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 #include <net/devlink.h>
44 
45 #include <linux/mlx4/driver.h>
46 #include <linux/mlx4/device.h>
47 #include <linux/mlx4/cmd.h>
48 #include <linux/mlx4/cq.h>
49 
50 #include "mlx4_en.h"
51 #include "en_port.h"
52 
53 int mlx4_en_setup_tc(struct net_device *dev, u8 up)
54 {
55 	struct mlx4_en_priv *priv = netdev_priv(dev);
56 	int i;
57 	unsigned int offset = 0;
58 
59 	if (up && up != MLX4_EN_NUM_UP)
60 		return -EINVAL;
61 
62 	netdev_set_num_tc(dev, up);
63 
64 	/* Partition Tx queues evenly amongst UP's */
65 	for (i = 0; i < up; i++) {
66 		netdev_set_tc_queue(dev, i, priv->num_tx_rings_p_up, offset);
67 		offset += priv->num_tx_rings_p_up;
68 	}
69 
70 	return 0;
71 }
72 
73 static int __mlx4_en_setup_tc(struct net_device *dev, u32 handle, __be16 proto,
74 			      struct tc_to_netdev *tc)
75 {
76 	if (tc->type != TC_SETUP_MQPRIO)
77 		return -EINVAL;
78 
79 	return mlx4_en_setup_tc(dev, tc->tc);
80 }
81 
82 #ifdef CONFIG_RFS_ACCEL
83 
84 struct mlx4_en_filter {
85 	struct list_head next;
86 	struct work_struct work;
87 
88 	u8     ip_proto;
89 	__be32 src_ip;
90 	__be32 dst_ip;
91 	__be16 src_port;
92 	__be16 dst_port;
93 
94 	int rxq_index;
95 	struct mlx4_en_priv *priv;
96 	u32 flow_id;			/* RFS infrastructure id */
97 	int id;				/* mlx4_en driver id */
98 	u64 reg_id;			/* Flow steering API id */
99 	u8 activated;			/* Used to prevent expiry before filter
100 					 * is attached
101 					 */
102 	struct hlist_node filter_chain;
103 };
104 
105 static void mlx4_en_filter_rfs_expire(struct mlx4_en_priv *priv);
106 
107 static enum mlx4_net_trans_rule_id mlx4_ip_proto_to_trans_rule_id(u8 ip_proto)
108 {
109 	switch (ip_proto) {
110 	case IPPROTO_UDP:
111 		return MLX4_NET_TRANS_RULE_ID_UDP;
112 	case IPPROTO_TCP:
113 		return MLX4_NET_TRANS_RULE_ID_TCP;
114 	default:
115 		return MLX4_NET_TRANS_RULE_NUM;
116 	}
117 };
118 
119 static void mlx4_en_filter_work(struct work_struct *work)
120 {
121 	struct mlx4_en_filter *filter = container_of(work,
122 						     struct mlx4_en_filter,
123 						     work);
124 	struct mlx4_en_priv *priv = filter->priv;
125 	struct mlx4_spec_list spec_tcp_udp = {
126 		.id = mlx4_ip_proto_to_trans_rule_id(filter->ip_proto),
127 		{
128 			.tcp_udp = {
129 				.dst_port = filter->dst_port,
130 				.dst_port_msk = (__force __be16)-1,
131 				.src_port = filter->src_port,
132 				.src_port_msk = (__force __be16)-1,
133 			},
134 		},
135 	};
136 	struct mlx4_spec_list spec_ip = {
137 		.id = MLX4_NET_TRANS_RULE_ID_IPV4,
138 		{
139 			.ipv4 = {
140 				.dst_ip = filter->dst_ip,
141 				.dst_ip_msk = (__force __be32)-1,
142 				.src_ip = filter->src_ip,
143 				.src_ip_msk = (__force __be32)-1,
144 			},
145 		},
146 	};
147 	struct mlx4_spec_list spec_eth = {
148 		.id = MLX4_NET_TRANS_RULE_ID_ETH,
149 	};
150 	struct mlx4_net_trans_rule rule = {
151 		.list = LIST_HEAD_INIT(rule.list),
152 		.queue_mode = MLX4_NET_TRANS_Q_LIFO,
153 		.exclusive = 1,
154 		.allow_loopback = 1,
155 		.promisc_mode = MLX4_FS_REGULAR,
156 		.port = priv->port,
157 		.priority = MLX4_DOMAIN_RFS,
158 	};
159 	int rc;
160 	__be64 mac_mask = cpu_to_be64(MLX4_MAC_MASK << 16);
161 
162 	if (spec_tcp_udp.id >= MLX4_NET_TRANS_RULE_NUM) {
163 		en_warn(priv, "RFS: ignoring unsupported ip protocol (%d)\n",
164 			filter->ip_proto);
165 		goto ignore;
166 	}
167 	list_add_tail(&spec_eth.list, &rule.list);
168 	list_add_tail(&spec_ip.list, &rule.list);
169 	list_add_tail(&spec_tcp_udp.list, &rule.list);
170 
171 	rule.qpn = priv->rss_map.qps[filter->rxq_index].qpn;
172 	memcpy(spec_eth.eth.dst_mac, priv->dev->dev_addr, ETH_ALEN);
173 	memcpy(spec_eth.eth.dst_mac_msk, &mac_mask, ETH_ALEN);
174 
175 	filter->activated = 0;
176 
177 	if (filter->reg_id) {
178 		rc = mlx4_flow_detach(priv->mdev->dev, filter->reg_id);
179 		if (rc && rc != -ENOENT)
180 			en_err(priv, "Error detaching flow. rc = %d\n", rc);
181 	}
182 
183 	rc = mlx4_flow_attach(priv->mdev->dev, &rule, &filter->reg_id);
184 	if (rc)
185 		en_err(priv, "Error attaching flow. err = %d\n", rc);
186 
187 ignore:
188 	mlx4_en_filter_rfs_expire(priv);
189 
190 	filter->activated = 1;
191 }
192 
193 static inline struct hlist_head *
194 filter_hash_bucket(struct mlx4_en_priv *priv, __be32 src_ip, __be32 dst_ip,
195 		   __be16 src_port, __be16 dst_port)
196 {
197 	unsigned long l;
198 	int bucket_idx;
199 
200 	l = (__force unsigned long)src_port |
201 	    ((__force unsigned long)dst_port << 2);
202 	l ^= (__force unsigned long)(src_ip ^ dst_ip);
203 
204 	bucket_idx = hash_long(l, MLX4_EN_FILTER_HASH_SHIFT);
205 
206 	return &priv->filter_hash[bucket_idx];
207 }
208 
209 static struct mlx4_en_filter *
210 mlx4_en_filter_alloc(struct mlx4_en_priv *priv, int rxq_index, __be32 src_ip,
211 		     __be32 dst_ip, u8 ip_proto, __be16 src_port,
212 		     __be16 dst_port, u32 flow_id)
213 {
214 	struct mlx4_en_filter *filter = NULL;
215 
216 	filter = kzalloc(sizeof(struct mlx4_en_filter), GFP_ATOMIC);
217 	if (!filter)
218 		return NULL;
219 
220 	filter->priv = priv;
221 	filter->rxq_index = rxq_index;
222 	INIT_WORK(&filter->work, mlx4_en_filter_work);
223 
224 	filter->src_ip = src_ip;
225 	filter->dst_ip = dst_ip;
226 	filter->ip_proto = ip_proto;
227 	filter->src_port = src_port;
228 	filter->dst_port = dst_port;
229 
230 	filter->flow_id = flow_id;
231 
232 	filter->id = priv->last_filter_id++ % RPS_NO_FILTER;
233 
234 	list_add_tail(&filter->next, &priv->filters);
235 	hlist_add_head(&filter->filter_chain,
236 		       filter_hash_bucket(priv, src_ip, dst_ip, src_port,
237 					  dst_port));
238 
239 	return filter;
240 }
241 
242 static void mlx4_en_filter_free(struct mlx4_en_filter *filter)
243 {
244 	struct mlx4_en_priv *priv = filter->priv;
245 	int rc;
246 
247 	list_del(&filter->next);
248 
249 	rc = mlx4_flow_detach(priv->mdev->dev, filter->reg_id);
250 	if (rc && rc != -ENOENT)
251 		en_err(priv, "Error detaching flow. rc = %d\n", rc);
252 
253 	kfree(filter);
254 }
255 
256 static inline struct mlx4_en_filter *
257 mlx4_en_filter_find(struct mlx4_en_priv *priv, __be32 src_ip, __be32 dst_ip,
258 		    u8 ip_proto, __be16 src_port, __be16 dst_port)
259 {
260 	struct mlx4_en_filter *filter;
261 	struct mlx4_en_filter *ret = NULL;
262 
263 	hlist_for_each_entry(filter,
264 			     filter_hash_bucket(priv, src_ip, dst_ip,
265 						src_port, dst_port),
266 			     filter_chain) {
267 		if (filter->src_ip == src_ip &&
268 		    filter->dst_ip == dst_ip &&
269 		    filter->ip_proto == ip_proto &&
270 		    filter->src_port == src_port &&
271 		    filter->dst_port == dst_port) {
272 			ret = filter;
273 			break;
274 		}
275 	}
276 
277 	return ret;
278 }
279 
280 static int
281 mlx4_en_filter_rfs(struct net_device *net_dev, const struct sk_buff *skb,
282 		   u16 rxq_index, u32 flow_id)
283 {
284 	struct mlx4_en_priv *priv = netdev_priv(net_dev);
285 	struct mlx4_en_filter *filter;
286 	const struct iphdr *ip;
287 	const __be16 *ports;
288 	u8 ip_proto;
289 	__be32 src_ip;
290 	__be32 dst_ip;
291 	__be16 src_port;
292 	__be16 dst_port;
293 	int nhoff = skb_network_offset(skb);
294 	int ret = 0;
295 
296 	if (skb->protocol != htons(ETH_P_IP))
297 		return -EPROTONOSUPPORT;
298 
299 	ip = (const struct iphdr *)(skb->data + nhoff);
300 	if (ip_is_fragment(ip))
301 		return -EPROTONOSUPPORT;
302 
303 	if ((ip->protocol != IPPROTO_TCP) && (ip->protocol != IPPROTO_UDP))
304 		return -EPROTONOSUPPORT;
305 	ports = (const __be16 *)(skb->data + nhoff + 4 * ip->ihl);
306 
307 	ip_proto = ip->protocol;
308 	src_ip = ip->saddr;
309 	dst_ip = ip->daddr;
310 	src_port = ports[0];
311 	dst_port = ports[1];
312 
313 	spin_lock_bh(&priv->filters_lock);
314 	filter = mlx4_en_filter_find(priv, src_ip, dst_ip, ip_proto,
315 				     src_port, dst_port);
316 	if (filter) {
317 		if (filter->rxq_index == rxq_index)
318 			goto out;
319 
320 		filter->rxq_index = rxq_index;
321 	} else {
322 		filter = mlx4_en_filter_alloc(priv, rxq_index,
323 					      src_ip, dst_ip, ip_proto,
324 					      src_port, dst_port, flow_id);
325 		if (!filter) {
326 			ret = -ENOMEM;
327 			goto err;
328 		}
329 	}
330 
331 	queue_work(priv->mdev->workqueue, &filter->work);
332 
333 out:
334 	ret = filter->id;
335 err:
336 	spin_unlock_bh(&priv->filters_lock);
337 
338 	return ret;
339 }
340 
341 void mlx4_en_cleanup_filters(struct mlx4_en_priv *priv)
342 {
343 	struct mlx4_en_filter *filter, *tmp;
344 	LIST_HEAD(del_list);
345 
346 	spin_lock_bh(&priv->filters_lock);
347 	list_for_each_entry_safe(filter, tmp, &priv->filters, next) {
348 		list_move(&filter->next, &del_list);
349 		hlist_del(&filter->filter_chain);
350 	}
351 	spin_unlock_bh(&priv->filters_lock);
352 
353 	list_for_each_entry_safe(filter, tmp, &del_list, next) {
354 		cancel_work_sync(&filter->work);
355 		mlx4_en_filter_free(filter);
356 	}
357 }
358 
359 static void mlx4_en_filter_rfs_expire(struct mlx4_en_priv *priv)
360 {
361 	struct mlx4_en_filter *filter = NULL, *tmp, *last_filter = NULL;
362 	LIST_HEAD(del_list);
363 	int i = 0;
364 
365 	spin_lock_bh(&priv->filters_lock);
366 	list_for_each_entry_safe(filter, tmp, &priv->filters, next) {
367 		if (i > MLX4_EN_FILTER_EXPIRY_QUOTA)
368 			break;
369 
370 		if (filter->activated &&
371 		    !work_pending(&filter->work) &&
372 		    rps_may_expire_flow(priv->dev,
373 					filter->rxq_index, filter->flow_id,
374 					filter->id)) {
375 			list_move(&filter->next, &del_list);
376 			hlist_del(&filter->filter_chain);
377 		} else
378 			last_filter = filter;
379 
380 		i++;
381 	}
382 
383 	if (last_filter && (&last_filter->next != priv->filters.next))
384 		list_move(&priv->filters, &last_filter->next);
385 
386 	spin_unlock_bh(&priv->filters_lock);
387 
388 	list_for_each_entry_safe(filter, tmp, &del_list, next)
389 		mlx4_en_filter_free(filter);
390 }
391 #endif
392 
393 static int mlx4_en_vlan_rx_add_vid(struct net_device *dev,
394 				   __be16 proto, u16 vid)
395 {
396 	struct mlx4_en_priv *priv = netdev_priv(dev);
397 	struct mlx4_en_dev *mdev = priv->mdev;
398 	int err;
399 	int idx;
400 
401 	en_dbg(HW, priv, "adding VLAN:%d\n", vid);
402 
403 	set_bit(vid, priv->active_vlans);
404 
405 	/* Add VID to port VLAN filter */
406 	mutex_lock(&mdev->state_lock);
407 	if (mdev->device_up && priv->port_up) {
408 		err = mlx4_SET_VLAN_FLTR(mdev->dev, priv);
409 		if (err)
410 			en_err(priv, "Failed configuring VLAN filter\n");
411 	}
412 	if (mlx4_register_vlan(mdev->dev, priv->port, vid, &idx))
413 		en_dbg(HW, priv, "failed adding vlan %d\n", vid);
414 	mutex_unlock(&mdev->state_lock);
415 
416 	return 0;
417 }
418 
419 static int mlx4_en_vlan_rx_kill_vid(struct net_device *dev,
420 				    __be16 proto, u16 vid)
421 {
422 	struct mlx4_en_priv *priv = netdev_priv(dev);
423 	struct mlx4_en_dev *mdev = priv->mdev;
424 	int err;
425 
426 	en_dbg(HW, priv, "Killing VID:%d\n", vid);
427 
428 	clear_bit(vid, priv->active_vlans);
429 
430 	/* Remove VID from port VLAN filter */
431 	mutex_lock(&mdev->state_lock);
432 	mlx4_unregister_vlan(mdev->dev, priv->port, vid);
433 
434 	if (mdev->device_up && priv->port_up) {
435 		err = mlx4_SET_VLAN_FLTR(mdev->dev, priv);
436 		if (err)
437 			en_err(priv, "Failed configuring VLAN filter\n");
438 	}
439 	mutex_unlock(&mdev->state_lock);
440 
441 	return 0;
442 }
443 
444 static void mlx4_en_u64_to_mac(unsigned char dst_mac[ETH_ALEN + 2], u64 src_mac)
445 {
446 	int i;
447 	for (i = ETH_ALEN - 1; i >= 0; --i) {
448 		dst_mac[i] = src_mac & 0xff;
449 		src_mac >>= 8;
450 	}
451 	memset(&dst_mac[ETH_ALEN], 0, 2);
452 }
453 
454 
455 static int mlx4_en_tunnel_steer_add(struct mlx4_en_priv *priv, unsigned char *addr,
456 				    int qpn, u64 *reg_id)
457 {
458 	int err;
459 
460 	if (priv->mdev->dev->caps.tunnel_offload_mode != MLX4_TUNNEL_OFFLOAD_MODE_VXLAN ||
461 	    priv->mdev->dev->caps.dmfs_high_steer_mode == MLX4_STEERING_DMFS_A0_STATIC)
462 		return 0; /* do nothing */
463 
464 	err = mlx4_tunnel_steer_add(priv->mdev->dev, addr, priv->port, qpn,
465 				    MLX4_DOMAIN_NIC, reg_id);
466 	if (err) {
467 		en_err(priv, "failed to add vxlan steering rule, err %d\n", err);
468 		return err;
469 	}
470 	en_dbg(DRV, priv, "added vxlan steering rule, mac %pM reg_id %llx\n", addr, *reg_id);
471 	return 0;
472 }
473 
474 
475 static int mlx4_en_uc_steer_add(struct mlx4_en_priv *priv,
476 				unsigned char *mac, int *qpn, u64 *reg_id)
477 {
478 	struct mlx4_en_dev *mdev = priv->mdev;
479 	struct mlx4_dev *dev = mdev->dev;
480 	int err;
481 
482 	switch (dev->caps.steering_mode) {
483 	case MLX4_STEERING_MODE_B0: {
484 		struct mlx4_qp qp;
485 		u8 gid[16] = {0};
486 
487 		qp.qpn = *qpn;
488 		memcpy(&gid[10], mac, ETH_ALEN);
489 		gid[5] = priv->port;
490 
491 		err = mlx4_unicast_attach(dev, &qp, gid, 0, MLX4_PROT_ETH);
492 		break;
493 	}
494 	case MLX4_STEERING_MODE_DEVICE_MANAGED: {
495 		struct mlx4_spec_list spec_eth = { {NULL} };
496 		__be64 mac_mask = cpu_to_be64(MLX4_MAC_MASK << 16);
497 
498 		struct mlx4_net_trans_rule rule = {
499 			.queue_mode = MLX4_NET_TRANS_Q_FIFO,
500 			.exclusive = 0,
501 			.allow_loopback = 1,
502 			.promisc_mode = MLX4_FS_REGULAR,
503 			.priority = MLX4_DOMAIN_NIC,
504 		};
505 
506 		rule.port = priv->port;
507 		rule.qpn = *qpn;
508 		INIT_LIST_HEAD(&rule.list);
509 
510 		spec_eth.id = MLX4_NET_TRANS_RULE_ID_ETH;
511 		memcpy(spec_eth.eth.dst_mac, mac, ETH_ALEN);
512 		memcpy(spec_eth.eth.dst_mac_msk, &mac_mask, ETH_ALEN);
513 		list_add_tail(&spec_eth.list, &rule.list);
514 
515 		err = mlx4_flow_attach(dev, &rule, reg_id);
516 		break;
517 	}
518 	default:
519 		return -EINVAL;
520 	}
521 	if (err)
522 		en_warn(priv, "Failed Attaching Unicast\n");
523 
524 	return err;
525 }
526 
527 static void mlx4_en_uc_steer_release(struct mlx4_en_priv *priv,
528 				     unsigned char *mac, int qpn, u64 reg_id)
529 {
530 	struct mlx4_en_dev *mdev = priv->mdev;
531 	struct mlx4_dev *dev = mdev->dev;
532 
533 	switch (dev->caps.steering_mode) {
534 	case MLX4_STEERING_MODE_B0: {
535 		struct mlx4_qp qp;
536 		u8 gid[16] = {0};
537 
538 		qp.qpn = qpn;
539 		memcpy(&gid[10], mac, ETH_ALEN);
540 		gid[5] = priv->port;
541 
542 		mlx4_unicast_detach(dev, &qp, gid, MLX4_PROT_ETH);
543 		break;
544 	}
545 	case MLX4_STEERING_MODE_DEVICE_MANAGED: {
546 		mlx4_flow_detach(dev, reg_id);
547 		break;
548 	}
549 	default:
550 		en_err(priv, "Invalid steering mode.\n");
551 	}
552 }
553 
554 static int mlx4_en_get_qp(struct mlx4_en_priv *priv)
555 {
556 	struct mlx4_en_dev *mdev = priv->mdev;
557 	struct mlx4_dev *dev = mdev->dev;
558 	int index = 0;
559 	int err = 0;
560 	int *qpn = &priv->base_qpn;
561 	u64 mac = mlx4_mac_to_u64(priv->dev->dev_addr);
562 
563 	en_dbg(DRV, priv, "Registering MAC: %pM for adding\n",
564 	       priv->dev->dev_addr);
565 	index = mlx4_register_mac(dev, priv->port, mac);
566 	if (index < 0) {
567 		err = index;
568 		en_err(priv, "Failed adding MAC: %pM\n",
569 		       priv->dev->dev_addr);
570 		return err;
571 	}
572 
573 	if (dev->caps.steering_mode == MLX4_STEERING_MODE_A0) {
574 		int base_qpn = mlx4_get_base_qpn(dev, priv->port);
575 		*qpn = base_qpn + index;
576 		return 0;
577 	}
578 
579 	err = mlx4_qp_reserve_range(dev, 1, 1, qpn, MLX4_RESERVE_A0_QP);
580 	en_dbg(DRV, priv, "Reserved qp %d\n", *qpn);
581 	if (err) {
582 		en_err(priv, "Failed to reserve qp for mac registration\n");
583 		mlx4_unregister_mac(dev, priv->port, mac);
584 		return err;
585 	}
586 
587 	return 0;
588 }
589 
590 static void mlx4_en_put_qp(struct mlx4_en_priv *priv)
591 {
592 	struct mlx4_en_dev *mdev = priv->mdev;
593 	struct mlx4_dev *dev = mdev->dev;
594 	int qpn = priv->base_qpn;
595 
596 	if (dev->caps.steering_mode == MLX4_STEERING_MODE_A0) {
597 		u64 mac = mlx4_mac_to_u64(priv->dev->dev_addr);
598 		en_dbg(DRV, priv, "Registering MAC: %pM for deleting\n",
599 		       priv->dev->dev_addr);
600 		mlx4_unregister_mac(dev, priv->port, mac);
601 	} else {
602 		en_dbg(DRV, priv, "Releasing qp: port %d, qpn %d\n",
603 		       priv->port, qpn);
604 		mlx4_qp_release_range(dev, qpn, 1);
605 		priv->flags &= ~MLX4_EN_FLAG_FORCE_PROMISC;
606 	}
607 }
608 
609 static int mlx4_en_replace_mac(struct mlx4_en_priv *priv, int qpn,
610 			       unsigned char *new_mac, unsigned char *prev_mac)
611 {
612 	struct mlx4_en_dev *mdev = priv->mdev;
613 	struct mlx4_dev *dev = mdev->dev;
614 	int err = 0;
615 	u64 new_mac_u64 = mlx4_mac_to_u64(new_mac);
616 
617 	if (dev->caps.steering_mode != MLX4_STEERING_MODE_A0) {
618 		struct hlist_head *bucket;
619 		unsigned int mac_hash;
620 		struct mlx4_mac_entry *entry;
621 		struct hlist_node *tmp;
622 		u64 prev_mac_u64 = mlx4_mac_to_u64(prev_mac);
623 
624 		bucket = &priv->mac_hash[prev_mac[MLX4_EN_MAC_HASH_IDX]];
625 		hlist_for_each_entry_safe(entry, tmp, bucket, hlist) {
626 			if (ether_addr_equal_64bits(entry->mac, prev_mac)) {
627 				mlx4_en_uc_steer_release(priv, entry->mac,
628 							 qpn, entry->reg_id);
629 				mlx4_unregister_mac(dev, priv->port,
630 						    prev_mac_u64);
631 				hlist_del_rcu(&entry->hlist);
632 				synchronize_rcu();
633 				memcpy(entry->mac, new_mac, ETH_ALEN);
634 				entry->reg_id = 0;
635 				mac_hash = new_mac[MLX4_EN_MAC_HASH_IDX];
636 				hlist_add_head_rcu(&entry->hlist,
637 						   &priv->mac_hash[mac_hash]);
638 				mlx4_register_mac(dev, priv->port, new_mac_u64);
639 				err = mlx4_en_uc_steer_add(priv, new_mac,
640 							   &qpn,
641 							   &entry->reg_id);
642 				if (err)
643 					return err;
644 				if (priv->tunnel_reg_id) {
645 					mlx4_flow_detach(priv->mdev->dev, priv->tunnel_reg_id);
646 					priv->tunnel_reg_id = 0;
647 				}
648 				err = mlx4_en_tunnel_steer_add(priv, new_mac, qpn,
649 							       &priv->tunnel_reg_id);
650 				return err;
651 			}
652 		}
653 		return -EINVAL;
654 	}
655 
656 	return __mlx4_replace_mac(dev, priv->port, qpn, new_mac_u64);
657 }
658 
659 static int mlx4_en_do_set_mac(struct mlx4_en_priv *priv,
660 			      unsigned char new_mac[ETH_ALEN + 2])
661 {
662 	int err = 0;
663 
664 	if (priv->port_up) {
665 		/* Remove old MAC and insert the new one */
666 		err = mlx4_en_replace_mac(priv, priv->base_qpn,
667 					  new_mac, priv->current_mac);
668 		if (err)
669 			en_err(priv, "Failed changing HW MAC address\n");
670 	} else
671 		en_dbg(HW, priv, "Port is down while registering mac, exiting...\n");
672 
673 	if (!err)
674 		memcpy(priv->current_mac, new_mac, sizeof(priv->current_mac));
675 
676 	return err;
677 }
678 
679 static int mlx4_en_set_mac(struct net_device *dev, void *addr)
680 {
681 	struct mlx4_en_priv *priv = netdev_priv(dev);
682 	struct mlx4_en_dev *mdev = priv->mdev;
683 	struct sockaddr *saddr = addr;
684 	unsigned char new_mac[ETH_ALEN + 2];
685 	int err;
686 
687 	if (!is_valid_ether_addr(saddr->sa_data))
688 		return -EADDRNOTAVAIL;
689 
690 	mutex_lock(&mdev->state_lock);
691 	memcpy(new_mac, saddr->sa_data, ETH_ALEN);
692 	err = mlx4_en_do_set_mac(priv, new_mac);
693 	if (!err)
694 		memcpy(dev->dev_addr, saddr->sa_data, ETH_ALEN);
695 	mutex_unlock(&mdev->state_lock);
696 
697 	return err;
698 }
699 
700 static void mlx4_en_clear_list(struct net_device *dev)
701 {
702 	struct mlx4_en_priv *priv = netdev_priv(dev);
703 	struct mlx4_en_mc_list *tmp, *mc_to_del;
704 
705 	list_for_each_entry_safe(mc_to_del, tmp, &priv->mc_list, list) {
706 		list_del(&mc_to_del->list);
707 		kfree(mc_to_del);
708 	}
709 }
710 
711 static void mlx4_en_cache_mclist(struct net_device *dev)
712 {
713 	struct mlx4_en_priv *priv = netdev_priv(dev);
714 	struct netdev_hw_addr *ha;
715 	struct mlx4_en_mc_list *tmp;
716 
717 	mlx4_en_clear_list(dev);
718 	netdev_for_each_mc_addr(ha, dev) {
719 		tmp = kzalloc(sizeof(struct mlx4_en_mc_list), GFP_ATOMIC);
720 		if (!tmp) {
721 			mlx4_en_clear_list(dev);
722 			return;
723 		}
724 		memcpy(tmp->addr, ha->addr, ETH_ALEN);
725 		list_add_tail(&tmp->list, &priv->mc_list);
726 	}
727 }
728 
729 static void update_mclist_flags(struct mlx4_en_priv *priv,
730 				struct list_head *dst,
731 				struct list_head *src)
732 {
733 	struct mlx4_en_mc_list *dst_tmp, *src_tmp, *new_mc;
734 	bool found;
735 
736 	/* Find all the entries that should be removed from dst,
737 	 * These are the entries that are not found in src
738 	 */
739 	list_for_each_entry(dst_tmp, dst, list) {
740 		found = false;
741 		list_for_each_entry(src_tmp, src, list) {
742 			if (ether_addr_equal(dst_tmp->addr, src_tmp->addr)) {
743 				found = true;
744 				break;
745 			}
746 		}
747 		if (!found)
748 			dst_tmp->action = MCLIST_REM;
749 	}
750 
751 	/* Add entries that exist in src but not in dst
752 	 * mark them as need to add
753 	 */
754 	list_for_each_entry(src_tmp, src, list) {
755 		found = false;
756 		list_for_each_entry(dst_tmp, dst, list) {
757 			if (ether_addr_equal(dst_tmp->addr, src_tmp->addr)) {
758 				dst_tmp->action = MCLIST_NONE;
759 				found = true;
760 				break;
761 			}
762 		}
763 		if (!found) {
764 			new_mc = kmemdup(src_tmp,
765 					 sizeof(struct mlx4_en_mc_list),
766 					 GFP_KERNEL);
767 			if (!new_mc)
768 				return;
769 
770 			new_mc->action = MCLIST_ADD;
771 			list_add_tail(&new_mc->list, dst);
772 		}
773 	}
774 }
775 
776 static void mlx4_en_set_rx_mode(struct net_device *dev)
777 {
778 	struct mlx4_en_priv *priv = netdev_priv(dev);
779 
780 	if (!priv->port_up)
781 		return;
782 
783 	queue_work(priv->mdev->workqueue, &priv->rx_mode_task);
784 }
785 
786 static void mlx4_en_set_promisc_mode(struct mlx4_en_priv *priv,
787 				     struct mlx4_en_dev *mdev)
788 {
789 	int err = 0;
790 
791 	if (!(priv->flags & MLX4_EN_FLAG_PROMISC)) {
792 		if (netif_msg_rx_status(priv))
793 			en_warn(priv, "Entering promiscuous mode\n");
794 		priv->flags |= MLX4_EN_FLAG_PROMISC;
795 
796 		/* Enable promiscouos mode */
797 		switch (mdev->dev->caps.steering_mode) {
798 		case MLX4_STEERING_MODE_DEVICE_MANAGED:
799 			err = mlx4_flow_steer_promisc_add(mdev->dev,
800 							  priv->port,
801 							  priv->base_qpn,
802 							  MLX4_FS_ALL_DEFAULT);
803 			if (err)
804 				en_err(priv, "Failed enabling promiscuous mode\n");
805 			priv->flags |= MLX4_EN_FLAG_MC_PROMISC;
806 			break;
807 
808 		case MLX4_STEERING_MODE_B0:
809 			err = mlx4_unicast_promisc_add(mdev->dev,
810 						       priv->base_qpn,
811 						       priv->port);
812 			if (err)
813 				en_err(priv, "Failed enabling unicast promiscuous mode\n");
814 
815 			/* Add the default qp number as multicast
816 			 * promisc
817 			 */
818 			if (!(priv->flags & MLX4_EN_FLAG_MC_PROMISC)) {
819 				err = mlx4_multicast_promisc_add(mdev->dev,
820 								 priv->base_qpn,
821 								 priv->port);
822 				if (err)
823 					en_err(priv, "Failed enabling multicast promiscuous mode\n");
824 				priv->flags |= MLX4_EN_FLAG_MC_PROMISC;
825 			}
826 			break;
827 
828 		case MLX4_STEERING_MODE_A0:
829 			err = mlx4_SET_PORT_qpn_calc(mdev->dev,
830 						     priv->port,
831 						     priv->base_qpn,
832 						     1);
833 			if (err)
834 				en_err(priv, "Failed enabling promiscuous mode\n");
835 			break;
836 		}
837 
838 		/* Disable port multicast filter (unconditionally) */
839 		err = mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, 0,
840 					  0, MLX4_MCAST_DISABLE);
841 		if (err)
842 			en_err(priv, "Failed disabling multicast filter\n");
843 	}
844 }
845 
846 static void mlx4_en_clear_promisc_mode(struct mlx4_en_priv *priv,
847 				       struct mlx4_en_dev *mdev)
848 {
849 	int err = 0;
850 
851 	if (netif_msg_rx_status(priv))
852 		en_warn(priv, "Leaving promiscuous mode\n");
853 	priv->flags &= ~MLX4_EN_FLAG_PROMISC;
854 
855 	/* Disable promiscouos mode */
856 	switch (mdev->dev->caps.steering_mode) {
857 	case MLX4_STEERING_MODE_DEVICE_MANAGED:
858 		err = mlx4_flow_steer_promisc_remove(mdev->dev,
859 						     priv->port,
860 						     MLX4_FS_ALL_DEFAULT);
861 		if (err)
862 			en_err(priv, "Failed disabling promiscuous mode\n");
863 		priv->flags &= ~MLX4_EN_FLAG_MC_PROMISC;
864 		break;
865 
866 	case MLX4_STEERING_MODE_B0:
867 		err = mlx4_unicast_promisc_remove(mdev->dev,
868 						  priv->base_qpn,
869 						  priv->port);
870 		if (err)
871 			en_err(priv, "Failed disabling unicast promiscuous mode\n");
872 		/* Disable Multicast promisc */
873 		if (priv->flags & MLX4_EN_FLAG_MC_PROMISC) {
874 			err = mlx4_multicast_promisc_remove(mdev->dev,
875 							    priv->base_qpn,
876 							    priv->port);
877 			if (err)
878 				en_err(priv, "Failed disabling multicast promiscuous mode\n");
879 			priv->flags &= ~MLX4_EN_FLAG_MC_PROMISC;
880 		}
881 		break;
882 
883 	case MLX4_STEERING_MODE_A0:
884 		err = mlx4_SET_PORT_qpn_calc(mdev->dev,
885 					     priv->port,
886 					     priv->base_qpn, 0);
887 		if (err)
888 			en_err(priv, "Failed disabling promiscuous mode\n");
889 		break;
890 	}
891 }
892 
893 static void mlx4_en_do_multicast(struct mlx4_en_priv *priv,
894 				 struct net_device *dev,
895 				 struct mlx4_en_dev *mdev)
896 {
897 	struct mlx4_en_mc_list *mclist, *tmp;
898 	u64 mcast_addr = 0;
899 	u8 mc_list[16] = {0};
900 	int err = 0;
901 
902 	/* Enable/disable the multicast filter according to IFF_ALLMULTI */
903 	if (dev->flags & IFF_ALLMULTI) {
904 		err = mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, 0,
905 					  0, MLX4_MCAST_DISABLE);
906 		if (err)
907 			en_err(priv, "Failed disabling multicast filter\n");
908 
909 		/* Add the default qp number as multicast promisc */
910 		if (!(priv->flags & MLX4_EN_FLAG_MC_PROMISC)) {
911 			switch (mdev->dev->caps.steering_mode) {
912 			case MLX4_STEERING_MODE_DEVICE_MANAGED:
913 				err = mlx4_flow_steer_promisc_add(mdev->dev,
914 								  priv->port,
915 								  priv->base_qpn,
916 								  MLX4_FS_MC_DEFAULT);
917 				break;
918 
919 			case MLX4_STEERING_MODE_B0:
920 				err = mlx4_multicast_promisc_add(mdev->dev,
921 								 priv->base_qpn,
922 								 priv->port);
923 				break;
924 
925 			case MLX4_STEERING_MODE_A0:
926 				break;
927 			}
928 			if (err)
929 				en_err(priv, "Failed entering multicast promisc mode\n");
930 			priv->flags |= MLX4_EN_FLAG_MC_PROMISC;
931 		}
932 	} else {
933 		/* Disable Multicast promisc */
934 		if (priv->flags & MLX4_EN_FLAG_MC_PROMISC) {
935 			switch (mdev->dev->caps.steering_mode) {
936 			case MLX4_STEERING_MODE_DEVICE_MANAGED:
937 				err = mlx4_flow_steer_promisc_remove(mdev->dev,
938 								     priv->port,
939 								     MLX4_FS_MC_DEFAULT);
940 				break;
941 
942 			case MLX4_STEERING_MODE_B0:
943 				err = mlx4_multicast_promisc_remove(mdev->dev,
944 								    priv->base_qpn,
945 								    priv->port);
946 				break;
947 
948 			case MLX4_STEERING_MODE_A0:
949 				break;
950 			}
951 			if (err)
952 				en_err(priv, "Failed disabling multicast promiscuous mode\n");
953 			priv->flags &= ~MLX4_EN_FLAG_MC_PROMISC;
954 		}
955 
956 		err = mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, 0,
957 					  0, MLX4_MCAST_DISABLE);
958 		if (err)
959 			en_err(priv, "Failed disabling multicast filter\n");
960 
961 		/* Flush mcast filter and init it with broadcast address */
962 		mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, ETH_BCAST,
963 				    1, MLX4_MCAST_CONFIG);
964 
965 		/* Update multicast list - we cache all addresses so they won't
966 		 * change while HW is updated holding the command semaphor */
967 		netif_addr_lock_bh(dev);
968 		mlx4_en_cache_mclist(dev);
969 		netif_addr_unlock_bh(dev);
970 		list_for_each_entry(mclist, &priv->mc_list, list) {
971 			mcast_addr = mlx4_mac_to_u64(mclist->addr);
972 			mlx4_SET_MCAST_FLTR(mdev->dev, priv->port,
973 					    mcast_addr, 0, MLX4_MCAST_CONFIG);
974 		}
975 		err = mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, 0,
976 					  0, MLX4_MCAST_ENABLE);
977 		if (err)
978 			en_err(priv, "Failed enabling multicast filter\n");
979 
980 		update_mclist_flags(priv, &priv->curr_list, &priv->mc_list);
981 		list_for_each_entry_safe(mclist, tmp, &priv->curr_list, list) {
982 			if (mclist->action == MCLIST_REM) {
983 				/* detach this address and delete from list */
984 				memcpy(&mc_list[10], mclist->addr, ETH_ALEN);
985 				mc_list[5] = priv->port;
986 				err = mlx4_multicast_detach(mdev->dev,
987 							    &priv->rss_map.indir_qp,
988 							    mc_list,
989 							    MLX4_PROT_ETH,
990 							    mclist->reg_id);
991 				if (err)
992 					en_err(priv, "Fail to detach multicast address\n");
993 
994 				if (mclist->tunnel_reg_id) {
995 					err = mlx4_flow_detach(priv->mdev->dev, mclist->tunnel_reg_id);
996 					if (err)
997 						en_err(priv, "Failed to detach multicast address\n");
998 				}
999 
1000 				/* remove from list */
1001 				list_del(&mclist->list);
1002 				kfree(mclist);
1003 			} else if (mclist->action == MCLIST_ADD) {
1004 				/* attach the address */
1005 				memcpy(&mc_list[10], mclist->addr, ETH_ALEN);
1006 				/* needed for B0 steering support */
1007 				mc_list[5] = priv->port;
1008 				err = mlx4_multicast_attach(mdev->dev,
1009 							    &priv->rss_map.indir_qp,
1010 							    mc_list,
1011 							    priv->port, 0,
1012 							    MLX4_PROT_ETH,
1013 							    &mclist->reg_id);
1014 				if (err)
1015 					en_err(priv, "Fail to attach multicast address\n");
1016 
1017 				err = mlx4_en_tunnel_steer_add(priv, &mc_list[10], priv->base_qpn,
1018 							       &mclist->tunnel_reg_id);
1019 				if (err)
1020 					en_err(priv, "Failed to attach multicast address\n");
1021 			}
1022 		}
1023 	}
1024 }
1025 
1026 static void mlx4_en_do_uc_filter(struct mlx4_en_priv *priv,
1027 				 struct net_device *dev,
1028 				 struct mlx4_en_dev *mdev)
1029 {
1030 	struct netdev_hw_addr *ha;
1031 	struct mlx4_mac_entry *entry;
1032 	struct hlist_node *tmp;
1033 	bool found;
1034 	u64 mac;
1035 	int err = 0;
1036 	struct hlist_head *bucket;
1037 	unsigned int i;
1038 	int removed = 0;
1039 	u32 prev_flags;
1040 
1041 	/* Note that we do not need to protect our mac_hash traversal with rcu,
1042 	 * since all modification code is protected by mdev->state_lock
1043 	 */
1044 
1045 	/* find what to remove */
1046 	for (i = 0; i < MLX4_EN_MAC_HASH_SIZE; ++i) {
1047 		bucket = &priv->mac_hash[i];
1048 		hlist_for_each_entry_safe(entry, tmp, bucket, hlist) {
1049 			found = false;
1050 			netdev_for_each_uc_addr(ha, dev) {
1051 				if (ether_addr_equal_64bits(entry->mac,
1052 							    ha->addr)) {
1053 					found = true;
1054 					break;
1055 				}
1056 			}
1057 
1058 			/* MAC address of the port is not in uc list */
1059 			if (ether_addr_equal_64bits(entry->mac,
1060 						    priv->current_mac))
1061 				found = true;
1062 
1063 			if (!found) {
1064 				mac = mlx4_mac_to_u64(entry->mac);
1065 				mlx4_en_uc_steer_release(priv, entry->mac,
1066 							 priv->base_qpn,
1067 							 entry->reg_id);
1068 				mlx4_unregister_mac(mdev->dev, priv->port, mac);
1069 
1070 				hlist_del_rcu(&entry->hlist);
1071 				kfree_rcu(entry, rcu);
1072 				en_dbg(DRV, priv, "Removed MAC %pM on port:%d\n",
1073 				       entry->mac, priv->port);
1074 				++removed;
1075 			}
1076 		}
1077 	}
1078 
1079 	/* if we didn't remove anything, there is no use in trying to add
1080 	 * again once we are in a forced promisc mode state
1081 	 */
1082 	if ((priv->flags & MLX4_EN_FLAG_FORCE_PROMISC) && 0 == removed)
1083 		return;
1084 
1085 	prev_flags = priv->flags;
1086 	priv->flags &= ~MLX4_EN_FLAG_FORCE_PROMISC;
1087 
1088 	/* find what to add */
1089 	netdev_for_each_uc_addr(ha, dev) {
1090 		found = false;
1091 		bucket = &priv->mac_hash[ha->addr[MLX4_EN_MAC_HASH_IDX]];
1092 		hlist_for_each_entry(entry, bucket, hlist) {
1093 			if (ether_addr_equal_64bits(entry->mac, ha->addr)) {
1094 				found = true;
1095 				break;
1096 			}
1097 		}
1098 
1099 		if (!found) {
1100 			entry = kmalloc(sizeof(*entry), GFP_KERNEL);
1101 			if (!entry) {
1102 				en_err(priv, "Failed adding MAC %pM on port:%d (out of memory)\n",
1103 				       ha->addr, priv->port);
1104 				priv->flags |= MLX4_EN_FLAG_FORCE_PROMISC;
1105 				break;
1106 			}
1107 			mac = mlx4_mac_to_u64(ha->addr);
1108 			memcpy(entry->mac, ha->addr, ETH_ALEN);
1109 			err = mlx4_register_mac(mdev->dev, priv->port, mac);
1110 			if (err < 0) {
1111 				en_err(priv, "Failed registering MAC %pM on port %d: %d\n",
1112 				       ha->addr, priv->port, err);
1113 				kfree(entry);
1114 				priv->flags |= MLX4_EN_FLAG_FORCE_PROMISC;
1115 				break;
1116 			}
1117 			err = mlx4_en_uc_steer_add(priv, ha->addr,
1118 						   &priv->base_qpn,
1119 						   &entry->reg_id);
1120 			if (err) {
1121 				en_err(priv, "Failed adding MAC %pM on port %d: %d\n",
1122 				       ha->addr, priv->port, err);
1123 				mlx4_unregister_mac(mdev->dev, priv->port, mac);
1124 				kfree(entry);
1125 				priv->flags |= MLX4_EN_FLAG_FORCE_PROMISC;
1126 				break;
1127 			} else {
1128 				unsigned int mac_hash;
1129 				en_dbg(DRV, priv, "Added MAC %pM on port:%d\n",
1130 				       ha->addr, priv->port);
1131 				mac_hash = ha->addr[MLX4_EN_MAC_HASH_IDX];
1132 				bucket = &priv->mac_hash[mac_hash];
1133 				hlist_add_head_rcu(&entry->hlist, bucket);
1134 			}
1135 		}
1136 	}
1137 
1138 	if (priv->flags & MLX4_EN_FLAG_FORCE_PROMISC) {
1139 		en_warn(priv, "Forcing promiscuous mode on port:%d\n",
1140 			priv->port);
1141 	} else if (prev_flags & MLX4_EN_FLAG_FORCE_PROMISC) {
1142 		en_warn(priv, "Stop forcing promiscuous mode on port:%d\n",
1143 			priv->port);
1144 	}
1145 }
1146 
1147 static void mlx4_en_do_set_rx_mode(struct work_struct *work)
1148 {
1149 	struct mlx4_en_priv *priv = container_of(work, struct mlx4_en_priv,
1150 						 rx_mode_task);
1151 	struct mlx4_en_dev *mdev = priv->mdev;
1152 	struct net_device *dev = priv->dev;
1153 
1154 	mutex_lock(&mdev->state_lock);
1155 	if (!mdev->device_up) {
1156 		en_dbg(HW, priv, "Card is not up, ignoring rx mode change.\n");
1157 		goto out;
1158 	}
1159 	if (!priv->port_up) {
1160 		en_dbg(HW, priv, "Port is down, ignoring rx mode change.\n");
1161 		goto out;
1162 	}
1163 
1164 	if (!netif_carrier_ok(dev)) {
1165 		if (!mlx4_en_QUERY_PORT(mdev, priv->port)) {
1166 			if (priv->port_state.link_state) {
1167 				priv->last_link_state = MLX4_DEV_EVENT_PORT_UP;
1168 				netif_carrier_on(dev);
1169 				en_dbg(LINK, priv, "Link Up\n");
1170 			}
1171 		}
1172 	}
1173 
1174 	if (dev->priv_flags & IFF_UNICAST_FLT)
1175 		mlx4_en_do_uc_filter(priv, dev, mdev);
1176 
1177 	/* Promsicuous mode: disable all filters */
1178 	if ((dev->flags & IFF_PROMISC) ||
1179 	    (priv->flags & MLX4_EN_FLAG_FORCE_PROMISC)) {
1180 		mlx4_en_set_promisc_mode(priv, mdev);
1181 		goto out;
1182 	}
1183 
1184 	/* Not in promiscuous mode */
1185 	if (priv->flags & MLX4_EN_FLAG_PROMISC)
1186 		mlx4_en_clear_promisc_mode(priv, mdev);
1187 
1188 	mlx4_en_do_multicast(priv, dev, mdev);
1189 out:
1190 	mutex_unlock(&mdev->state_lock);
1191 }
1192 
1193 #ifdef CONFIG_NET_POLL_CONTROLLER
1194 static void mlx4_en_netpoll(struct net_device *dev)
1195 {
1196 	struct mlx4_en_priv *priv = netdev_priv(dev);
1197 	struct mlx4_en_cq *cq;
1198 	int i;
1199 
1200 	for (i = 0; i < priv->rx_ring_num; i++) {
1201 		cq = priv->rx_cq[i];
1202 		napi_schedule(&cq->napi);
1203 	}
1204 }
1205 #endif
1206 
1207 static int mlx4_en_set_rss_steer_rules(struct mlx4_en_priv *priv)
1208 {
1209 	u64 reg_id;
1210 	int err = 0;
1211 	int *qpn = &priv->base_qpn;
1212 	struct mlx4_mac_entry *entry;
1213 
1214 	err = mlx4_en_uc_steer_add(priv, priv->dev->dev_addr, qpn, &reg_id);
1215 	if (err)
1216 		return err;
1217 
1218 	err = mlx4_en_tunnel_steer_add(priv, priv->dev->dev_addr, *qpn,
1219 				       &priv->tunnel_reg_id);
1220 	if (err)
1221 		goto tunnel_err;
1222 
1223 	entry = kmalloc(sizeof(*entry), GFP_KERNEL);
1224 	if (!entry) {
1225 		err = -ENOMEM;
1226 		goto alloc_err;
1227 	}
1228 
1229 	memcpy(entry->mac, priv->dev->dev_addr, sizeof(entry->mac));
1230 	memcpy(priv->current_mac, entry->mac, sizeof(priv->current_mac));
1231 	entry->reg_id = reg_id;
1232 	hlist_add_head_rcu(&entry->hlist,
1233 			   &priv->mac_hash[entry->mac[MLX4_EN_MAC_HASH_IDX]]);
1234 
1235 	return 0;
1236 
1237 alloc_err:
1238 	if (priv->tunnel_reg_id)
1239 		mlx4_flow_detach(priv->mdev->dev, priv->tunnel_reg_id);
1240 
1241 tunnel_err:
1242 	mlx4_en_uc_steer_release(priv, priv->dev->dev_addr, *qpn, reg_id);
1243 	return err;
1244 }
1245 
1246 static void mlx4_en_delete_rss_steer_rules(struct mlx4_en_priv *priv)
1247 {
1248 	u64 mac;
1249 	unsigned int i;
1250 	int qpn = priv->base_qpn;
1251 	struct hlist_head *bucket;
1252 	struct hlist_node *tmp;
1253 	struct mlx4_mac_entry *entry;
1254 
1255 	for (i = 0; i < MLX4_EN_MAC_HASH_SIZE; ++i) {
1256 		bucket = &priv->mac_hash[i];
1257 		hlist_for_each_entry_safe(entry, tmp, bucket, hlist) {
1258 			mac = mlx4_mac_to_u64(entry->mac);
1259 			en_dbg(DRV, priv, "Registering MAC:%pM for deleting\n",
1260 			       entry->mac);
1261 			mlx4_en_uc_steer_release(priv, entry->mac,
1262 						 qpn, entry->reg_id);
1263 
1264 			mlx4_unregister_mac(priv->mdev->dev, priv->port, mac);
1265 			hlist_del_rcu(&entry->hlist);
1266 			kfree_rcu(entry, rcu);
1267 		}
1268 	}
1269 
1270 	if (priv->tunnel_reg_id) {
1271 		mlx4_flow_detach(priv->mdev->dev, priv->tunnel_reg_id);
1272 		priv->tunnel_reg_id = 0;
1273 	}
1274 }
1275 
1276 static void mlx4_en_tx_timeout(struct net_device *dev)
1277 {
1278 	struct mlx4_en_priv *priv = netdev_priv(dev);
1279 	struct mlx4_en_dev *mdev = priv->mdev;
1280 	int i;
1281 
1282 	if (netif_msg_timer(priv))
1283 		en_warn(priv, "Tx timeout called on port:%d\n", priv->port);
1284 
1285 	for (i = 0; i < priv->tx_ring_num; i++) {
1286 		if (!netif_tx_queue_stopped(netdev_get_tx_queue(dev, i)))
1287 			continue;
1288 		en_warn(priv, "TX timeout on queue: %d, QP: 0x%x, CQ: 0x%x, Cons: 0x%x, Prod: 0x%x\n",
1289 			i, priv->tx_ring[i]->qpn, priv->tx_ring[i]->cqn,
1290 			priv->tx_ring[i]->cons, priv->tx_ring[i]->prod);
1291 	}
1292 
1293 	priv->port_stats.tx_timeout++;
1294 	en_dbg(DRV, priv, "Scheduling watchdog\n");
1295 	queue_work(mdev->workqueue, &priv->watchdog_task);
1296 }
1297 
1298 
1299 static struct rtnl_link_stats64 *
1300 mlx4_en_get_stats64(struct net_device *dev, struct rtnl_link_stats64 *stats)
1301 {
1302 	struct mlx4_en_priv *priv = netdev_priv(dev);
1303 
1304 	spin_lock_bh(&priv->stats_lock);
1305 	netdev_stats_to_stats64(stats, &dev->stats);
1306 	spin_unlock_bh(&priv->stats_lock);
1307 
1308 	return stats;
1309 }
1310 
1311 static void mlx4_en_set_default_moderation(struct mlx4_en_priv *priv)
1312 {
1313 	struct mlx4_en_cq *cq;
1314 	int i;
1315 
1316 	/* If we haven't received a specific coalescing setting
1317 	 * (module param), we set the moderation parameters as follows:
1318 	 * - moder_cnt is set to the number of mtu sized packets to
1319 	 *   satisfy our coalescing target.
1320 	 * - moder_time is set to a fixed value.
1321 	 */
1322 	priv->rx_frames = MLX4_EN_RX_COAL_TARGET;
1323 	priv->rx_usecs = MLX4_EN_RX_COAL_TIME;
1324 	priv->tx_frames = MLX4_EN_TX_COAL_PKTS;
1325 	priv->tx_usecs = MLX4_EN_TX_COAL_TIME;
1326 	en_dbg(INTR, priv, "Default coalesing params for mtu:%d - rx_frames:%d rx_usecs:%d\n",
1327 	       priv->dev->mtu, priv->rx_frames, priv->rx_usecs);
1328 
1329 	/* Setup cq moderation params */
1330 	for (i = 0; i < priv->rx_ring_num; i++) {
1331 		cq = priv->rx_cq[i];
1332 		cq->moder_cnt = priv->rx_frames;
1333 		cq->moder_time = priv->rx_usecs;
1334 		priv->last_moder_time[i] = MLX4_EN_AUTO_CONF;
1335 		priv->last_moder_packets[i] = 0;
1336 		priv->last_moder_bytes[i] = 0;
1337 	}
1338 
1339 	for (i = 0; i < priv->tx_ring_num; i++) {
1340 		cq = priv->tx_cq[i];
1341 		cq->moder_cnt = priv->tx_frames;
1342 		cq->moder_time = priv->tx_usecs;
1343 	}
1344 
1345 	/* Reset auto-moderation params */
1346 	priv->pkt_rate_low = MLX4_EN_RX_RATE_LOW;
1347 	priv->rx_usecs_low = MLX4_EN_RX_COAL_TIME_LOW;
1348 	priv->pkt_rate_high = MLX4_EN_RX_RATE_HIGH;
1349 	priv->rx_usecs_high = MLX4_EN_RX_COAL_TIME_HIGH;
1350 	priv->sample_interval = MLX4_EN_SAMPLE_INTERVAL;
1351 	priv->adaptive_rx_coal = 1;
1352 	priv->last_moder_jiffies = 0;
1353 	priv->last_moder_tx_packets = 0;
1354 }
1355 
1356 static void mlx4_en_auto_moderation(struct mlx4_en_priv *priv)
1357 {
1358 	unsigned long period = (unsigned long) (jiffies - priv->last_moder_jiffies);
1359 	struct mlx4_en_cq *cq;
1360 	unsigned long packets;
1361 	unsigned long rate;
1362 	unsigned long avg_pkt_size;
1363 	unsigned long rx_packets;
1364 	unsigned long rx_bytes;
1365 	unsigned long rx_pkt_diff;
1366 	int moder_time;
1367 	int ring, err;
1368 
1369 	if (!priv->adaptive_rx_coal || period < priv->sample_interval * HZ)
1370 		return;
1371 
1372 	for (ring = 0; ring < priv->rx_ring_num; ring++) {
1373 		spin_lock_bh(&priv->stats_lock);
1374 		rx_packets = priv->rx_ring[ring]->packets;
1375 		rx_bytes = priv->rx_ring[ring]->bytes;
1376 		spin_unlock_bh(&priv->stats_lock);
1377 
1378 		rx_pkt_diff = ((unsigned long) (rx_packets -
1379 				priv->last_moder_packets[ring]));
1380 		packets = rx_pkt_diff;
1381 		rate = packets * HZ / period;
1382 		avg_pkt_size = packets ? ((unsigned long) (rx_bytes -
1383 				priv->last_moder_bytes[ring])) / packets : 0;
1384 
1385 		/* Apply auto-moderation only when packet rate
1386 		 * exceeds a rate that it matters */
1387 		if (rate > (MLX4_EN_RX_RATE_THRESH / priv->rx_ring_num) &&
1388 		    avg_pkt_size > MLX4_EN_AVG_PKT_SMALL) {
1389 			if (rate < priv->pkt_rate_low)
1390 				moder_time = priv->rx_usecs_low;
1391 			else if (rate > priv->pkt_rate_high)
1392 				moder_time = priv->rx_usecs_high;
1393 			else
1394 				moder_time = (rate - priv->pkt_rate_low) *
1395 					(priv->rx_usecs_high - priv->rx_usecs_low) /
1396 					(priv->pkt_rate_high - priv->pkt_rate_low) +
1397 					priv->rx_usecs_low;
1398 		} else {
1399 			moder_time = priv->rx_usecs_low;
1400 		}
1401 
1402 		if (moder_time != priv->last_moder_time[ring]) {
1403 			priv->last_moder_time[ring] = moder_time;
1404 			cq = priv->rx_cq[ring];
1405 			cq->moder_time = moder_time;
1406 			cq->moder_cnt = priv->rx_frames;
1407 			err = mlx4_en_set_cq_moder(priv, cq);
1408 			if (err)
1409 				en_err(priv, "Failed modifying moderation for cq:%d\n",
1410 				       ring);
1411 		}
1412 		priv->last_moder_packets[ring] = rx_packets;
1413 		priv->last_moder_bytes[ring] = rx_bytes;
1414 	}
1415 
1416 	priv->last_moder_jiffies = jiffies;
1417 }
1418 
1419 static void mlx4_en_do_get_stats(struct work_struct *work)
1420 {
1421 	struct delayed_work *delay = to_delayed_work(work);
1422 	struct mlx4_en_priv *priv = container_of(delay, struct mlx4_en_priv,
1423 						 stats_task);
1424 	struct mlx4_en_dev *mdev = priv->mdev;
1425 	int err;
1426 
1427 	mutex_lock(&mdev->state_lock);
1428 	if (mdev->device_up) {
1429 		if (priv->port_up) {
1430 			err = mlx4_en_DUMP_ETH_STATS(mdev, priv->port, 0);
1431 			if (err)
1432 				en_dbg(HW, priv, "Could not update stats\n");
1433 
1434 			mlx4_en_auto_moderation(priv);
1435 		}
1436 
1437 		queue_delayed_work(mdev->workqueue, &priv->stats_task, STATS_DELAY);
1438 	}
1439 	if (mdev->mac_removed[MLX4_MAX_PORTS + 1 - priv->port]) {
1440 		mlx4_en_do_set_mac(priv, priv->current_mac);
1441 		mdev->mac_removed[MLX4_MAX_PORTS + 1 - priv->port] = 0;
1442 	}
1443 	mutex_unlock(&mdev->state_lock);
1444 }
1445 
1446 /* mlx4_en_service_task - Run service task for tasks that needed to be done
1447  * periodically
1448  */
1449 static void mlx4_en_service_task(struct work_struct *work)
1450 {
1451 	struct delayed_work *delay = to_delayed_work(work);
1452 	struct mlx4_en_priv *priv = container_of(delay, struct mlx4_en_priv,
1453 						 service_task);
1454 	struct mlx4_en_dev *mdev = priv->mdev;
1455 
1456 	mutex_lock(&mdev->state_lock);
1457 	if (mdev->device_up) {
1458 		if (mdev->dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_TS)
1459 			mlx4_en_ptp_overflow_check(mdev);
1460 
1461 		mlx4_en_recover_from_oom(priv);
1462 		queue_delayed_work(mdev->workqueue, &priv->service_task,
1463 				   SERVICE_TASK_DELAY);
1464 	}
1465 	mutex_unlock(&mdev->state_lock);
1466 }
1467 
1468 static void mlx4_en_linkstate(struct work_struct *work)
1469 {
1470 	struct mlx4_en_priv *priv = container_of(work, struct mlx4_en_priv,
1471 						 linkstate_task);
1472 	struct mlx4_en_dev *mdev = priv->mdev;
1473 	int linkstate = priv->link_state;
1474 
1475 	mutex_lock(&mdev->state_lock);
1476 	/* If observable port state changed set carrier state and
1477 	 * report to system log */
1478 	if (priv->last_link_state != linkstate) {
1479 		if (linkstate == MLX4_DEV_EVENT_PORT_DOWN) {
1480 			en_info(priv, "Link Down\n");
1481 			netif_carrier_off(priv->dev);
1482 		} else {
1483 			en_info(priv, "Link Up\n");
1484 			netif_carrier_on(priv->dev);
1485 		}
1486 	}
1487 	priv->last_link_state = linkstate;
1488 	mutex_unlock(&mdev->state_lock);
1489 }
1490 
1491 static int mlx4_en_init_affinity_hint(struct mlx4_en_priv *priv, int ring_idx)
1492 {
1493 	struct mlx4_en_rx_ring *ring = priv->rx_ring[ring_idx];
1494 	int numa_node = priv->mdev->dev->numa_node;
1495 
1496 	if (!zalloc_cpumask_var(&ring->affinity_mask, GFP_KERNEL))
1497 		return -ENOMEM;
1498 
1499 	cpumask_set_cpu(cpumask_local_spread(ring_idx, numa_node),
1500 			ring->affinity_mask);
1501 	return 0;
1502 }
1503 
1504 static void mlx4_en_free_affinity_hint(struct mlx4_en_priv *priv, int ring_idx)
1505 {
1506 	free_cpumask_var(priv->rx_ring[ring_idx]->affinity_mask);
1507 }
1508 
1509 int mlx4_en_start_port(struct net_device *dev)
1510 {
1511 	struct mlx4_en_priv *priv = netdev_priv(dev);
1512 	struct mlx4_en_dev *mdev = priv->mdev;
1513 	struct mlx4_en_cq *cq;
1514 	struct mlx4_en_tx_ring *tx_ring;
1515 	int rx_index = 0;
1516 	int tx_index = 0;
1517 	int err = 0;
1518 	int i;
1519 	int j;
1520 	u8 mc_list[16] = {0};
1521 
1522 	if (priv->port_up) {
1523 		en_dbg(DRV, priv, "start port called while port already up\n");
1524 		return 0;
1525 	}
1526 
1527 	INIT_LIST_HEAD(&priv->mc_list);
1528 	INIT_LIST_HEAD(&priv->curr_list);
1529 	INIT_LIST_HEAD(&priv->ethtool_list);
1530 	memset(&priv->ethtool_rules[0], 0,
1531 	       sizeof(struct ethtool_flow_id) * MAX_NUM_OF_FS_RULES);
1532 
1533 	/* Calculate Rx buf size */
1534 	dev->mtu = min(dev->mtu, priv->max_mtu);
1535 	mlx4_en_calc_rx_buf(dev);
1536 	en_dbg(DRV, priv, "Rx buf size:%d\n", priv->rx_skb_size);
1537 
1538 	/* Configure rx cq's and rings */
1539 	err = mlx4_en_activate_rx_rings(priv);
1540 	if (err) {
1541 		en_err(priv, "Failed to activate RX rings\n");
1542 		return err;
1543 	}
1544 	for (i = 0; i < priv->rx_ring_num; i++) {
1545 		cq = priv->rx_cq[i];
1546 
1547 		err = mlx4_en_init_affinity_hint(priv, i);
1548 		if (err) {
1549 			en_err(priv, "Failed preparing IRQ affinity hint\n");
1550 			goto cq_err;
1551 		}
1552 
1553 		err = mlx4_en_activate_cq(priv, cq, i);
1554 		if (err) {
1555 			en_err(priv, "Failed activating Rx CQ\n");
1556 			mlx4_en_free_affinity_hint(priv, i);
1557 			goto cq_err;
1558 		}
1559 
1560 		for (j = 0; j < cq->size; j++) {
1561 			struct mlx4_cqe *cqe = NULL;
1562 
1563 			cqe = mlx4_en_get_cqe(cq->buf, j, priv->cqe_size) +
1564 			      priv->cqe_factor;
1565 			cqe->owner_sr_opcode = MLX4_CQE_OWNER_MASK;
1566 		}
1567 
1568 		err = mlx4_en_set_cq_moder(priv, cq);
1569 		if (err) {
1570 			en_err(priv, "Failed setting cq moderation parameters\n");
1571 			mlx4_en_deactivate_cq(priv, cq);
1572 			mlx4_en_free_affinity_hint(priv, i);
1573 			goto cq_err;
1574 		}
1575 		mlx4_en_arm_cq(priv, cq);
1576 		priv->rx_ring[i]->cqn = cq->mcq.cqn;
1577 		++rx_index;
1578 	}
1579 
1580 	/* Set qp number */
1581 	en_dbg(DRV, priv, "Getting qp number for port %d\n", priv->port);
1582 	err = mlx4_en_get_qp(priv);
1583 	if (err) {
1584 		en_err(priv, "Failed getting eth qp\n");
1585 		goto cq_err;
1586 	}
1587 	mdev->mac_removed[priv->port] = 0;
1588 
1589 	priv->counter_index =
1590 			mlx4_get_default_counter_index(mdev->dev, priv->port);
1591 
1592 	err = mlx4_en_config_rss_steer(priv);
1593 	if (err) {
1594 		en_err(priv, "Failed configuring rss steering\n");
1595 		goto mac_err;
1596 	}
1597 
1598 	err = mlx4_en_create_drop_qp(priv);
1599 	if (err)
1600 		goto rss_err;
1601 
1602 	/* Configure tx cq's and rings */
1603 	for (i = 0; i < priv->tx_ring_num; i++) {
1604 		/* Configure cq */
1605 		cq = priv->tx_cq[i];
1606 		err = mlx4_en_activate_cq(priv, cq, i);
1607 		if (err) {
1608 			en_err(priv, "Failed allocating Tx CQ\n");
1609 			goto tx_err;
1610 		}
1611 		err = mlx4_en_set_cq_moder(priv, cq);
1612 		if (err) {
1613 			en_err(priv, "Failed setting cq moderation parameters\n");
1614 			mlx4_en_deactivate_cq(priv, cq);
1615 			goto tx_err;
1616 		}
1617 		en_dbg(DRV, priv, "Resetting index of collapsed CQ:%d to -1\n", i);
1618 		cq->buf->wqe_index = cpu_to_be16(0xffff);
1619 
1620 		/* Configure ring */
1621 		tx_ring = priv->tx_ring[i];
1622 		err = mlx4_en_activate_tx_ring(priv, tx_ring, cq->mcq.cqn,
1623 			i / priv->num_tx_rings_p_up);
1624 		if (err) {
1625 			en_err(priv, "Failed allocating Tx ring\n");
1626 			mlx4_en_deactivate_cq(priv, cq);
1627 			goto tx_err;
1628 		}
1629 		tx_ring->tx_queue = netdev_get_tx_queue(dev, i);
1630 
1631 		/* Arm CQ for TX completions */
1632 		mlx4_en_arm_cq(priv, cq);
1633 
1634 		/* Set initial ownership of all Tx TXBBs to SW (1) */
1635 		for (j = 0; j < tx_ring->buf_size; j += STAMP_STRIDE)
1636 			*((u32 *) (tx_ring->buf + j)) = 0xffffffff;
1637 		++tx_index;
1638 	}
1639 
1640 	/* Configure port */
1641 	err = mlx4_SET_PORT_general(mdev->dev, priv->port,
1642 				    priv->rx_skb_size + ETH_FCS_LEN,
1643 				    priv->prof->tx_pause,
1644 				    priv->prof->tx_ppp,
1645 				    priv->prof->rx_pause,
1646 				    priv->prof->rx_ppp);
1647 	if (err) {
1648 		en_err(priv, "Failed setting port general configurations for port %d, with error %d\n",
1649 		       priv->port, err);
1650 		goto tx_err;
1651 	}
1652 	/* Set default qp number */
1653 	err = mlx4_SET_PORT_qpn_calc(mdev->dev, priv->port, priv->base_qpn, 0);
1654 	if (err) {
1655 		en_err(priv, "Failed setting default qp numbers\n");
1656 		goto tx_err;
1657 	}
1658 
1659 	if (mdev->dev->caps.tunnel_offload_mode == MLX4_TUNNEL_OFFLOAD_MODE_VXLAN) {
1660 		err = mlx4_SET_PORT_VXLAN(mdev->dev, priv->port, VXLAN_STEER_BY_OUTER_MAC, 1);
1661 		if (err) {
1662 			en_err(priv, "Failed setting port L2 tunnel configuration, err %d\n",
1663 			       err);
1664 			goto tx_err;
1665 		}
1666 	}
1667 
1668 	/* Init port */
1669 	en_dbg(HW, priv, "Initializing port\n");
1670 	err = mlx4_INIT_PORT(mdev->dev, priv->port);
1671 	if (err) {
1672 		en_err(priv, "Failed Initializing port\n");
1673 		goto tx_err;
1674 	}
1675 
1676 	/* Set Unicast and VXLAN steering rules */
1677 	if (mdev->dev->caps.steering_mode != MLX4_STEERING_MODE_A0 &&
1678 	    mlx4_en_set_rss_steer_rules(priv))
1679 		mlx4_warn(mdev, "Failed setting steering rules\n");
1680 
1681 	/* Attach rx QP to bradcast address */
1682 	eth_broadcast_addr(&mc_list[10]);
1683 	mc_list[5] = priv->port; /* needed for B0 steering support */
1684 	if (mlx4_multicast_attach(mdev->dev, &priv->rss_map.indir_qp, mc_list,
1685 				  priv->port, 0, MLX4_PROT_ETH,
1686 				  &priv->broadcast_id))
1687 		mlx4_warn(mdev, "Failed Attaching Broadcast\n");
1688 
1689 	/* Must redo promiscuous mode setup. */
1690 	priv->flags &= ~(MLX4_EN_FLAG_PROMISC | MLX4_EN_FLAG_MC_PROMISC);
1691 
1692 	/* Schedule multicast task to populate multicast list */
1693 	queue_work(mdev->workqueue, &priv->rx_mode_task);
1694 
1695 #ifdef CONFIG_MLX4_EN_VXLAN
1696 	if (priv->mdev->dev->caps.tunnel_offload_mode == MLX4_TUNNEL_OFFLOAD_MODE_VXLAN)
1697 		vxlan_get_rx_port(dev);
1698 #endif
1699 	priv->port_up = true;
1700 	netif_tx_start_all_queues(dev);
1701 	netif_device_attach(dev);
1702 
1703 	return 0;
1704 
1705 tx_err:
1706 	while (tx_index--) {
1707 		mlx4_en_deactivate_tx_ring(priv, priv->tx_ring[tx_index]);
1708 		mlx4_en_deactivate_cq(priv, priv->tx_cq[tx_index]);
1709 	}
1710 	mlx4_en_destroy_drop_qp(priv);
1711 rss_err:
1712 	mlx4_en_release_rss_steer(priv);
1713 mac_err:
1714 	mlx4_en_put_qp(priv);
1715 cq_err:
1716 	while (rx_index--) {
1717 		mlx4_en_deactivate_cq(priv, priv->rx_cq[rx_index]);
1718 		mlx4_en_free_affinity_hint(priv, rx_index);
1719 	}
1720 	for (i = 0; i < priv->rx_ring_num; i++)
1721 		mlx4_en_deactivate_rx_ring(priv, priv->rx_ring[i]);
1722 
1723 	return err; /* need to close devices */
1724 }
1725 
1726 
1727 void mlx4_en_stop_port(struct net_device *dev, int detach)
1728 {
1729 	struct mlx4_en_priv *priv = netdev_priv(dev);
1730 	struct mlx4_en_dev *mdev = priv->mdev;
1731 	struct mlx4_en_mc_list *mclist, *tmp;
1732 	struct ethtool_flow_id *flow, *tmp_flow;
1733 	int i;
1734 	u8 mc_list[16] = {0};
1735 
1736 	if (!priv->port_up) {
1737 		en_dbg(DRV, priv, "stop port called while port already down\n");
1738 		return;
1739 	}
1740 
1741 	/* close port*/
1742 	mlx4_CLOSE_PORT(mdev->dev, priv->port);
1743 
1744 	/* Synchronize with tx routine */
1745 	netif_tx_lock_bh(dev);
1746 	if (detach)
1747 		netif_device_detach(dev);
1748 	netif_tx_stop_all_queues(dev);
1749 	netif_tx_unlock_bh(dev);
1750 
1751 	netif_tx_disable(dev);
1752 
1753 	/* Set port as not active */
1754 	priv->port_up = false;
1755 	priv->counter_index = MLX4_SINK_COUNTER_INDEX(mdev->dev);
1756 
1757 	/* Promsicuous mode */
1758 	if (mdev->dev->caps.steering_mode ==
1759 	    MLX4_STEERING_MODE_DEVICE_MANAGED) {
1760 		priv->flags &= ~(MLX4_EN_FLAG_PROMISC |
1761 				 MLX4_EN_FLAG_MC_PROMISC);
1762 		mlx4_flow_steer_promisc_remove(mdev->dev,
1763 					       priv->port,
1764 					       MLX4_FS_ALL_DEFAULT);
1765 		mlx4_flow_steer_promisc_remove(mdev->dev,
1766 					       priv->port,
1767 					       MLX4_FS_MC_DEFAULT);
1768 	} else if (priv->flags & MLX4_EN_FLAG_PROMISC) {
1769 		priv->flags &= ~MLX4_EN_FLAG_PROMISC;
1770 
1771 		/* Disable promiscouos mode */
1772 		mlx4_unicast_promisc_remove(mdev->dev, priv->base_qpn,
1773 					    priv->port);
1774 
1775 		/* Disable Multicast promisc */
1776 		if (priv->flags & MLX4_EN_FLAG_MC_PROMISC) {
1777 			mlx4_multicast_promisc_remove(mdev->dev, priv->base_qpn,
1778 						      priv->port);
1779 			priv->flags &= ~MLX4_EN_FLAG_MC_PROMISC;
1780 		}
1781 	}
1782 
1783 	/* Detach All multicasts */
1784 	eth_broadcast_addr(&mc_list[10]);
1785 	mc_list[5] = priv->port; /* needed for B0 steering support */
1786 	mlx4_multicast_detach(mdev->dev, &priv->rss_map.indir_qp, mc_list,
1787 			      MLX4_PROT_ETH, priv->broadcast_id);
1788 	list_for_each_entry(mclist, &priv->curr_list, list) {
1789 		memcpy(&mc_list[10], mclist->addr, ETH_ALEN);
1790 		mc_list[5] = priv->port;
1791 		mlx4_multicast_detach(mdev->dev, &priv->rss_map.indir_qp,
1792 				      mc_list, MLX4_PROT_ETH, mclist->reg_id);
1793 		if (mclist->tunnel_reg_id)
1794 			mlx4_flow_detach(mdev->dev, mclist->tunnel_reg_id);
1795 	}
1796 	mlx4_en_clear_list(dev);
1797 	list_for_each_entry_safe(mclist, tmp, &priv->curr_list, list) {
1798 		list_del(&mclist->list);
1799 		kfree(mclist);
1800 	}
1801 
1802 	/* Flush multicast filter */
1803 	mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, 0, 1, MLX4_MCAST_CONFIG);
1804 
1805 	/* Remove flow steering rules for the port*/
1806 	if (mdev->dev->caps.steering_mode ==
1807 	    MLX4_STEERING_MODE_DEVICE_MANAGED) {
1808 		ASSERT_RTNL();
1809 		list_for_each_entry_safe(flow, tmp_flow,
1810 					 &priv->ethtool_list, list) {
1811 			mlx4_flow_detach(mdev->dev, flow->id);
1812 			list_del(&flow->list);
1813 		}
1814 	}
1815 
1816 	mlx4_en_destroy_drop_qp(priv);
1817 
1818 	/* Free TX Rings */
1819 	for (i = 0; i < priv->tx_ring_num; i++) {
1820 		mlx4_en_deactivate_tx_ring(priv, priv->tx_ring[i]);
1821 		mlx4_en_deactivate_cq(priv, priv->tx_cq[i]);
1822 	}
1823 	msleep(10);
1824 
1825 	for (i = 0; i < priv->tx_ring_num; i++)
1826 		mlx4_en_free_tx_buf(dev, priv->tx_ring[i]);
1827 
1828 	if (mdev->dev->caps.steering_mode != MLX4_STEERING_MODE_A0)
1829 		mlx4_en_delete_rss_steer_rules(priv);
1830 
1831 	/* Free RSS qps */
1832 	mlx4_en_release_rss_steer(priv);
1833 
1834 	/* Unregister Mac address for the port */
1835 	mlx4_en_put_qp(priv);
1836 	if (!(mdev->dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_REASSIGN_MAC_EN))
1837 		mdev->mac_removed[priv->port] = 1;
1838 
1839 	/* Free RX Rings */
1840 	for (i = 0; i < priv->rx_ring_num; i++) {
1841 		struct mlx4_en_cq *cq = priv->rx_cq[i];
1842 
1843 		napi_synchronize(&cq->napi);
1844 		mlx4_en_deactivate_rx_ring(priv, priv->rx_ring[i]);
1845 		mlx4_en_deactivate_cq(priv, cq);
1846 
1847 		mlx4_en_free_affinity_hint(priv, i);
1848 	}
1849 }
1850 
1851 static void mlx4_en_restart(struct work_struct *work)
1852 {
1853 	struct mlx4_en_priv *priv = container_of(work, struct mlx4_en_priv,
1854 						 watchdog_task);
1855 	struct mlx4_en_dev *mdev = priv->mdev;
1856 	struct net_device *dev = priv->dev;
1857 
1858 	en_dbg(DRV, priv, "Watchdog task called for port %d\n", priv->port);
1859 
1860 	rtnl_lock();
1861 	mutex_lock(&mdev->state_lock);
1862 	if (priv->port_up) {
1863 		mlx4_en_stop_port(dev, 1);
1864 		if (mlx4_en_start_port(dev))
1865 			en_err(priv, "Failed restarting port %d\n", priv->port);
1866 	}
1867 	mutex_unlock(&mdev->state_lock);
1868 	rtnl_unlock();
1869 }
1870 
1871 static void mlx4_en_clear_stats(struct net_device *dev)
1872 {
1873 	struct mlx4_en_priv *priv = netdev_priv(dev);
1874 	struct mlx4_en_dev *mdev = priv->mdev;
1875 	int i;
1876 
1877 	if (mlx4_en_DUMP_ETH_STATS(mdev, priv->port, 1))
1878 		en_dbg(HW, priv, "Failed dumping statistics\n");
1879 
1880 	memset(&priv->pstats, 0, sizeof(priv->pstats));
1881 	memset(&priv->pkstats, 0, sizeof(priv->pkstats));
1882 	memset(&priv->port_stats, 0, sizeof(priv->port_stats));
1883 	memset(&priv->rx_flowstats, 0, sizeof(priv->rx_flowstats));
1884 	memset(&priv->tx_flowstats, 0, sizeof(priv->tx_flowstats));
1885 	memset(&priv->rx_priority_flowstats, 0,
1886 	       sizeof(priv->rx_priority_flowstats));
1887 	memset(&priv->tx_priority_flowstats, 0,
1888 	       sizeof(priv->tx_priority_flowstats));
1889 	memset(&priv->pf_stats, 0, sizeof(priv->pf_stats));
1890 
1891 	for (i = 0; i < priv->tx_ring_num; i++) {
1892 		priv->tx_ring[i]->bytes = 0;
1893 		priv->tx_ring[i]->packets = 0;
1894 		priv->tx_ring[i]->tx_csum = 0;
1895 		priv->tx_ring[i]->tx_dropped = 0;
1896 		priv->tx_ring[i]->queue_stopped = 0;
1897 		priv->tx_ring[i]->wake_queue = 0;
1898 		priv->tx_ring[i]->tso_packets = 0;
1899 		priv->tx_ring[i]->xmit_more = 0;
1900 	}
1901 	for (i = 0; i < priv->rx_ring_num; i++) {
1902 		priv->rx_ring[i]->bytes = 0;
1903 		priv->rx_ring[i]->packets = 0;
1904 		priv->rx_ring[i]->csum_ok = 0;
1905 		priv->rx_ring[i]->csum_none = 0;
1906 		priv->rx_ring[i]->csum_complete = 0;
1907 	}
1908 }
1909 
1910 static int mlx4_en_open(struct net_device *dev)
1911 {
1912 	struct mlx4_en_priv *priv = netdev_priv(dev);
1913 	struct mlx4_en_dev *mdev = priv->mdev;
1914 	int err = 0;
1915 
1916 	mutex_lock(&mdev->state_lock);
1917 
1918 	if (!mdev->device_up) {
1919 		en_err(priv, "Cannot open - device down/disabled\n");
1920 		err = -EBUSY;
1921 		goto out;
1922 	}
1923 
1924 	/* Reset HW statistics and SW counters */
1925 	mlx4_en_clear_stats(dev);
1926 
1927 	err = mlx4_en_start_port(dev);
1928 	if (err)
1929 		en_err(priv, "Failed starting port:%d\n", priv->port);
1930 
1931 out:
1932 	mutex_unlock(&mdev->state_lock);
1933 	return err;
1934 }
1935 
1936 
1937 static int mlx4_en_close(struct net_device *dev)
1938 {
1939 	struct mlx4_en_priv *priv = netdev_priv(dev);
1940 	struct mlx4_en_dev *mdev = priv->mdev;
1941 
1942 	en_dbg(IFDOWN, priv, "Close port called\n");
1943 
1944 	mutex_lock(&mdev->state_lock);
1945 
1946 	mlx4_en_stop_port(dev, 0);
1947 	netif_carrier_off(dev);
1948 
1949 	mutex_unlock(&mdev->state_lock);
1950 	return 0;
1951 }
1952 
1953 void mlx4_en_free_resources(struct mlx4_en_priv *priv)
1954 {
1955 	int i;
1956 
1957 #ifdef CONFIG_RFS_ACCEL
1958 	priv->dev->rx_cpu_rmap = NULL;
1959 #endif
1960 
1961 	for (i = 0; i < priv->tx_ring_num; i++) {
1962 		if (priv->tx_ring && priv->tx_ring[i])
1963 			mlx4_en_destroy_tx_ring(priv, &priv->tx_ring[i]);
1964 		if (priv->tx_cq && priv->tx_cq[i])
1965 			mlx4_en_destroy_cq(priv, &priv->tx_cq[i]);
1966 	}
1967 
1968 	for (i = 0; i < priv->rx_ring_num; i++) {
1969 		if (priv->rx_ring[i])
1970 			mlx4_en_destroy_rx_ring(priv, &priv->rx_ring[i],
1971 				priv->prof->rx_ring_size, priv->stride);
1972 		if (priv->rx_cq[i])
1973 			mlx4_en_destroy_cq(priv, &priv->rx_cq[i]);
1974 	}
1975 
1976 }
1977 
1978 int mlx4_en_alloc_resources(struct mlx4_en_priv *priv)
1979 {
1980 	struct mlx4_en_port_profile *prof = priv->prof;
1981 	int i;
1982 	int node;
1983 
1984 	/* Create tx Rings */
1985 	for (i = 0; i < priv->tx_ring_num; i++) {
1986 		node = cpu_to_node(i % num_online_cpus());
1987 		if (mlx4_en_create_cq(priv, &priv->tx_cq[i],
1988 				      prof->tx_ring_size, i, TX, node))
1989 			goto err;
1990 
1991 		if (mlx4_en_create_tx_ring(priv, &priv->tx_ring[i],
1992 					   prof->tx_ring_size, TXBB_SIZE,
1993 					   node, i))
1994 			goto err;
1995 	}
1996 
1997 	/* Create rx Rings */
1998 	for (i = 0; i < priv->rx_ring_num; i++) {
1999 		node = cpu_to_node(i % num_online_cpus());
2000 		if (mlx4_en_create_cq(priv, &priv->rx_cq[i],
2001 				      prof->rx_ring_size, i, RX, node))
2002 			goto err;
2003 
2004 		if (mlx4_en_create_rx_ring(priv, &priv->rx_ring[i],
2005 					   prof->rx_ring_size, priv->stride,
2006 					   node))
2007 			goto err;
2008 	}
2009 
2010 #ifdef CONFIG_RFS_ACCEL
2011 	priv->dev->rx_cpu_rmap = mlx4_get_cpu_rmap(priv->mdev->dev, priv->port);
2012 #endif
2013 
2014 	return 0;
2015 
2016 err:
2017 	en_err(priv, "Failed to allocate NIC resources\n");
2018 	for (i = 0; i < priv->rx_ring_num; i++) {
2019 		if (priv->rx_ring[i])
2020 			mlx4_en_destroy_rx_ring(priv, &priv->rx_ring[i],
2021 						prof->rx_ring_size,
2022 						priv->stride);
2023 		if (priv->rx_cq[i])
2024 			mlx4_en_destroy_cq(priv, &priv->rx_cq[i]);
2025 	}
2026 	for (i = 0; i < priv->tx_ring_num; i++) {
2027 		if (priv->tx_ring[i])
2028 			mlx4_en_destroy_tx_ring(priv, &priv->tx_ring[i]);
2029 		if (priv->tx_cq[i])
2030 			mlx4_en_destroy_cq(priv, &priv->tx_cq[i]);
2031 	}
2032 	return -ENOMEM;
2033 }
2034 
2035 
2036 void mlx4_en_destroy_netdev(struct net_device *dev)
2037 {
2038 	struct mlx4_en_priv *priv = netdev_priv(dev);
2039 	struct mlx4_en_dev *mdev = priv->mdev;
2040 
2041 	en_dbg(DRV, priv, "Destroying netdev on port:%d\n", priv->port);
2042 
2043 	/* Unregister device - this will close the port if it was up */
2044 	if (priv->registered) {
2045 		devlink_port_type_clear(mlx4_get_devlink_port(mdev->dev,
2046 							      priv->port));
2047 		unregister_netdev(dev);
2048 	}
2049 
2050 	if (priv->allocated)
2051 		mlx4_free_hwq_res(mdev->dev, &priv->res, MLX4_EN_PAGE_SIZE);
2052 
2053 	cancel_delayed_work(&priv->stats_task);
2054 	cancel_delayed_work(&priv->service_task);
2055 	/* flush any pending task for this netdev */
2056 	flush_workqueue(mdev->workqueue);
2057 
2058 	if (mdev->dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_TS)
2059 		mlx4_en_remove_timestamp(mdev);
2060 
2061 	/* Detach the netdev so tasks would not attempt to access it */
2062 	mutex_lock(&mdev->state_lock);
2063 	mdev->pndev[priv->port] = NULL;
2064 	mdev->upper[priv->port] = NULL;
2065 	mutex_unlock(&mdev->state_lock);
2066 
2067 	mlx4_en_free_resources(priv);
2068 
2069 	kfree(priv->tx_ring);
2070 	kfree(priv->tx_cq);
2071 
2072 	free_netdev(dev);
2073 }
2074 
2075 static int mlx4_en_change_mtu(struct net_device *dev, int new_mtu)
2076 {
2077 	struct mlx4_en_priv *priv = netdev_priv(dev);
2078 	struct mlx4_en_dev *mdev = priv->mdev;
2079 	int err = 0;
2080 
2081 	en_dbg(DRV, priv, "Change MTU called - current:%d new:%d\n",
2082 		 dev->mtu, new_mtu);
2083 
2084 	if ((new_mtu < MLX4_EN_MIN_MTU) || (new_mtu > priv->max_mtu)) {
2085 		en_err(priv, "Bad MTU size:%d.\n", new_mtu);
2086 		return -EPERM;
2087 	}
2088 	dev->mtu = new_mtu;
2089 
2090 	if (netif_running(dev)) {
2091 		mutex_lock(&mdev->state_lock);
2092 		if (!mdev->device_up) {
2093 			/* NIC is probably restarting - let watchdog task reset
2094 			 * the port */
2095 			en_dbg(DRV, priv, "Change MTU called with card down!?\n");
2096 		} else {
2097 			mlx4_en_stop_port(dev, 1);
2098 			err = mlx4_en_start_port(dev);
2099 			if (err) {
2100 				en_err(priv, "Failed restarting port:%d\n",
2101 					 priv->port);
2102 				queue_work(mdev->workqueue, &priv->watchdog_task);
2103 			}
2104 		}
2105 		mutex_unlock(&mdev->state_lock);
2106 	}
2107 	return 0;
2108 }
2109 
2110 static int mlx4_en_hwtstamp_set(struct net_device *dev, struct ifreq *ifr)
2111 {
2112 	struct mlx4_en_priv *priv = netdev_priv(dev);
2113 	struct mlx4_en_dev *mdev = priv->mdev;
2114 	struct hwtstamp_config config;
2115 
2116 	if (copy_from_user(&config, ifr->ifr_data, sizeof(config)))
2117 		return -EFAULT;
2118 
2119 	/* reserved for future extensions */
2120 	if (config.flags)
2121 		return -EINVAL;
2122 
2123 	/* device doesn't support time stamping */
2124 	if (!(mdev->dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_TS))
2125 		return -EINVAL;
2126 
2127 	/* TX HW timestamp */
2128 	switch (config.tx_type) {
2129 	case HWTSTAMP_TX_OFF:
2130 	case HWTSTAMP_TX_ON:
2131 		break;
2132 	default:
2133 		return -ERANGE;
2134 	}
2135 
2136 	/* RX HW timestamp */
2137 	switch (config.rx_filter) {
2138 	case HWTSTAMP_FILTER_NONE:
2139 		break;
2140 	case HWTSTAMP_FILTER_ALL:
2141 	case HWTSTAMP_FILTER_SOME:
2142 	case HWTSTAMP_FILTER_PTP_V1_L4_EVENT:
2143 	case HWTSTAMP_FILTER_PTP_V1_L4_SYNC:
2144 	case HWTSTAMP_FILTER_PTP_V1_L4_DELAY_REQ:
2145 	case HWTSTAMP_FILTER_PTP_V2_L4_EVENT:
2146 	case HWTSTAMP_FILTER_PTP_V2_L4_SYNC:
2147 	case HWTSTAMP_FILTER_PTP_V2_L4_DELAY_REQ:
2148 	case HWTSTAMP_FILTER_PTP_V2_L2_EVENT:
2149 	case HWTSTAMP_FILTER_PTP_V2_L2_SYNC:
2150 	case HWTSTAMP_FILTER_PTP_V2_L2_DELAY_REQ:
2151 	case HWTSTAMP_FILTER_PTP_V2_EVENT:
2152 	case HWTSTAMP_FILTER_PTP_V2_SYNC:
2153 	case HWTSTAMP_FILTER_PTP_V2_DELAY_REQ:
2154 		config.rx_filter = HWTSTAMP_FILTER_ALL;
2155 		break;
2156 	default:
2157 		return -ERANGE;
2158 	}
2159 
2160 	if (mlx4_en_reset_config(dev, config, dev->features)) {
2161 		config.tx_type = HWTSTAMP_TX_OFF;
2162 		config.rx_filter = HWTSTAMP_FILTER_NONE;
2163 	}
2164 
2165 	return copy_to_user(ifr->ifr_data, &config,
2166 			    sizeof(config)) ? -EFAULT : 0;
2167 }
2168 
2169 static int mlx4_en_hwtstamp_get(struct net_device *dev, struct ifreq *ifr)
2170 {
2171 	struct mlx4_en_priv *priv = netdev_priv(dev);
2172 
2173 	return copy_to_user(ifr->ifr_data, &priv->hwtstamp_config,
2174 			    sizeof(priv->hwtstamp_config)) ? -EFAULT : 0;
2175 }
2176 
2177 static int mlx4_en_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
2178 {
2179 	switch (cmd) {
2180 	case SIOCSHWTSTAMP:
2181 		return mlx4_en_hwtstamp_set(dev, ifr);
2182 	case SIOCGHWTSTAMP:
2183 		return mlx4_en_hwtstamp_get(dev, ifr);
2184 	default:
2185 		return -EOPNOTSUPP;
2186 	}
2187 }
2188 
2189 static netdev_features_t mlx4_en_fix_features(struct net_device *netdev,
2190 					      netdev_features_t features)
2191 {
2192 	struct mlx4_en_priv *en_priv = netdev_priv(netdev);
2193 	struct mlx4_en_dev *mdev = en_priv->mdev;
2194 
2195 	/* Since there is no support for separate RX C-TAG/S-TAG vlan accel
2196 	 * enable/disable make sure S-TAG flag is always in same state as
2197 	 * C-TAG.
2198 	 */
2199 	if (features & NETIF_F_HW_VLAN_CTAG_RX &&
2200 	    !(mdev->dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_SKIP_OUTER_VLAN))
2201 		features |= NETIF_F_HW_VLAN_STAG_RX;
2202 	else
2203 		features &= ~NETIF_F_HW_VLAN_STAG_RX;
2204 
2205 	return features;
2206 }
2207 
2208 static int mlx4_en_set_features(struct net_device *netdev,
2209 		netdev_features_t features)
2210 {
2211 	struct mlx4_en_priv *priv = netdev_priv(netdev);
2212 	bool reset = false;
2213 	int ret = 0;
2214 
2215 	if (DEV_FEATURE_CHANGED(netdev, features, NETIF_F_RXFCS)) {
2216 		en_info(priv, "Turn %s RX-FCS\n",
2217 			(features & NETIF_F_RXFCS) ? "ON" : "OFF");
2218 		reset = true;
2219 	}
2220 
2221 	if (DEV_FEATURE_CHANGED(netdev, features, NETIF_F_RXALL)) {
2222 		u8 ignore_fcs_value = (features & NETIF_F_RXALL) ? 1 : 0;
2223 
2224 		en_info(priv, "Turn %s RX-ALL\n",
2225 			ignore_fcs_value ? "ON" : "OFF");
2226 		ret = mlx4_SET_PORT_fcs_check(priv->mdev->dev,
2227 					      priv->port, ignore_fcs_value);
2228 		if (ret)
2229 			return ret;
2230 	}
2231 
2232 	if (DEV_FEATURE_CHANGED(netdev, features, NETIF_F_HW_VLAN_CTAG_RX)) {
2233 		en_info(priv, "Turn %s RX vlan strip offload\n",
2234 			(features & NETIF_F_HW_VLAN_CTAG_RX) ? "ON" : "OFF");
2235 		reset = true;
2236 	}
2237 
2238 	if (DEV_FEATURE_CHANGED(netdev, features, NETIF_F_HW_VLAN_CTAG_TX))
2239 		en_info(priv, "Turn %s TX vlan strip offload\n",
2240 			(features & NETIF_F_HW_VLAN_CTAG_TX) ? "ON" : "OFF");
2241 
2242 	if (DEV_FEATURE_CHANGED(netdev, features, NETIF_F_HW_VLAN_STAG_TX))
2243 		en_info(priv, "Turn %s TX S-VLAN strip offload\n",
2244 			(features & NETIF_F_HW_VLAN_STAG_TX) ? "ON" : "OFF");
2245 
2246 	if (DEV_FEATURE_CHANGED(netdev, features, NETIF_F_LOOPBACK)) {
2247 		en_info(priv, "Turn %s loopback\n",
2248 			(features & NETIF_F_LOOPBACK) ? "ON" : "OFF");
2249 		mlx4_en_update_loopback_state(netdev, features);
2250 	}
2251 
2252 	if (reset) {
2253 		ret = mlx4_en_reset_config(netdev, priv->hwtstamp_config,
2254 					   features);
2255 		if (ret)
2256 			return ret;
2257 	}
2258 
2259 	return 0;
2260 }
2261 
2262 static int mlx4_en_set_vf_mac(struct net_device *dev, int queue, u8 *mac)
2263 {
2264 	struct mlx4_en_priv *en_priv = netdev_priv(dev);
2265 	struct mlx4_en_dev *mdev = en_priv->mdev;
2266 	u64 mac_u64 = mlx4_mac_to_u64(mac);
2267 
2268 	if (is_multicast_ether_addr(mac))
2269 		return -EINVAL;
2270 
2271 	return mlx4_set_vf_mac(mdev->dev, en_priv->port, queue, mac_u64);
2272 }
2273 
2274 static int mlx4_en_set_vf_vlan(struct net_device *dev, int vf, u16 vlan, u8 qos)
2275 {
2276 	struct mlx4_en_priv *en_priv = netdev_priv(dev);
2277 	struct mlx4_en_dev *mdev = en_priv->mdev;
2278 
2279 	return mlx4_set_vf_vlan(mdev->dev, en_priv->port, vf, vlan, qos);
2280 }
2281 
2282 static int mlx4_en_set_vf_rate(struct net_device *dev, int vf, int min_tx_rate,
2283 			       int max_tx_rate)
2284 {
2285 	struct mlx4_en_priv *en_priv = netdev_priv(dev);
2286 	struct mlx4_en_dev *mdev = en_priv->mdev;
2287 
2288 	return mlx4_set_vf_rate(mdev->dev, en_priv->port, vf, min_tx_rate,
2289 				max_tx_rate);
2290 }
2291 
2292 static int mlx4_en_set_vf_spoofchk(struct net_device *dev, int vf, bool setting)
2293 {
2294 	struct mlx4_en_priv *en_priv = netdev_priv(dev);
2295 	struct mlx4_en_dev *mdev = en_priv->mdev;
2296 
2297 	return mlx4_set_vf_spoofchk(mdev->dev, en_priv->port, vf, setting);
2298 }
2299 
2300 static int mlx4_en_get_vf_config(struct net_device *dev, int vf, struct ifla_vf_info *ivf)
2301 {
2302 	struct mlx4_en_priv *en_priv = netdev_priv(dev);
2303 	struct mlx4_en_dev *mdev = en_priv->mdev;
2304 
2305 	return mlx4_get_vf_config(mdev->dev, en_priv->port, vf, ivf);
2306 }
2307 
2308 static int mlx4_en_set_vf_link_state(struct net_device *dev, int vf, int link_state)
2309 {
2310 	struct mlx4_en_priv *en_priv = netdev_priv(dev);
2311 	struct mlx4_en_dev *mdev = en_priv->mdev;
2312 
2313 	return mlx4_set_vf_link_state(mdev->dev, en_priv->port, vf, link_state);
2314 }
2315 
2316 static int mlx4_en_get_vf_stats(struct net_device *dev, int vf,
2317 				struct ifla_vf_stats *vf_stats)
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_stats(mdev->dev, en_priv->port, vf, vf_stats);
2323 }
2324 
2325 #define PORT_ID_BYTE_LEN 8
2326 static int mlx4_en_get_phys_port_id(struct net_device *dev,
2327 				    struct netdev_phys_item_id *ppid)
2328 {
2329 	struct mlx4_en_priv *priv = netdev_priv(dev);
2330 	struct mlx4_dev *mdev = priv->mdev->dev;
2331 	int i;
2332 	u64 phys_port_id = mdev->caps.phys_port_id[priv->port];
2333 
2334 	if (!phys_port_id)
2335 		return -EOPNOTSUPP;
2336 
2337 	ppid->id_len = sizeof(phys_port_id);
2338 	for (i = PORT_ID_BYTE_LEN - 1; i >= 0; --i) {
2339 		ppid->id[i] =  phys_port_id & 0xff;
2340 		phys_port_id >>= 8;
2341 	}
2342 	return 0;
2343 }
2344 
2345 #ifdef CONFIG_MLX4_EN_VXLAN
2346 static void mlx4_en_add_vxlan_offloads(struct work_struct *work)
2347 {
2348 	int ret;
2349 	struct mlx4_en_priv *priv = container_of(work, struct mlx4_en_priv,
2350 						 vxlan_add_task);
2351 
2352 	ret = mlx4_config_vxlan_port(priv->mdev->dev, priv->vxlan_port);
2353 	if (ret)
2354 		goto out;
2355 
2356 	ret = mlx4_SET_PORT_VXLAN(priv->mdev->dev, priv->port,
2357 				  VXLAN_STEER_BY_OUTER_MAC, 1);
2358 out:
2359 	if (ret) {
2360 		en_err(priv, "failed setting L2 tunnel configuration ret %d\n", ret);
2361 		return;
2362 	}
2363 
2364 	/* set offloads */
2365 	priv->dev->hw_enc_features |= NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM |
2366 				      NETIF_F_RXCSUM |
2367 				      NETIF_F_TSO | NETIF_F_TSO6 |
2368 				      NETIF_F_GSO_UDP_TUNNEL |
2369 				      NETIF_F_GSO_UDP_TUNNEL_CSUM |
2370 				      NETIF_F_GSO_PARTIAL;
2371 }
2372 
2373 static void mlx4_en_del_vxlan_offloads(struct work_struct *work)
2374 {
2375 	int ret;
2376 	struct mlx4_en_priv *priv = container_of(work, struct mlx4_en_priv,
2377 						 vxlan_del_task);
2378 	/* unset offloads */
2379 	priv->dev->hw_enc_features &= ~(NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM |
2380 					NETIF_F_RXCSUM |
2381 					NETIF_F_TSO | NETIF_F_TSO6 |
2382 					NETIF_F_GSO_UDP_TUNNEL |
2383 					NETIF_F_GSO_UDP_TUNNEL_CSUM |
2384 					NETIF_F_GSO_PARTIAL);
2385 
2386 	ret = mlx4_SET_PORT_VXLAN(priv->mdev->dev, priv->port,
2387 				  VXLAN_STEER_BY_OUTER_MAC, 0);
2388 	if (ret)
2389 		en_err(priv, "failed setting L2 tunnel configuration ret %d\n", ret);
2390 
2391 	priv->vxlan_port = 0;
2392 }
2393 
2394 static void mlx4_en_add_vxlan_port(struct  net_device *dev,
2395 				   sa_family_t sa_family, __be16 port)
2396 {
2397 	struct mlx4_en_priv *priv = netdev_priv(dev);
2398 	__be16 current_port;
2399 
2400 	if (priv->mdev->dev->caps.tunnel_offload_mode != MLX4_TUNNEL_OFFLOAD_MODE_VXLAN)
2401 		return;
2402 
2403 	if (sa_family == AF_INET6)
2404 		return;
2405 
2406 	current_port = priv->vxlan_port;
2407 	if (current_port && current_port != port) {
2408 		en_warn(priv, "vxlan port %d configured, can't add port %d\n",
2409 			ntohs(current_port), ntohs(port));
2410 		return;
2411 	}
2412 
2413 	priv->vxlan_port = port;
2414 	queue_work(priv->mdev->workqueue, &priv->vxlan_add_task);
2415 }
2416 
2417 static void mlx4_en_del_vxlan_port(struct  net_device *dev,
2418 				   sa_family_t sa_family, __be16 port)
2419 {
2420 	struct mlx4_en_priv *priv = netdev_priv(dev);
2421 	__be16 current_port;
2422 
2423 	if (priv->mdev->dev->caps.tunnel_offload_mode != MLX4_TUNNEL_OFFLOAD_MODE_VXLAN)
2424 		return;
2425 
2426 	if (sa_family == AF_INET6)
2427 		return;
2428 
2429 	current_port = priv->vxlan_port;
2430 	if (current_port != port) {
2431 		en_dbg(DRV, priv, "vxlan port %d isn't configured, ignoring\n", ntohs(port));
2432 		return;
2433 	}
2434 
2435 	queue_work(priv->mdev->workqueue, &priv->vxlan_del_task);
2436 }
2437 
2438 static netdev_features_t mlx4_en_features_check(struct sk_buff *skb,
2439 						struct net_device *dev,
2440 						netdev_features_t features)
2441 {
2442 	features = vlan_features_check(skb, features);
2443 	features = vxlan_features_check(skb, features);
2444 
2445 	/* The ConnectX-3 doesn't support outer IPv6 checksums but it does
2446 	 * support inner IPv6 checksums and segmentation so  we need to
2447 	 * strip that feature if this is an IPv6 encapsulated frame.
2448 	 */
2449 	if (skb->encapsulation &&
2450 	    (skb->ip_summed == CHECKSUM_PARTIAL) &&
2451 	    (ip_hdr(skb)->version != 4))
2452 		features &= ~(NETIF_F_CSUM_MASK | NETIF_F_GSO_MASK);
2453 
2454 	return features;
2455 }
2456 #endif
2457 
2458 static int mlx4_en_set_tx_maxrate(struct net_device *dev, int queue_index, u32 maxrate)
2459 {
2460 	struct mlx4_en_priv *priv = netdev_priv(dev);
2461 	struct mlx4_en_tx_ring *tx_ring = priv->tx_ring[queue_index];
2462 	struct mlx4_update_qp_params params;
2463 	int err;
2464 
2465 	if (!(priv->mdev->dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_QP_RATE_LIMIT))
2466 		return -EOPNOTSUPP;
2467 
2468 	/* rate provided to us in Mbs, check if it fits into 12 bits, if not use Gbs */
2469 	if (maxrate >> 12) {
2470 		params.rate_unit = MLX4_QP_RATE_LIMIT_GBS;
2471 		params.rate_val  = maxrate / 1000;
2472 	} else if (maxrate) {
2473 		params.rate_unit = MLX4_QP_RATE_LIMIT_MBS;
2474 		params.rate_val  = maxrate;
2475 	} else { /* zero serves to revoke the QP rate-limitation */
2476 		params.rate_unit = 0;
2477 		params.rate_val  = 0;
2478 	}
2479 
2480 	err = mlx4_update_qp(priv->mdev->dev, tx_ring->qpn, MLX4_UPDATE_QP_RATE_LIMIT,
2481 			     &params);
2482 	return err;
2483 }
2484 
2485 static const struct net_device_ops mlx4_netdev_ops = {
2486 	.ndo_open		= mlx4_en_open,
2487 	.ndo_stop		= mlx4_en_close,
2488 	.ndo_start_xmit		= mlx4_en_xmit,
2489 	.ndo_select_queue	= mlx4_en_select_queue,
2490 	.ndo_get_stats64	= mlx4_en_get_stats64,
2491 	.ndo_set_rx_mode	= mlx4_en_set_rx_mode,
2492 	.ndo_set_mac_address	= mlx4_en_set_mac,
2493 	.ndo_validate_addr	= eth_validate_addr,
2494 	.ndo_change_mtu		= mlx4_en_change_mtu,
2495 	.ndo_do_ioctl		= mlx4_en_ioctl,
2496 	.ndo_tx_timeout		= mlx4_en_tx_timeout,
2497 	.ndo_vlan_rx_add_vid	= mlx4_en_vlan_rx_add_vid,
2498 	.ndo_vlan_rx_kill_vid	= mlx4_en_vlan_rx_kill_vid,
2499 #ifdef CONFIG_NET_POLL_CONTROLLER
2500 	.ndo_poll_controller	= mlx4_en_netpoll,
2501 #endif
2502 	.ndo_set_features	= mlx4_en_set_features,
2503 	.ndo_fix_features	= mlx4_en_fix_features,
2504 	.ndo_setup_tc		= __mlx4_en_setup_tc,
2505 #ifdef CONFIG_RFS_ACCEL
2506 	.ndo_rx_flow_steer	= mlx4_en_filter_rfs,
2507 #endif
2508 	.ndo_get_phys_port_id	= mlx4_en_get_phys_port_id,
2509 #ifdef CONFIG_MLX4_EN_VXLAN
2510 	.ndo_add_vxlan_port	= mlx4_en_add_vxlan_port,
2511 	.ndo_del_vxlan_port	= mlx4_en_del_vxlan_port,
2512 	.ndo_features_check	= mlx4_en_features_check,
2513 #endif
2514 	.ndo_set_tx_maxrate	= mlx4_en_set_tx_maxrate,
2515 };
2516 
2517 static const struct net_device_ops mlx4_netdev_ops_master = {
2518 	.ndo_open		= mlx4_en_open,
2519 	.ndo_stop		= mlx4_en_close,
2520 	.ndo_start_xmit		= mlx4_en_xmit,
2521 	.ndo_select_queue	= mlx4_en_select_queue,
2522 	.ndo_get_stats64	= mlx4_en_get_stats64,
2523 	.ndo_set_rx_mode	= mlx4_en_set_rx_mode,
2524 	.ndo_set_mac_address	= mlx4_en_set_mac,
2525 	.ndo_validate_addr	= eth_validate_addr,
2526 	.ndo_change_mtu		= mlx4_en_change_mtu,
2527 	.ndo_tx_timeout		= mlx4_en_tx_timeout,
2528 	.ndo_vlan_rx_add_vid	= mlx4_en_vlan_rx_add_vid,
2529 	.ndo_vlan_rx_kill_vid	= mlx4_en_vlan_rx_kill_vid,
2530 	.ndo_set_vf_mac		= mlx4_en_set_vf_mac,
2531 	.ndo_set_vf_vlan	= mlx4_en_set_vf_vlan,
2532 	.ndo_set_vf_rate	= mlx4_en_set_vf_rate,
2533 	.ndo_set_vf_spoofchk	= mlx4_en_set_vf_spoofchk,
2534 	.ndo_set_vf_link_state	= mlx4_en_set_vf_link_state,
2535 	.ndo_get_vf_stats       = mlx4_en_get_vf_stats,
2536 	.ndo_get_vf_config	= mlx4_en_get_vf_config,
2537 #ifdef CONFIG_NET_POLL_CONTROLLER
2538 	.ndo_poll_controller	= mlx4_en_netpoll,
2539 #endif
2540 	.ndo_set_features	= mlx4_en_set_features,
2541 	.ndo_fix_features	= mlx4_en_fix_features,
2542 	.ndo_setup_tc		= __mlx4_en_setup_tc,
2543 #ifdef CONFIG_RFS_ACCEL
2544 	.ndo_rx_flow_steer	= mlx4_en_filter_rfs,
2545 #endif
2546 	.ndo_get_phys_port_id	= mlx4_en_get_phys_port_id,
2547 #ifdef CONFIG_MLX4_EN_VXLAN
2548 	.ndo_add_vxlan_port	= mlx4_en_add_vxlan_port,
2549 	.ndo_del_vxlan_port	= mlx4_en_del_vxlan_port,
2550 	.ndo_features_check	= mlx4_en_features_check,
2551 #endif
2552 	.ndo_set_tx_maxrate	= mlx4_en_set_tx_maxrate,
2553 };
2554 
2555 struct mlx4_en_bond {
2556 	struct work_struct work;
2557 	struct mlx4_en_priv *priv;
2558 	int is_bonded;
2559 	struct mlx4_port_map port_map;
2560 };
2561 
2562 static void mlx4_en_bond_work(struct work_struct *work)
2563 {
2564 	struct mlx4_en_bond *bond = container_of(work,
2565 						     struct mlx4_en_bond,
2566 						     work);
2567 	int err = 0;
2568 	struct mlx4_dev *dev = bond->priv->mdev->dev;
2569 
2570 	if (bond->is_bonded) {
2571 		if (!mlx4_is_bonded(dev)) {
2572 			err = mlx4_bond(dev);
2573 			if (err)
2574 				en_err(bond->priv, "Fail to bond device\n");
2575 		}
2576 		if (!err) {
2577 			err = mlx4_port_map_set(dev, &bond->port_map);
2578 			if (err)
2579 				en_err(bond->priv, "Fail to set port map [%d][%d]: %d\n",
2580 				       bond->port_map.port1,
2581 				       bond->port_map.port2,
2582 				       err);
2583 		}
2584 	} else if (mlx4_is_bonded(dev)) {
2585 		err = mlx4_unbond(dev);
2586 		if (err)
2587 			en_err(bond->priv, "Fail to unbond device\n");
2588 	}
2589 	dev_put(bond->priv->dev);
2590 	kfree(bond);
2591 }
2592 
2593 static int mlx4_en_queue_bond_work(struct mlx4_en_priv *priv, int is_bonded,
2594 				   u8 v2p_p1, u8 v2p_p2)
2595 {
2596 	struct mlx4_en_bond *bond = NULL;
2597 
2598 	bond = kzalloc(sizeof(*bond), GFP_ATOMIC);
2599 	if (!bond)
2600 		return -ENOMEM;
2601 
2602 	INIT_WORK(&bond->work, mlx4_en_bond_work);
2603 	bond->priv = priv;
2604 	bond->is_bonded = is_bonded;
2605 	bond->port_map.port1 = v2p_p1;
2606 	bond->port_map.port2 = v2p_p2;
2607 	dev_hold(priv->dev);
2608 	queue_work(priv->mdev->workqueue, &bond->work);
2609 	return 0;
2610 }
2611 
2612 int mlx4_en_netdev_event(struct notifier_block *this,
2613 			 unsigned long event, void *ptr)
2614 {
2615 	struct net_device *ndev = netdev_notifier_info_to_dev(ptr);
2616 	u8 port = 0;
2617 	struct mlx4_en_dev *mdev;
2618 	struct mlx4_dev *dev;
2619 	int i, num_eth_ports = 0;
2620 	bool do_bond = true;
2621 	struct mlx4_en_priv *priv;
2622 	u8 v2p_port1 = 0;
2623 	u8 v2p_port2 = 0;
2624 
2625 	if (!net_eq(dev_net(ndev), &init_net))
2626 		return NOTIFY_DONE;
2627 
2628 	mdev = container_of(this, struct mlx4_en_dev, nb);
2629 	dev = mdev->dev;
2630 
2631 	/* Go into this mode only when two network devices set on two ports
2632 	 * of the same mlx4 device are slaves of the same bonding master
2633 	 */
2634 	mlx4_foreach_port(i, dev, MLX4_PORT_TYPE_ETH) {
2635 		++num_eth_ports;
2636 		if (!port && (mdev->pndev[i] == ndev))
2637 			port = i;
2638 		mdev->upper[i] = mdev->pndev[i] ?
2639 			netdev_master_upper_dev_get(mdev->pndev[i]) : NULL;
2640 		/* condition not met: network device is a slave */
2641 		if (!mdev->upper[i])
2642 			do_bond = false;
2643 		if (num_eth_ports < 2)
2644 			continue;
2645 		/* condition not met: same master */
2646 		if (mdev->upper[i] != mdev->upper[i-1])
2647 			do_bond = false;
2648 	}
2649 	/* condition not met: 2 salves */
2650 	do_bond = (num_eth_ports ==  2) ? do_bond : false;
2651 
2652 	/* handle only events that come with enough info */
2653 	if ((do_bond && (event != NETDEV_BONDING_INFO)) || !port)
2654 		return NOTIFY_DONE;
2655 
2656 	priv = netdev_priv(ndev);
2657 	if (do_bond) {
2658 		struct netdev_notifier_bonding_info *notifier_info = ptr;
2659 		struct netdev_bonding_info *bonding_info =
2660 			&notifier_info->bonding_info;
2661 
2662 		/* required mode 1, 2 or 4 */
2663 		if ((bonding_info->master.bond_mode != BOND_MODE_ACTIVEBACKUP) &&
2664 		    (bonding_info->master.bond_mode != BOND_MODE_XOR) &&
2665 		    (bonding_info->master.bond_mode != BOND_MODE_8023AD))
2666 			do_bond = false;
2667 
2668 		/* require exactly 2 slaves */
2669 		if (bonding_info->master.num_slaves != 2)
2670 			do_bond = false;
2671 
2672 		/* calc v2p */
2673 		if (do_bond) {
2674 			if (bonding_info->master.bond_mode ==
2675 			    BOND_MODE_ACTIVEBACKUP) {
2676 				/* in active-backup mode virtual ports are
2677 				 * mapped to the physical port of the active
2678 				 * slave */
2679 				if (bonding_info->slave.state ==
2680 				    BOND_STATE_BACKUP) {
2681 					if (port == 1) {
2682 						v2p_port1 = 2;
2683 						v2p_port2 = 2;
2684 					} else {
2685 						v2p_port1 = 1;
2686 						v2p_port2 = 1;
2687 					}
2688 				} else { /* BOND_STATE_ACTIVE */
2689 					if (port == 1) {
2690 						v2p_port1 = 1;
2691 						v2p_port2 = 1;
2692 					} else {
2693 						v2p_port1 = 2;
2694 						v2p_port2 = 2;
2695 					}
2696 				}
2697 			} else { /* Active-Active */
2698 				/* in active-active mode a virtual port is
2699 				 * mapped to the native physical port if and only
2700 				 * if the physical port is up */
2701 				__s8 link = bonding_info->slave.link;
2702 
2703 				if (port == 1)
2704 					v2p_port2 = 2;
2705 				else
2706 					v2p_port1 = 1;
2707 				if ((link == BOND_LINK_UP) ||
2708 				    (link == BOND_LINK_FAIL)) {
2709 					if (port == 1)
2710 						v2p_port1 = 1;
2711 					else
2712 						v2p_port2 = 2;
2713 				} else { /* BOND_LINK_DOWN || BOND_LINK_BACK */
2714 					if (port == 1)
2715 						v2p_port1 = 2;
2716 					else
2717 						v2p_port2 = 1;
2718 				}
2719 			}
2720 		}
2721 	}
2722 
2723 	mlx4_en_queue_bond_work(priv, do_bond,
2724 				v2p_port1, v2p_port2);
2725 
2726 	return NOTIFY_DONE;
2727 }
2728 
2729 void mlx4_en_update_pfc_stats_bitmap(struct mlx4_dev *dev,
2730 				     struct mlx4_en_stats_bitmap *stats_bitmap,
2731 				     u8 rx_ppp, u8 rx_pause,
2732 				     u8 tx_ppp, u8 tx_pause)
2733 {
2734 	int last_i = NUM_MAIN_STATS + NUM_PORT_STATS + NUM_PF_STATS;
2735 
2736 	if (!mlx4_is_slave(dev) &&
2737 	    (dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_FLOWSTATS_EN)) {
2738 		mutex_lock(&stats_bitmap->mutex);
2739 		bitmap_clear(stats_bitmap->bitmap, last_i, NUM_FLOW_STATS);
2740 
2741 		if (rx_ppp)
2742 			bitmap_set(stats_bitmap->bitmap, last_i,
2743 				   NUM_FLOW_PRIORITY_STATS_RX);
2744 		last_i += NUM_FLOW_PRIORITY_STATS_RX;
2745 
2746 		if (rx_pause && !(rx_ppp))
2747 			bitmap_set(stats_bitmap->bitmap, last_i,
2748 				   NUM_FLOW_STATS_RX);
2749 		last_i += NUM_FLOW_STATS_RX;
2750 
2751 		if (tx_ppp)
2752 			bitmap_set(stats_bitmap->bitmap, last_i,
2753 				   NUM_FLOW_PRIORITY_STATS_TX);
2754 		last_i += NUM_FLOW_PRIORITY_STATS_TX;
2755 
2756 		if (tx_pause && !(tx_ppp))
2757 			bitmap_set(stats_bitmap->bitmap, last_i,
2758 				   NUM_FLOW_STATS_TX);
2759 		last_i += NUM_FLOW_STATS_TX;
2760 
2761 		mutex_unlock(&stats_bitmap->mutex);
2762 	}
2763 }
2764 
2765 void mlx4_en_set_stats_bitmap(struct mlx4_dev *dev,
2766 			      struct mlx4_en_stats_bitmap *stats_bitmap,
2767 			      u8 rx_ppp, u8 rx_pause,
2768 			      u8 tx_ppp, u8 tx_pause)
2769 {
2770 	int last_i = 0;
2771 
2772 	mutex_init(&stats_bitmap->mutex);
2773 	bitmap_zero(stats_bitmap->bitmap, NUM_ALL_STATS);
2774 
2775 	if (mlx4_is_slave(dev)) {
2776 		bitmap_set(stats_bitmap->bitmap, last_i +
2777 					 MLX4_FIND_NETDEV_STAT(rx_packets), 1);
2778 		bitmap_set(stats_bitmap->bitmap, last_i +
2779 					 MLX4_FIND_NETDEV_STAT(tx_packets), 1);
2780 		bitmap_set(stats_bitmap->bitmap, last_i +
2781 					 MLX4_FIND_NETDEV_STAT(rx_bytes), 1);
2782 		bitmap_set(stats_bitmap->bitmap, last_i +
2783 					 MLX4_FIND_NETDEV_STAT(tx_bytes), 1);
2784 		bitmap_set(stats_bitmap->bitmap, last_i +
2785 					 MLX4_FIND_NETDEV_STAT(rx_dropped), 1);
2786 		bitmap_set(stats_bitmap->bitmap, last_i +
2787 					 MLX4_FIND_NETDEV_STAT(tx_dropped), 1);
2788 	} else {
2789 		bitmap_set(stats_bitmap->bitmap, last_i, NUM_MAIN_STATS);
2790 	}
2791 	last_i += NUM_MAIN_STATS;
2792 
2793 	bitmap_set(stats_bitmap->bitmap, last_i, NUM_PORT_STATS);
2794 	last_i += NUM_PORT_STATS;
2795 
2796 	if (mlx4_is_master(dev))
2797 		bitmap_set(stats_bitmap->bitmap, last_i,
2798 			   NUM_PF_STATS);
2799 	last_i += NUM_PF_STATS;
2800 
2801 	mlx4_en_update_pfc_stats_bitmap(dev, stats_bitmap,
2802 					rx_ppp, rx_pause,
2803 					tx_ppp, tx_pause);
2804 	last_i += NUM_FLOW_STATS;
2805 
2806 	if (!mlx4_is_slave(dev))
2807 		bitmap_set(stats_bitmap->bitmap, last_i, NUM_PKT_STATS);
2808 }
2809 
2810 int mlx4_en_init_netdev(struct mlx4_en_dev *mdev, int port,
2811 			struct mlx4_en_port_profile *prof)
2812 {
2813 	struct net_device *dev;
2814 	struct mlx4_en_priv *priv;
2815 	int i;
2816 	int err;
2817 
2818 	dev = alloc_etherdev_mqs(sizeof(struct mlx4_en_priv),
2819 				 MAX_TX_RINGS, MAX_RX_RINGS);
2820 	if (dev == NULL)
2821 		return -ENOMEM;
2822 
2823 	netif_set_real_num_tx_queues(dev, prof->tx_ring_num);
2824 	netif_set_real_num_rx_queues(dev, prof->rx_ring_num);
2825 
2826 	SET_NETDEV_DEV(dev, &mdev->dev->persist->pdev->dev);
2827 	dev->dev_port = port - 1;
2828 
2829 	/*
2830 	 * Initialize driver private data
2831 	 */
2832 
2833 	priv = netdev_priv(dev);
2834 	memset(priv, 0, sizeof(struct mlx4_en_priv));
2835 	priv->counter_index = MLX4_SINK_COUNTER_INDEX(mdev->dev);
2836 	spin_lock_init(&priv->stats_lock);
2837 	INIT_WORK(&priv->rx_mode_task, mlx4_en_do_set_rx_mode);
2838 	INIT_WORK(&priv->watchdog_task, mlx4_en_restart);
2839 	INIT_WORK(&priv->linkstate_task, mlx4_en_linkstate);
2840 	INIT_DELAYED_WORK(&priv->stats_task, mlx4_en_do_get_stats);
2841 	INIT_DELAYED_WORK(&priv->service_task, mlx4_en_service_task);
2842 #ifdef CONFIG_MLX4_EN_VXLAN
2843 	INIT_WORK(&priv->vxlan_add_task, mlx4_en_add_vxlan_offloads);
2844 	INIT_WORK(&priv->vxlan_del_task, mlx4_en_del_vxlan_offloads);
2845 #endif
2846 #ifdef CONFIG_RFS_ACCEL
2847 	INIT_LIST_HEAD(&priv->filters);
2848 	spin_lock_init(&priv->filters_lock);
2849 #endif
2850 
2851 	priv->dev = dev;
2852 	priv->mdev = mdev;
2853 	priv->ddev = &mdev->pdev->dev;
2854 	priv->prof = prof;
2855 	priv->port = port;
2856 	priv->port_up = false;
2857 	priv->flags = prof->flags;
2858 	priv->pflags = MLX4_EN_PRIV_FLAGS_BLUEFLAME;
2859 	priv->ctrl_flags = cpu_to_be32(MLX4_WQE_CTRL_CQ_UPDATE |
2860 			MLX4_WQE_CTRL_SOLICITED);
2861 	priv->num_tx_rings_p_up = mdev->profile.num_tx_rings_p_up;
2862 	priv->tx_ring_num = prof->tx_ring_num;
2863 	priv->tx_work_limit = MLX4_EN_DEFAULT_TX_WORK;
2864 	netdev_rss_key_fill(priv->rss_key, sizeof(priv->rss_key));
2865 
2866 	priv->tx_ring = kzalloc(sizeof(struct mlx4_en_tx_ring *) * MAX_TX_RINGS,
2867 				GFP_KERNEL);
2868 	if (!priv->tx_ring) {
2869 		err = -ENOMEM;
2870 		goto out;
2871 	}
2872 	priv->tx_cq = kzalloc(sizeof(struct mlx4_en_cq *) * MAX_TX_RINGS,
2873 			      GFP_KERNEL);
2874 	if (!priv->tx_cq) {
2875 		err = -ENOMEM;
2876 		goto out;
2877 	}
2878 	priv->rx_ring_num = prof->rx_ring_num;
2879 	priv->cqe_factor = (mdev->dev->caps.cqe_size == 64) ? 1 : 0;
2880 	priv->cqe_size = mdev->dev->caps.cqe_size;
2881 	priv->mac_index = -1;
2882 	priv->msg_enable = MLX4_EN_MSG_LEVEL;
2883 #ifdef CONFIG_MLX4_EN_DCB
2884 	if (!mlx4_is_slave(priv->mdev->dev)) {
2885 		if (mdev->dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_ETS_CFG) {
2886 			dev->dcbnl_ops = &mlx4_en_dcbnl_ops;
2887 		} else {
2888 			en_info(priv, "enabling only PFC DCB ops\n");
2889 			dev->dcbnl_ops = &mlx4_en_dcbnl_pfc_ops;
2890 		}
2891 	}
2892 #endif
2893 
2894 	for (i = 0; i < MLX4_EN_MAC_HASH_SIZE; ++i)
2895 		INIT_HLIST_HEAD(&priv->mac_hash[i]);
2896 
2897 	/* Query for default mac and max mtu */
2898 	priv->max_mtu = mdev->dev->caps.eth_mtu_cap[priv->port];
2899 
2900 	if (mdev->dev->caps.rx_checksum_flags_port[priv->port] &
2901 	    MLX4_RX_CSUM_MODE_VAL_NON_TCP_UDP)
2902 		priv->flags |= MLX4_EN_FLAG_RX_CSUM_NON_TCP_UDP;
2903 
2904 	/* Set default MAC */
2905 	dev->addr_len = ETH_ALEN;
2906 	mlx4_en_u64_to_mac(dev->dev_addr, mdev->dev->caps.def_mac[priv->port]);
2907 	if (!is_valid_ether_addr(dev->dev_addr)) {
2908 		en_err(priv, "Port: %d, invalid mac burned: %pM, quiting\n",
2909 		       priv->port, dev->dev_addr);
2910 		err = -EINVAL;
2911 		goto out;
2912 	} else if (mlx4_is_slave(priv->mdev->dev) &&
2913 		   (priv->mdev->dev->port_random_macs & 1 << priv->port)) {
2914 		/* Random MAC was assigned in mlx4_slave_cap
2915 		 * in mlx4_core module
2916 		 */
2917 		dev->addr_assign_type |= NET_ADDR_RANDOM;
2918 		en_warn(priv, "Assigned random MAC address %pM\n", dev->dev_addr);
2919 	}
2920 
2921 	memcpy(priv->current_mac, dev->dev_addr, sizeof(priv->current_mac));
2922 
2923 	priv->stride = roundup_pow_of_two(sizeof(struct mlx4_en_rx_desc) +
2924 					  DS_SIZE * MLX4_EN_MAX_RX_FRAGS);
2925 	err = mlx4_en_alloc_resources(priv);
2926 	if (err)
2927 		goto out;
2928 
2929 	/* Initialize time stamping config */
2930 	priv->hwtstamp_config.flags = 0;
2931 	priv->hwtstamp_config.tx_type = HWTSTAMP_TX_OFF;
2932 	priv->hwtstamp_config.rx_filter = HWTSTAMP_FILTER_NONE;
2933 
2934 	/* Allocate page for receive rings */
2935 	err = mlx4_alloc_hwq_res(mdev->dev, &priv->res,
2936 				MLX4_EN_PAGE_SIZE);
2937 	if (err) {
2938 		en_err(priv, "Failed to allocate page for rx qps\n");
2939 		goto out;
2940 	}
2941 	priv->allocated = 1;
2942 
2943 	/*
2944 	 * Initialize netdev entry points
2945 	 */
2946 	if (mlx4_is_master(priv->mdev->dev))
2947 		dev->netdev_ops = &mlx4_netdev_ops_master;
2948 	else
2949 		dev->netdev_ops = &mlx4_netdev_ops;
2950 	dev->watchdog_timeo = MLX4_EN_WATCHDOG_TIMEOUT;
2951 	netif_set_real_num_tx_queues(dev, priv->tx_ring_num);
2952 	netif_set_real_num_rx_queues(dev, priv->rx_ring_num);
2953 
2954 	dev->ethtool_ops = &mlx4_en_ethtool_ops;
2955 
2956 	/*
2957 	 * Set driver features
2958 	 */
2959 	dev->hw_features = NETIF_F_SG | NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM;
2960 	if (mdev->LSO_support)
2961 		dev->hw_features |= NETIF_F_TSO | NETIF_F_TSO6;
2962 
2963 	dev->vlan_features = dev->hw_features;
2964 
2965 	dev->hw_features |= NETIF_F_RXCSUM | NETIF_F_RXHASH;
2966 	dev->features = dev->hw_features | NETIF_F_HIGHDMA |
2967 			NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_CTAG_RX |
2968 			NETIF_F_HW_VLAN_CTAG_FILTER;
2969 	dev->hw_features |= NETIF_F_LOOPBACK |
2970 			NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_CTAG_RX;
2971 
2972 	if (!(mdev->dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_SKIP_OUTER_VLAN)) {
2973 		dev->features |= NETIF_F_HW_VLAN_STAG_RX |
2974 			NETIF_F_HW_VLAN_STAG_FILTER;
2975 		dev->hw_features |= NETIF_F_HW_VLAN_STAG_RX;
2976 	}
2977 
2978 	if (mlx4_is_slave(mdev->dev)) {
2979 		int phv;
2980 
2981 		err = get_phv_bit(mdev->dev, port, &phv);
2982 		if (!err && phv) {
2983 			dev->hw_features |= NETIF_F_HW_VLAN_STAG_TX;
2984 			priv->pflags |= MLX4_EN_PRIV_FLAGS_PHV;
2985 		}
2986 	} else {
2987 		if (mdev->dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_PHV_EN &&
2988 		    !(mdev->dev->caps.flags2 &
2989 		      MLX4_DEV_CAP_FLAG2_SKIP_OUTER_VLAN))
2990 			dev->hw_features |= NETIF_F_HW_VLAN_STAG_TX;
2991 	}
2992 
2993 	if (mdev->dev->caps.flags & MLX4_DEV_CAP_FLAG_FCS_KEEP)
2994 		dev->hw_features |= NETIF_F_RXFCS;
2995 
2996 	if (mdev->dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_IGNORE_FCS)
2997 		dev->hw_features |= NETIF_F_RXALL;
2998 
2999 	if (mdev->dev->caps.steering_mode ==
3000 	    MLX4_STEERING_MODE_DEVICE_MANAGED &&
3001 	    mdev->dev->caps.dmfs_high_steer_mode != MLX4_STEERING_DMFS_A0_STATIC)
3002 		dev->hw_features |= NETIF_F_NTUPLE;
3003 
3004 	if (mdev->dev->caps.steering_mode != MLX4_STEERING_MODE_A0)
3005 		dev->priv_flags |= IFF_UNICAST_FLT;
3006 
3007 	/* Setting a default hash function value */
3008 	if (mdev->dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_RSS_TOP) {
3009 		priv->rss_hash_fn = ETH_RSS_HASH_TOP;
3010 	} else if (mdev->dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_RSS_XOR) {
3011 		priv->rss_hash_fn = ETH_RSS_HASH_XOR;
3012 	} else {
3013 		en_warn(priv,
3014 			"No RSS hash capabilities exposed, using Toeplitz\n");
3015 		priv->rss_hash_fn = ETH_RSS_HASH_TOP;
3016 	}
3017 
3018 	if (mdev->dev->caps.tunnel_offload_mode == MLX4_TUNNEL_OFFLOAD_MODE_VXLAN) {
3019 		dev->hw_features |= NETIF_F_GSO_UDP_TUNNEL |
3020 				    NETIF_F_GSO_UDP_TUNNEL_CSUM |
3021 				    NETIF_F_GSO_PARTIAL;
3022 		dev->features    |= NETIF_F_GSO_UDP_TUNNEL |
3023 				    NETIF_F_GSO_UDP_TUNNEL_CSUM |
3024 				    NETIF_F_GSO_PARTIAL;
3025 		dev->gso_partial_features = NETIF_F_GSO_UDP_TUNNEL_CSUM;
3026 	}
3027 
3028 	mdev->pndev[port] = dev;
3029 	mdev->upper[port] = NULL;
3030 
3031 	netif_carrier_off(dev);
3032 	mlx4_en_set_default_moderation(priv);
3033 
3034 	en_warn(priv, "Using %d TX rings\n", prof->tx_ring_num);
3035 	en_warn(priv, "Using %d RX rings\n", prof->rx_ring_num);
3036 
3037 	mlx4_en_update_loopback_state(priv->dev, priv->dev->features);
3038 
3039 	/* Configure port */
3040 	mlx4_en_calc_rx_buf(dev);
3041 	err = mlx4_SET_PORT_general(mdev->dev, priv->port,
3042 				    priv->rx_skb_size + ETH_FCS_LEN,
3043 				    prof->tx_pause, prof->tx_ppp,
3044 				    prof->rx_pause, prof->rx_ppp);
3045 	if (err) {
3046 		en_err(priv, "Failed setting port general configurations for port %d, with error %d\n",
3047 		       priv->port, err);
3048 		goto out;
3049 	}
3050 
3051 	if (mdev->dev->caps.tunnel_offload_mode == MLX4_TUNNEL_OFFLOAD_MODE_VXLAN) {
3052 		err = mlx4_SET_PORT_VXLAN(mdev->dev, priv->port, VXLAN_STEER_BY_OUTER_MAC, 1);
3053 		if (err) {
3054 			en_err(priv, "Failed setting port L2 tunnel configuration, err %d\n",
3055 			       err);
3056 			goto out;
3057 		}
3058 	}
3059 
3060 	/* Init port */
3061 	en_warn(priv, "Initializing port\n");
3062 	err = mlx4_INIT_PORT(mdev->dev, priv->port);
3063 	if (err) {
3064 		en_err(priv, "Failed Initializing port\n");
3065 		goto out;
3066 	}
3067 	queue_delayed_work(mdev->workqueue, &priv->stats_task, STATS_DELAY);
3068 
3069 	/* Initialize time stamp mechanism */
3070 	if (mdev->dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_TS)
3071 		mlx4_en_init_timestamp(mdev);
3072 
3073 	queue_delayed_work(mdev->workqueue, &priv->service_task,
3074 			   SERVICE_TASK_DELAY);
3075 
3076 	mlx4_en_set_stats_bitmap(mdev->dev, &priv->stats_bitmap,
3077 				 mdev->profile.prof[priv->port].rx_ppp,
3078 				 mdev->profile.prof[priv->port].rx_pause,
3079 				 mdev->profile.prof[priv->port].tx_ppp,
3080 				 mdev->profile.prof[priv->port].tx_pause);
3081 
3082 	err = register_netdev(dev);
3083 	if (err) {
3084 		en_err(priv, "Netdev registration failed for port %d\n", port);
3085 		goto out;
3086 	}
3087 
3088 	priv->registered = 1;
3089 	devlink_port_type_eth_set(mlx4_get_devlink_port(mdev->dev, priv->port),
3090 				  dev);
3091 
3092 	return 0;
3093 
3094 out:
3095 	mlx4_en_destroy_netdev(dev);
3096 	return err;
3097 }
3098 
3099 int mlx4_en_reset_config(struct net_device *dev,
3100 			 struct hwtstamp_config ts_config,
3101 			 netdev_features_t features)
3102 {
3103 	struct mlx4_en_priv *priv = netdev_priv(dev);
3104 	struct mlx4_en_dev *mdev = priv->mdev;
3105 	int port_up = 0;
3106 	int err = 0;
3107 
3108 	if (priv->hwtstamp_config.tx_type == ts_config.tx_type &&
3109 	    priv->hwtstamp_config.rx_filter == ts_config.rx_filter &&
3110 	    !DEV_FEATURE_CHANGED(dev, features, NETIF_F_HW_VLAN_CTAG_RX) &&
3111 	    !DEV_FEATURE_CHANGED(dev, features, NETIF_F_RXFCS))
3112 		return 0; /* Nothing to change */
3113 
3114 	if (DEV_FEATURE_CHANGED(dev, features, NETIF_F_HW_VLAN_CTAG_RX) &&
3115 	    (features & NETIF_F_HW_VLAN_CTAG_RX) &&
3116 	    (priv->hwtstamp_config.rx_filter != HWTSTAMP_FILTER_NONE)) {
3117 		en_warn(priv, "Can't turn ON rx vlan offload while time-stamping rx filter is ON\n");
3118 		return -EINVAL;
3119 	}
3120 
3121 	mutex_lock(&mdev->state_lock);
3122 	if (priv->port_up) {
3123 		port_up = 1;
3124 		mlx4_en_stop_port(dev, 1);
3125 	}
3126 
3127 	mlx4_en_free_resources(priv);
3128 
3129 	en_warn(priv, "Changing device configuration rx filter(%x) rx vlan(%x)\n",
3130 		ts_config.rx_filter, !!(features & NETIF_F_HW_VLAN_CTAG_RX));
3131 
3132 	priv->hwtstamp_config.tx_type = ts_config.tx_type;
3133 	priv->hwtstamp_config.rx_filter = ts_config.rx_filter;
3134 
3135 	if (DEV_FEATURE_CHANGED(dev, features, NETIF_F_HW_VLAN_CTAG_RX)) {
3136 		if (features & NETIF_F_HW_VLAN_CTAG_RX)
3137 			dev->features |= NETIF_F_HW_VLAN_CTAG_RX;
3138 		else
3139 			dev->features &= ~NETIF_F_HW_VLAN_CTAG_RX;
3140 	} else if (ts_config.rx_filter == HWTSTAMP_FILTER_NONE) {
3141 		/* RX time-stamping is OFF, update the RX vlan offload
3142 		 * to the latest wanted state
3143 		 */
3144 		if (dev->wanted_features & NETIF_F_HW_VLAN_CTAG_RX)
3145 			dev->features |= NETIF_F_HW_VLAN_CTAG_RX;
3146 		else
3147 			dev->features &= ~NETIF_F_HW_VLAN_CTAG_RX;
3148 	}
3149 
3150 	if (DEV_FEATURE_CHANGED(dev, features, NETIF_F_RXFCS)) {
3151 		if (features & NETIF_F_RXFCS)
3152 			dev->features |= NETIF_F_RXFCS;
3153 		else
3154 			dev->features &= ~NETIF_F_RXFCS;
3155 	}
3156 
3157 	/* RX vlan offload and RX time-stamping can't co-exist !
3158 	 * Regardless of the caller's choice,
3159 	 * Turn Off RX vlan offload in case of time-stamping is ON
3160 	 */
3161 	if (ts_config.rx_filter != HWTSTAMP_FILTER_NONE) {
3162 		if (dev->features & NETIF_F_HW_VLAN_CTAG_RX)
3163 			en_warn(priv, "Turning off RX vlan offload since RX time-stamping is ON\n");
3164 		dev->features &= ~NETIF_F_HW_VLAN_CTAG_RX;
3165 	}
3166 
3167 	err = mlx4_en_alloc_resources(priv);
3168 	if (err) {
3169 		en_err(priv, "Failed reallocating port resources\n");
3170 		goto out;
3171 	}
3172 	if (port_up) {
3173 		err = mlx4_en_start_port(dev);
3174 		if (err)
3175 			en_err(priv, "Failed starting port\n");
3176 	}
3177 
3178 out:
3179 	mutex_unlock(&mdev->state_lock);
3180 	netdev_features_change(dev);
3181 	return err;
3182 }
3183