BY_OUTER_MAC, 1);
1680 		if (err) {
1681 			en_err(priv, "Failed setting port L2 tunnel configuration, err %d\n",
1682 			       err);
1683 			goto tx_err;
1684 		}
1685 	}
1686 
1687 	/* Init port */
1688 	en_dbg(HW, priv, "Initializing port\n");
1689 	err = mlx4_INIT_PORT(mdev->dev, priv->port);
1690 	if (err) {
1691 		en_err(priv, "Failed Initializing port\n");
1692 		goto tx_err;
1693 	}
1694 
1695 	/* Set Unicast and VXLAN steering rules */
1696 	if (mdev->dev->caps.steering_mode != MLX4_STEERING_MODE_A0 &&
1697 	    mlx4_en_set_rss_steer_rules(priv))
1698 		mlx4_warn(mdev, "Failed setting steering rules\n");
1699 
1700 	/* Attach rx QP to bradcast address */
1701 	eth_broadcast_addr(&mc_list[10]);
1702 	mc_list[5] = priv->port; /* needed for B0 steering support */
1703 	if (mlx4_multicast_attach(mdev->dev, &priv->rss_map.indir_qp, mc_list,
1704 				  priv->port, 0, MLX4_PROT_ETH,
1705 				  &priv->broadcast_id))
1706 		mlx4_warn(mdev, "Failed Attaching Broadcast\n");
1707 
1708 	/* Must redo promiscuous mode setup. */
1709 	priv->flags &= ~(MLX4_EN_FLAG_PROMISC | MLX4_EN_FLAG_MC_PROMISC);
1710 
1711 	/* Schedule multicast task to populate multicast list */
1712 	queue_work(mdev->workqueue, &priv->rx_mode_task);
1713 
1714 #ifdef CONFIG_MLX4_EN_VXLAN
1715 	if (priv->mdev->dev->caps.tunnel_offload_mode == MLX4_TUNNEL_OFFLOAD_MODE_VXLAN)
1716 		vxlan_get_rx_port(dev);
1717 #endif
1718 	priv->port_up = true;
1719 	netif_tx_start_all_queues(dev);
1720 	netif_device_attach(dev);
1721 
1722 	return 0;
1723 
1724 tx_err:
1725 	while (tx_index--) {
1726 		mlx4_en_deactivate_tx_ring(priv, priv->tx_ring[tx_index]);
1727 		mlx4_en_deactivate_cq(priv, priv->tx_cq[tx_index]);
1728 	}
1729 	mlx4_en_destroy_drop_qp(priv);
1730 rss_err:
1731 	mlx4_en_release_rss_steer(priv);
1732 mac_err:
1733 	mlx4_en_put_qp(priv);
1734 cq_err:
1735 	while (rx_index--) {
1736 		mlx4_en_deactivate_cq(priv, priv->rx_cq[rx_index]);
1737 		mlx4_en_free_affinity_hint(priv, rx_index);
1738 	}
1739 	for (i = 0; i < priv->rx_ring_num; i++)
1740 		mlx4_en_deactivate_rx_ring(priv, priv->rx_ring[i]);
1741 
1742 	return err; /* need to close devices */
1743 }
1744 
1745 
1746 void mlx4_en_stop_port(struct net_device *dev, int detach)
1747 {
1748 	struct mlx4_en_priv *priv = netdev_priv(dev);
1749 	struct mlx4_en_dev *mdev = priv->mdev;
1750 	struct mlx4_en_mc_list *mclist, *tmp;
1751 	struct ethtool_flow_id *flow, *tmp_flow;
1752 	int i;
1753 	u8 mc_list[16] = {0};
1754 
1755 	if (!priv->port_up) {
1756 		en_dbg(DRV, priv, "stop port called while port already down\n");
1757 		return;
1758 	}
1759 
1760 	/* close port*/
1761 	mlx4_CLOSE_PORT(mdev->dev, priv->port);
1762 
1763 	/* Synchronize with tx routine */
1764 	netif_tx_lock_bh(dev);
1765 	if (detach)
1766 		netif_device_detach(dev);
1767 	netif_tx_stop_all_queues(dev);
1768 	netif_tx_unlock_bh(dev);
1769 
1770 	netif_tx_disable(dev);
1771 
1772 	/* Set port as not active */
1773 	priv->port_up = false;
1774 	priv->counter_index = MLX4_SINK_COUNTER_INDEX(mdev->dev);
1775 
1776 	/* Promsicuous mode */
1777 	if (mdev->dev->caps.steering_mode ==
1778 	    MLX4_STEERING_MODE_DEVICE_MANAGED) {
1779 		priv->flags &= ~(MLX4_EN_FLAG_PROMISC |
1780 				 MLX4_EN_FLAG_MC_PROMISC);
1781 		mlx4_flow_steer_promisc_remove(mdev->dev,
1782 					       priv->port,
1783 					       MLX4_FS_ALL_DEFAULT);
1784 		mlx4_flow_steer_promisc_remove(mdev->dev,
1785 					       priv->port,
1786 					       MLX4_FS_MC_DEFAULT);
1787 	} else if (priv->flags & MLX4_EN_FLAG_PROMISC) {
1788 		priv->flags &= ~MLX4_EN_FLAG_PROMISC;
1789 
1790 		/* Disable promiscouos mode */
1791 		mlx4_unicast_promisc_remove(mdev->dev, priv->base_qpn,
1792 					    priv->port);
1793 
1794 		/* Disable Multicast promisc */
1795 		if (priv->flags & MLX4_EN_FLAG_MC_PROMISC) {
1796 			mlx4_multicast_promisc_remove(mdev->dev, priv->base_qpn,
1797 						      priv->port);
1798 			priv->flags &= ~MLX4_EN_FLAG_MC_PROMISC;
1799 		}
1800 	}
1801 
1802 	/* Detach All multicasts */
1803 	eth_broadcast_addr(&mc_list[10]);
1804 	mc_list[5] = priv->port; /* needed for B0 steering support */
1805 	mlx4_multicast_detach(mdev->dev, &priv->rss_map.indir_qp, mc_list,
1806 			      MLX4_PROT_ETH, priv->broadcast_id);
1807 	list_for_each_entry(mclist, &priv->curr_list, list) {
1808 		memcpy(&mc_list[10], mclist->addr, ETH_ALEN);
1809 		mc_list[5] = priv->port;
1810 		mlx4_multicast_detach(mdev->dev, &priv->rss_map.indir_qp,
1811 				      mc_list, MLX4_PROT_ETH, mclist->reg_id);
1812 		if (mclist->tunnel_reg_id)
1813 			mlx4_flow_detach(mdev->dev, mclist->tunnel_reg_id);
1814 	}
1815 	mlx4_en_clear_list(dev);
1816 	list_for_each_entry_safe(mclist, tmp, &priv->curr_list, list) {
1817 		list_del(&mclist->list);
1818 		kfree(mclist);
1819 	}
1820 
1821 	/* Flush multicast filter */
1822 	mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, 0, 1, MLX4_MCAST_CONFIG);
1823 
1824 	/* Remove flow steering rules for the port*/
1825 	if (mdev->dev->caps.steering_mode ==
1826 	    MLX4_STEERING_MODE_DEVICE_MANAGED) {
1827 		ASSERT_RTNL();
1828 		list_for_each_entry_safe(flow, tmp_flow,
1829 					 &priv->ethtool_list, list) {
1830 			mlx4_flow_detach(mdev->dev, flow->id);
1831 			list_del(&flow->list);
1832 		}
1833 	}
1834 
1835 	mlx4_en_destroy_drop_qp(priv);
1836 
1837 	/* Free TX Rings */
1838 	for (i = 0; i < priv->tx_ring_num; i++) {
1839 		mlx4_en_deactivate_tx_ring(priv, priv->tx_ring[i]);
1840 		mlx4_en_deactivate_cq(priv, priv->tx_cq[i]);
1841 	}
1842 	msleep(10);
1843 
1844 	for (i = 0; i < priv->tx_ring_num; i++)
1845 		mlx4_en_free_tx_buf(dev, priv->tx_ring[i]);
1846 
1847 	if (mdev->dev->caps.steering_mode != MLX4_STEERING_MODE_A0)
1848 		mlx4_en_delete_rss_steer_rules(priv);
1849 
1850 	/* Free RSS qps */
1851 	mlx4_en_release_rss_steer(priv);
1852 
1853 	/* Unregister Mac address for the port */
1854 	mlx4_en_put_qp(priv);
1855 	if (!(mdev->dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_REASSIGN_MAC_EN))
1856 		mdev->mac_removed[priv->port] = 1;
1857 
1858 	/* Free RX Rings */
1859 	for (i = 0; i < priv->rx_ring_num; i++) {
1860 		struct mlx4_en_cq *cq = priv->rx_cq[i];
1861 
1862 		local_bh_disable();
1863 		while (!mlx4_en_cq_lock_napi(cq)) {
1864 			pr_info("CQ %d locked\n", i);
1865 			mdelay(1);
1866 		}
1867 		local_bh_enable();
1868 
1869 		napi_synchronize(&cq->napi);
1870 		mlx4_en_deactivate_rx_ring(priv, priv->rx_ring[i]);
1871 		mlx4_en_deactivate_cq(priv, cq);
1872 
1873 		mlx4_en_free_affinity_hint(priv, i);
1874 	}
1875 }
1876 
1877 static void mlx4_en_restart(struct work_struct *work)
1878 {
1879 	struct mlx4_en_priv *priv = container_of(work, struct mlx4_en_priv,
1880 						 watchdog_task);
1881 	struct mlx4_en_dev *mdev = priv->mdev;
1882 	struct net_device *dev = priv->dev;
1883 
1884 	en_dbg(DRV, priv, "Watchdog task called for port %d\n", priv->port);
1885 
1886 	mutex_lock(&mdev->state_lock);
1887 	if (priv->port_up) {
1888 		mlx4_en_stop_port(dev, 1);
1889 		if (mlx4_en_start_port(dev))
1890 			en_err(priv, "Failed restarting port %d\n", priv->port);
1891 	}
1892 	mutex_unlock(&mdev->state_lock);
1893 }
1894 
1895 static void mlx4_en_clear_stats(struct net_device *dev)
1896 {
1897 	struct mlx4_en_priv *priv = netdev_priv(dev);
1898 	struct mlx4_en_dev *mdev = priv->mdev;
1899 	int i;
1900 
1901 	if (mlx4_en_DUMP_ETH_STATS(mdev, priv->port, 1))
1902 		en_dbg(HW, priv, "Failed dumping statistics\n");
1903 
1904 	memset(&priv->stats, 0, sizeof(priv->stats));
1905 	memset(&priv->pstats, 0, sizeof(priv->pstats));
1906 	memset(&priv->pkstats, 0, sizeof(priv->pkstats));
1907 	memset(&priv->port_stats, 0, sizeof(priv->port_stats));
1908 	memset(&priv->rx_flowstats, 0, sizeof(priv->rx_flowstats));
1909 	memset(&priv->tx_flowstats, 0, sizeof(priv->tx_flowstats));
1910 	memset(&priv->rx_priority_flowstats, 0,
1911 	       sizeof(priv->rx_priority_flowstats));
1912 	memset(&priv->tx_priority_flowstats, 0,
1913 	       sizeof(priv->tx_priority_flowstats));
1914 	memset(&priv->pf_stats, 0, sizeof(priv->pf_stats));
1915 
1916 	for (i = 0; i < priv->tx_ring_num; i++) {
1917 		priv->tx_ring[i]->bytes = 0;
1918 		priv->tx_ring[i]->packets = 0;
1919 		priv->tx_ring[i]->tx_csum = 0;
1920 	}
1921 	for (i = 0; i < priv->rx_ring_num; i++) {
1922 		priv->rx_ring[i]->bytes = 0;
1923 		priv->rx_ring[i]->packets = 0;
1924 		priv->rx_ring[i]->csum_ok = 0;
1925 		priv->rx_ring[i]->csum_none = 0;
1926 		priv->rx_ring[i]->csum_complete = 0;
1927 	}
1928 }
1929 
1930 static int mlx4_en_open(struct net_device *dev)
1931 {
1932 	struct mlx4_en_priv *priv = netdev_priv(dev);
1933 	struct mlx4_en_dev *mdev = priv->mdev;
1934 	int err = 0;
1935 
1936 	mutex_lock(&mdev->state_lock);
1937 
1938 	if (!mdev->device_up) {
1939 		en_err(priv, "Cannot open - device down/disabled\n");
1940 		err = -EBUSY;
1941 		goto out;
1942 	}
1943 
1944 	/* Reset HW statistics and SW counters */
1945 	mlx4_en_clear_stats(dev);
1946 
1947 	err = mlx4_en_start_port(dev);
1948 	if (err)
1949 		en_err(priv, "Failed starting port:%d\n", priv->port);
1950 
1951 out:
1952 	mutex_unlock(&mdev->state_lock);
1953 	return err;
1954 }
1955 
1956 
1957 static int mlx4_en_close(struct net_device *dev)
1958 {
1959 	struct mlx4_en_priv *priv = netdev_priv(dev);
1960 	struct mlx4_en_dev *mdev = priv->mdev;
1961 
1962 	en_dbg(IFDOWN, priv, "Close port called\n");
1963 
1964 	mutex_lock(&mdev->state_lock);
1965 
1966 	mlx4_en_stop_port(dev, 0);
1967 	netif_carrier_off(dev);
1968 
1969 	mutex_unlock(&mdev->state_lock);
1970 	return 0;
1971 }
1972 
1973 void mlx4_en_free_resources(struct mlx4_en_priv *priv)
1974 {
1975 	int i;
1976 
1977 #ifdef CONFIG_RFS_ACCEL
1978 	priv->dev->rx_cpu_rmap = NULL;
1979 #endif
1980 
1981 	for (i = 0; i < priv->tx_ring_num; i++) {
1982 		if (priv->tx_ring && priv->tx_ring[i])
1983 			mlx4_en_destroy_tx_ring(priv, &priv->tx_ring[i]);
1984 		if (priv->tx_cq && priv->tx_cq[i])
1985 			mlx4_en_destroy_cq(priv, &priv->tx_cq[i]);
1986 	}
1987 
1988 	for (i = 0; i < priv->rx_ring_num; i++) {
1989 		if (priv->rx_ring[i])
1990 			mlx4_en_destroy_rx_ring(priv, &priv->rx_ring[i],
1991 				priv->prof->rx_ring_size, priv->stride);
1992 		if (priv->rx_cq[i])
1993 			mlx4_en_destroy_cq(priv, &priv->rx_cq[i]);
1994 	}
1995 
1996 }
1997 
1998 int mlx4_en_alloc_resources(struct mlx4_en_priv *priv)
1999 {
2000 	struct mlx4_en_port_profile *prof = priv->prof;
2001 	int i;
2002 	int node;
2003 
2004 	/* Create tx Rings */
2005 	for (i = 0; i < priv->tx_ring_num; i++) {
2006 		node = cpu_to_node(i % num_online_cpus());
2007 		if (mlx4_en_create_cq(priv, &priv->tx_cq[i],
2008 				      prof->tx_ring_size, i, TX, node))
2009 			goto err;
2010 
2011 		if (mlx4_en_create_tx_ring(priv, &priv->tx_ring[i],
2012 					   prof->tx_ring_size, TXBB_SIZE,
2013 					   node, i))
2014 			goto err;
2015 	}
2016 
2017 	/* Create rx Rings */
2018 	for (i = 0; i < priv->rx_ring_num; i++) {
2019 		node = cpu_to_node(i % num_online_cpus());
2020 		if (mlx4_en_create_cq(priv, &priv->rx_cq[i],
2021 				      prof->rx_ring_size, i, RX, node))
2022 			goto err;
2023 
2024 		if (mlx4_en_create_rx_ring(priv, &priv->rx_ring[i],
2025 					   prof->rx_ring_size, priv->stride,
2026 					   node))
2027 			goto err;
2028 	}
2029 
2030 #ifdef CONFIG_RFS_ACCEL
2031 	priv->dev->rx_cpu_rmap = mlx4_get_cpu_rmap(priv->mdev->dev, priv->port);
2032 #endif
2033 
2034 	return 0;
2035 
2036 err:
2037 	en_err(priv, "Failed to allocate NIC resources\n");
2038 	for (i = 0; i < priv->rx_ring_num; i++) {
2039 		if (priv->rx_ring[i])
2040 			mlx4_en_destroy_rx_ring(priv, &priv->rx_ring[i],
2041 						prof->rx_ring_size,
2042 						priv->stride);
2043 		if (priv->rx_cq[i])
2044 			mlx4_en_destroy_cq(priv, &priv->rx_cq[i]);
2045 	}
2046 	for (i = 0; i < priv->tx_ring_num; i++) {
2047 		if (priv->tx_ring[i])
2048 			mlx4_en_destroy_tx_ring(priv, &priv->tx_ring[i]);
2049 		if (priv->tx_cq[i])
2050 			mlx4_en_destroy_cq(priv, &priv->tx_cq[i]);
2051 	}
2052 	return -ENOMEM;
2053 }
2054 
2055 
2056 void mlx4_en_destroy_netdev(struct net_device *dev)
2057 {
2058 	struct mlx4_en_priv *priv = netdev_priv(dev);
2059 	struct mlx4_en_dev *mdev = priv->mdev;
2060 
2061 	en_dbg(DRV, priv, "Destroying netdev on port:%d\n", priv->port);
2062 
2063 	/* Unregister device - this will close the port if it was up */
2064 	if (priv->registered)
2065 		unregister_netdev(dev);
2066 
2067 	if (priv->allocated)
2068 		mlx4_free_hwq_res(mdev->dev, &priv->res, MLX4_EN_PAGE_SIZE);
2069 
2070 	cancel_delayed_work(&priv->stats_task);
2071 	cancel_delayed_work(&priv->service_task);
2072 	/* flush any pending task for this netdev */
2073 	flush_workqueue(mdev->workqueue);
2074 
2075 	if (mdev->dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_TS)
2076 		mlx4_en_remove_timestamp(mdev);
2077 
2078 	/* Detach the netdev so tasks would not attempt to access it */
2079 	mutex_lock(&mdev->state_lock);
2080 	mdev->pndev[priv->port] = NULL;
2081 	mdev->upper[priv->port] = NULL;
2082 	mutex_unlock(&mdev->state_lock);
2083 
2084 	mlx4_en_free_resources(priv);
2085 
2086 	kfree(priv->tx_ring);
2087 	kfree(priv->tx_cq);
2088 
2089 	free_netdev(dev);
2090 }
2091 
2092 static int mlx4_en_change_mtu(struct net_device *dev, int new_mtu)
2093 {
2094 	struct mlx4_en_priv *priv = netdev_priv(dev);
2095 	struct mlx4_en_dev *mdev = priv->mdev;
2096 	int err = 0;
2097 
2098 	en_dbg(DRV, priv, "Change MTU called - current:%d new:%d\n",
2099 		 dev->mtu, new_mtu);
2100 
2101 	if ((new_mtu < MLX4_EN_MIN_MTU) || (new_mtu > priv->max_mtu)) {
2102 		en_err(priv, "Bad MTU size:%d.\n", new_mtu);
2103 		return -EPERM;
2104 	}
2105 	dev->mtu = new_mtu;
2106 
2107 	if (netif_running(dev)) {
2108 		mutex_lock(&mdev->state_lock);
2109 		if (!mdev->device_up) {
2110 			/* NIC is probably restarting - let watchdog task reset
2111 			 * the port */
2112 			en_dbg(DRV, priv, "Change MTU called with card down!?\n");
2113 		} else {
2114 			mlx4_en_stop_port(dev, 1);
2115 			err = mlx4_en_start_port(dev);
2116 			if (err) {
2117 				en_err(priv, "Failed restarting port:%d\n",
2118 					 priv->port);
2119 				queue_work(mdev->workqueue, &priv->watchdog_task);
2120 			}
2121 		}
2122 		mutex_unlock(&mdev->state_lock);
2123 	}
2124 	return 0;
2125 }
2126 
2127 static int mlx4_en_hwtstamp_set(struct net_device *dev, struct ifreq *ifr)
2128 {
2129 	struct mlx4_en_priv *priv = netdev_priv(dev);
2130 	struct mlx4_en_dev *mdev = priv->mdev;
2131 	struct hwtstamp_config config;
2132 
2133 	if (copy_from_user(&config, ifr->ifr_data, sizeof(config)))
2134 		return -EFAULT;
2135 
2136 	/* reserved for future extensions */
2137 	if (config.flags)
2138 		return -EINVAL;
2139 
2140 	/* device doesn't support time stamping */
2141 	if (!(mdev->dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_TS))
2142 		return -EINVAL;
2143 
2144 	/* TX HW timestamp */
2145 	switch (config.tx_type) {
2146 	case HWTSTAMP_TX_OFF:
2147 	case HWTSTAMP_TX_ON:
2148 		break;
2149 	default:
2150 		return -ERANGE;
2151 	}
2152 
2153 	/* RX HW timestamp */
2154 	switch (config.rx_filter) {
2155 	case HWTSTAMP_FILTER_NONE:
2156 		break;
2157 	case HWTSTAMP_FILTER_ALL:
2158 	case HWTSTAMP_FILTER_SOME:
2159 	case HWTSTAMP_FILTER_PTP_V1_L4_EVENT:
2160 	case HWTSTAMP_FILTER_PTP_V1_L4_SYNC:
2161 	case HWTSTAMP_FILTER_PTP_V1_L4_DELAY_REQ:
2162 	case HWTSTAMP_FILTER_PTP_V2_L4_EVENT:
2163 	case HWTSTAMP_FILTER_PTP_V2_L4_SYNC:
2164 	case HWTSTAMP_FILTER_PTP_V2_L4_DELAY_REQ:
2165 	case HWTSTAMP_FILTER_PTP_V2_L2_EVENT:
2166 	case HWTSTAMP_FILTER_PTP_V2_L2_SYNC:
2167 	case HWTSTAMP_FILTER_PTP_V2_L2_DELAY_REQ:
2168 	case HWTSTAMP_FILTER_PTP_V2_EVENT:
2169 	case HWTSTAMP_FILTER_PTP_V2_SYNC:
2170 	case HWTSTAMP_FILTER_PTP_V2_DELAY_REQ:
2171 		config.rx_filter = HWTSTAMP_FILTER_ALL;
2172 		break;
2173 	default:
2174 		return -ERANGE;
2175 	}
2176 
2177 	if (mlx4_en_reset_config(dev, config, dev->features)) {
2178 		config.tx_type = HWTSTAMP_TX_OFF;
2179 		config.rx_filter = HWTSTAMP_FILTER_NONE;
2180 	}
2181 
2182 	return copy_to_user(ifr->ifr_data, &config,
2183 			    sizeof(config)) ? -EFAULT : 0;
2184 }
2185 
2186 static int mlx4_en_hwtstamp_get(struct net_device *dev, struct ifreq *ifr)
2187 {
2188 	struct mlx4_en_priv *priv = netdev_priv(dev);
2189 
2190 	return copy_to_user(ifr->ifr_data, &priv->hwtstamp_config,
2191 			    sizeof(priv->hwtstamp_config)) ? -EFAULT : 0;
2192 }
2193 
2194 static int mlx4_en_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
2195 {
2196 	switch (cmd) {
2197 	case SIOCSHWTSTAMP:
2198 		return mlx4_en_hwtstamp_set(dev, ifr);
2199 	case SIOCGHWTSTAMP:
2200 		return mlx4_en_hwtstamp_get(dev, ifr);
2201 	default:
2202 		return -EOPNOTSUPP;
2203 	}
2204 }
2205 
2206 static netdev_features_t mlx4_en_fix_features(struct net_device *netdev,
2207 					      netdev_features_t features)
2208 {
2209 	struct mlx4_en_priv *en_priv = netdev_priv(netdev);
2210 	struct mlx4_en_dev *mdev = en_priv->mdev;
2211 
2212 	/* Since there is no support for separate RX C-TAG/S-TAG vlan accel
2213 	 * enable/disable make sure S-TAG flag is always in same state as
2214 	 * C-TAG.
2215 	 */
2216 	if (features & NETIF_F_HW_VLAN_CTAG_RX &&
2217 	    !(mdev->dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_SKIP_OUTER_VLAN))
2218 		features |= NETIF_F_HW_VLAN_STAG_RX;
2219 	else
2220 		features &= ~NETIF_F_HW_VLAN_STAG_RX;
2221 
2222 	return features;
2223 }
2224 
2225 static int mlx4_en_set_features(struct net_device *netdev,
2226 		netdev_features_t features)
2227 {
2228 	struct mlx4_en_priv *priv = netdev_priv(netdev);
2229 	bool reset = false;
2230 	int ret = 0;
2231 
2232 	if (DEV_FEATURE_CHANGED(netdev, features, NETIF_F_RXFCS)) {
2233 		en_info(priv, "Turn %s RX-FCS\n",
2234 			(features & NETIF_F_RXFCS) ? "ON" : "OFF");
2235 		reset = true;
2236 	}
2237 
2238 	if (DEV_FEATURE_CHANGED(netdev, features, NETIF_F_RXALL)) {
2239 		u8 ignore_fcs_value = (features & NETIF_F_RXALL) ? 1 : 0;
2240 
2241 		en_info(priv, "Turn %s RX-ALL\n",
2242 			ignore_fcs_value ? "ON" : "OFF");
2243 		ret = mlx4_SET_PORT_fcs_check(priv->mdev->dev,
2244 					      priv->port, ignore_fcs_value);
2245 		if (ret)
2246 			return ret;
2247 	}
2248 
2249 	if (DEV_FEATURE_CHANGED(netdev, features, NETIF_F_HW_VLAN_CTAG_RX)) {
2250 		en_info(priv, "Turn %s RX vlan strip offload\n",
2251 			(features & NETIF_F_HW_VLAN_CTAG_RX) ? "ON" : "OFF");
2252 		reset = true;
2253 	}
2254 
2255 	if (DEV_FEATURE_CHANGED(netdev, features, NETIF_F_HW_VLAN_CTAG_TX))
2256 		en_info(priv, "Turn %s TX vlan strip offload\n",
2257 			(features & NETIF_F_HW_VLAN_CTAG_TX) ? "ON" : "OFF");
2258 
2259 	if (DEV_FEATURE_CHANGED(netdev, features, NETIF_F_HW_VLAN_STAG_TX))
2260 		en_info(priv, "Turn %s TX S-VLAN strip offload\n",
2261 			(features & NETIF_F_HW_VLAN_STAG_TX) ? "ON" : "OFF");
2262 
2263 	if (DEV_FEATURE_CHANGED(netdev, features, NETIF_F_LOOPBACK)) {
2264 		en_info(priv, "Turn %s loopback\n",
2265 			(features & NETIF_F_LOOPBACK) ? "ON" : "OFF");
2266 		mlx4_en_update_loopback_state(netdev, features);
2267 	}
2268 
2269 	if (reset) {
2270 		ret = mlx4_en_reset_config(netdev, priv->hwtstamp_config,
2271 					   features);
2272 		if (ret)
2273 			return ret;
2274 	}
2275 
2276 	return 0;
2277 }
2278 
2279 static int mlx4_en_set_vf_mac(struct net_device *dev, int queue, u8 *mac)
2280 {
2281 	struct mlx4_en_priv *en_priv = netdev_priv(dev);
2282 	struct mlx4_en_dev *mdev = en_priv->mdev;
2283 	u64 mac_u64 = mlx4_mac_to_u64(mac);
2284 
2285 	if (!is_valid_ether_addr(mac))
2286 		return -EINVAL;
2287 
2288 	return mlx4_set_vf_mac(mdev->dev, en_priv->port, queue, mac_u64);
2289 }
2290 
2291 static int mlx4_en_set_vf_vlan(struct net_device *dev, int vf, u16 vlan, u8 qos)
2292 {
2293 	struct mlx4_en_priv *en_priv = netdev_priv(dev);
2294 	struct mlx4_en_dev *mdev = en_priv->mdev;
2295 
2296 	return mlx4_set_vf_vlan(mdev->dev, en_priv->port, vf, vlan, qos);
2297 }
2298 
2299 static int mlx4_en_set_vf_rate(struct net_device *dev, int vf, int min_tx_rate,
2300 			       int max_tx_rate)
2301 {
2302 	struct mlx4_en_priv *en_priv = netdev_priv(dev);
2303 	struct mlx4_en_dev *mdev = en_priv->mdev;
2304 
2305 	return mlx4_set_vf_rate(mdev->dev, en_priv->port, vf, min_tx_rate,
2306 				max_tx_rate);
2307 }
2308 
2309 static int mlx4_en_set_vf_spoofchk(struct net_device *dev, int vf, bool setting)
2310 {
2311 	struct mlx4_en_priv *en_priv = netdev_priv(dev);
2312 	struct mlx4_en_dev *mdev = en_priv->mdev;
2313 
2314 	return mlx4_set_vf_spoofchk(mdev->dev, en_priv->port, vf, setting);
2315 }
2316 
2317 static int mlx4_en_get_vf_config(struct net_device *dev, int vf, struct ifla_vf_info *ivf)
2318 {
2319 	struct mlx4_en_priv *en_priv = netdev_priv(dev);
2320 	struct mlx4_en_dev *mdev = en_priv->mdev;
2321 
2322 	return mlx4_get_vf_config(mdev->dev, en_priv->port, vf, ivf);
2323 }
2324 
2325 static int mlx4_en_set_vf_link_state(struct net_device *dev, int vf, int link_state)
2326 {
2327 	struct mlx4_en_priv *en_priv = netdev_priv(dev);
2328 	struct mlx4_en_dev *mdev = en_priv->mdev;
2329 
2330 	return mlx4_set_vf_link_state(mdev->dev, en_priv->port, vf, link_state);
2331 }
2332 
2333 static int mlx4_en_get_vf_stats(struct net_device *dev, int vf,
2334 				struct ifla_vf_stats *vf_stats)
2335 {
2336 	struct mlx4_en_priv *en_priv = netdev_priv(dev);
2337 	struct mlx4_en_dev *mdev = en_priv->mdev;
2338 
2339 	return mlx4_get_vf_stats(mdev->dev, en_priv->port, vf, vf_stats);
2340 }
2341 
2342 #define PORT_ID_BYTE_LEN 8
2343 static int mlx4_en_get_phys_port_id(struct net_device *dev,
2344 				    struct netdev_phys_item_id *ppid)
2345 {
2346 	struct mlx4_en_priv *priv = netdev_priv(dev);
2347 	struct mlx4_dev *mdev = priv->mdev->dev;
2348 	int i;
2349 	u64 phys_port_id = mdev->caps.phys_port_id[priv->port];
2350 
2351 	if (!phys_port_id)
2352 		return -EOPNOTSUPP;
2353 
2354 	ppid->id_len = sizeof(phys_port_id);
2355 	for (i = PORT_ID_BYTE_LEN - 1; i >= 0; --i) {
2356 		ppid->id[i] =  phys_port_id & 0xff;
2357 		phys_port_id >>= 8;
2358 	}
2359 	return 0;
2360 }
2361 
2362 #ifdef CONFIG_MLX4_EN_VXLAN
2363 static void mlx4_en_add_vxlan_offloads(struct work_struct *work)
2364 {
2365 	int ret;
2366 	struct mlx4_en_priv *priv = container_of(work, struct mlx4_en_priv,
2367 						 vxlan_add_task);
2368 
2369 	ret = mlx4_config_vxlan_port(priv->mdev->dev, priv->vxlan_port);
2370 	if (ret)
2371 		goto out;
2372 
2373 	ret = mlx4_SET_PORT_VXLAN(priv->mdev->dev, priv->port,
2374 				  VXLAN_STEER_BY_OUTER_MAC, 1);
2375 out:
2376 	if (ret) {
2377 		en_err(priv, "failed setting L2 tunnel configuration ret %d\n", ret);
2378 		return;
2379 	}
2380 
2381 	/* set offloads */
2382 	priv->dev->hw_enc_features |= NETIF_F_IP_CSUM | NETIF_F_RXCSUM |
2383 				      NETIF_F_TSO | NETIF_F_GSO_UDP_TUNNEL;
2384 	priv->dev->hw_features |= NETIF_F_GSO_UDP_TUNNEL;
2385 	priv->dev->features    |= NETIF_F_GSO_UDP_TUNNEL;
2386 }
2387 
2388 static void mlx4_en_del_vxlan_offloads(struct work_struct *work)
2389 {
2390 	int ret;
2391 	struct mlx4_en_priv *priv = container_of(work, struct mlx4_en_priv,
2392 						 vxlan_del_task);
2393 	/* unset offloads */
2394 	priv->dev->hw_enc_features &= ~(NETIF_F_IP_CSUM | NETIF_F_RXCSUM |
2395 				      NETIF_F_TSO | NETIF_F_GSO_UDP_TUNNEL);
2396 	priv->dev->hw_features &= ~NETIF_F_GSO_UDP_TUNNEL;
2397 	priv->dev->features    &= ~NETIF_F_GSO_UDP_TUNNEL;
2398 
2399 	ret = mlx4_SET_PORT_VXLAN(priv->mdev->dev, priv->port,
2400 				  VXLAN_STEER_BY_OUTER_MAC, 0);
2401 	if (ret)
2402 		en_err(priv, "failed setting L2 tunnel configuration ret %d\n", ret);
2403 
2404 	priv->vxlan_port = 0;
2405 }
2406 
2407 static void mlx4_en_add_vxlan_port(struct  net_device *dev,
2408 				   sa_family_t sa_family, __be16 port)
2409 {
2410 	struct mlx4_en_priv *priv = netdev_priv(dev);
2411 	__be16 current_port;
2412 
2413 	if (priv->mdev->dev->caps.tunnel_offload_mode != MLX4_TUNNEL_OFFLOAD_MODE_VXLAN)
2414 		return;
2415 
2416 	if (sa_family == AF_INET6)
2417 		return;
2418 
2419 	current_port = priv->vxlan_port;
2420 	if (current_port && current_port != port) {
2421 		en_warn(priv, "vxlan port %d configured, can't add port %d\n",
2422 			ntohs(current_port), ntohs(port));
2423 		return;
2424 	}
2425 
2426 	priv->vxlan_port = port;
2427 	queue_work(priv->mdev->workqueue, &priv->vxlan_add_task);
2428 }
2429 
2430 static void mlx4_en_del_vxlan_port(struct  net_device *dev,
2431 				   sa_family_t sa_family, __be16 port)
2432 {
2433 	struct mlx4_en_priv *priv = netdev_priv(dev);
2434 	__be16 current_port;
2435 
2436 	if (priv->mdev->dev->caps.tunnel_offload_mode != MLX4_TUNNEL_OFFLOAD_MODE_VXLAN)
2437 		return;
2438 
2439 	if (sa_family == AF_INET6)
2440 		return;
2441 
2442 	current_port = priv->vxlan_port;
2443 	if (current_port != port) {
2444 		en_dbg(DRV, priv, "vxlan port %d isn't configured, ignoring\n", ntohs(port));
2445 		return;
2446 	}
2447 
2448 	queue_work(priv->mdev->workqueue, &priv->vxlan_del_task);
2449 }
2450 
2451 static netdev_features_t mlx4_en_features_check(struct sk_buff *skb,
2452 						struct net_device *dev,
2453 						netdev_features_t features)
2454 {
2455 	features = vlan_features_check(skb, features);
2456 	return vxlan_features_check(skb, features);
2457 }
2458 #endif
2459 
2460 static int mlx4_en_set_tx_maxrate(struct net_device *dev, int queue_index, u32 maxrate)
2461 {
2462 	struct mlx4_en_priv *priv = netdev_priv(dev);
2463 	struct mlx4_en_tx_ring *tx_ring = priv->tx_ring[queue_index];
2464 	struct mlx4_update_qp_params params;
2465 	int err;
2466 
2467 	if (!(priv->mdev->dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_QP_RATE_LIMIT))
2468 		return -EOPNOTSUPP;
2469 
2470 	/* rate provided to us in Mbs, check if it fits into 12 bits, if not use Gbs */
2471 	if (maxrate >> 12) {
2472 		params.rate_unit = MLX4_QP_RATE_LIMIT_GBS;
2473 		params.rate_val  = maxrate / 1000;
2474 	} else if (maxrate) {
2475 		params.rate_unit = MLX4_QP_RATE_LIMIT_MBS;
2476 		params.rate_val  = maxrate;
2477 	} else { /* zero serves to revoke the QP rate-limitation */
2478 		params.rate_unit = 0;
2479 		params.rate_val  = 0;
2480 	}
2481 
2482 	err = mlx4_update_qp(priv->mdev->dev, tx_ring->qpn, MLX4_UPDATE_QP_RATE_LIMIT,
2483 			     &params);
2484 	return err;
2485 }
2486 
2487 static const struct net_device_ops mlx4_netdev_ops = {
2488 	.ndo_open		= mlx4_en_open,
2489 	.ndo_stop		= mlx4_en_close,
2490 	.ndo_start_xmit		= mlx4_en_xmit,
2491 	.ndo_select_queue	= mlx4_en_select_queue,
2492 	.ndo_get_stats		= mlx4_en_get_stats,
2493 	.ndo_set_rx_mode	= mlx4_en_set_rx_mode,
2494 	.ndo_set_mac_address	= mlx4_en_set_mac,
2495 	.ndo_validate_addr	= eth_validate_addr,
2496 	.ndo_change_mtu		= mlx4_en_change_mtu,
2497 	.ndo_do_ioctl		= mlx4_en_ioctl,
2498 	.ndo_tx_timeout		= mlx4_en_tx_timeout,
2499 	.ndo_vlan_rx_add_vid	= mlx4_en_vlan_rx_add_vid,
2500 	.ndo_vlan_rx_kill_vid	= mlx4_en_vlan_rx_kill_vid,
2501 #ifdef CONFIG_NET_POLL_CONTROLLER
2502 	.ndo_poll_controller	= mlx4_en_netpoll,
2503 #endif
2504 	.ndo_set_features	= mlx4_en_set_features,
2505 	.ndo_fix_features	= mlx4_en_fix_features,
2506 	.ndo_setup_tc		= mlx4_en_setup_tc,
2507 #ifdef CONFIG_RFS_ACCEL
2508 	.ndo_rx_flow_steer	= mlx4_en_filter_rfs,
2509 #endif
2510 #ifdef CONFIG_NET_RX_BUSY_POLL
2511 	.ndo_busy_poll		= mlx4_en_low_latency_recv,
2512 #endif
2513 	.ndo_get_phys_port_id	= mlx4_en_get_phys_port_id,
2514 #ifdef CONFIG_MLX4_EN_VXLAN
2515 	.ndo_add_vxlan_port	= mlx4_en_add_vxlan_port,
2516 	.ndo_del_vxlan_port	= mlx4_en_del_vxlan_port,
2517 	.ndo_features_check	= mlx4_en_features_check,
2518 #endif
2519 	.ndo_set_tx_maxrate	= mlx4_en_set_tx_maxrate,
2520 };
2521 
2522 static const struct net_device_ops mlx4_netdev_ops_master = {
2523 	.ndo_open		= mlx4_en_open,
2524 	.ndo_stop		= mlx4_en_close,
2525 	.ndo_start_xmit		= mlx4_en_xmit,
2526 	.ndo_select_queue	= mlx4_en_select_queue,
2527 	.ndo_get_stats		= mlx4_en_get_stats,
2528 	.ndo_set_rx_mode	= mlx4_en_set_rx_mode,
2529 	.ndo_set_mac_address	= mlx4_en_set_mac,
2530 	.ndo_validate_addr	= eth_validate_addr,
2531 	.ndo_change_mtu		= mlx4_en_change_mtu,
2532 	.ndo_tx_timeout		= mlx4_en_tx_timeout,
2533 	.ndo_vlan_rx_add_vid	= mlx4_en_vlan_rx_add_vid,
2534 	.ndo_vlan_rx_kill_vid	= mlx4_en_vlan_rx_kill_vid,
2535 	.ndo_set_vf_mac		= mlx4_en_set_vf_mac,
2536 	.ndo_set_vf_vlan	= mlx4_en_set_vf_vlan,
2537 	.ndo_set_vf_rate	= mlx4_en_set_vf_rate,
2538 	.ndo_set_vf_spoofchk	= mlx4_en_set_vf_spoofchk,
2539 	.ndo_set_vf_link_state	= mlx4_en_set_vf_link_state,
2540 	.ndo_get_vf_stats       = mlx4_en_get_vf_stats,
2541 	.ndo_get_vf_config	= mlx4_en_get_vf_config,
2542 #ifdef CONFIG_NET_POLL_CONTROLLER
2543 	.ndo_poll_controller	= mlx4_en_netpoll,
2544 #endif
2545 	.ndo_set_features	= mlx4_en_set_features,
2546 	.ndo_fix_features	= mlx4_en_fix_features,
2547 	.ndo_setup_tc		= mlx4_en_setup_tc,
2548 #ifdef CONFIG_RFS_ACCEL
2549 	.ndo_rx_flow_steer	= mlx4_en_filter_rfs,
2550 #endif
2551 	.ndo_get_phys_port_id	= mlx4_en_get_phys_port_id,
2552 #ifdef CONFIG_MLX4_EN_VXLAN
2553 	.ndo_add_vxlan_port	= mlx4_en_add_vxlan_port,
2554 	.ndo_del_vxlan_port	= mlx4_en_del_vxlan_port,
2555 	.ndo_features_check	= mlx4_en_features_check,
2556 #endif
2557 	.ndo_set_tx_maxrate	= mlx4_en_set_tx_maxrate,
2558 };
2559 
2560 struct mlx4_en_bond {
2561 	struct work_struct work;
2562 	struct mlx4_en_priv *priv;
2563 	int is_bonded;
2564 	struct mlx4_port_map port_map;
2565 };
2566 
2567 static void mlx4_en_bond_work(struct work_struct *work)
2568 {
2569 	struct mlx4_en_bond *bond = container_of(work,
2570 						     struct mlx4_en_bond,
2571 						     work);
2572 	int err = 0;
2573 	struct mlx4_dev *dev = bond->priv->mdev->dev;
2574 
2575 	if (bond->is_bonded) {
2576 		if (!mlx4_is_bonded(dev)) {
2577 			err = mlx4_bond(dev);
2578 			if (err)
2579 				en_err(bond->priv, "Fail to bond device\n");
2580 		}
2581 		if (!err) {
2582 			err = mlx4_port_map_set(dev, &bond->port_map);
2583 			if (err)
2584 				en_err(bond->priv, "Fail to set port map [%d][%d]: %d\n",
2585 				       bond->port_map.port1,
2586 				       bond->port_map.port2,
2587 				       err);
2588 		}
2589 	} else if (mlx4_is_bonded(dev)) {
2590 		err = mlx4_unbond(dev);
2591 		if (err)
2592 			en_err(bond->priv, "Fail to unbond device\n");
2593 	}
2594 	dev_put(bond->priv->dev);
2595 	kfree(bond);
2596 }
2597 
2598 static int mlx4_en_queue_bond_work(struct mlx4_en_priv *priv, int is_bonded,
2599 				   u8 v2p_p1, u8 v2p_p2)
2600 {
2601 	struct mlx4_en_bond *bond = NULL;
2602 
2603 	bond = kzalloc(sizeof(*bond), GFP_ATOMIC);
2604 	if (!bond)
2605 		return -ENOMEM;
2606 
2607 	INIT_WORK(&bond->work, mlx4_en_bond_work);
2608 	bond->priv = priv;
2609 	bond->is_bonded = is_bonded;
2610 	bond->port_map.port1 = v2p_p1;
2611 	bond->port_map.port2 = v2p_p2;
2612 	dev_hold(priv->dev);
2613 	queue_work(priv->mdev->workqueue, &bond->work);
2614 	return 0;
2615 }
2616 
2617 int mlx4_en_netdev_event(struct notifier_block *this,
2618 			 unsigned long event, void *ptr)
2619 {
2620 	struct net_device *ndev = netdev_notifier_info_to_dev(ptr);
2621 	u8 port = 0;
2622 	struct mlx4_en_dev *mdev;
2623 	struct mlx4_dev *dev;
2624 	int i, num_eth_ports = 0;
2625 	bool do_bond = true;
2626 	struct mlx4_en_priv *priv;
2627 	u8 v2p_port1 = 0;
2628 	u8 v2p_port2 = 0;
2629 
2630 	if (!net_eq(dev_net(ndev), &init_net))
2631 		return NOTIFY_DONE;
2632 
2633 	mdev = container_of(this, struct mlx4_en_dev, nb);
2634 	dev = mdev->dev;
2635 
2636 	/* Go into this mode only when two network devices set on two ports
2637 	 * of the same mlx4 device are slaves of the same bonding master
2638 	 */
2639 	mlx4_foreach_port(i, dev, MLX4_PORT_TYPE_ETH) {
2640 		++num_eth_ports;
2641 		if (!port && (mdev->pndev[i] == ndev))
2642 			port = i;
2643 		mdev->upper[i] = mdev->pndev[i] ?
2644 			netdev_master_upper_dev_get(mdev->pndev[i]) : NULL;
2645 		/* condition not met: network device is a slave */
2646 		if (!mdev->upper[i])
2647 			do_bond = false;
2648 		if (num_eth_ports < 2)
2649 			continue;
2650 		/* condition not met: same master */
2651 		if (mdev->upper[i] != mdev->upper[i-1])
2652 			do_bond = false;
2653 	}
2654 	/* condition not met: 2 salves */
2655 	do_bond = (num_eth_ports ==  2) ? do_bond : false;
2656 
2657 	/* handle only events that come with enough info */
2658 	if ((do_bond && (event != NETDEV_BONDING_INFO)) || !port)
2659 		return NOTIFY_DONE;
2660 
2661 	priv = netdev_priv(ndev);
2662 	if (do_bond) {
2663 		struct netdev_notifier_bonding_info *notifier_info = ptr;
2664 		struct netdev_bonding_info *bonding_info =
2665 			&notifier_info->bonding_info;
2666 
2667 		/* required mode 1, 2 or 4 */
2668 		if ((bonding_info->master.bond_mode != BOND_MODE_ACTIVEBACKUP) &&
2669 		    (bonding_info->master.bond_mode != BOND_MODE_XOR) &&
2670 		    (bonding_info->master.bond_mode != BOND_MODE_8023AD))
2671 			do_bond = false;
2672 
2673 		/* require exactly 2 slaves */
2674 		if (bonding_info->master.num_slaves != 2)
2675 			do_bond = false;
2676 
2677 		/* calc v2p */
2678 		if (do_bond) {
2679 			if (bonding_info->master.bond_mode ==
2680 			    BOND_MODE_ACTIVEBACKUP) {
2681 				/* in active-backup mode virtual ports are
2682 				 * mapped to the physical port of the active
2683 				 * slave */
2684 				if (bonding_info->slave.state ==
2685 				    BOND_STATE_BACKUP) {
2686 					if (port == 1) {
2687 						v2p_port1 = 2;
2688 						v2p_port2 = 2;
2689 					} else {
2690 						v2p_port1 = 1;
2691 						v2p_port2 = 1;
2692 					}
2693 				} else { /* BOND_STATE_ACTIVE */
2694 					if (port == 1) {
2695 						v2p_port1 = 1;
2696 						v2p_port2 = 1;
2697 					} else {
2698 						v2p_port1 = 2;
2699 						v2p_port2 = 2;
2700 					}
2701 				}
2702 			} else { /* Active-Active */
2703 				/* in active-active mode a virtual port is
2704 				 * mapped to the native physical port if and only
2705 				 * if the physical port is up */
2706 				__s8 link = bonding_info->slave.link;
2707 
2708 				if (port == 1)
2709 					v2p_port2 = 2;
2710 				else
2711 					v2p_port1 = 1;
2712 				if ((link == BOND_LINK_UP) ||
2713 				    (link == BOND_LINK_FAIL)) {
2714 					if (port == 1)
2715 						v2p_port1 = 1;
2716 					else
2717 						v2p_port2 = 2;
2718 				} else { /* BOND_LINK_DOWN || BOND_LINK_BACK */
2719 					if (port == 1)
2720 						v2p_port1 = 2;
2721 					else
2722 						v2p_port2 = 1;
2723 				}
2724 			}
2725 		}
2726 	}
2727 
2728 	mlx4_en_queue_bond_work(priv, do_bond,
2729 				v2p_port1, v2p_port2);
2730 
2731 	return NOTIFY_DONE;
2732 }
2733 
2734 void mlx4_en_update_pfc_stats_bitmap(struct mlx4_dev *dev,
2735 				     struct mlx4_en_stats_bitmap *stats_bitmap,
2736 				     u8 rx_ppp, u8 rx_pause,
2737 				     u8 tx_ppp, u8 tx_pause)
2738 {
2739 	int last_i = NUM_MAIN_STATS + NUM_PORT_STATS + NUM_PF_STATS;
2740 
2741 	if (!mlx4_is_slave(dev) &&
2742 	    (dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_FLOWSTATS_EN)) {
2743 		mutex_lock(&stats_bitmap->mutex);
2744 		bitmap_clear(stats_bitmap->bitmap, last_i, NUM_FLOW_STATS);
2745 
2746 		if (rx_ppp)
2747 			bitmap_set(stats_bitmap->bitmap, last_i,
2748 				   NUM_FLOW_PRIORITY_STATS_RX);
2749 		last_i += NUM_FLOW_PRIORITY_STATS_RX;
2750 
2751 		if (rx_pause && !(rx_ppp))
2752 			bitmap_set(stats_bitmap->bitmap, last_i,
2753 				   NUM_FLOW_STATS_RX);
2754 		last_i += NUM_FLOW_STATS_RX;
2755 
2756 		if (tx_ppp)
2757 			bitmap_set(stats_bitmap->bitmap, last_i,
2758 				   NUM_FLOW_PRIORITY_STATS_TX);
2759 		last_i += NUM_FLOW_PRIORITY_STATS_TX;
2760 
2761 		if (tx_pause && !(tx_ppp))
2762 			bitmap_set(stats_bitmap->bitmap, last_i,
2763 				   NUM_FLOW_STATS_TX);
2764 		last_i += NUM_FLOW_STATS_TX;
2765 
2766 		mutex_unlock(&stats_bitmap->mutex);
2767 	}
2768 }
2769 
2770 void mlx4_en_set_stats_bitmap(struct mlx4_dev *dev,
2771 			      struct mlx4_en_stats_bitmap *stats_bitmap,
2772 			      u8 rx_ppp, u8 rx_pause,
2773 			      u8 tx_ppp, u8 tx_pause)
2774 {
2775 	int last_i = 0;
2776 
2777 	mutex_init(&stats_bitmap->mutex);
2778 	bitmap_zero(stats_bitmap->bitmap, NUM_ALL_STATS);
2779 
2780 	if (mlx4_is_slave(dev)) {
2781 		bitmap_set(stats_bitmap->bitmap, last_i +
2782 					 MLX4_FIND_NETDEV_STAT(rx_packets), 1);
2783 		bitmap_set(stats_bitmap->bitmap, last_i +
2784 					 MLX4_FIND_NETDEV_STAT(tx_packets), 1);
2785 		bitmap_set(stats_bitmap->bitmap, last_i +
2786 					 MLX4_FIND_NETDEV_STAT(rx_bytes), 1);
2787 		bitmap_set(stats_bitmap->bitmap, last_i +
2788 					 MLX4_FIND_NETDEV_STAT(tx_bytes), 1);
2789 		bitmap_set(stats_bitmap->bitmap, last_i +
2790 					 MLX4_FIND_NETDEV_STAT(rx_dropped), 1);
2791 		bitmap_set(stats_bitmap->bitmap, last_i +
2792 					 MLX4_FIND_NETDEV_STAT(tx_dropped), 1);
2793 	} else {
2794 		bitmap_set(stats_bitmap->bitmap, last_i, NUM_MAIN_STATS);
2795 	}
2796 	last_i += NUM_MAIN_STATS;
2797 
2798 	bitmap_set(stats_bitmap->bitmap, last_i, NUM_PORT_STATS);
2799 	last_i += NUM_PORT_STATS;
2800 
2801 	if (mlx4_is_master(dev))
2802 		bitmap_set(stats_bitmap->bitmap, last_i,
2803 			   NUM_PF_STATS);
2804 	last_i += NUM_PF_STATS;
2805 
2806 	mlx4_en_update_pfc_stats_bitmap(dev, stats_bitmap,
2807 					rx_ppp, rx_pause,
2808 					tx_ppp, tx_pause);
2809 	last_i += NUM_FLOW_STATS;
2810 
2811 	if (!mlx4_is_slave(dev))
2812 		bitmap_set(stats_bitmap->bitmap, last_i, NUM_PKT_STATS);
2813 }
2814 
2815 int mlx4_en_init_netdev(struct mlx4_en_dev *mdev, int port,
2816 			struct mlx4_en_port_profile *prof)
2817 {
2818 	struct net_device *dev;
2819 	struct mlx4_en_priv *priv;
2820 	int i;
2821 	int err;
2822 
2823 	dev = alloc_etherdev_mqs(sizeof(struct mlx4_en_priv),
2824 				 MAX_TX_RINGS, MAX_RX_RINGS);
2825 	if (dev == NULL)
2826 		return -ENOMEM;
2827 
2828 	netif_set_real_num_tx_queues(dev, prof->tx_ring_num);
2829 	netif_set_real_num_rx_queues(dev, prof->rx_ring_num);
2830 
2831 	SET_NETDEV_DEV(dev, &mdev->dev->persist->pdev->dev);
2832 	dev->dev_port = port - 1;
2833 
2834 	/*
2835 	 * Initialize driver private data
2836 	 */
2837 
2838 	priv = netdev_priv(dev);
2839 	memset(priv, 0, sizeof(struct mlx4_en_priv));
2840 	priv->counter_index = MLX4_SINK_COUNTER_INDEX(mdev->dev);
2841 	spin_lock_init(&priv->stats_lock);
2842 	INIT_WORK(&priv->rx_mode_task, mlx4_en_do_set_rx_mode);
2843 	INIT_WORK(&priv->watchdog_task, mlx4_en_restart);
2844 	INIT_WORK(&priv->linkstate_task, mlx4_en_linkstate);
2845 	INIT_DELAYED_WORK(&priv->stats_task, mlx4_en_do_get_stats);
2846 	INIT_DELAYED_WORK(&priv->service_task, mlx4_en_service_task);
2847 #ifdef CONFIG_MLX4_EN_VXLAN
2848 	INIT_WORK(&priv->vxlan_add_task, mlx4_en_add_vxlan_offloads);
2849 	INIT_WORK(&priv->vxlan_del_task, mlx4_en_del_vxlan_offloads);
2850 #endif
2851 #ifdef CONFIG_RFS_ACCEL
2852 	INIT_LIST_HEAD(&priv->filters);
2853 	spin_lock_init(&priv->filters_lock);
2854 #endif
2855 
2856 	priv->dev = dev;
2857 	priv->mdev = mdev;
2858 	priv->ddev = &mdev->pdev->dev;
2859 	priv->prof = prof;
2860 	priv->port = port;
2861 	priv->port_up = false;
2862 	priv->flags = prof->flags;
2863 	priv->pflags = MLX4_EN_PRIV_FLAGS_BLUEFLAME;
2864 	priv->ctrl_flags = cpu_to_be32(MLX4_WQE_CTRL_CQ_UPDATE |
2865 			MLX4_WQE_CTRL_SOLICITED);
2866 	priv->num_tx_rings_p_up = mdev->profile.num_tx_rings_p_up;
2867 	priv->tx_ring_num = prof->tx_ring_num;
2868 	priv->tx_work_limit = MLX4_EN_DEFAULT_TX_WORK;
2869 	netdev_rss_key_fill(priv->rss_key, sizeof(priv->rss_key));
2870 
2871 	priv->tx_ring = kzalloc(sizeof(struct mlx4_en_tx_ring *) * MAX_TX_RINGS,
2872 				GFP_KERNEL);
2873 	if (!priv->tx_ring) {
2874 		err = -ENOMEM;
2875 		goto out;
2876 	}
2877 	priv->tx_cq = kzalloc(sizeof(struct mlx4_en_cq *) * MAX_TX_RINGS,
2878 			      GFP_KERNEL);
2879 	if (!priv->tx_cq) {
2880 		err = -ENOMEM;
2881 		goto out;
2882 	}
2883 	priv->rx_ring_num = prof->rx_ring_num;
2884 	priv->cqe_factor = (mdev->dev->caps.cqe_size == 64) ? 1 : 0;
2885 	priv->cqe_size = mdev->dev->caps.cqe_size;
2886 	priv->mac_index = -1;
2887 	priv->msg_enable = MLX4_EN_MSG_LEVEL;
2888 #ifdef CONFIG_MLX4_EN_DCB
2889 	if (!mlx4_is_slave(priv->mdev->dev)) {
2890 		if (mdev->dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_ETS_CFG) {
2891 			dev->dcbnl_ops = &mlx4_en_dcbnl_ops;
2892 		} else {
2893 			en_info(priv, "enabling only PFC DCB ops\n");
2894 			dev->dcbnl_ops = &mlx4_en_dcbnl_pfc_ops;
2895 		}
2896 	}
2897 #endif
2898 
2899 	for (i = 0; i < MLX4_EN_MAC_HASH_SIZE; ++i)
2900 		INIT_HLIST_HEAD(&priv->mac_hash[i]);
2901 
2902 	/* Query for default mac and max mtu */
2903 	priv->max_mtu = mdev->dev->caps.eth_mtu_cap[priv->port];
2904 
2905 	if (mdev->dev->caps.rx_checksum_flags_port[priv->port] &
2906 	    MLX4_RX_CSUM_MODE_VAL_NON_TCP_UDP)
2907 		priv->flags |= MLX4_EN_FLAG_RX_CSUM_NON_TCP_UDP;
2908 
2909 	/* Set default MAC */
2910 	dev->addr_len = ETH_ALEN;
2911 	mlx4_en_u64_to_mac(dev->dev_addr, mdev->dev->caps.def_mac[priv->port]);
2912 	if (!is_valid_ether_addr(dev->dev_addr)) {
2913 		en_err(priv, "Port: %d, invalid mac burned: %pM, quiting\n",
2914 		       priv->port, dev->dev_addr);
2915 		err = -EINVAL;
2916 		goto out;
2917 	} else if (mlx4_is_slave(priv->mdev->dev) &&
2918 		   (priv->mdev->dev->port_random_macs & 1 << priv->port)) {
2919 		/* Random MAC was assigned in mlx4_slave_cap
2920 		 * in mlx4_core module
2921 		 */
2922 		dev->addr_assign_type |= NET_ADDR_RANDOM;
2923 		en_warn(priv, "Assigned random MAC address %pM\n", dev->dev_addr);
2924 	}
2925 
2926 	memcpy(priv->current_mac, dev->dev_addr, sizeof(priv->current_mac));
2927 
2928 	priv->stride = roundup_pow_of_two(sizeof(struct mlx4_en_rx_desc) +
2929 					  DS_SIZE * MLX4_EN_MAX_RX_FRAGS);
2930 	err = mlx4_en_alloc_resources(priv);
2931 	if (err)
2932 		goto out;
2933 
2934 	/* Initialize time stamping config */
2935 	priv->hwtstamp_config.flags = 0;
2936 	priv->hwtstamp_config.tx_type = HWTSTAMP_TX_OFF;
2937 	priv->hwtstamp_config.rx_filter = HWTSTAMP_FILTER_NONE;
2938 
2939 	/* Allocate page for receive rings */
2940 	err = mlx4_alloc_hwq_res(mdev->dev, &priv->res,
2941 				MLX4_EN_PAGE_SIZE, MLX4_EN_PAGE_SIZE);
2942 	if (err) {
2943 		en_err(priv, "Failed to allocate page for rx qps\n");
2944 		goto out;
2945 	}
2946 	priv->allocated = 1;
2947 
2948 	/*
2949 	 * Initialize netdev entry points
2950 	 */
2951 	if (mlx4_is_master(priv->mdev->dev))
2952 		dev->netdev_ops = &mlx4_netdev_ops_master;
2953 	else
2954 		dev->netdev_ops = &mlx4_netdev_ops;
2955 	dev->watchdog_timeo = MLX4_EN_WATCHDOG_TIMEOUT;
2956 	netif_set_real_num_tx_queues(dev, priv->tx_ring_num);
2957 	netif_set_real_num_rx_queues(dev, priv->rx_ring_num);
2958 
2959 	dev->ethtool_ops = &mlx4_en_ethtool_ops;
2960 
2961 	/*
2962 	 * Set driver features
2963 	 */
2964 	dev->hw_features = NETIF_F_SG | NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM;
2965 	if (mdev->LSO_support)
2966 		dev->hw_features |= NETIF_F_TSO | NETIF_F_TSO6;
2967 
2968 	dev->vlan_features = dev->hw_features;
2969 
2970 	dev->hw_features |= NETIF_F_RXCSUM | NETIF_F_RXHASH;
2971 	dev->features = dev->hw_features | NETIF_F_HIGHDMA |
2972 			NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_CTAG_RX |
2973 			NETIF_F_HW_VLAN_CTAG_FILTER;
2974 	dev->hw_features |= NETIF_F_LOOPBACK |
2975 			NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_CTAG_RX;
2976 
2977 	if (!(mdev->dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_SKIP_OUTER_VLAN)) {
2978 		dev->features |= NETIF_F_HW_VLAN_STAG_RX |
2979 			NETIF_F_HW_VLAN_STAG_FILTER;
2980 		dev->hw_features |= NETIF_F_HW_VLAN_STAG_RX;
2981 	}
2982 
2983 	if (mlx4_is_slave(mdev->dev)) {
2984 		int phv;
2985 
2986 		err = get_phv_bit(mdev->dev, port, &phv);
2987 		if (!err && phv) {
2988 			dev->hw_features |= NETIF_F_HW_VLAN_STAG_TX;
2989 			priv->pflags |= MLX4_EN_PRIV_FLAGS_PHV;
2990 		}
2991 	} else {
2992 		if (mdev->dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_PHV_EN &&
2993 		    !(mdev->dev->caps.flags2 &
2994 		      MLX4_DEV_CAP_FLAG2_SKIP_OUTER_VLAN))
2995 			dev->hw_features |= NETIF_F_HW_VLAN_STAG_TX;
2996 	}
2997 
2998 	if (mdev->dev->caps.flags & MLX4_DEV_CAP_FLAG_FCS_KEEP)
2999 		dev->hw_features |= NETIF_F_RXFCS;
3000 
3001 	if (mdev->dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_IGNORE_FCS)
3002 		dev->hw_features |= NETIF_F_RXALL;
3003 
3004 	if (mdev->dev->caps.steering_mode ==
3005 	    MLX4_STEERING_MODE_DEVICE_MANAGED &&
3006 	    mdev->dev->caps.dmfs_high_steer_mode != MLX4_STEERING_DMFS_A0_STATIC)
3007 		dev->hw_features |= NETIF_F_NTUPLE;
3008 
3009 	if (mdev->dev->caps.steering_mode != MLX4_STEERING_MODE_A0)
3010 		dev->priv_flags |= IFF_UNICAST_FLT;
3011 
3012 	/* Setting a default hash function value */
3013 	if (mdev->dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_RSS_TOP) {
3014 		priv->rss_hash_fn = ETH_RSS_HASH_TOP;
3015 	} else if (mdev->dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_RSS_XOR) {
3016 		priv->rss_hash_fn = ETH_RSS_HASH_XOR;
3017 	} else {
3018 		en_warn(priv,
3019 			"No RSS hash capabilities exposed, using Toeplitz\n");
3020 		priv->rss_hash_fn = ETH_RSS_HASH_TOP;
3021 	}
3022 
3023 	mdev->pndev[port] = dev;
3024 	mdev->upper[port] = NULL;
3025 
3026 	netif_carrier_off(dev);
3027 	mlx4_en_set_default_moderation(priv);
3028 
3029 	en_warn(priv, "Using %d TX rings\n", prof->tx_ring_num);
3030 	en_warn(priv, "Using %d RX rings\n", prof->rx_ring_num);
3031 
3032 	mlx4_en_update_loopback_state(priv->dev, priv->dev->features);
3033 
3034 	/* Configure port */
3035 	mlx4_en_calc_rx_buf(dev);
3036 	err = mlx4_SET_PORT_general(mdev->dev, priv->port,
3037 				    priv->rx_skb_size + ETH_FCS_LEN,
3038 				    prof->tx_pause, prof->tx_ppp,
3039 				    prof->rx_pause, prof->rx_ppp);
3040 	if (err) {
3041 		en_err(priv, "Failed setting port general configurations for port %d, with error %d\n",
3042 		       priv->port, err);
3043 		goto out;
3044 	}
3045 
3046 	if (mdev->dev->caps.tunnel_offload_mode == MLX4_TUNNEL_OFFLOAD_MODE_VXLAN) {
3047 		err = mlx4_SET_PORT_VXLAN(mdev->dev, priv->port, VXLAN_STEER_BY_OUTER_MAC, 1);
3048 		if (err) {
3049 			en_err(priv, "Failed setting port L2 tunnel configuration, err %d\n",
3050 			       err);
3051 			goto out;
3052 		}
3053 	}
3054 
3055 	/* Init port */
3056 	en_warn(priv, "Initializing port\n");
3057 	err = mlx4_INIT_PORT(mdev->dev, priv->port);
3058 	if (err) {
3059 		en_err(priv, "Failed Initializing port\n");
3060 		goto out;
3061 	}
3062 	queue_delayed_work(mdev->workqueue, &priv->stats_task, STATS_DELAY);
3063 
3064 	/* Initialize time stamp mechanism */
3065 	if (mdev->dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_TS)
3066 		mlx4_en_init_timestamp(mdev);
3067 
3068 	queue_delayed_work(mdev->workqueue, &priv->service_task,
3069 			   SERVICE_TASK_DELAY);
3070 
3071 	mlx4_en_set_stats_bitmap(mdev->dev, &priv->stats_bitmap,
3072 				 mdev->profile.prof[priv->port].rx_ppp,
3073 				 mdev->profile.prof[priv->port].rx_pause,
3074 				 mdev->profile.prof[priv->port].tx_ppp,
3075 				 mdev->profile.prof[priv->port].tx_pause);
3076 
3077 	err = register_netdev(dev);
3078 	if (err) {
3079 		en_err(priv, "Netdev registration failed for port %d\n", port);
3080 		goto out;
3081 	}
3082 
3083 	priv->registered = 1;
3084 
3085 	return 0;
3086 
3087 out:
3088 	mlx4_en_destroy_netdev(dev);
3089 	return err;
3090 }
3091 
3092 int mlx4_en_reset_config(struct net_device *dev,
3093 			 struct hwtstamp_config ts_config,
3094 			 netdev_features_t features)
3095 {
3096 	struct mlx4_en_priv *priv = netdev_priv(dev);
3097 	struct mlx4_en_dev *mdev = priv->mdev;
3098 	int port_up = 0;
3099 	int err = 0;
3100 
3101 	if (priv->hwtstamp_config.tx_type == ts_config.tx_type &&
3102 	    priv->hwtstamp_config.rx_filter == ts_config.rx_filter &&
3103 	    !DEV_FEATURE_CHANGED(dev, features, NETIF_F_HW_VLAN_CTAG_RX) &&
3104 	    !DEV_FEATURE_CHANGED(dev, features, NETIF_F_RXFCS))
3105 		return 0; /* Nothing to change */
3106 
3107 	if (DEV_FEATURE_CHANGED(dev, features, NETIF_F_HW_VLAN_CTAG_RX) &&
3108 	    (features & NETIF_F_HW_VLAN_CTAG_RX) &&
3109 	    (priv->hwtstamp_config.rx_filter != HWTSTAMP_FILTER_NONE)) {
3110 		en_warn(priv, "Can't turn ON rx vlan offload while time-stamping rx filter is ON\n");
3111 		return -EINVAL;
3112 	}
3113 
3114 	mutex_lock(&mdev->state_lock);
3115 	if (priv->port_up) {
3116 		port_up = 1;
3117 		mlx4_en_stop_port(dev, 1);
3118 	}
3119 
3120 	mlx4_en_free_resources(priv);
3121 
3122 	en_warn(priv, "Changing device configuration rx filter(%x) rx vlan(%x)\n",
3123 		ts_config.rx_filter, !!(features & NETIF_F_HW_VLAN_CTAG_RX));
3124 
3125 	priv->hwtstamp_config.tx_type = ts_config.tx_type;
3126 	priv->hwtstamp_config.rx_filter = ts_config.rx_filter;
3127 
3128 	if (DEV_FEATURE_CHANGED(dev, features, NETIF_F_HW_VLAN_CTAG_RX)) {
3129 		if (features & NETIF_F_HW_VLAN_CTAG_RX)
3130 			dev->features |= NETIF_F_HW_VLAN_CTAG_RX;
3131 		else
3132 			dev->features &= ~NETIF_F_HW_VLAN_CTAG_RX;
3133 	} else if (ts_config.rx_filter == HWTSTAMP_FILTER_NONE) {
3134 		/* RX time-stamping is OFF, update the RX vlan offload
3135 		 * to the latest wanted state
3136 		 */
3137 		if (dev->wanted_features & NETIF_F_HW_VLAN_CTAG_RX)
3138 			dev->features |= NETIF_F_HW_VLAN_CTAG_RX;
3139 		else
3140 			dev->features &= ~NETIF_F_HW_VLAN_CTAG_RX;
3141 	}
3142 
3143 	if (DEV_FEATURE_CHANGED(dev, features, NETIF_F_RXFCS)) {
3144 		if (features & NETIF_F_RXFCS)
3145 			dev->features |= NETIF_F_RXFCS;
3146 		else
3147 			dev->features &= ~NETIF_F_RXFCS;
3148 	}
3149 
3150 	/* RX vlan offload and RX time-stamping can't co-exist !
3151 	 * Regardless of the caller's choice,
3152 	 * Turn Off RX vlan offload in case of time-stamping is ON
3153 	 */
3154 	if (ts_config.rx_filter != HWTSTAMP_FILTER_NONE) {
3155 		if (dev->features & NETIF_F_HW_VLAN_CTAG_RX)
3156 			en_warn(priv, "Turning off RX vlan offload since RX time-stamping is ON\n");
3157 		dev->features &= ~NETIF_F_HW_VLAN_CTAG_RX;
3158 	}
3159 
3160 	err = mlx4_en_alloc_resources(priv);
3161 	if (err) {
3162 		en_err(priv, "Failed reallocating port resources\n");
3163 		goto out;
3164 	}
3165 	if (port_up) {
3166 		err = mlx4_en_start_port(dev);
3167 		if (err)
3168 			en_err(priv, "Failed starting port\n");
3169 	}
3170 
3171 out:
3172 	mutex_unlock(&mdev->state_lock);
3173 	netdev_features_change(dev);
3174 	return err;
3175 }
3176