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/bpf.h>
35 #include <linux/etherdevice.h>
36 #include <linux/tcp.h>
37 #include <linux/if_vlan.h>
38 #include <linux/delay.h>
39 #include <linux/slab.h>
40 #include <linux/hash.h>
41 #include <net/ip.h>
42 #include <net/busy_poll.h>
43 #include <net/vxlan.h>
44 #include <net/devlink.h>
45 
46 #include <linux/mlx4/driver.h>
47 #include <linux/mlx4/device.h>
48 #include <linux/mlx4/cmd.h>
49 #include <linux/mlx4/cq.h>
50 
51 #include "mlx4_en.h"
52 #include "en_port.h"
53 
54 #define MLX4_EN_MAX_XDP_MTU ((int)(PAGE_SIZE - ETH_HLEN - (2 * VLAN_HLEN) - \
55 				   XDP_PACKET_HEADROOM))
56 
57 int mlx4_en_setup_tc(struct net_device *dev, u8 up)
58 {
59 	struct mlx4_en_priv *priv = netdev_priv(dev);
60 	int i;
61 	unsigned int offset = 0;
62 
63 	if (up && up != MLX4_EN_NUM_UP_HIGH)
64 		return -EINVAL;
65 
66 	netdev_set_num_tc(dev, up);
67 	netif_set_real_num_tx_queues(dev, priv->tx_ring_num[TX]);
68 	/* Partition Tx queues evenly amongst UP's */
69 	for (i = 0; i < up; i++) {
70 		netdev_set_tc_queue(dev, i, priv->num_tx_rings_p_up, offset);
71 		offset += priv->num_tx_rings_p_up;
72 	}
73 
74 #ifdef CONFIG_MLX4_EN_DCB
75 	if (!mlx4_is_slave(priv->mdev->dev)) {
76 		if (up) {
77 			if (priv->dcbx_cap)
78 				priv->flags |= MLX4_EN_FLAG_DCB_ENABLED;
79 		} else {
80 			priv->flags &= ~MLX4_EN_FLAG_DCB_ENABLED;
81 			priv->cee_config.pfc_state = false;
82 		}
83 	}
84 #endif /* CONFIG_MLX4_EN_DCB */
85 
86 	return 0;
87 }
88 
89 int mlx4_en_alloc_tx_queue_per_tc(struct net_device *dev, u8 tc)
90 {
91 	struct mlx4_en_priv *priv = netdev_priv(dev);
92 	struct mlx4_en_dev *mdev = priv->mdev;
93 	struct mlx4_en_port_profile new_prof;
94 	struct mlx4_en_priv *tmp;
95 	int port_up = 0;
96 	int err = 0;
97 
98 	tmp = kzalloc(sizeof(*tmp), GFP_KERNEL);
99 	if (!tmp)
100 		return -ENOMEM;
101 
102 	mutex_lock(&mdev->state_lock);
103 	memcpy(&new_prof, priv->prof, sizeof(struct mlx4_en_port_profile));
104 	new_prof.num_up = (tc == 0) ? MLX4_EN_NUM_UP_LOW :
105 				      MLX4_EN_NUM_UP_HIGH;
106 	new_prof.tx_ring_num[TX] = new_prof.num_tx_rings_p_up *
107 				   new_prof.num_up;
108 	err = mlx4_en_try_alloc_resources(priv, tmp, &new_prof, true);
109 	if (err)
110 		goto out;
111 
112 	if (priv->port_up) {
113 		port_up = 1;
114 		mlx4_en_stop_port(dev, 1);
115 	}
116 
117 	mlx4_en_safe_replace_resources(priv, tmp);
118 	if (port_up) {
119 		err = mlx4_en_start_port(dev);
120 		if (err) {
121 			en_err(priv, "Failed starting port for setup TC\n");
122 			goto out;
123 		}
124 	}
125 
126 	err = mlx4_en_setup_tc(dev, tc);
127 out:
128 	mutex_unlock(&mdev->state_lock);
129 	kfree(tmp);
130 	return err;
131 }
132 
133 static int __mlx4_en_setup_tc(struct net_device *dev, enum tc_setup_type type,
134 			      void *type_data)
135 {
136 	struct tc_mqprio_qopt *mqprio = type_data;
137 
138 	if (type != TC_SETUP_QDISC_MQPRIO)
139 		return -EOPNOTSUPP;
140 
141 	if (mqprio->num_tc && mqprio->num_tc != MLX4_EN_NUM_UP_HIGH)
142 		return -EINVAL;
143 
144 	mqprio->hw = TC_MQPRIO_HW_OFFLOAD_TCS;
145 
146 	return mlx4_en_alloc_tx_queue_per_tc(dev, mqprio->num_tc);
147 }
148 
149 #ifdef CONFIG_RFS_ACCEL
150 
151 struct mlx4_en_filter {
152 	struct list_head next;
153 	struct work_struct work;
154 
155 	u8     ip_proto;
156 	__be32 src_ip;
157 	__be32 dst_ip;
158 	__be16 src_port;
159 	__be16 dst_port;
160 
161 	int rxq_index;
162 	struct mlx4_en_priv *priv;
163 	u32 flow_id;			/* RFS infrastructure id */
164 	int id;				/* mlx4_en driver id */
165 	u64 reg_id;			/* Flow steering API id */
166 	u8 activated;			/* Used to prevent expiry before filter
167 					 * is attached
168 					 */
169 	struct hlist_node filter_chain;
170 };
171 
172 static void mlx4_en_filter_rfs_expire(struct mlx4_en_priv *priv);
173 
174 static enum mlx4_net_trans_rule_id mlx4_ip_proto_to_trans_rule_id(u8 ip_proto)
175 {
176 	switch (ip_proto) {
177 	case IPPROTO_UDP:
178 		return MLX4_NET_TRANS_RULE_ID_UDP;
179 	case IPPROTO_TCP:
180 		return MLX4_NET_TRANS_RULE_ID_TCP;
181 	default:
182 		return MLX4_NET_TRANS_RULE_NUM;
183 	}
184 };
185 
186 /* Must not acquire state_lock, as its corresponding work_sync
187  * is done under it.
188  */
189 static void mlx4_en_filter_work(struct work_struct *work)
190 {
191 	struct mlx4_en_filter *filter = container_of(work,
192 						     struct mlx4_en_filter,
193 						     work);
194 	struct mlx4_en_priv *priv = filter->priv;
195 	struct mlx4_spec_list spec_tcp_udp = {
196 		.id = mlx4_ip_proto_to_trans_rule_id(filter->ip_proto),
197 		{
198 			.tcp_udp = {
199 				.dst_port = filter->dst_port,
200 				.dst_port_msk = (__force __be16)-1,
201 				.src_port = filter->src_port,
202 				.src_port_msk = (__force __be16)-1,
203 			},
204 		},
205 	};
206 	struct mlx4_spec_list spec_ip = {
207 		.id = MLX4_NET_TRANS_RULE_ID_IPV4,
208 		{
209 			.ipv4 = {
210 				.dst_ip = filter->dst_ip,
211 				.dst_ip_msk = (__force __be32)-1,
212 				.src_ip = filter->src_ip,
213 				.src_ip_msk = (__force __be32)-1,
214 			},
215 		},
216 	};
217 	struct mlx4_spec_list spec_eth = {
218 		.id = MLX4_NET_TRANS_RULE_ID_ETH,
219 	};
220 	struct mlx4_net_trans_rule rule = {
221 		.list = LIST_HEAD_INIT(rule.list),
222 		.queue_mode = MLX4_NET_TRANS_Q_LIFO,
223 		.exclusive = 1,
224 		.allow_loopback = 1,
225 		.promisc_mode = MLX4_FS_REGULAR,
226 		.port = priv->port,
227 		.priority = MLX4_DOMAIN_RFS,
228 	};
229 	int rc;
230 	__be64 mac_mask = cpu_to_be64(MLX4_MAC_MASK << 16);
231 
232 	if (spec_tcp_udp.id >= MLX4_NET_TRANS_RULE_NUM) {
233 		en_warn(priv, "RFS: ignoring unsupported ip protocol (%d)\n",
234 			filter->ip_proto);
235 		goto ignore;
236 	}
237 	list_add_tail(&spec_eth.list, &rule.list);
238 	list_add_tail(&spec_ip.list, &rule.list);
239 	list_add_tail(&spec_tcp_udp.list, &rule.list);
240 
241 	rule.qpn = priv->rss_map.qps[filter->rxq_index].qpn;
242 	memcpy(spec_eth.eth.dst_mac, priv->dev->dev_addr, ETH_ALEN);
243 	memcpy(spec_eth.eth.dst_mac_msk, &mac_mask, ETH_ALEN);
244 
245 	filter->activated = 0;
246 
247 	if (filter->reg_id) {
248 		rc = mlx4_flow_detach(priv->mdev->dev, filter->reg_id);
249 		if (rc && rc != -ENOENT)
250 			en_err(priv, "Error detaching flow. rc = %d\n", rc);
251 	}
252 
253 	rc = mlx4_flow_attach(priv->mdev->dev, &rule, &filter->reg_id);
254 	if (rc)
255 		en_err(priv, "Error attaching flow. err = %d\n", rc);
256 
257 ignore:
258 	mlx4_en_filter_rfs_expire(priv);
259 
260 	filter->activated = 1;
261 }
262 
263 static inline struct hlist_head *
264 filter_hash_bucket(struct mlx4_en_priv *priv, __be32 src_ip, __be32 dst_ip,
265 		   __be16 src_port, __be16 dst_port)
266 {
267 	unsigned long l;
268 	int bucket_idx;
269 
270 	l = (__force unsigned long)src_port |
271 	    ((__force unsigned long)dst_port << 2);
272 	l ^= (__force unsigned long)(src_ip ^ dst_ip);
273 
274 	bucket_idx = hash_long(l, MLX4_EN_FILTER_HASH_SHIFT);
275 
276 	return &priv->filter_hash[bucket_idx];
277 }
278 
279 static struct mlx4_en_filter *
280 mlx4_en_filter_alloc(struct mlx4_en_priv *priv, int rxq_index, __be32 src_ip,
281 		     __be32 dst_ip, u8 ip_proto, __be16 src_port,
282 		     __be16 dst_port, u32 flow_id)
283 {
284 	struct mlx4_en_filter *filter = NULL;
285 
286 	filter = kzalloc(sizeof(struct mlx4_en_filter), GFP_ATOMIC);
287 	if (!filter)
288 		return NULL;
289 
290 	filter->priv = priv;
291 	filter->rxq_index = rxq_index;
292 	INIT_WORK(&filter->work, mlx4_en_filter_work);
293 
294 	filter->src_ip = src_ip;
295 	filter->dst_ip = dst_ip;
296 	filter->ip_proto = ip_proto;
297 	filter->src_port = src_port;
298 	filter->dst_port = dst_port;
299 
300 	filter->flow_id = flow_id;
301 
302 	filter->id = priv->last_filter_id++ % RPS_NO_FILTER;
303 
304 	list_add_tail(&filter->next, &priv->filters);
305 	hlist_add_head(&filter->filter_chain,
306 		       filter_hash_bucket(priv, src_ip, dst_ip, src_port,
307 					  dst_port));
308 
309 	return filter;
310 }
311 
312 static void mlx4_en_filter_free(struct mlx4_en_filter *filter)
313 {
314 	struct mlx4_en_priv *priv = filter->priv;
315 	int rc;
316 
317 	list_del(&filter->next);
318 
319 	rc = mlx4_flow_detach(priv->mdev->dev, filter->reg_id);
320 	if (rc && rc != -ENOENT)
321 		en_err(priv, "Error detaching flow. rc = %d\n", rc);
322 
323 	kfree(filter);
324 }
325 
326 static inline struct mlx4_en_filter *
327 mlx4_en_filter_find(struct mlx4_en_priv *priv, __be32 src_ip, __be32 dst_ip,
328 		    u8 ip_proto, __be16 src_port, __be16 dst_port)
329 {
330 	struct mlx4_en_filter *filter;
331 	struct mlx4_en_filter *ret = NULL;
332 
333 	hlist_for_each_entry(filter,
334 			     filter_hash_bucket(priv, src_ip, dst_ip,
335 						src_port, dst_port),
336 			     filter_chain) {
337 		if (filter->src_ip == src_ip &&
338 		    filter->dst_ip == dst_ip &&
339 		    filter->ip_proto == ip_proto &&
340 		    filter->src_port == src_port &&
341 		    filter->dst_port == dst_port) {
342 			ret = filter;
343 			break;
344 		}
345 	}
346 
347 	return ret;
348 }
349 
350 static int
351 mlx4_en_filter_rfs(struct net_device *net_dev, const struct sk_buff *skb,
352 		   u16 rxq_index, u32 flow_id)
353 {
354 	struct mlx4_en_priv *priv = netdev_priv(net_dev);
355 	struct mlx4_en_filter *filter;
356 	const struct iphdr *ip;
357 	const __be16 *ports;
358 	u8 ip_proto;
359 	__be32 src_ip;
360 	__be32 dst_ip;
361 	__be16 src_port;
362 	__be16 dst_port;
363 	int nhoff = skb_network_offset(skb);
364 	int ret = 0;
365 
366 	if (skb->protocol != htons(ETH_P_IP))
367 		return -EPROTONOSUPPORT;
368 
369 	ip = (const struct iphdr *)(skb->data + nhoff);
370 	if (ip_is_fragment(ip))
371 		return -EPROTONOSUPPORT;
372 
373 	if ((ip->protocol != IPPROTO_TCP) && (ip->protocol != IPPROTO_UDP))
374 		return -EPROTONOSUPPORT;
375 	ports = (const __be16 *)(skb->data + nhoff + 4 * ip->ihl);
376 
377 	ip_proto = ip->protocol;
378 	src_ip = ip->saddr;
379 	dst_ip = ip->daddr;
380 	src_port = ports[0];
381 	dst_port = ports[1];
382 
383 	spin_lock_bh(&priv->filters_lock);
384 	filter = mlx4_en_filter_find(priv, src_ip, dst_ip, ip_proto,
385 				     src_port, dst_port);
386 	if (filter) {
387 		if (filter->rxq_index == rxq_index)
388 			goto out;
389 
390 		filter->rxq_index = rxq_index;
391 	} else {
392 		filter = mlx4_en_filter_alloc(priv, rxq_index,
393 					      src_ip, dst_ip, ip_proto,
394 					      src_port, dst_port, flow_id);
395 		if (!filter) {
396 			ret = -ENOMEM;
397 			goto err;
398 		}
399 	}
400 
401 	queue_work(priv->mdev->workqueue, &filter->work);
402 
403 out:
404 	ret = filter->id;
405 err:
406 	spin_unlock_bh(&priv->filters_lock);
407 
408 	return ret;
409 }
410 
411 void mlx4_en_cleanup_filters(struct mlx4_en_priv *priv)
412 {
413 	struct mlx4_en_filter *filter, *tmp;
414 	LIST_HEAD(del_list);
415 
416 	spin_lock_bh(&priv->filters_lock);
417 	list_for_each_entry_safe(filter, tmp, &priv->filters, next) {
418 		list_move(&filter->next, &del_list);
419 		hlist_del(&filter->filter_chain);
420 	}
421 	spin_unlock_bh(&priv->filters_lock);
422 
423 	list_for_each_entry_safe(filter, tmp, &del_list, next) {
424 		cancel_work_sync(&filter->work);
425 		mlx4_en_filter_free(filter);
426 	}
427 }
428 
429 static void mlx4_en_filter_rfs_expire(struct mlx4_en_priv *priv)
430 {
431 	struct mlx4_en_filter *filter = NULL, *tmp, *last_filter = NULL;
432 	LIST_HEAD(del_list);
433 	int i = 0;
434 
435 	spin_lock_bh(&priv->filters_lock);
436 	list_for_each_entry_safe(filter, tmp, &priv->filters, next) {
437 		if (i > MLX4_EN_FILTER_EXPIRY_QUOTA)
438 			break;
439 
440 		if (filter->activated &&
441 		    !work_pending(&filter->work) &&
442 		    rps_may_expire_flow(priv->dev,
443 					filter->rxq_index, filter->flow_id,
444 					filter->id)) {
445 			list_move(&filter->next, &del_list);
446 			hlist_del(&filter->filter_chain);
447 		} else
448 			last_filter = filter;
449 
450 		i++;
451 	}
452 
453 	if (last_filter && (&last_filter->next != priv->filters.next))
454 		list_move(&priv->filters, &last_filter->next);
455 
456 	spin_unlock_bh(&priv->filters_lock);
457 
458 	list_for_each_entry_safe(filter, tmp, &del_list, next)
459 		mlx4_en_filter_free(filter);
460 }
461 #endif
462 
463 static int mlx4_en_vlan_rx_add_vid(struct net_device *dev,
464 				   __be16 proto, u16 vid)
465 {
466 	struct mlx4_en_priv *priv = netdev_priv(dev);
467 	struct mlx4_en_dev *mdev = priv->mdev;
468 	int err;
469 	int idx;
470 
471 	en_dbg(HW, priv, "adding VLAN:%d\n", vid);
472 
473 	set_bit(vid, priv->active_vlans);
474 
475 	/* Add VID to port VLAN filter */
476 	mutex_lock(&mdev->state_lock);
477 	if (mdev->device_up && priv->port_up) {
478 		err = mlx4_SET_VLAN_FLTR(mdev->dev, priv);
479 		if (err) {
480 			en_err(priv, "Failed configuring VLAN filter\n");
481 			goto out;
482 		}
483 	}
484 	err = mlx4_register_vlan(mdev->dev, priv->port, vid, &idx);
485 	if (err)
486 		en_dbg(HW, priv, "Failed adding vlan %d\n", vid);
487 
488 out:
489 	mutex_unlock(&mdev->state_lock);
490 	return err;
491 }
492 
493 static int mlx4_en_vlan_rx_kill_vid(struct net_device *dev,
494 				    __be16 proto, u16 vid)
495 {
496 	struct mlx4_en_priv *priv = netdev_priv(dev);
497 	struct mlx4_en_dev *mdev = priv->mdev;
498 	int err = 0;
499 
500 	en_dbg(HW, priv, "Killing VID:%d\n", vid);
501 
502 	clear_bit(vid, priv->active_vlans);
503 
504 	/* Remove VID from port VLAN filter */
505 	mutex_lock(&mdev->state_lock);
506 	mlx4_unregister_vlan(mdev->dev, priv->port, vid);
507 
508 	if (mdev->device_up && priv->port_up) {
509 		err = mlx4_SET_VLAN_FLTR(mdev->dev, priv);
510 		if (err)
511 			en_err(priv, "Failed configuring VLAN filter\n");
512 	}
513 	mutex_unlock(&mdev->state_lock);
514 
515 	return err;
516 }
517 
518 static void mlx4_en_u64_to_mac(unsigned char dst_mac[ETH_ALEN + 2], u64 src_mac)
519 {
520 	int i;
521 	for (i = ETH_ALEN - 1; i >= 0; --i) {
522 		dst_mac[i] = src_mac & 0xff;
523 		src_mac >>= 8;
524 	}
525 	memset(&dst_mac[ETH_ALEN], 0, 2);
526 }
527 
528 
529 static int mlx4_en_tunnel_steer_add(struct mlx4_en_priv *priv, unsigned char *addr,
530 				    int qpn, u64 *reg_id)
531 {
532 	int err;
533 
534 	if (priv->mdev->dev->caps.tunnel_offload_mode != MLX4_TUNNEL_OFFLOAD_MODE_VXLAN ||
535 	    priv->mdev->dev->caps.dmfs_high_steer_mode == MLX4_STEERING_DMFS_A0_STATIC)
536 		return 0; /* do nothing */
537 
538 	err = mlx4_tunnel_steer_add(priv->mdev->dev, addr, priv->port, qpn,
539 				    MLX4_DOMAIN_NIC, reg_id);
540 	if (err) {
541 		en_err(priv, "failed to add vxlan steering rule, err %d\n", err);
542 		return err;
543 	}
544 	en_dbg(DRV, priv, "added vxlan steering rule, mac %pM reg_id %llx\n", addr, *reg_id);
545 	return 0;
546 }
547 
548 
549 static int mlx4_en_uc_steer_add(struct mlx4_en_priv *priv,
550 				unsigned char *mac, int *qpn, u64 *reg_id)
551 {
552 	struct mlx4_en_dev *mdev = priv->mdev;
553 	struct mlx4_dev *dev = mdev->dev;
554 	int err;
555 
556 	switch (dev->caps.steering_mode) {
557 	case MLX4_STEERING_MODE_B0: {
558 		struct mlx4_qp qp;
559 		u8 gid[16] = {0};
560 
561 		qp.qpn = *qpn;
562 		memcpy(&gid[10], mac, ETH_ALEN);
563 		gid[5] = priv->port;
564 
565 		err = mlx4_unicast_attach(dev, &qp, gid, 0, MLX4_PROT_ETH);
566 		break;
567 	}
568 	case MLX4_STEERING_MODE_DEVICE_MANAGED: {
569 		struct mlx4_spec_list spec_eth = { {NULL} };
570 		__be64 mac_mask = cpu_to_be64(MLX4_MAC_MASK << 16);
571 
572 		struct mlx4_net_trans_rule rule = {
573 			.queue_mode = MLX4_NET_TRANS_Q_FIFO,
574 			.exclusive = 0,
575 			.allow_loopback = 1,
576 			.promisc_mode = MLX4_FS_REGULAR,
577 			.priority = MLX4_DOMAIN_NIC,
578 		};
579 
580 		rule.port = priv->port;
581 		rule.qpn = *qpn;
582 		INIT_LIST_HEAD(&rule.list);
583 
584 		spec_eth.id = MLX4_NET_TRANS_RULE_ID_ETH;
585 		memcpy(spec_eth.eth.dst_mac, mac, ETH_ALEN);
586 		memcpy(spec_eth.eth.dst_mac_msk, &mac_mask, ETH_ALEN);
587 		list_add_tail(&spec_eth.list, &rule.list);
588 
589 		err = mlx4_flow_attach(dev, &rule, reg_id);
590 		break;
591 	}
592 	default:
593 		return -EINVAL;
594 	}
595 	if (err)
596 		en_warn(priv, "Failed Attaching Unicast\n");
597 
598 	return err;
599 }
600 
601 static void mlx4_en_uc_steer_release(struct mlx4_en_priv *priv,
602 				     unsigned char *mac, int qpn, u64 reg_id)
603 {
604 	struct mlx4_en_dev *mdev = priv->mdev;
605 	struct mlx4_dev *dev = mdev->dev;
606 
607 	switch (dev->caps.steering_mode) {
608 	case MLX4_STEERING_MODE_B0: {
609 		struct mlx4_qp qp;
610 		u8 gid[16] = {0};
611 
612 		qp.qpn = qpn;
613 		memcpy(&gid[10], mac, ETH_ALEN);
614 		gid[5] = priv->port;
615 
616 		mlx4_unicast_detach(dev, &qp, gid, MLX4_PROT_ETH);
617 		break;
618 	}
619 	case MLX4_STEERING_MODE_DEVICE_MANAGED: {
620 		mlx4_flow_detach(dev, reg_id);
621 		break;
622 	}
623 	default:
624 		en_err(priv, "Invalid steering mode.\n");
625 	}
626 }
627 
628 static int mlx4_en_get_qp(struct mlx4_en_priv *priv)
629 {
630 	struct mlx4_en_dev *mdev = priv->mdev;
631 	struct mlx4_dev *dev = mdev->dev;
632 	int index = 0;
633 	int err = 0;
634 	int *qpn = &priv->base_qpn;
635 	u64 mac = mlx4_mac_to_u64(priv->dev->dev_addr);
636 
637 	en_dbg(DRV, priv, "Registering MAC: %pM for adding\n",
638 	       priv->dev->dev_addr);
639 	index = mlx4_register_mac(dev, priv->port, mac);
640 	if (index < 0) {
641 		err = index;
642 		en_err(priv, "Failed adding MAC: %pM\n",
643 		       priv->dev->dev_addr);
644 		return err;
645 	}
646 
647 	en_info(priv, "Steering Mode %d\n", dev->caps.steering_mode);
648 
649 	if (dev->caps.steering_mode == MLX4_STEERING_MODE_A0) {
650 		int base_qpn = mlx4_get_base_qpn(dev, priv->port);
651 		*qpn = base_qpn + index;
652 		return 0;
653 	}
654 
655 	err = mlx4_qp_reserve_range(dev, 1, 1, qpn, MLX4_RESERVE_A0_QP,
656 				    MLX4_RES_USAGE_DRIVER);
657 	en_dbg(DRV, priv, "Reserved qp %d\n", *qpn);
658 	if (err) {
659 		en_err(priv, "Failed to reserve qp for mac registration\n");
660 		mlx4_unregister_mac(dev, priv->port, mac);
661 		return err;
662 	}
663 
664 	return 0;
665 }
666 
667 static void mlx4_en_put_qp(struct mlx4_en_priv *priv)
668 {
669 	struct mlx4_en_dev *mdev = priv->mdev;
670 	struct mlx4_dev *dev = mdev->dev;
671 	int qpn = priv->base_qpn;
672 
673 	if (dev->caps.steering_mode == MLX4_STEERING_MODE_A0) {
674 		u64 mac = mlx4_mac_to_u64(priv->dev->dev_addr);
675 		en_dbg(DRV, priv, "Registering MAC: %pM for deleting\n",
676 		       priv->dev->dev_addr);
677 		mlx4_unregister_mac(dev, priv->port, mac);
678 	} else {
679 		en_dbg(DRV, priv, "Releasing qp: port %d, qpn %d\n",
680 		       priv->port, qpn);
681 		mlx4_qp_release_range(dev, qpn, 1);
682 		priv->flags &= ~MLX4_EN_FLAG_FORCE_PROMISC;
683 	}
684 }
685 
686 static int mlx4_en_replace_mac(struct mlx4_en_priv *priv, int qpn,
687 			       unsigned char *new_mac, unsigned char *prev_mac)
688 {
689 	struct mlx4_en_dev *mdev = priv->mdev;
690 	struct mlx4_dev *dev = mdev->dev;
691 	int err = 0;
692 	u64 new_mac_u64 = mlx4_mac_to_u64(new_mac);
693 
694 	if (dev->caps.steering_mode != MLX4_STEERING_MODE_A0) {
695 		struct hlist_head *bucket;
696 		unsigned int mac_hash;
697 		struct mlx4_mac_entry *entry;
698 		struct hlist_node *tmp;
699 		u64 prev_mac_u64 = mlx4_mac_to_u64(prev_mac);
700 
701 		bucket = &priv->mac_hash[prev_mac[MLX4_EN_MAC_HASH_IDX]];
702 		hlist_for_each_entry_safe(entry, tmp, bucket, hlist) {
703 			if (ether_addr_equal_64bits(entry->mac, prev_mac)) {
704 				mlx4_en_uc_steer_release(priv, entry->mac,
705 							 qpn, entry->reg_id);
706 				mlx4_unregister_mac(dev, priv->port,
707 						    prev_mac_u64);
708 				hlist_del_rcu(&entry->hlist);
709 				synchronize_rcu();
710 				memcpy(entry->mac, new_mac, ETH_ALEN);
711 				entry->reg_id = 0;
712 				mac_hash = new_mac[MLX4_EN_MAC_HASH_IDX];
713 				hlist_add_head_rcu(&entry->hlist,
714 						   &priv->mac_hash[mac_hash]);
715 				mlx4_register_mac(dev, priv->port, new_mac_u64);
716 				err = mlx4_en_uc_steer_add(priv, new_mac,
717 							   &qpn,
718 							   &entry->reg_id);
719 				if (err)
720 					return err;
721 				if (priv->tunnel_reg_id) {
722 					mlx4_flow_detach(priv->mdev->dev, priv->tunnel_reg_id);
723 					priv->tunnel_reg_id = 0;
724 				}
725 				err = mlx4_en_tunnel_steer_add(priv, new_mac, qpn,
726 							       &priv->tunnel_reg_id);
727 				return err;
728 			}
729 		}
730 		return -EINVAL;
731 	}
732 
733 	return __mlx4_replace_mac(dev, priv->port, qpn, new_mac_u64);
734 }
735 
736 static void mlx4_en_update_user_mac(struct mlx4_en_priv *priv,
737 				    unsigned char new_mac[ETH_ALEN + 2])
738 {
739 	struct mlx4_en_dev *mdev = priv->mdev;
740 	int err;
741 
742 	if (!(mdev->dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_USER_MAC_EN))
743 		return;
744 
745 	err = mlx4_SET_PORT_user_mac(mdev->dev, priv->port, new_mac);
746 	if (err)
747 		en_err(priv, "Failed to pass user MAC(%pM) to Firmware for port %d, with error %d\n",
748 		       new_mac, priv->port, err);
749 }
750 
751 static int mlx4_en_do_set_mac(struct mlx4_en_priv *priv,
752 			      unsigned char new_mac[ETH_ALEN + 2])
753 {
754 	int err = 0;
755 
756 	if (priv->port_up) {
757 		/* Remove old MAC and insert the new one */
758 		err = mlx4_en_replace_mac(priv, priv->base_qpn,
759 					  new_mac, priv->current_mac);
760 		if (err)
761 			en_err(priv, "Failed changing HW MAC address\n");
762 	} else
763 		en_dbg(HW, priv, "Port is down while registering mac, exiting...\n");
764 
765 	if (!err)
766 		memcpy(priv->current_mac, new_mac, sizeof(priv->current_mac));
767 
768 	return err;
769 }
770 
771 static int mlx4_en_set_mac(struct net_device *dev, void *addr)
772 {
773 	struct mlx4_en_priv *priv = netdev_priv(dev);
774 	struct mlx4_en_dev *mdev = priv->mdev;
775 	struct sockaddr *saddr = addr;
776 	unsigned char new_mac[ETH_ALEN + 2];
777 	int err;
778 
779 	if (!is_valid_ether_addr(saddr->sa_data))
780 		return -EADDRNOTAVAIL;
781 
782 	mutex_lock(&mdev->state_lock);
783 	memcpy(new_mac, saddr->sa_data, ETH_ALEN);
784 	err = mlx4_en_do_set_mac(priv, new_mac);
785 	if (err)
786 		goto out;
787 
788 	memcpy(dev->dev_addr, saddr->sa_data, ETH_ALEN);
789 	mlx4_en_update_user_mac(priv, new_mac);
790 out:
791 	mutex_unlock(&mdev->state_lock);
792 
793 	return err;
794 }
795 
796 static void mlx4_en_clear_list(struct net_device *dev)
797 {
798 	struct mlx4_en_priv *priv = netdev_priv(dev);
799 	struct mlx4_en_mc_list *tmp, *mc_to_del;
800 
801 	list_for_each_entry_safe(mc_to_del, tmp, &priv->mc_list, list) {
802 		list_del(&mc_to_del->list);
803 		kfree(mc_to_del);
804 	}
805 }
806 
807 static void mlx4_en_cache_mclist(struct net_device *dev)
808 {
809 	struct mlx4_en_priv *priv = netdev_priv(dev);
810 	struct netdev_hw_addr *ha;
811 	struct mlx4_en_mc_list *tmp;
812 
813 	mlx4_en_clear_list(dev);
814 	netdev_for_each_mc_addr(ha, dev) {
815 		tmp = kzalloc(sizeof(struct mlx4_en_mc_list), GFP_ATOMIC);
816 		if (!tmp) {
817 			mlx4_en_clear_list(dev);
818 			return;
819 		}
820 		memcpy(tmp->addr, ha->addr, ETH_ALEN);
821 		list_add_tail(&tmp->list, &priv->mc_list);
822 	}
823 }
824 
825 static void update_mclist_flags(struct mlx4_en_priv *priv,
826 				struct list_head *dst,
827 				struct list_head *src)
828 {
829 	struct mlx4_en_mc_list *dst_tmp, *src_tmp, *new_mc;
830 	bool found;
831 
832 	/* Find all the entries that should be removed from dst,
833 	 * These are the entries that are not found in src
834 	 */
835 	list_for_each_entry(dst_tmp, dst, list) {
836 		found = false;
837 		list_for_each_entry(src_tmp, src, list) {
838 			if (ether_addr_equal(dst_tmp->addr, src_tmp->addr)) {
839 				found = true;
840 				break;
841 			}
842 		}
843 		if (!found)
844 			dst_tmp->action = MCLIST_REM;
845 	}
846 
847 	/* Add entries that exist in src but not in dst
848 	 * mark them as need to add
849 	 */
850 	list_for_each_entry(src_tmp, src, list) {
851 		found = false;
852 		list_for_each_entry(dst_tmp, dst, list) {
853 			if (ether_addr_equal(dst_tmp->addr, src_tmp->addr)) {
854 				dst_tmp->action = MCLIST_NONE;
855 				found = true;
856 				break;
857 			}
858 		}
859 		if (!found) {
860 			new_mc = kmemdup(src_tmp,
861 					 sizeof(struct mlx4_en_mc_list),
862 					 GFP_KERNEL);
863 			if (!new_mc)
864 				return;
865 
866 			new_mc->action = MCLIST_ADD;
867 			list_add_tail(&new_mc->list, dst);
868 		}
869 	}
870 }
871 
872 static void mlx4_en_set_rx_mode(struct net_device *dev)
873 {
874 	struct mlx4_en_priv *priv = netdev_priv(dev);
875 
876 	if (!priv->port_up)
877 		return;
878 
879 	queue_work(priv->mdev->workqueue, &priv->rx_mode_task);
880 }
881 
882 static void mlx4_en_set_promisc_mode(struct mlx4_en_priv *priv,
883 				     struct mlx4_en_dev *mdev)
884 {
885 	int err = 0;
886 
887 	if (!(priv->flags & MLX4_EN_FLAG_PROMISC)) {
888 		if (netif_msg_rx_status(priv))
889 			en_warn(priv, "Entering promiscuous mode\n");
890 		priv->flags |= MLX4_EN_FLAG_PROMISC;
891 
892 		/* Enable promiscouos mode */
893 		switch (mdev->dev->caps.steering_mode) {
894 		case MLX4_STEERING_MODE_DEVICE_MANAGED:
895 			err = mlx4_flow_steer_promisc_add(mdev->dev,
896 							  priv->port,
897 							  priv->base_qpn,
898 							  MLX4_FS_ALL_DEFAULT);
899 			if (err)
900 				en_err(priv, "Failed enabling promiscuous mode\n");
901 			priv->flags |= MLX4_EN_FLAG_MC_PROMISC;
902 			break;
903 
904 		case MLX4_STEERING_MODE_B0:
905 			err = mlx4_unicast_promisc_add(mdev->dev,
906 						       priv->base_qpn,
907 						       priv->port);
908 			if (err)
909 				en_err(priv, "Failed enabling unicast promiscuous mode\n");
910 
911 			/* Add the default qp number as multicast
912 			 * promisc
913 			 */
914 			if (!(priv->flags & MLX4_EN_FLAG_MC_PROMISC)) {
915 				err = mlx4_multicast_promisc_add(mdev->dev,
916 								 priv->base_qpn,
917 								 priv->port);
918 				if (err)
919 					en_err(priv, "Failed enabling multicast promiscuous mode\n");
920 				priv->flags |= MLX4_EN_FLAG_MC_PROMISC;
921 			}
922 			break;
923 
924 		case MLX4_STEERING_MODE_A0:
925 			err = mlx4_SET_PORT_qpn_calc(mdev->dev,
926 						     priv->port,
927 						     priv->base_qpn,
928 						     1);
929 			if (err)
930 				en_err(priv, "Failed enabling promiscuous mode\n");
931 			break;
932 		}
933 
934 		/* Disable port multicast filter (unconditionally) */
935 		err = mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, 0,
936 					  0, MLX4_MCAST_DISABLE);
937 		if (err)
938 			en_err(priv, "Failed disabling multicast filter\n");
939 	}
940 }
941 
942 static void mlx4_en_clear_promisc_mode(struct mlx4_en_priv *priv,
943 				       struct mlx4_en_dev *mdev)
944 {
945 	int err = 0;
946 
947 	if (netif_msg_rx_status(priv))
948 		en_warn(priv, "Leaving promiscuous mode\n");
949 	priv->flags &= ~MLX4_EN_FLAG_PROMISC;
950 
951 	/* Disable promiscouos mode */
952 	switch (mdev->dev->caps.steering_mode) {
953 	case MLX4_STEERING_MODE_DEVICE_MANAGED:
954 		err = mlx4_flow_steer_promisc_remove(mdev->dev,
955 						     priv->port,
956 						     MLX4_FS_ALL_DEFAULT);
957 		if (err)
958 			en_err(priv, "Failed disabling promiscuous mode\n");
959 		priv->flags &= ~MLX4_EN_FLAG_MC_PROMISC;
960 		break;
961 
962 	case MLX4_STEERING_MODE_B0:
963 		err = mlx4_unicast_promisc_remove(mdev->dev,
964 						  priv->base_qpn,
965 						  priv->port);
966 		if (err)
967 			en_err(priv, "Failed disabling unicast promiscuous mode\n");
968 		/* Disable Multicast promisc */
969 		if (priv->flags & MLX4_EN_FLAG_MC_PROMISC) {
970 			err = mlx4_multicast_promisc_remove(mdev->dev,
971 							    priv->base_qpn,
972 							    priv->port);
973 			if (err)
974 				en_err(priv, "Failed disabling multicast promiscuous mode\n");
975 			priv->flags &= ~MLX4_EN_FLAG_MC_PROMISC;
976 		}
977 		break;
978 
979 	case MLX4_STEERING_MODE_A0:
980 		err = mlx4_SET_PORT_qpn_calc(mdev->dev,
981 					     priv->port,
982 					     priv->base_qpn, 0);
983 		if (err)
984 			en_err(priv, "Failed disabling promiscuous mode\n");
985 		break;
986 	}
987 }
988 
989 static void mlx4_en_do_multicast(struct mlx4_en_priv *priv,
990 				 struct net_device *dev,
991 				 struct mlx4_en_dev *mdev)
992 {
993 	struct mlx4_en_mc_list *mclist, *tmp;
994 	u64 mcast_addr = 0;
995 	u8 mc_list[16] = {0};
996 	int err = 0;
997 
998 	/* Enable/disable the multicast filter according to IFF_ALLMULTI */
999 	if (dev->flags & IFF_ALLMULTI) {
1000 		err = mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, 0,
1001 					  0, MLX4_MCAST_DISABLE);
1002 		if (err)
1003 			en_err(priv, "Failed disabling multicast filter\n");
1004 
1005 		/* Add the default qp number as multicast promisc */
1006 		if (!(priv->flags & MLX4_EN_FLAG_MC_PROMISC)) {
1007 			switch (mdev->dev->caps.steering_mode) {
1008 			case MLX4_STEERING_MODE_DEVICE_MANAGED:
1009 				err = mlx4_flow_steer_promisc_add(mdev->dev,
1010 								  priv->port,
1011 								  priv->base_qpn,
1012 								  MLX4_FS_MC_DEFAULT);
1013 				break;
1014 
1015 			case MLX4_STEERING_MODE_B0:
1016 				err = mlx4_multicast_promisc_add(mdev->dev,
1017 								 priv->base_qpn,
1018 								 priv->port);
1019 				break;
1020 
1021 			case MLX4_STEERING_MODE_A0:
1022 				break;
1023 			}
1024 			if (err)
1025 				en_err(priv, "Failed entering multicast promisc mode\n");
1026 			priv->flags |= MLX4_EN_FLAG_MC_PROMISC;
1027 		}
1028 	} else {
1029 		/* Disable Multicast promisc */
1030 		if (priv->flags & MLX4_EN_FLAG_MC_PROMISC) {
1031 			switch (mdev->dev->caps.steering_mode) {
1032 			case MLX4_STEERING_MODE_DEVICE_MANAGED:
1033 				err = mlx4_flow_steer_promisc_remove(mdev->dev,
1034 								     priv->port,
1035 								     MLX4_FS_MC_DEFAULT);
1036 				break;
1037 
1038 			case MLX4_STEERING_MODE_B0:
1039 				err = mlx4_multicast_promisc_remove(mdev->dev,
1040 								    priv->base_qpn,
1041 								    priv->port);
1042 				break;
1043 
1044 			case MLX4_STEERING_MODE_A0:
1045 				break;
1046 			}
1047 			if (err)
1048 				en_err(priv, "Failed disabling multicast promiscuous mode\n");
1049 			priv->flags &= ~MLX4_EN_FLAG_MC_PROMISC;
1050 		}
1051 
1052 		err = mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, 0,
1053 					  0, MLX4_MCAST_DISABLE);
1054 		if (err)
1055 			en_err(priv, "Failed disabling multicast filter\n");
1056 
1057 		/* Flush mcast filter and init it with broadcast address */
1058 		mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, ETH_BCAST,
1059 				    1, MLX4_MCAST_CONFIG);
1060 
1061 		/* Update multicast list - we cache all addresses so they won't
1062 		 * change while HW is updated holding the command semaphor */
1063 		netif_addr_lock_bh(dev);
1064 		mlx4_en_cache_mclist(dev);
1065 		netif_addr_unlock_bh(dev);
1066 		list_for_each_entry(mclist, &priv->mc_list, list) {
1067 			mcast_addr = mlx4_mac_to_u64(mclist->addr);
1068 			mlx4_SET_MCAST_FLTR(mdev->dev, priv->port,
1069 					    mcast_addr, 0, MLX4_MCAST_CONFIG);
1070 		}
1071 		err = mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, 0,
1072 					  0, MLX4_MCAST_ENABLE);
1073 		if (err)
1074 			en_err(priv, "Failed enabling multicast filter\n");
1075 
1076 		update_mclist_flags(priv, &priv->curr_list, &priv->mc_list);
1077 		list_for_each_entry_safe(mclist, tmp, &priv->curr_list, list) {
1078 			if (mclist->action == MCLIST_REM) {
1079 				/* detach this address and delete from list */
1080 				memcpy(&mc_list[10], mclist->addr, ETH_ALEN);
1081 				mc_list[5] = priv->port;
1082 				err = mlx4_multicast_detach(mdev->dev,
1083 							    priv->rss_map.indir_qp,
1084 							    mc_list,
1085 							    MLX4_PROT_ETH,
1086 							    mclist->reg_id);
1087 				if (err)
1088 					en_err(priv, "Fail to detach multicast address\n");
1089 
1090 				if (mclist->tunnel_reg_id) {
1091 					err = mlx4_flow_detach(priv->mdev->dev, mclist->tunnel_reg_id);
1092 					if (err)
1093 						en_err(priv, "Failed to detach multicast address\n");
1094 				}
1095 
1096 				/* remove from list */
1097 				list_del(&mclist->list);
1098 				kfree(mclist);
1099 			} else if (mclist->action == MCLIST_ADD) {
1100 				/* attach the address */
1101 				memcpy(&mc_list[10], mclist->addr, ETH_ALEN);
1102 				/* needed for B0 steering support */
1103 				mc_list[5] = priv->port;
1104 				err = mlx4_multicast_attach(mdev->dev,
1105 							    priv->rss_map.indir_qp,
1106 							    mc_list,
1107 							    priv->port, 0,
1108 							    MLX4_PROT_ETH,
1109 							    &mclist->reg_id);
1110 				if (err)
1111 					en_err(priv, "Fail to attach multicast address\n");
1112 
1113 				err = mlx4_en_tunnel_steer_add(priv, &mc_list[10], priv->base_qpn,
1114 							       &mclist->tunnel_reg_id);
1115 				if (err)
1116 					en_err(priv, "Failed to attach multicast address\n");
1117 			}
1118 		}
1119 	}
1120 }
1121 
1122 static void mlx4_en_do_uc_filter(struct mlx4_en_priv *priv,
1123 				 struct net_device *dev,
1124 				 struct mlx4_en_dev *mdev)
1125 {
1126 	struct netdev_hw_addr *ha;
1127 	struct mlx4_mac_entry *entry;
1128 	struct hlist_node *tmp;
1129 	bool found;
1130 	u64 mac;
1131 	int err = 0;
1132 	struct hlist_head *bucket;
1133 	unsigned int i;
1134 	int removed = 0;
1135 	u32 prev_flags;
1136 
1137 	/* Note that we do not need to protect our mac_hash traversal with rcu,
1138 	 * since all modification code is protected by mdev->state_lock
1139 	 */
1140 
1141 	/* find what to remove */
1142 	for (i = 0; i < MLX4_EN_MAC_HASH_SIZE; ++i) {
1143 		bucket = &priv->mac_hash[i];
1144 		hlist_for_each_entry_safe(entry, tmp, bucket, hlist) {
1145 			found = false;
1146 			netdev_for_each_uc_addr(ha, dev) {
1147 				if (ether_addr_equal_64bits(entry->mac,
1148 							    ha->addr)) {
1149 					found = true;
1150 					break;
1151 				}
1152 			}
1153 
1154 			/* MAC address of the port is not in uc list */
1155 			if (ether_addr_equal_64bits(entry->mac,
1156 						    priv->current_mac))
1157 				found = true;
1158 
1159 			if (!found) {
1160 				mac = mlx4_mac_to_u64(entry->mac);
1161 				mlx4_en_uc_steer_release(priv, entry->mac,
1162 							 priv->base_qpn,
1163 							 entry->reg_id);
1164 				mlx4_unregister_mac(mdev->dev, priv->port, mac);
1165 
1166 				hlist_del_rcu(&entry->hlist);
1167 				kfree_rcu(entry, rcu);
1168 				en_dbg(DRV, priv, "Removed MAC %pM on port:%d\n",
1169 				       entry->mac, priv->port);
1170 				++removed;
1171 			}
1172 		}
1173 	}
1174 
1175 	/* if we didn't remove anything, there is no use in trying to add
1176 	 * again once we are in a forced promisc mode state
1177 	 */
1178 	if ((priv->flags & MLX4_EN_FLAG_FORCE_PROMISC) && 0 == removed)
1179 		return;
1180 
1181 	prev_flags = priv->flags;
1182 	priv->flags &= ~MLX4_EN_FLAG_FORCE_PROMISC;
1183 
1184 	/* find what to add */
1185 	netdev_for_each_uc_addr(ha, dev) {
1186 		found = false;
1187 		bucket = &priv->mac_hash[ha->addr[MLX4_EN_MAC_HASH_IDX]];
1188 		hlist_for_each_entry(entry, bucket, hlist) {
1189 			if (ether_addr_equal_64bits(entry->mac, ha->addr)) {
1190 				found = true;
1191 				break;
1192 			}
1193 		}
1194 
1195 		if (!found) {
1196 			entry = kmalloc(sizeof(*entry), GFP_KERNEL);
1197 			if (!entry) {
1198 				en_err(priv, "Failed adding MAC %pM on port:%d (out of memory)\n",
1199 				       ha->addr, priv->port);
1200 				priv->flags |= MLX4_EN_FLAG_FORCE_PROMISC;
1201 				break;
1202 			}
1203 			mac = mlx4_mac_to_u64(ha->addr);
1204 			memcpy(entry->mac, ha->addr, ETH_ALEN);
1205 			err = mlx4_register_mac(mdev->dev, priv->port, mac);
1206 			if (err < 0) {
1207 				en_err(priv, "Failed registering MAC %pM on port %d: %d\n",
1208 				       ha->addr, priv->port, err);
1209 				kfree(entry);
1210 				priv->flags |= MLX4_EN_FLAG_FORCE_PROMISC;
1211 				break;
1212 			}
1213 			err = mlx4_en_uc_steer_add(priv, ha->addr,
1214 						   &priv->base_qpn,
1215 						   &entry->reg_id);
1216 			if (err) {
1217 				en_err(priv, "Failed adding MAC %pM on port %d: %d\n",
1218 				       ha->addr, priv->port, err);
1219 				mlx4_unregister_mac(mdev->dev, priv->port, mac);
1220 				kfree(entry);
1221 				priv->flags |= MLX4_EN_FLAG_FORCE_PROMISC;
1222 				break;
1223 			} else {
1224 				unsigned int mac_hash;
1225 				en_dbg(DRV, priv, "Added MAC %pM on port:%d\n",
1226 				       ha->addr, priv->port);
1227 				mac_hash = ha->addr[MLX4_EN_MAC_HASH_IDX];
1228 				bucket = &priv->mac_hash[mac_hash];
1229 				hlist_add_head_rcu(&entry->hlist, bucket);
1230 			}
1231 		}
1232 	}
1233 
1234 	if (priv->flags & MLX4_EN_FLAG_FORCE_PROMISC) {
1235 		en_warn(priv, "Forcing promiscuous mode on port:%d\n",
1236 			priv->port);
1237 	} else if (prev_flags & MLX4_EN_FLAG_FORCE_PROMISC) {
1238 		en_warn(priv, "Stop forcing promiscuous mode on port:%d\n",
1239 			priv->port);
1240 	}
1241 }
1242 
1243 static void mlx4_en_do_set_rx_mode(struct work_struct *work)
1244 {
1245 	struct mlx4_en_priv *priv = container_of(work, struct mlx4_en_priv,
1246 						 rx_mode_task);
1247 	struct mlx4_en_dev *mdev = priv->mdev;
1248 	struct net_device *dev = priv->dev;
1249 
1250 	mutex_lock(&mdev->state_lock);
1251 	if (!mdev->device_up) {
1252 		en_dbg(HW, priv, "Card is not up, ignoring rx mode change.\n");
1253 		goto out;
1254 	}
1255 	if (!priv->port_up) {
1256 		en_dbg(HW, priv, "Port is down, ignoring rx mode change.\n");
1257 		goto out;
1258 	}
1259 
1260 	if (!netif_carrier_ok(dev)) {
1261 		if (!mlx4_en_QUERY_PORT(mdev, priv->port)) {
1262 			if (priv->port_state.link_state) {
1263 				priv->last_link_state = MLX4_DEV_EVENT_PORT_UP;
1264 				netif_carrier_on(dev);
1265 				en_dbg(LINK, priv, "Link Up\n");
1266 			}
1267 		}
1268 	}
1269 
1270 	if (dev->priv_flags & IFF_UNICAST_FLT)
1271 		mlx4_en_do_uc_filter(priv, dev, mdev);
1272 
1273 	/* Promsicuous mode: disable all filters */
1274 	if ((dev->flags & IFF_PROMISC) ||
1275 	    (priv->flags & MLX4_EN_FLAG_FORCE_PROMISC)) {
1276 		mlx4_en_set_promisc_mode(priv, mdev);
1277 		goto out;
1278 	}
1279 
1280 	/* Not in promiscuous mode */
1281 	if (priv->flags & MLX4_EN_FLAG_PROMISC)
1282 		mlx4_en_clear_promisc_mode(priv, mdev);
1283 
1284 	mlx4_en_do_multicast(priv, dev, mdev);
1285 out:
1286 	mutex_unlock(&mdev->state_lock);
1287 }
1288 
1289 #ifdef CONFIG_NET_POLL_CONTROLLER
1290 static void mlx4_en_netpoll(struct net_device *dev)
1291 {
1292 	struct mlx4_en_priv *priv = netdev_priv(dev);
1293 	struct mlx4_en_cq *cq;
1294 	int i;
1295 
1296 	for (i = 0; i < priv->tx_ring_num[TX]; i++) {
1297 		cq = priv->tx_cq[TX][i];
1298 		napi_schedule(&cq->napi);
1299 	}
1300 }
1301 #endif
1302 
1303 static int mlx4_en_set_rss_steer_rules(struct mlx4_en_priv *priv)
1304 {
1305 	u64 reg_id;
1306 	int err = 0;
1307 	int *qpn = &priv->base_qpn;
1308 	struct mlx4_mac_entry *entry;
1309 
1310 	err = mlx4_en_uc_steer_add(priv, priv->dev->dev_addr, qpn, &reg_id);
1311 	if (err)
1312 		return err;
1313 
1314 	err = mlx4_en_tunnel_steer_add(priv, priv->dev->dev_addr, *qpn,
1315 				       &priv->tunnel_reg_id);
1316 	if (err)
1317 		goto tunnel_err;
1318 
1319 	entry = kmalloc(sizeof(*entry), GFP_KERNEL);
1320 	if (!entry) {
1321 		err = -ENOMEM;
1322 		goto alloc_err;
1323 	}
1324 
1325 	memcpy(entry->mac, priv->dev->dev_addr, sizeof(entry->mac));
1326 	memcpy(priv->current_mac, entry->mac, sizeof(priv->current_mac));
1327 	entry->reg_id = reg_id;
1328 	hlist_add_head_rcu(&entry->hlist,
1329 			   &priv->mac_hash[entry->mac[MLX4_EN_MAC_HASH_IDX]]);
1330 
1331 	return 0;
1332 
1333 alloc_err:
1334 	if (priv->tunnel_reg_id)
1335 		mlx4_flow_detach(priv->mdev->dev, priv->tunnel_reg_id);
1336 
1337 tunnel_err:
1338 	mlx4_en_uc_steer_release(priv, priv->dev->dev_addr, *qpn, reg_id);
1339 	return err;
1340 }
1341 
1342 static void mlx4_en_delete_rss_steer_rules(struct mlx4_en_priv *priv)
1343 {
1344 	u64 mac;
1345 	unsigned int i;
1346 	int qpn = priv->base_qpn;
1347 	struct hlist_head *bucket;
1348 	struct hlist_node *tmp;
1349 	struct mlx4_mac_entry *entry;
1350 
1351 	for (i = 0; i < MLX4_EN_MAC_HASH_SIZE; ++i) {
1352 		bucket = &priv->mac_hash[i];
1353 		hlist_for_each_entry_safe(entry, tmp, bucket, hlist) {
1354 			mac = mlx4_mac_to_u64(entry->mac);
1355 			en_dbg(DRV, priv, "Registering MAC:%pM for deleting\n",
1356 			       entry->mac);
1357 			mlx4_en_uc_steer_release(priv, entry->mac,
1358 						 qpn, entry->reg_id);
1359 
1360 			mlx4_unregister_mac(priv->mdev->dev, priv->port, mac);
1361 			hlist_del_rcu(&entry->hlist);
1362 			kfree_rcu(entry, rcu);
1363 		}
1364 	}
1365 
1366 	if (priv->tunnel_reg_id) {
1367 		mlx4_flow_detach(priv->mdev->dev, priv->tunnel_reg_id);
1368 		priv->tunnel_reg_id = 0;
1369 	}
1370 }
1371 
1372 static void mlx4_en_tx_timeout(struct net_device *dev)
1373 {
1374 	struct mlx4_en_priv *priv = netdev_priv(dev);
1375 	struct mlx4_en_dev *mdev = priv->mdev;
1376 	int i;
1377 
1378 	if (netif_msg_timer(priv))
1379 		en_warn(priv, "Tx timeout called on port:%d\n", priv->port);
1380 
1381 	for (i = 0; i < priv->tx_ring_num[TX]; i++) {
1382 		struct mlx4_en_tx_ring *tx_ring = priv->tx_ring[TX][i];
1383 
1384 		if (!netif_tx_queue_stopped(netdev_get_tx_queue(dev, i)))
1385 			continue;
1386 		en_warn(priv, "TX timeout on queue: %d, QP: 0x%x, CQ: 0x%x, Cons: 0x%x, Prod: 0x%x\n",
1387 			i, tx_ring->qpn, tx_ring->sp_cqn,
1388 			tx_ring->cons, tx_ring->prod);
1389 	}
1390 
1391 	priv->port_stats.tx_timeout++;
1392 	en_dbg(DRV, priv, "Scheduling watchdog\n");
1393 	queue_work(mdev->workqueue, &priv->watchdog_task);
1394 }
1395 
1396 
1397 static void
1398 mlx4_en_get_stats64(struct net_device *dev, struct rtnl_link_stats64 *stats)
1399 {
1400 	struct mlx4_en_priv *priv = netdev_priv(dev);
1401 
1402 	spin_lock_bh(&priv->stats_lock);
1403 	mlx4_en_fold_software_stats(dev);
1404 	netdev_stats_to_stats64(stats, &dev->stats);
1405 	spin_unlock_bh(&priv->stats_lock);
1406 }
1407 
1408 static void mlx4_en_set_default_moderation(struct mlx4_en_priv *priv)
1409 {
1410 	struct mlx4_en_cq *cq;
1411 	int i, t;
1412 
1413 	/* If we haven't received a specific coalescing setting
1414 	 * (module param), we set the moderation parameters as follows:
1415 	 * - moder_cnt is set to the number of mtu sized packets to
1416 	 *   satisfy our coalescing target.
1417 	 * - moder_time is set to a fixed value.
1418 	 */
1419 	priv->rx_frames = MLX4_EN_RX_COAL_TARGET;
1420 	priv->rx_usecs = MLX4_EN_RX_COAL_TIME;
1421 	priv->tx_frames = MLX4_EN_TX_COAL_PKTS;
1422 	priv->tx_usecs = MLX4_EN_TX_COAL_TIME;
1423 	en_dbg(INTR, priv, "Default coalescing params for mtu:%d - rx_frames:%d rx_usecs:%d\n",
1424 	       priv->dev->mtu, priv->rx_frames, priv->rx_usecs);
1425 
1426 	/* Setup cq moderation params */
1427 	for (i = 0; i < priv->rx_ring_num; i++) {
1428 		cq = priv->rx_cq[i];
1429 		cq->moder_cnt = priv->rx_frames;
1430 		cq->moder_time = priv->rx_usecs;
1431 		priv->last_moder_time[i] = MLX4_EN_AUTO_CONF;
1432 		priv->last_moder_packets[i] = 0;
1433 		priv->last_moder_bytes[i] = 0;
1434 	}
1435 
1436 	for (t = 0 ; t < MLX4_EN_NUM_TX_TYPES; t++) {
1437 		for (i = 0; i < priv->tx_ring_num[t]; i++) {
1438 			cq = priv->tx_cq[t][i];
1439 			cq->moder_cnt = priv->tx_frames;
1440 			cq->moder_time = priv->tx_usecs;
1441 		}
1442 	}
1443 
1444 	/* Reset auto-moderation params */
1445 	priv->pkt_rate_low = MLX4_EN_RX_RATE_LOW;
1446 	priv->rx_usecs_low = MLX4_EN_RX_COAL_TIME_LOW;
1447 	priv->pkt_rate_high = MLX4_EN_RX_RATE_HIGH;
1448 	priv->rx_usecs_high = MLX4_EN_RX_COAL_TIME_HIGH;
1449 	priv->sample_interval = MLX4_EN_SAMPLE_INTERVAL;
1450 	priv->adaptive_rx_coal = 1;
1451 	priv->last_moder_jiffies = 0;
1452 	priv->last_moder_tx_packets = 0;
1453 }
1454 
1455 static void mlx4_en_auto_moderation(struct mlx4_en_priv *priv)
1456 {
1457 	unsigned long period = (unsigned long) (jiffies - priv->last_moder_jiffies);
1458 	u32 pkt_rate_high, pkt_rate_low;
1459 	struct mlx4_en_cq *cq;
1460 	unsigned long packets;
1461 	unsigned long rate;
1462 	unsigned long avg_pkt_size;
1463 	unsigned long rx_packets;
1464 	unsigned long rx_bytes;
1465 	unsigned long rx_pkt_diff;
1466 	int moder_time;
1467 	int ring, err;
1468 
1469 	if (!priv->adaptive_rx_coal || period < priv->sample_interval * HZ)
1470 		return;
1471 
1472 	pkt_rate_low = READ_ONCE(priv->pkt_rate_low);
1473 	pkt_rate_high = READ_ONCE(priv->pkt_rate_high);
1474 
1475 	for (ring = 0; ring < priv->rx_ring_num; ring++) {
1476 		rx_packets = READ_ONCE(priv->rx_ring[ring]->packets);
1477 		rx_bytes = READ_ONCE(priv->rx_ring[ring]->bytes);
1478 
1479 		rx_pkt_diff = rx_packets - priv->last_moder_packets[ring];
1480 		packets = rx_pkt_diff;
1481 		rate = packets * HZ / period;
1482 		avg_pkt_size = packets ? (rx_bytes -
1483 				priv->last_moder_bytes[ring]) / packets : 0;
1484 
1485 		/* Apply auto-moderation only when packet rate
1486 		 * exceeds a rate that it matters */
1487 		if (rate > (MLX4_EN_RX_RATE_THRESH / priv->rx_ring_num) &&
1488 		    avg_pkt_size > MLX4_EN_AVG_PKT_SMALL) {
1489 			if (rate <= pkt_rate_low)
1490 				moder_time = priv->rx_usecs_low;
1491 			else if (rate >= pkt_rate_high)
1492 				moder_time = priv->rx_usecs_high;
1493 			else
1494 				moder_time = (rate - pkt_rate_low) *
1495 					(priv->rx_usecs_high - priv->rx_usecs_low) /
1496 					(pkt_rate_high - pkt_rate_low) +
1497 					priv->rx_usecs_low;
1498 		} else {
1499 			moder_time = priv->rx_usecs_low;
1500 		}
1501 
1502 		cq = priv->rx_cq[ring];
1503 		if (moder_time != priv->last_moder_time[ring] ||
1504 		    cq->moder_cnt != priv->rx_frames) {
1505 			priv->last_moder_time[ring] = moder_time;
1506 			cq->moder_time = moder_time;
1507 			cq->moder_cnt = priv->rx_frames;
1508 			err = mlx4_en_set_cq_moder(priv, cq);
1509 			if (err)
1510 				en_err(priv, "Failed modifying moderation for cq:%d\n",
1511 				       ring);
1512 		}
1513 		priv->last_moder_packets[ring] = rx_packets;
1514 		priv->last_moder_bytes[ring] = rx_bytes;
1515 	}
1516 
1517 	priv->last_moder_jiffies = jiffies;
1518 }
1519 
1520 static void mlx4_en_do_get_stats(struct work_struct *work)
1521 {
1522 	struct delayed_work *delay = to_delayed_work(work);
1523 	struct mlx4_en_priv *priv = container_of(delay, struct mlx4_en_priv,
1524 						 stats_task);
1525 	struct mlx4_en_dev *mdev = priv->mdev;
1526 	int err;
1527 
1528 	mutex_lock(&mdev->state_lock);
1529 	if (mdev->device_up) {
1530 		if (priv->port_up) {
1531 			err = mlx4_en_DUMP_ETH_STATS(mdev, priv->port, 0);
1532 			if (err)
1533 				en_dbg(HW, priv, "Could not update stats\n");
1534 
1535 			mlx4_en_auto_moderation(priv);
1536 		}
1537 
1538 		queue_delayed_work(mdev->workqueue, &priv->stats_task, STATS_DELAY);
1539 	}
1540 	if (mdev->mac_removed[MLX4_MAX_PORTS + 1 - priv->port]) {
1541 		mlx4_en_do_set_mac(priv, priv->current_mac);
1542 		mdev->mac_removed[MLX4_MAX_PORTS + 1 - priv->port] = 0;
1543 	}
1544 	mutex_unlock(&mdev->state_lock);
1545 }
1546 
1547 /* mlx4_en_service_task - Run service task for tasks that needed to be done
1548  * periodically
1549  */
1550 static void mlx4_en_service_task(struct work_struct *work)
1551 {
1552 	struct delayed_work *delay = to_delayed_work(work);
1553 	struct mlx4_en_priv *priv = container_of(delay, struct mlx4_en_priv,
1554 						 service_task);
1555 	struct mlx4_en_dev *mdev = priv->mdev;
1556 
1557 	mutex_lock(&mdev->state_lock);
1558 	if (mdev->device_up) {
1559 		if (mdev->dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_TS)
1560 			mlx4_en_ptp_overflow_check(mdev);
1561 
1562 		mlx4_en_recover_from_oom(priv);
1563 		queue_delayed_work(mdev->workqueue, &priv->service_task,
1564 				   SERVICE_TASK_DELAY);
1565 	}
1566 	mutex_unlock(&mdev->state_lock);
1567 }
1568 
1569 static void mlx4_en_linkstate(struct work_struct *work)
1570 {
1571 	struct mlx4_en_priv *priv = container_of(work, struct mlx4_en_priv,
1572 						 linkstate_task);
1573 	struct mlx4_en_dev *mdev = priv->mdev;
1574 	int linkstate = priv->link_state;
1575 
1576 	mutex_lock(&mdev->state_lock);
1577 	/* If observable port state changed set carrier state and
1578 	 * report to system log */
1579 	if (priv->last_link_state != linkstate) {
1580 		if (linkstate == MLX4_DEV_EVENT_PORT_DOWN) {
1581 			en_info(priv, "Link Down\n");
1582 			netif_carrier_off(priv->dev);
1583 		} else {
1584 			en_info(priv, "Link Up\n");
1585 			netif_carrier_on(priv->dev);
1586 		}
1587 	}
1588 	priv->last_link_state = linkstate;
1589 	mutex_unlock(&mdev->state_lock);
1590 }
1591 
1592 static int mlx4_en_init_affinity_hint(struct mlx4_en_priv *priv, int ring_idx)
1593 {
1594 	struct mlx4_en_rx_ring *ring = priv->rx_ring[ring_idx];
1595 	int numa_node = priv->mdev->dev->numa_node;
1596 
1597 	if (!zalloc_cpumask_var(&ring->affinity_mask, GFP_KERNEL))
1598 		return -ENOMEM;
1599 
1600 	cpumask_set_cpu(cpumask_local_spread(ring_idx, numa_node),
1601 			ring->affinity_mask);
1602 	return 0;
1603 }
1604 
1605 static void mlx4_en_free_affinity_hint(struct mlx4_en_priv *priv, int ring_idx)
1606 {
1607 	free_cpumask_var(priv->rx_ring[ring_idx]->affinity_mask);
1608 }
1609 
1610 static void mlx4_en_init_recycle_ring(struct mlx4_en_priv *priv,
1611 				      int tx_ring_idx)
1612 {
1613 	struct mlx4_en_tx_ring *tx_ring = priv->tx_ring[TX_XDP][tx_ring_idx];
1614 	int rr_index = tx_ring_idx;
1615 
1616 	tx_ring->free_tx_desc = mlx4_en_recycle_tx_desc;
1617 	tx_ring->recycle_ring = priv->rx_ring[rr_index];
1618 	en_dbg(DRV, priv, "Set tx_ring[%d][%d]->recycle_ring = rx_ring[%d]\n",
1619 	       TX_XDP, tx_ring_idx, rr_index);
1620 }
1621 
1622 int mlx4_en_start_port(struct net_device *dev)
1623 {
1624 	struct mlx4_en_priv *priv = netdev_priv(dev);
1625 	struct mlx4_en_dev *mdev = priv->mdev;
1626 	struct mlx4_en_cq *cq;
1627 	struct mlx4_en_tx_ring *tx_ring;
1628 	int rx_index = 0;
1629 	int err = 0;
1630 	int i, t;
1631 	int j;
1632 	u8 mc_list[16] = {0};
1633 
1634 	if (priv->port_up) {
1635 		en_dbg(DRV, priv, "start port called while port already up\n");
1636 		return 0;
1637 	}
1638 
1639 	INIT_LIST_HEAD(&priv->mc_list);
1640 	INIT_LIST_HEAD(&priv->curr_list);
1641 	INIT_LIST_HEAD(&priv->ethtool_list);
1642 	memset(&priv->ethtool_rules[0], 0,
1643 	       sizeof(struct ethtool_flow_id) * MAX_NUM_OF_FS_RULES);
1644 
1645 	/* Calculate Rx buf size */
1646 	dev->mtu = min(dev->mtu, priv->max_mtu);
1647 	mlx4_en_calc_rx_buf(dev);
1648 	en_dbg(DRV, priv, "Rx buf size:%d\n", priv->rx_skb_size);
1649 
1650 	/* Configure rx cq's and rings */
1651 	err = mlx4_en_activate_rx_rings(priv);
1652 	if (err) {
1653 		en_err(priv, "Failed to activate RX rings\n");
1654 		return err;
1655 	}
1656 	for (i = 0; i < priv->rx_ring_num; i++) {
1657 		cq = priv->rx_cq[i];
1658 
1659 		err = mlx4_en_init_affinity_hint(priv, i);
1660 		if (err) {
1661 			en_err(priv, "Failed preparing IRQ affinity hint\n");
1662 			goto cq_err;
1663 		}
1664 
1665 		err = mlx4_en_activate_cq(priv, cq, i);
1666 		if (err) {
1667 			en_err(priv, "Failed activating Rx CQ\n");
1668 			mlx4_en_free_affinity_hint(priv, i);
1669 			goto cq_err;
1670 		}
1671 
1672 		for (j = 0; j < cq->size; j++) {
1673 			struct mlx4_cqe *cqe = NULL;
1674 
1675 			cqe = mlx4_en_get_cqe(cq->buf, j, priv->cqe_size) +
1676 			      priv->cqe_factor;
1677 			cqe->owner_sr_opcode = MLX4_CQE_OWNER_MASK;
1678 		}
1679 
1680 		err = mlx4_en_set_cq_moder(priv, cq);
1681 		if (err) {
1682 			en_err(priv, "Failed setting cq moderation parameters\n");
1683 			mlx4_en_deactivate_cq(priv, cq);
1684 			mlx4_en_free_affinity_hint(priv, i);
1685 			goto cq_err;
1686 		}
1687 		mlx4_en_arm_cq(priv, cq);
1688 		priv->rx_ring[i]->cqn = cq->mcq.cqn;
1689 		++rx_index;
1690 	}
1691 
1692 	/* Set qp number */
1693 	en_dbg(DRV, priv, "Getting qp number for port %d\n", priv->port);
1694 	err = mlx4_en_get_qp(priv);
1695 	if (err) {
1696 		en_err(priv, "Failed getting eth qp\n");
1697 		goto cq_err;
1698 	}
1699 	mdev->mac_removed[priv->port] = 0;
1700 
1701 	priv->counter_index =
1702 			mlx4_get_default_counter_index(mdev->dev, priv->port);
1703 
1704 	err = mlx4_en_config_rss_steer(priv);
1705 	if (err) {
1706 		en_err(priv, "Failed configuring rss steering\n");
1707 		goto mac_err;
1708 	}
1709 
1710 	err = mlx4_en_create_drop_qp(priv);
1711 	if (err)
1712 		goto rss_err;
1713 
1714 	/* Configure tx cq's and rings */
1715 	for (t = 0 ; t < MLX4_EN_NUM_TX_TYPES; t++) {
1716 		u8 num_tx_rings_p_up = t == TX ?
1717 			priv->num_tx_rings_p_up : priv->tx_ring_num[t];
1718 
1719 		for (i = 0; i < priv->tx_ring_num[t]; i++) {
1720 			/* Configure cq */
1721 			cq = priv->tx_cq[t][i];
1722 			err = mlx4_en_activate_cq(priv, cq, i);
1723 			if (err) {
1724 				en_err(priv, "Failed allocating Tx CQ\n");
1725 				goto tx_err;
1726 			}
1727 			err = mlx4_en_set_cq_moder(priv, cq);
1728 			if (err) {
1729 				en_err(priv, "Failed setting cq moderation parameters\n");
1730 				mlx4_en_deactivate_cq(priv, cq);
1731 				goto tx_err;
1732 			}
1733 			en_dbg(DRV, priv,
1734 			       "Resetting index of collapsed CQ:%d to -1\n", i);
1735 			cq->buf->wqe_index = cpu_to_be16(0xffff);
1736 
1737 			/* Configure ring */
1738 			tx_ring = priv->tx_ring[t][i];
1739 			err = mlx4_en_activate_tx_ring(priv, tx_ring,
1740 						       cq->mcq.cqn,
1741 						       i / num_tx_rings_p_up);
1742 			if (err) {
1743 				en_err(priv, "Failed allocating Tx ring\n");
1744 				mlx4_en_deactivate_cq(priv, cq);
1745 				goto tx_err;
1746 			}
1747 			if (t != TX_XDP) {
1748 				tx_ring->tx_queue = netdev_get_tx_queue(dev, i);
1749 				tx_ring->recycle_ring = NULL;
1750 
1751 				/* Arm CQ for TX completions */
1752 				mlx4_en_arm_cq(priv, cq);
1753 
1754 			} else {
1755 				mlx4_en_init_tx_xdp_ring_descs(priv, tx_ring);
1756 				mlx4_en_init_recycle_ring(priv, i);
1757 				/* XDP TX CQ should never be armed */
1758 			}
1759 
1760 			/* Set initial ownership of all Tx TXBBs to SW (1) */
1761 			for (j = 0; j < tx_ring->buf_size; j += STAMP_STRIDE)
1762 				*((u32 *)(tx_ring->buf + j)) = 0xffffffff;
1763 		}
1764 	}
1765 
1766 	/* Configure port */
1767 	err = mlx4_SET_PORT_general(mdev->dev, priv->port,
1768 				    priv->rx_skb_size + ETH_FCS_LEN,
1769 				    priv->prof->tx_pause,
1770 				    priv->prof->tx_ppp,
1771 				    priv->prof->rx_pause,
1772 				    priv->prof->rx_ppp);
1773 	if (err) {
1774 		en_err(priv, "Failed setting port general configurations for port %d, with error %d\n",
1775 		       priv->port, err);
1776 		goto tx_err;
1777 	}
1778 
1779 	err = mlx4_SET_PORT_user_mtu(mdev->dev, priv->port, dev->mtu);
1780 	if (err) {
1781 		en_err(priv, "Failed to pass user MTU(%d) to Firmware for port %d, with error %d\n",
1782 		       dev->mtu, priv->port, err);
1783 		goto tx_err;
1784 	}
1785 
1786 	/* Set default qp number */
1787 	err = mlx4_SET_PORT_qpn_calc(mdev->dev, priv->port, priv->base_qpn, 0);
1788 	if (err) {
1789 		en_err(priv, "Failed setting default qp numbers\n");
1790 		goto tx_err;
1791 	}
1792 
1793 	if (mdev->dev->caps.tunnel_offload_mode == MLX4_TUNNEL_OFFLOAD_MODE_VXLAN) {
1794 		err = mlx4_SET_PORT_VXLAN(mdev->dev, priv->port, VXLAN_STEER_BY_OUTER_MAC, 1);
1795 		if (err) {
1796 			en_err(priv, "Failed setting port L2 tunnel configuration, err %d\n",
1797 			       err);
1798 			goto tx_err;
1799 		}
1800 	}
1801 
1802 	/* Init port */
1803 	en_dbg(HW, priv, "Initializing port\n");
1804 	err = mlx4_INIT_PORT(mdev->dev, priv->port);
1805 	if (err) {
1806 		en_err(priv, "Failed Initializing port\n");
1807 		goto tx_err;
1808 	}
1809 
1810 	/* Set Unicast and VXLAN steering rules */
1811 	if (mdev->dev->caps.steering_mode != MLX4_STEERING_MODE_A0 &&
1812 	    mlx4_en_set_rss_steer_rules(priv))
1813 		mlx4_warn(mdev, "Failed setting steering rules\n");
1814 
1815 	/* Attach rx QP to bradcast address */
1816 	eth_broadcast_addr(&mc_list[10]);
1817 	mc_list[5] = priv->port; /* needed for B0 steering support */
1818 	if (mlx4_multicast_attach(mdev->dev, priv->rss_map.indir_qp, mc_list,
1819 				  priv->port, 0, MLX4_PROT_ETH,
1820 				  &priv->broadcast_id))
1821 		mlx4_warn(mdev, "Failed Attaching Broadcast\n");
1822 
1823 	/* Must redo promiscuous mode setup. */
1824 	priv->flags &= ~(MLX4_EN_FLAG_PROMISC | MLX4_EN_FLAG_MC_PROMISC);
1825 
1826 	/* Schedule multicast task to populate multicast list */
1827 	queue_work(mdev->workqueue, &priv->rx_mode_task);
1828 
1829 	if (priv->mdev->dev->caps.tunnel_offload_mode == MLX4_TUNNEL_OFFLOAD_MODE_VXLAN)
1830 		udp_tunnel_get_rx_info(dev);
1831 
1832 	priv->port_up = true;
1833 
1834 	/* Process all completions if exist to prevent
1835 	 * the queues freezing if they are full
1836 	 */
1837 	for (i = 0; i < priv->rx_ring_num; i++) {
1838 		local_bh_disable();
1839 		napi_schedule(&priv->rx_cq[i]->napi);
1840 		local_bh_enable();
1841 	}
1842 
1843 	netif_tx_start_all_queues(dev);
1844 	netif_device_attach(dev);
1845 
1846 	return 0;
1847 
1848 tx_err:
1849 	if (t == MLX4_EN_NUM_TX_TYPES) {
1850 		t--;
1851 		i = priv->tx_ring_num[t];
1852 	}
1853 	while (t >= 0) {
1854 		while (i--) {
1855 			mlx4_en_deactivate_tx_ring(priv, priv->tx_ring[t][i]);
1856 			mlx4_en_deactivate_cq(priv, priv->tx_cq[t][i]);
1857 		}
1858 		if (!t--)
1859 			break;
1860 		i = priv->tx_ring_num[t];
1861 	}
1862 	mlx4_en_destroy_drop_qp(priv);
1863 rss_err:
1864 	mlx4_en_release_rss_steer(priv);
1865 mac_err:
1866 	mlx4_en_put_qp(priv);
1867 cq_err:
1868 	while (rx_index--) {
1869 		mlx4_en_deactivate_cq(priv, priv->rx_cq[rx_index]);
1870 		mlx4_en_free_affinity_hint(priv, rx_index);
1871 	}
1872 	for (i = 0; i < priv->rx_ring_num; i++)
1873 		mlx4_en_deactivate_rx_ring(priv, priv->rx_ring[i]);
1874 
1875 	return err; /* need to close devices */
1876 }
1877 
1878 
1879 void mlx4_en_stop_port(struct net_device *dev, int detach)
1880 {
1881 	struct mlx4_en_priv *priv = netdev_priv(dev);
1882 	struct mlx4_en_dev *mdev = priv->mdev;
1883 	struct mlx4_en_mc_list *mclist, *tmp;
1884 	struct ethtool_flow_id *flow, *tmp_flow;
1885 	int i, t;
1886 	u8 mc_list[16] = {0};
1887 
1888 	if (!priv->port_up) {
1889 		en_dbg(DRV, priv, "stop port called while port already down\n");
1890 		return;
1891 	}
1892 
1893 	/* close port*/
1894 	mlx4_CLOSE_PORT(mdev->dev, priv->port);
1895 
1896 	/* Synchronize with tx routine */
1897 	netif_tx_lock_bh(dev);
1898 	if (detach)
1899 		netif_device_detach(dev);
1900 	netif_tx_stop_all_queues(dev);
1901 	netif_tx_unlock_bh(dev);
1902 
1903 	netif_tx_disable(dev);
1904 
1905 	spin_lock_bh(&priv->stats_lock);
1906 	mlx4_en_fold_software_stats(dev);
1907 	/* Set port as not active */
1908 	priv->port_up = false;
1909 	spin_unlock_bh(&priv->stats_lock);
1910 
1911 	priv->counter_index = MLX4_SINK_COUNTER_INDEX(mdev->dev);
1912 
1913 	/* Promsicuous mode */
1914 	if (mdev->dev->caps.steering_mode ==
1915 	    MLX4_STEERING_MODE_DEVICE_MANAGED) {
1916 		priv->flags &= ~(MLX4_EN_FLAG_PROMISC |
1917 				 MLX4_EN_FLAG_MC_PROMISC);
1918 		mlx4_flow_steer_promisc_remove(mdev->dev,
1919 					       priv->port,
1920 					       MLX4_FS_ALL_DEFAULT);
1921 		mlx4_flow_steer_promisc_remove(mdev->dev,
1922 					       priv->port,
1923 					       MLX4_FS_MC_DEFAULT);
1924 	} else if (priv->flags & MLX4_EN_FLAG_PROMISC) {
1925 		priv->flags &= ~MLX4_EN_FLAG_PROMISC;
1926 
1927 		/* Disable promiscouos mode */
1928 		mlx4_unicast_promisc_remove(mdev->dev, priv->base_qpn,
1929 					    priv->port);
1930 
1931 		/* Disable Multicast promisc */
1932 		if (priv->flags & MLX4_EN_FLAG_MC_PROMISC) {
1933 			mlx4_multicast_promisc_remove(mdev->dev, priv->base_qpn,
1934 						      priv->port);
1935 			priv->flags &= ~MLX4_EN_FLAG_MC_PROMISC;
1936 		}
1937 	}
1938 
1939 	/* Detach All multicasts */
1940 	eth_broadcast_addr(&mc_list[10]);
1941 	mc_list[5] = priv->port; /* needed for B0 steering support */
1942 	mlx4_multicast_detach(mdev->dev, priv->rss_map.indir_qp, mc_list,
1943 			      MLX4_PROT_ETH, priv->broadcast_id);
1944 	list_for_each_entry(mclist, &priv->curr_list, list) {
1945 		memcpy(&mc_list[10], mclist->addr, ETH_ALEN);
1946 		mc_list[5] = priv->port;
1947 		mlx4_multicast_detach(mdev->dev, priv->rss_map.indir_qp,
1948 				      mc_list, MLX4_PROT_ETH, mclist->reg_id);
1949 		if (mclist->tunnel_reg_id)
1950 			mlx4_flow_detach(mdev->dev, mclist->tunnel_reg_id);
1951 	}
1952 	mlx4_en_clear_list(dev);
1953 	list_for_each_entry_safe(mclist, tmp, &priv->curr_list, list) {
1954 		list_del(&mclist->list);
1955 		kfree(mclist);
1956 	}
1957 
1958 	/* Flush multicast filter */
1959 	mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, 0, 1, MLX4_MCAST_CONFIG);
1960 
1961 	/* Remove flow steering rules for the port*/
1962 	if (mdev->dev->caps.steering_mode ==
1963 	    MLX4_STEERING_MODE_DEVICE_MANAGED) {
1964 		ASSERT_RTNL();
1965 		list_for_each_entry_safe(flow, tmp_flow,
1966 					 &priv->ethtool_list, list) {
1967 			mlx4_flow_detach(mdev->dev, flow->id);
1968 			list_del(&flow->list);
1969 		}
1970 	}
1971 
1972 	mlx4_en_destroy_drop_qp(priv);
1973 
1974 	/* Free TX Rings */
1975 	for (t = 0; t < MLX4_EN_NUM_TX_TYPES; t++) {
1976 		for (i = 0; i < priv->tx_ring_num[t]; i++) {
1977 			mlx4_en_deactivate_tx_ring(priv, priv->tx_ring[t][i]);
1978 			mlx4_en_deactivate_cq(priv, priv->tx_cq[t][i]);
1979 		}
1980 	}
1981 	msleep(10);
1982 
1983 	for (t = 0; t < MLX4_EN_NUM_TX_TYPES; t++)
1984 		for (i = 0; i < priv->tx_ring_num[t]; i++)
1985 			mlx4_en_free_tx_buf(dev, priv->tx_ring[t][i]);
1986 
1987 	if (mdev->dev->caps.steering_mode != MLX4_STEERING_MODE_A0)
1988 		mlx4_en_delete_rss_steer_rules(priv);
1989 
1990 	/* Free RSS qps */
1991 	mlx4_en_release_rss_steer(priv);
1992 
1993 	/* Unregister Mac address for the port */
1994 	mlx4_en_put_qp(priv);
1995 	if (!(mdev->dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_REASSIGN_MAC_EN))
1996 		mdev->mac_removed[priv->port] = 1;
1997 
1998 	/* Free RX Rings */
1999 	for (i = 0; i < priv->rx_ring_num; i++) {
2000 		struct mlx4_en_cq *cq = priv->rx_cq[i];
2001 
2002 		napi_synchronize(&cq->napi);
2003 		mlx4_en_deactivate_rx_ring(priv, priv->rx_ring[i]);
2004 		mlx4_en_deactivate_cq(priv, cq);
2005 
2006 		mlx4_en_free_affinity_hint(priv, i);
2007 	}
2008 }
2009 
2010 static void mlx4_en_restart(struct work_struct *work)
2011 {
2012 	struct mlx4_en_priv *priv = container_of(work, struct mlx4_en_priv,
2013 						 watchdog_task);
2014 	struct mlx4_en_dev *mdev = priv->mdev;
2015 	struct net_device *dev = priv->dev;
2016 
2017 	en_dbg(DRV, priv, "Watchdog task called for port %d\n", priv->port);
2018 
2019 	rtnl_lock();
2020 	mutex_lock(&mdev->state_lock);
2021 	if (priv->port_up) {
2022 		mlx4_en_stop_port(dev, 1);
2023 		if (mlx4_en_start_port(dev))
2024 			en_err(priv, "Failed restarting port %d\n", priv->port);
2025 	}
2026 	mutex_unlock(&mdev->state_lock);
2027 	rtnl_unlock();
2028 }
2029 
2030 static void mlx4_en_clear_stats(struct net_device *dev)
2031 {
2032 	struct mlx4_en_priv *priv = netdev_priv(dev);
2033 	struct mlx4_en_dev *mdev = priv->mdev;
2034 	struct mlx4_en_tx_ring **tx_ring;
2035 	int i;
2036 
2037 	if (!mlx4_is_slave(mdev->dev))
2038 		if (mlx4_en_DUMP_ETH_STATS(mdev, priv->port, 1))
2039 			en_dbg(HW, priv, "Failed dumping statistics\n");
2040 
2041 	memset(&priv->pstats, 0, sizeof(priv->pstats));
2042 	memset(&priv->pkstats, 0, sizeof(priv->pkstats));
2043 	memset(&priv->port_stats, 0, sizeof(priv->port_stats));
2044 	memset(&priv->rx_flowstats, 0, sizeof(priv->rx_flowstats));
2045 	memset(&priv->tx_flowstats, 0, sizeof(priv->tx_flowstats));
2046 	memset(&priv->rx_priority_flowstats, 0,
2047 	       sizeof(priv->rx_priority_flowstats));
2048 	memset(&priv->tx_priority_flowstats, 0,
2049 	       sizeof(priv->tx_priority_flowstats));
2050 	memset(&priv->pf_stats, 0, sizeof(priv->pf_stats));
2051 
2052 	tx_ring = priv->tx_ring[TX];
2053 	for (i = 0; i < priv->tx_ring_num[TX]; i++) {
2054 		tx_ring[i]->bytes = 0;
2055 		tx_ring[i]->packets = 0;
2056 		tx_ring[i]->tx_csum = 0;
2057 		tx_ring[i]->tx_dropped = 0;
2058 		tx_ring[i]->queue_stopped = 0;
2059 		tx_ring[i]->wake_queue = 0;
2060 		tx_ring[i]->tso_packets = 0;
2061 		tx_ring[i]->xmit_more = 0;
2062 	}
2063 	for (i = 0; i < priv->rx_ring_num; i++) {
2064 		priv->rx_ring[i]->bytes = 0;
2065 		priv->rx_ring[i]->packets = 0;
2066 		priv->rx_ring[i]->csum_ok = 0;
2067 		priv->rx_ring[i]->csum_none = 0;
2068 		priv->rx_ring[i]->csum_complete = 0;
2069 	}
2070 }
2071 
2072 static int mlx4_en_open(struct net_device *dev)
2073 {
2074 	struct mlx4_en_priv *priv = netdev_priv(dev);
2075 	struct mlx4_en_dev *mdev = priv->mdev;
2076 	int err = 0;
2077 
2078 	mutex_lock(&mdev->state_lock);
2079 
2080 	if (!mdev->device_up) {
2081 		en_err(priv, "Cannot open - device down/disabled\n");
2082 		err = -EBUSY;
2083 		goto out;
2084 	}
2085 
2086 	/* Reset HW statistics and SW counters */
2087 	mlx4_en_clear_stats(dev);
2088 
2089 	err = mlx4_en_start_port(dev);
2090 	if (err)
2091 		en_err(priv, "Failed starting port:%d\n", priv->port);
2092 
2093 out:
2094 	mutex_unlock(&mdev->state_lock);
2095 	return err;
2096 }
2097 
2098 
2099 static int mlx4_en_close(struct net_device *dev)
2100 {
2101 	struct mlx4_en_priv *priv = netdev_priv(dev);
2102 	struct mlx4_en_dev *mdev = priv->mdev;
2103 
2104 	en_dbg(IFDOWN, priv, "Close port called\n");
2105 
2106 	mutex_lock(&mdev->state_lock);
2107 
2108 	mlx4_en_stop_port(dev, 0);
2109 	netif_carrier_off(dev);
2110 
2111 	mutex_unlock(&mdev->state_lock);
2112 	return 0;
2113 }
2114 
2115 static void mlx4_en_free_resources(struct mlx4_en_priv *priv)
2116 {
2117 	int i, t;
2118 
2119 #ifdef CONFIG_RFS_ACCEL
2120 	priv->dev->rx_cpu_rmap = NULL;
2121 #endif
2122 
2123 	for (t = 0; t < MLX4_EN_NUM_TX_TYPES; t++) {
2124 		for (i = 0; i < priv->tx_ring_num[t]; i++) {
2125 			if (priv->tx_ring[t] && priv->tx_ring[t][i])
2126 				mlx4_en_destroy_tx_ring(priv,
2127 							&priv->tx_ring[t][i]);
2128 			if (priv->tx_cq[t] && priv->tx_cq[t][i])
2129 				mlx4_en_destroy_cq(priv, &priv->tx_cq[t][i]);
2130 		}
2131 		kfree(priv->tx_ring[t]);
2132 		kfree(priv->tx_cq[t]);
2133 	}
2134 
2135 	for (i = 0; i < priv->rx_ring_num; i++) {
2136 		if (priv->rx_ring[i])
2137 			mlx4_en_destroy_rx_ring(priv, &priv->rx_ring[i],
2138 				priv->prof->rx_ring_size, priv->stride);
2139 		if (priv->rx_cq[i])
2140 			mlx4_en_destroy_cq(priv, &priv->rx_cq[i]);
2141 	}
2142 
2143 }
2144 
2145 static int mlx4_en_alloc_resources(struct mlx4_en_priv *priv)
2146 {
2147 	struct mlx4_en_port_profile *prof = priv->prof;
2148 	int i, t;
2149 	int node;
2150 
2151 	/* Create tx Rings */
2152 	for (t = 0; t < MLX4_EN_NUM_TX_TYPES; t++) {
2153 		for (i = 0; i < priv->tx_ring_num[t]; i++) {
2154 			node = cpu_to_node(i % num_online_cpus());
2155 			if (mlx4_en_create_cq(priv, &priv->tx_cq[t][i],
2156 					      prof->tx_ring_size, i, t, node))
2157 				goto err;
2158 
2159 			if (mlx4_en_create_tx_ring(priv, &priv->tx_ring[t][i],
2160 						   prof->tx_ring_size,
2161 						   TXBB_SIZE, node, i))
2162 				goto err;
2163 		}
2164 	}
2165 
2166 	/* Create rx Rings */
2167 	for (i = 0; i < priv->rx_ring_num; i++) {
2168 		node = cpu_to_node(i % num_online_cpus());
2169 		if (mlx4_en_create_cq(priv, &priv->rx_cq[i],
2170 				      prof->rx_ring_size, i, RX, node))
2171 			goto err;
2172 
2173 		if (mlx4_en_create_rx_ring(priv, &priv->rx_ring[i],
2174 					   prof->rx_ring_size, priv->stride,
2175 					   node, i))
2176 			goto err;
2177 
2178 	}
2179 
2180 #ifdef CONFIG_RFS_ACCEL
2181 	priv->dev->rx_cpu_rmap = mlx4_get_cpu_rmap(priv->mdev->dev, priv->port);
2182 #endif
2183 
2184 	return 0;
2185 
2186 err:
2187 	en_err(priv, "Failed to allocate NIC resources\n");
2188 	for (i = 0; i < priv->rx_ring_num; i++) {
2189 		if (priv->rx_ring[i])
2190 			mlx4_en_destroy_rx_ring(priv, &priv->rx_ring[i],
2191 						prof->rx_ring_size,
2192 						priv->stride);
2193 		if (priv->rx_cq[i])
2194 			mlx4_en_destroy_cq(priv, &priv->rx_cq[i]);
2195 	}
2196 	for (t = 0; t < MLX4_EN_NUM_TX_TYPES; t++) {
2197 		for (i = 0; i < priv->tx_ring_num[t]; i++) {
2198 			if (priv->tx_ring[t][i])
2199 				mlx4_en_destroy_tx_ring(priv,
2200 							&priv->tx_ring[t][i]);
2201 			if (priv->tx_cq[t][i])
2202 				mlx4_en_destroy_cq(priv, &priv->tx_cq[t][i]);
2203 		}
2204 	}
2205 	return -ENOMEM;
2206 }
2207 
2208 
2209 static int mlx4_en_copy_priv(struct mlx4_en_priv *dst,
2210 			     struct mlx4_en_priv *src,
2211 			     struct mlx4_en_port_profile *prof)
2212 {
2213 	int t;
2214 
2215 	memcpy(&dst->hwtstamp_config, &prof->hwtstamp_config,
2216 	       sizeof(dst->hwtstamp_config));
2217 	dst->num_tx_rings_p_up = prof->num_tx_rings_p_up;
2218 	dst->rx_ring_num = prof->rx_ring_num;
2219 	dst->flags = prof->flags;
2220 	dst->mdev = src->mdev;
2221 	dst->port = src->port;
2222 	dst->dev = src->dev;
2223 	dst->prof = prof;
2224 	dst->stride = roundup_pow_of_two(sizeof(struct mlx4_en_rx_desc) +
2225 					 DS_SIZE * MLX4_EN_MAX_RX_FRAGS);
2226 
2227 	for (t = 0; t < MLX4_EN_NUM_TX_TYPES; t++) {
2228 		dst->tx_ring_num[t] = prof->tx_ring_num[t];
2229 		if (!dst->tx_ring_num[t])
2230 			continue;
2231 
2232 		dst->tx_ring[t] = kcalloc(MAX_TX_RINGS,
2233 					  sizeof(struct mlx4_en_tx_ring *),
2234 					  GFP_KERNEL);
2235 		if (!dst->tx_ring[t])
2236 			goto err_free_tx;
2237 
2238 		dst->tx_cq[t] = kcalloc(MAX_TX_RINGS,
2239 					sizeof(struct mlx4_en_cq *),
2240 					GFP_KERNEL);
2241 		if (!dst->tx_cq[t]) {
2242 			kfree(dst->tx_ring[t]);
2243 			goto err_free_tx;
2244 		}
2245 	}
2246 
2247 	return 0;
2248 
2249 err_free_tx:
2250 	while (t--) {
2251 		kfree(dst->tx_ring[t]);
2252 		kfree(dst->tx_cq[t]);
2253 	}
2254 	return -ENOMEM;
2255 }
2256 
2257 static void mlx4_en_update_priv(struct mlx4_en_priv *dst,
2258 				struct mlx4_en_priv *src)
2259 {
2260 	int t;
2261 	memcpy(dst->rx_ring, src->rx_ring,
2262 	       sizeof(struct mlx4_en_rx_ring *) * src->rx_ring_num);
2263 	memcpy(dst->rx_cq, src->rx_cq,
2264 	       sizeof(struct mlx4_en_cq *) * src->rx_ring_num);
2265 	memcpy(&dst->hwtstamp_config, &src->hwtstamp_config,
2266 	       sizeof(dst->hwtstamp_config));
2267 	for (t = 0; t < MLX4_EN_NUM_TX_TYPES; t++) {
2268 		dst->tx_ring_num[t] = src->tx_ring_num[t];
2269 		dst->tx_ring[t] = src->tx_ring[t];
2270 		dst->tx_cq[t] = src->tx_cq[t];
2271 	}
2272 	dst->num_tx_rings_p_up = src->num_tx_rings_p_up;
2273 	dst->rx_ring_num = src->rx_ring_num;
2274 	memcpy(dst->prof, src->prof, sizeof(struct mlx4_en_port_profile));
2275 }
2276 
2277 int mlx4_en_try_alloc_resources(struct mlx4_en_priv *priv,
2278 				struct mlx4_en_priv *tmp,
2279 				struct mlx4_en_port_profile *prof,
2280 				bool carry_xdp_prog)
2281 {
2282 	struct bpf_prog *xdp_prog;
2283 	int i, t;
2284 
2285 	mlx4_en_copy_priv(tmp, priv, prof);
2286 
2287 	if (mlx4_en_alloc_resources(tmp)) {
2288 		en_warn(priv,
2289 			"%s: Resource allocation failed, using previous configuration\n",
2290 			__func__);
2291 		for (t = 0; t < MLX4_EN_NUM_TX_TYPES; t++) {
2292 			kfree(tmp->tx_ring[t]);
2293 			kfree(tmp->tx_cq[t]);
2294 		}
2295 		return -ENOMEM;
2296 	}
2297 
2298 	/* All rx_rings has the same xdp_prog.  Pick the first one. */
2299 	xdp_prog = rcu_dereference_protected(
2300 		priv->rx_ring[0]->xdp_prog,
2301 		lockdep_is_held(&priv->mdev->state_lock));
2302 
2303 	if (xdp_prog && carry_xdp_prog) {
2304 		xdp_prog = bpf_prog_add(xdp_prog, tmp->rx_ring_num);
2305 		if (IS_ERR(xdp_prog)) {
2306 			mlx4_en_free_resources(tmp);
2307 			return PTR_ERR(xdp_prog);
2308 		}
2309 		for (i = 0; i < tmp->rx_ring_num; i++)
2310 			rcu_assign_pointer(tmp->rx_ring[i]->xdp_prog,
2311 					   xdp_prog);
2312 	}
2313 
2314 	return 0;
2315 }
2316 
2317 void mlx4_en_safe_replace_resources(struct mlx4_en_priv *priv,
2318 				    struct mlx4_en_priv *tmp)
2319 {
2320 	mlx4_en_free_resources(priv);
2321 	mlx4_en_update_priv(priv, tmp);
2322 }
2323 
2324 void mlx4_en_destroy_netdev(struct net_device *dev)
2325 {
2326 	struct mlx4_en_priv *priv = netdev_priv(dev);
2327 	struct mlx4_en_dev *mdev = priv->mdev;
2328 
2329 	en_dbg(DRV, priv, "Destroying netdev on port:%d\n", priv->port);
2330 
2331 	/* Unregister device - this will close the port if it was up */
2332 	if (priv->registered) {
2333 		devlink_port_type_clear(mlx4_get_devlink_port(mdev->dev,
2334 							      priv->port));
2335 		unregister_netdev(dev);
2336 	}
2337 
2338 	if (priv->allocated)
2339 		mlx4_free_hwq_res(mdev->dev, &priv->res, MLX4_EN_PAGE_SIZE);
2340 
2341 	cancel_delayed_work(&priv->stats_task);
2342 	cancel_delayed_work(&priv->service_task);
2343 	/* flush any pending task for this netdev */
2344 	flush_workqueue(mdev->workqueue);
2345 
2346 	if (mdev->dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_TS)
2347 		mlx4_en_remove_timestamp(mdev);
2348 
2349 	/* Detach the netdev so tasks would not attempt to access it */
2350 	mutex_lock(&mdev->state_lock);
2351 	mdev->pndev[priv->port] = NULL;
2352 	mdev->upper[priv->port] = NULL;
2353 
2354 #ifdef CONFIG_RFS_ACCEL
2355 	mlx4_en_cleanup_filters(priv);
2356 #endif
2357 
2358 	mlx4_en_free_resources(priv);
2359 	mutex_unlock(&mdev->state_lock);
2360 
2361 	free_netdev(dev);
2362 }
2363 
2364 static bool mlx4_en_check_xdp_mtu(struct net_device *dev, int mtu)
2365 {
2366 	struct mlx4_en_priv *priv = netdev_priv(dev);
2367 
2368 	if (mtu > MLX4_EN_MAX_XDP_MTU) {
2369 		en_err(priv, "mtu:%d > max:%d when XDP prog is attached\n",
2370 		       mtu, MLX4_EN_MAX_XDP_MTU);
2371 		return false;
2372 	}
2373 
2374 	return true;
2375 }
2376 
2377 static int mlx4_en_change_mtu(struct net_device *dev, int new_mtu)
2378 {
2379 	struct mlx4_en_priv *priv = netdev_priv(dev);
2380 	struct mlx4_en_dev *mdev = priv->mdev;
2381 	int err = 0;
2382 
2383 	en_dbg(DRV, priv, "Change MTU called - current:%d new:%d\n",
2384 		 dev->mtu, new_mtu);
2385 
2386 	if (priv->tx_ring_num[TX_XDP] &&
2387 	    !mlx4_en_check_xdp_mtu(dev, new_mtu))
2388 		return -EOPNOTSUPP;
2389 
2390 	dev->mtu = new_mtu;
2391 
2392 	if (netif_running(dev)) {
2393 		mutex_lock(&mdev->state_lock);
2394 		if (!mdev->device_up) {
2395 			/* NIC is probably restarting - let watchdog task reset
2396 			 * the port */
2397 			en_dbg(DRV, priv, "Change MTU called with card down!?\n");
2398 		} else {
2399 			mlx4_en_stop_port(dev, 1);
2400 			err = mlx4_en_start_port(dev);
2401 			if (err) {
2402 				en_err(priv, "Failed restarting port:%d\n",
2403 					 priv->port);
2404 				queue_work(mdev->workqueue, &priv->watchdog_task);
2405 			}
2406 		}
2407 		mutex_unlock(&mdev->state_lock);
2408 	}
2409 	return 0;
2410 }
2411 
2412 static int mlx4_en_hwtstamp_set(struct net_device *dev, struct ifreq *ifr)
2413 {
2414 	struct mlx4_en_priv *priv = netdev_priv(dev);
2415 	struct mlx4_en_dev *mdev = priv->mdev;
2416 	struct hwtstamp_config config;
2417 
2418 	if (copy_from_user(&config, ifr->ifr_data, sizeof(config)))
2419 		return -EFAULT;
2420 
2421 	/* reserved for future extensions */
2422 	if (config.flags)
2423 		return -EINVAL;
2424 
2425 	/* device doesn't support time stamping */
2426 	if (!(mdev->dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_TS))
2427 		return -EINVAL;
2428 
2429 	/* TX HW timestamp */
2430 	switch (config.tx_type) {
2431 	case HWTSTAMP_TX_OFF:
2432 	case HWTSTAMP_TX_ON:
2433 		break;
2434 	default:
2435 		return -ERANGE;
2436 	}
2437 
2438 	/* RX HW timestamp */
2439 	switch (config.rx_filter) {
2440 	case HWTSTAMP_FILTER_NONE:
2441 		break;
2442 	case HWTSTAMP_FILTER_ALL:
2443 	case HWTSTAMP_FILTER_SOME:
2444 	case HWTSTAMP_FILTER_PTP_V1_L4_EVENT:
2445 	case HWTSTAMP_FILTER_PTP_V1_L4_SYNC:
2446 	case HWTSTAMP_FILTER_PTP_V1_L4_DELAY_REQ:
2447 	case HWTSTAMP_FILTER_PTP_V2_L4_EVENT:
2448 	case HWTSTAMP_FILTER_PTP_V2_L4_SYNC:
2449 	case HWTSTAMP_FILTER_PTP_V2_L4_DELAY_REQ:
2450 	case HWTSTAMP_FILTER_PTP_V2_L2_EVENT:
2451 	case HWTSTAMP_FILTER_PTP_V2_L2_SYNC:
2452 	case HWTSTAMP_FILTER_PTP_V2_L2_DELAY_REQ:
2453 	case HWTSTAMP_FILTER_PTP_V2_EVENT:
2454 	case HWTSTAMP_FILTER_PTP_V2_SYNC:
2455 	case HWTSTAMP_FILTER_PTP_V2_DELAY_REQ:
2456 	case HWTSTAMP_FILTER_NTP_ALL:
2457 		config.rx_filter = HWTSTAMP_FILTER_ALL;
2458 		break;
2459 	default:
2460 		return -ERANGE;
2461 	}
2462 
2463 	if (mlx4_en_reset_config(dev, config, dev->features)) {
2464 		config.tx_type = HWTSTAMP_TX_OFF;
2465 		config.rx_filter = HWTSTAMP_FILTER_NONE;
2466 	}
2467 
2468 	return copy_to_user(ifr->ifr_data, &config,
2469 			    sizeof(config)) ? -EFAULT : 0;
2470 }
2471 
2472 static int mlx4_en_hwtstamp_get(struct net_device *dev, struct ifreq *ifr)
2473 {
2474 	struct mlx4_en_priv *priv = netdev_priv(dev);
2475 
2476 	return copy_to_user(ifr->ifr_data, &priv->hwtstamp_config,
2477 			    sizeof(priv->hwtstamp_config)) ? -EFAULT : 0;
2478 }
2479 
2480 static int mlx4_en_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
2481 {
2482 	switch (cmd) {
2483 	case SIOCSHWTSTAMP:
2484 		return mlx4_en_hwtstamp_set(dev, ifr);
2485 	case SIOCGHWTSTAMP:
2486 		return mlx4_en_hwtstamp_get(dev, ifr);
2487 	default:
2488 		return -EOPNOTSUPP;
2489 	}
2490 }
2491 
2492 static netdev_features_t mlx4_en_fix_features(struct net_device *netdev,
2493 					      netdev_features_t features)
2494 {
2495 	struct mlx4_en_priv *en_priv = netdev_priv(netdev);
2496 	struct mlx4_en_dev *mdev = en_priv->mdev;
2497 
2498 	/* Since there is no support for separate RX C-TAG/S-TAG vlan accel
2499 	 * enable/disable make sure S-TAG flag is always in same state as
2500 	 * C-TAG.
2501 	 */
2502 	if (features & NETIF_F_HW_VLAN_CTAG_RX &&
2503 	    !(mdev->dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_SKIP_OUTER_VLAN))
2504 		features |= NETIF_F_HW_VLAN_STAG_RX;
2505 	else
2506 		features &= ~NETIF_F_HW_VLAN_STAG_RX;
2507 
2508 	return features;
2509 }
2510 
2511 static int mlx4_en_set_features(struct net_device *netdev,
2512 		netdev_features_t features)
2513 {
2514 	struct mlx4_en_priv *priv = netdev_priv(netdev);
2515 	bool reset = false;
2516 	int ret = 0;
2517 
2518 	if (DEV_FEATURE_CHANGED(netdev, features, NETIF_F_RXFCS)) {
2519 		en_info(priv, "Turn %s RX-FCS\n",
2520 			(features & NETIF_F_RXFCS) ? "ON" : "OFF");
2521 		reset = true;
2522 	}
2523 
2524 	if (DEV_FEATURE_CHANGED(netdev, features, NETIF_F_RXALL)) {
2525 		u8 ignore_fcs_value = (features & NETIF_F_RXALL) ? 1 : 0;
2526 
2527 		en_info(priv, "Turn %s RX-ALL\n",
2528 			ignore_fcs_value ? "ON" : "OFF");
2529 		ret = mlx4_SET_PORT_fcs_check(priv->mdev->dev,
2530 					      priv->port, ignore_fcs_value);
2531 		if (ret)
2532 			return ret;
2533 	}
2534 
2535 	if (DEV_FEATURE_CHANGED(netdev, features, NETIF_F_HW_VLAN_CTAG_RX)) {
2536 		en_info(priv, "Turn %s RX vlan strip offload\n",
2537 			(features & NETIF_F_HW_VLAN_CTAG_RX) ? "ON" : "OFF");
2538 		reset = true;
2539 	}
2540 
2541 	if (DEV_FEATURE_CHANGED(netdev, features, NETIF_F_HW_VLAN_CTAG_TX))
2542 		en_info(priv, "Turn %s TX vlan strip offload\n",
2543 			(features & NETIF_F_HW_VLAN_CTAG_TX) ? "ON" : "OFF");
2544 
2545 	if (DEV_FEATURE_CHANGED(netdev, features, NETIF_F_HW_VLAN_STAG_TX))
2546 		en_info(priv, "Turn %s TX S-VLAN strip offload\n",
2547 			(features & NETIF_F_HW_VLAN_STAG_TX) ? "ON" : "OFF");
2548 
2549 	if (DEV_FEATURE_CHANGED(netdev, features, NETIF_F_LOOPBACK)) {
2550 		en_info(priv, "Turn %s loopback\n",
2551 			(features & NETIF_F_LOOPBACK) ? "ON" : "OFF");
2552 		mlx4_en_update_loopback_state(netdev, features);
2553 	}
2554 
2555 	if (reset) {
2556 		ret = mlx4_en_reset_config(netdev, priv->hwtstamp_config,
2557 					   features);
2558 		if (ret)
2559 			return ret;
2560 	}
2561 
2562 	return 0;
2563 }
2564 
2565 static int mlx4_en_set_vf_mac(struct net_device *dev, int queue, u8 *mac)
2566 {
2567 	struct mlx4_en_priv *en_priv = netdev_priv(dev);
2568 	struct mlx4_en_dev *mdev = en_priv->mdev;
2569 
2570 	return mlx4_set_vf_mac(mdev->dev, en_priv->port, queue, mac);
2571 }
2572 
2573 static int mlx4_en_set_vf_vlan(struct net_device *dev, int vf, u16 vlan, u8 qos,
2574 			       __be16 vlan_proto)
2575 {
2576 	struct mlx4_en_priv *en_priv = netdev_priv(dev);
2577 	struct mlx4_en_dev *mdev = en_priv->mdev;
2578 
2579 	return mlx4_set_vf_vlan(mdev->dev, en_priv->port, vf, vlan, qos,
2580 				vlan_proto);
2581 }
2582 
2583 static int mlx4_en_set_vf_rate(struct net_device *dev, int vf, int min_tx_rate,
2584 			       int max_tx_rate)
2585 {
2586 	struct mlx4_en_priv *en_priv = netdev_priv(dev);
2587 	struct mlx4_en_dev *mdev = en_priv->mdev;
2588 
2589 	return mlx4_set_vf_rate(mdev->dev, en_priv->port, vf, min_tx_rate,
2590 				max_tx_rate);
2591 }
2592 
2593 static int mlx4_en_set_vf_spoofchk(struct net_device *dev, int vf, bool setting)
2594 {
2595 	struct mlx4_en_priv *en_priv = netdev_priv(dev);
2596 	struct mlx4_en_dev *mdev = en_priv->mdev;
2597 
2598 	return mlx4_set_vf_spoofchk(mdev->dev, en_priv->port, vf, setting);
2599 }
2600 
2601 static int mlx4_en_get_vf_config(struct net_device *dev, int vf, struct ifla_vf_info *ivf)
2602 {
2603 	struct mlx4_en_priv *en_priv = netdev_priv(dev);
2604 	struct mlx4_en_dev *mdev = en_priv->mdev;
2605 
2606 	return mlx4_get_vf_config(mdev->dev, en_priv->port, vf, ivf);
2607 }
2608 
2609 static int mlx4_en_set_vf_link_state(struct net_device *dev, int vf, int link_state)
2610 {
2611 	struct mlx4_en_priv *en_priv = netdev_priv(dev);
2612 	struct mlx4_en_dev *mdev = en_priv->mdev;
2613 
2614 	return mlx4_set_vf_link_state(mdev->dev, en_priv->port, vf, link_state);
2615 }
2616 
2617 static int mlx4_en_get_vf_stats(struct net_device *dev, int vf,
2618 				struct ifla_vf_stats *vf_stats)
2619 {
2620 	struct mlx4_en_priv *en_priv = netdev_priv(dev);
2621 	struct mlx4_en_dev *mdev = en_priv->mdev;
2622 
2623 	return mlx4_get_vf_stats(mdev->dev, en_priv->port, vf, vf_stats);
2624 }
2625 
2626 #define PORT_ID_BYTE_LEN 8
2627 static int mlx4_en_get_phys_port_id(struct net_device *dev,
2628 				    struct netdev_phys_item_id *ppid)
2629 {
2630 	struct mlx4_en_priv *priv = netdev_priv(dev);
2631 	struct mlx4_dev *mdev = priv->mdev->dev;
2632 	int i;
2633 	u64 phys_port_id = mdev->caps.phys_port_id[priv->port];
2634 
2635 	if (!phys_port_id)
2636 		return -EOPNOTSUPP;
2637 
2638 	ppid->id_len = sizeof(phys_port_id);
2639 	for (i = PORT_ID_BYTE_LEN - 1; i >= 0; --i) {
2640 		ppid->id[i] =  phys_port_id & 0xff;
2641 		phys_port_id >>= 8;
2642 	}
2643 	return 0;
2644 }
2645 
2646 static void mlx4_en_add_vxlan_offloads(struct work_struct *work)
2647 {
2648 	int ret;
2649 	struct mlx4_en_priv *priv = container_of(work, struct mlx4_en_priv,
2650 						 vxlan_add_task);
2651 
2652 	ret = mlx4_config_vxlan_port(priv->mdev->dev, priv->vxlan_port);
2653 	if (ret)
2654 		goto out;
2655 
2656 	ret = mlx4_SET_PORT_VXLAN(priv->mdev->dev, priv->port,
2657 				  VXLAN_STEER_BY_OUTER_MAC, 1);
2658 out:
2659 	if (ret) {
2660 		en_err(priv, "failed setting L2 tunnel configuration ret %d\n", ret);
2661 		return;
2662 	}
2663 
2664 	/* set offloads */
2665 	priv->dev->hw_enc_features |= NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM |
2666 				      NETIF_F_RXCSUM |
2667 				      NETIF_F_TSO | NETIF_F_TSO6 |
2668 				      NETIF_F_GSO_UDP_TUNNEL |
2669 				      NETIF_F_GSO_UDP_TUNNEL_CSUM |
2670 				      NETIF_F_GSO_PARTIAL;
2671 }
2672 
2673 static void mlx4_en_del_vxlan_offloads(struct work_struct *work)
2674 {
2675 	int ret;
2676 	struct mlx4_en_priv *priv = container_of(work, struct mlx4_en_priv,
2677 						 vxlan_del_task);
2678 	/* unset offloads */
2679 	priv->dev->hw_enc_features &= ~(NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM |
2680 					NETIF_F_RXCSUM |
2681 					NETIF_F_TSO | NETIF_F_TSO6 |
2682 					NETIF_F_GSO_UDP_TUNNEL |
2683 					NETIF_F_GSO_UDP_TUNNEL_CSUM |
2684 					NETIF_F_GSO_PARTIAL);
2685 
2686 	ret = mlx4_SET_PORT_VXLAN(priv->mdev->dev, priv->port,
2687 				  VXLAN_STEER_BY_OUTER_MAC, 0);
2688 	if (ret)
2689 		en_err(priv, "failed setting L2 tunnel configuration ret %d\n", ret);
2690 
2691 	priv->vxlan_port = 0;
2692 }
2693 
2694 static void mlx4_en_add_vxlan_port(struct  net_device *dev,
2695 				   struct udp_tunnel_info *ti)
2696 {
2697 	struct mlx4_en_priv *priv = netdev_priv(dev);
2698 	__be16 port = ti->port;
2699 	__be16 current_port;
2700 
2701 	if (ti->type != UDP_TUNNEL_TYPE_VXLAN)
2702 		return;
2703 
2704 	if (ti->sa_family != AF_INET)
2705 		return;
2706 
2707 	if (priv->mdev->dev->caps.tunnel_offload_mode != MLX4_TUNNEL_OFFLOAD_MODE_VXLAN)
2708 		return;
2709 
2710 	current_port = priv->vxlan_port;
2711 	if (current_port && current_port != port) {
2712 		en_warn(priv, "vxlan port %d configured, can't add port %d\n",
2713 			ntohs(current_port), ntohs(port));
2714 		return;
2715 	}
2716 
2717 	priv->vxlan_port = port;
2718 	queue_work(priv->mdev->workqueue, &priv->vxlan_add_task);
2719 }
2720 
2721 static void mlx4_en_del_vxlan_port(struct  net_device *dev,
2722 				   struct udp_tunnel_info *ti)
2723 {
2724 	struct mlx4_en_priv *priv = netdev_priv(dev);
2725 	__be16 port = ti->port;
2726 	__be16 current_port;
2727 
2728 	if (ti->type != UDP_TUNNEL_TYPE_VXLAN)
2729 		return;
2730 
2731 	if (ti->sa_family != AF_INET)
2732 		return;
2733 
2734 	if (priv->mdev->dev->caps.tunnel_offload_mode != MLX4_TUNNEL_OFFLOAD_MODE_VXLAN)
2735 		return;
2736 
2737 	current_port = priv->vxlan_port;
2738 	if (current_port != port) {
2739 		en_dbg(DRV, priv, "vxlan port %d isn't configured, ignoring\n", ntohs(port));
2740 		return;
2741 	}
2742 
2743 	queue_work(priv->mdev->workqueue, &priv->vxlan_del_task);
2744 }
2745 
2746 static netdev_features_t mlx4_en_features_check(struct sk_buff *skb,
2747 						struct net_device *dev,
2748 						netdev_features_t features)
2749 {
2750 	features = vlan_features_check(skb, features);
2751 	features = vxlan_features_check(skb, features);
2752 
2753 	/* The ConnectX-3 doesn't support outer IPv6 checksums but it does
2754 	 * support inner IPv6 checksums and segmentation so  we need to
2755 	 * strip that feature if this is an IPv6 encapsulated frame.
2756 	 */
2757 	if (skb->encapsulation &&
2758 	    (skb->ip_summed == CHECKSUM_PARTIAL)) {
2759 		struct mlx4_en_priv *priv = netdev_priv(dev);
2760 
2761 		if (!priv->vxlan_port ||
2762 		    (ip_hdr(skb)->version != 4) ||
2763 		    (udp_hdr(skb)->dest != priv->vxlan_port))
2764 			features &= ~(NETIF_F_CSUM_MASK | NETIF_F_GSO_MASK);
2765 	}
2766 
2767 	return features;
2768 }
2769 
2770 static int mlx4_en_set_tx_maxrate(struct net_device *dev, int queue_index, u32 maxrate)
2771 {
2772 	struct mlx4_en_priv *priv = netdev_priv(dev);
2773 	struct mlx4_en_tx_ring *tx_ring = priv->tx_ring[TX][queue_index];
2774 	struct mlx4_update_qp_params params;
2775 	int err;
2776 
2777 	if (!(priv->mdev->dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_QP_RATE_LIMIT))
2778 		return -EOPNOTSUPP;
2779 
2780 	/* rate provided to us in Mbs, check if it fits into 12 bits, if not use Gbs */
2781 	if (maxrate >> 12) {
2782 		params.rate_unit = MLX4_QP_RATE_LIMIT_GBS;
2783 		params.rate_val  = maxrate / 1000;
2784 	} else if (maxrate) {
2785 		params.rate_unit = MLX4_QP_RATE_LIMIT_MBS;
2786 		params.rate_val  = maxrate;
2787 	} else { /* zero serves to revoke the QP rate-limitation */
2788 		params.rate_unit = 0;
2789 		params.rate_val  = 0;
2790 	}
2791 
2792 	err = mlx4_update_qp(priv->mdev->dev, tx_ring->qpn, MLX4_UPDATE_QP_RATE_LIMIT,
2793 			     &params);
2794 	return err;
2795 }
2796 
2797 static int mlx4_xdp_set(struct net_device *dev, struct bpf_prog *prog)
2798 {
2799 	struct mlx4_en_priv *priv = netdev_priv(dev);
2800 	struct mlx4_en_dev *mdev = priv->mdev;
2801 	struct mlx4_en_port_profile new_prof;
2802 	struct bpf_prog *old_prog;
2803 	struct mlx4_en_priv *tmp;
2804 	int tx_changed = 0;
2805 	int xdp_ring_num;
2806 	int port_up = 0;
2807 	int err;
2808 	int i;
2809 
2810 	xdp_ring_num = prog ? priv->rx_ring_num : 0;
2811 
2812 	/* No need to reconfigure buffers when simply swapping the
2813 	 * program for a new one.
2814 	 */
2815 	if (priv->tx_ring_num[TX_XDP] == xdp_ring_num) {
2816 		if (prog) {
2817 			prog = bpf_prog_add(prog, priv->rx_ring_num - 1);
2818 			if (IS_ERR(prog))
2819 				return PTR_ERR(prog);
2820 		}
2821 		mutex_lock(&mdev->state_lock);
2822 		for (i = 0; i < priv->rx_ring_num; i++) {
2823 			old_prog = rcu_dereference_protected(
2824 					priv->rx_ring[i]->xdp_prog,
2825 					lockdep_is_held(&mdev->state_lock));
2826 			rcu_assign_pointer(priv->rx_ring[i]->xdp_prog, prog);
2827 			if (old_prog)
2828 				bpf_prog_put(old_prog);
2829 		}
2830 		mutex_unlock(&mdev->state_lock);
2831 		return 0;
2832 	}
2833 
2834 	if (!mlx4_en_check_xdp_mtu(dev, dev->mtu))
2835 		return -EOPNOTSUPP;
2836 
2837 	tmp = kzalloc(sizeof(*tmp), GFP_KERNEL);
2838 	if (!tmp)
2839 		return -ENOMEM;
2840 
2841 	if (prog) {
2842 		prog = bpf_prog_add(prog, priv->rx_ring_num - 1);
2843 		if (IS_ERR(prog)) {
2844 			err = PTR_ERR(prog);
2845 			goto out;
2846 		}
2847 	}
2848 
2849 	mutex_lock(&mdev->state_lock);
2850 	memcpy(&new_prof, priv->prof, sizeof(struct mlx4_en_port_profile));
2851 	new_prof.tx_ring_num[TX_XDP] = xdp_ring_num;
2852 
2853 	if (priv->tx_ring_num[TX] + xdp_ring_num > MAX_TX_RINGS) {
2854 		tx_changed = 1;
2855 		new_prof.tx_ring_num[TX] =
2856 			MAX_TX_RINGS - ALIGN(xdp_ring_num, priv->prof->num_up);
2857 		en_warn(priv, "Reducing the number of TX rings, to not exceed the max total rings number.\n");
2858 	}
2859 
2860 	err = mlx4_en_try_alloc_resources(priv, tmp, &new_prof, false);
2861 	if (err) {
2862 		if (prog)
2863 			bpf_prog_sub(prog, priv->rx_ring_num - 1);
2864 		goto unlock_out;
2865 	}
2866 
2867 	if (priv->port_up) {
2868 		port_up = 1;
2869 		mlx4_en_stop_port(dev, 1);
2870 	}
2871 
2872 	mlx4_en_safe_replace_resources(priv, tmp);
2873 	if (tx_changed)
2874 		netif_set_real_num_tx_queues(dev, priv->tx_ring_num[TX]);
2875 
2876 	for (i = 0; i < priv->rx_ring_num; i++) {
2877 		old_prog = rcu_dereference_protected(
2878 					priv->rx_ring[i]->xdp_prog,
2879 					lockdep_is_held(&mdev->state_lock));
2880 		rcu_assign_pointer(priv->rx_ring[i]->xdp_prog, prog);
2881 		if (old_prog)
2882 			bpf_prog_put(old_prog);
2883 	}
2884 
2885 	if (port_up) {
2886 		err = mlx4_en_start_port(dev);
2887 		if (err) {
2888 			en_err(priv, "Failed starting port %d for XDP change\n",
2889 			       priv->port);
2890 			queue_work(mdev->workqueue, &priv->watchdog_task);
2891 		}
2892 	}
2893 
2894 unlock_out:
2895 	mutex_unlock(&mdev->state_lock);
2896 out:
2897 	kfree(tmp);
2898 	return err;
2899 }
2900 
2901 static u32 mlx4_xdp_query(struct net_device *dev)
2902 {
2903 	struct mlx4_en_priv *priv = netdev_priv(dev);
2904 	struct mlx4_en_dev *mdev = priv->mdev;
2905 	const struct bpf_prog *xdp_prog;
2906 	u32 prog_id = 0;
2907 
2908 	if (!priv->tx_ring_num[TX_XDP])
2909 		return prog_id;
2910 
2911 	mutex_lock(&mdev->state_lock);
2912 	xdp_prog = rcu_dereference_protected(
2913 		priv->rx_ring[0]->xdp_prog,
2914 		lockdep_is_held(&mdev->state_lock));
2915 	if (xdp_prog)
2916 		prog_id = xdp_prog->aux->id;
2917 	mutex_unlock(&mdev->state_lock);
2918 
2919 	return prog_id;
2920 }
2921 
2922 static int mlx4_xdp(struct net_device *dev, struct netdev_bpf *xdp)
2923 {
2924 	switch (xdp->command) {
2925 	case XDP_SETUP_PROG:
2926 		return mlx4_xdp_set(dev, xdp->prog);
2927 	case XDP_QUERY_PROG:
2928 		xdp->prog_id = mlx4_xdp_query(dev);
2929 		return 0;
2930 	default:
2931 		return -EINVAL;
2932 	}
2933 }
2934 
2935 static const struct net_device_ops mlx4_netdev_ops = {
2936 	.ndo_open		= mlx4_en_open,
2937 	.ndo_stop		= mlx4_en_close,
2938 	.ndo_start_xmit		= mlx4_en_xmit,
2939 	.ndo_select_queue	= mlx4_en_select_queue,
2940 	.ndo_get_stats64	= mlx4_en_get_stats64,
2941 	.ndo_set_rx_mode	= mlx4_en_set_rx_mode,
2942 	.ndo_set_mac_address	= mlx4_en_set_mac,
2943 	.ndo_validate_addr	= eth_validate_addr,
2944 	.ndo_change_mtu		= mlx4_en_change_mtu,
2945 	.ndo_do_ioctl		= mlx4_en_ioctl,
2946 	.ndo_tx_timeout		= mlx4_en_tx_timeout,
2947 	.ndo_vlan_rx_add_vid	= mlx4_en_vlan_rx_add_vid,
2948 	.ndo_vlan_rx_kill_vid	= mlx4_en_vlan_rx_kill_vid,
2949 #ifdef CONFIG_NET_POLL_CONTROLLER
2950 	.ndo_poll_controller	= mlx4_en_netpoll,
2951 #endif
2952 	.ndo_set_features	= mlx4_en_set_features,
2953 	.ndo_fix_features	= mlx4_en_fix_features,
2954 	.ndo_setup_tc		= __mlx4_en_setup_tc,
2955 #ifdef CONFIG_RFS_ACCEL
2956 	.ndo_rx_flow_steer	= mlx4_en_filter_rfs,
2957 #endif
2958 	.ndo_get_phys_port_id	= mlx4_en_get_phys_port_id,
2959 	.ndo_udp_tunnel_add	= mlx4_en_add_vxlan_port,
2960 	.ndo_udp_tunnel_del	= mlx4_en_del_vxlan_port,
2961 	.ndo_features_check	= mlx4_en_features_check,
2962 	.ndo_set_tx_maxrate	= mlx4_en_set_tx_maxrate,
2963 	.ndo_bpf		= mlx4_xdp,
2964 };
2965 
2966 static const struct net_device_ops mlx4_netdev_ops_master = {
2967 	.ndo_open		= mlx4_en_open,
2968 	.ndo_stop		= mlx4_en_close,
2969 	.ndo_start_xmit		= mlx4_en_xmit,
2970 	.ndo_select_queue	= mlx4_en_select_queue,
2971 	.ndo_get_stats64	= mlx4_en_get_stats64,
2972 	.ndo_set_rx_mode	= mlx4_en_set_rx_mode,
2973 	.ndo_set_mac_address	= mlx4_en_set_mac,
2974 	.ndo_validate_addr	= eth_validate_addr,
2975 	.ndo_change_mtu		= mlx4_en_change_mtu,
2976 	.ndo_tx_timeout		= mlx4_en_tx_timeout,
2977 	.ndo_vlan_rx_add_vid	= mlx4_en_vlan_rx_add_vid,
2978 	.ndo_vlan_rx_kill_vid	= mlx4_en_vlan_rx_kill_vid,
2979 	.ndo_set_vf_mac		= mlx4_en_set_vf_mac,
2980 	.ndo_set_vf_vlan	= mlx4_en_set_vf_vlan,
2981 	.ndo_set_vf_rate	= mlx4_en_set_vf_rate,
2982 	.ndo_set_vf_spoofchk	= mlx4_en_set_vf_spoofchk,
2983 	.ndo_set_vf_link_state	= mlx4_en_set_vf_link_state,
2984 	.ndo_get_vf_stats       = mlx4_en_get_vf_stats,
2985 	.ndo_get_vf_config	= mlx4_en_get_vf_config,
2986 #ifdef CONFIG_NET_POLL_CONTROLLER
2987 	.ndo_poll_controller	= mlx4_en_netpoll,
2988 #endif
2989 	.ndo_set_features	= mlx4_en_set_features,
2990 	.ndo_fix_features	= mlx4_en_fix_features,
2991 	.ndo_setup_tc		= __mlx4_en_setup_tc,
2992 #ifdef CONFIG_RFS_ACCEL
2993 	.ndo_rx_flow_steer	= mlx4_en_filter_rfs,
2994 #endif
2995 	.ndo_get_phys_port_id	= mlx4_en_get_phys_port_id,
2996 	.ndo_udp_tunnel_add	= mlx4_en_add_vxlan_port,
2997 	.ndo_udp_tunnel_del	= mlx4_en_del_vxlan_port,
2998 	.ndo_features_check	= mlx4_en_features_check,
2999 	.ndo_set_tx_maxrate	= mlx4_en_set_tx_maxrate,
3000 	.ndo_bpf		= mlx4_xdp,
3001 };
3002 
3003 struct mlx4_en_bond {
3004 	struct work_struct work;
3005 	struct mlx4_en_priv *priv;
3006 	int is_bonded;
3007 	struct mlx4_port_map port_map;
3008 };
3009 
3010 static void mlx4_en_bond_work(struct work_struct *work)
3011 {
3012 	struct mlx4_en_bond *bond = container_of(work,
3013 						     struct mlx4_en_bond,
3014 						     work);
3015 	int err = 0;
3016 	struct mlx4_dev *dev = bond->priv->mdev->dev;
3017 
3018 	if (bond->is_bonded) {
3019 		if (!mlx4_is_bonded(dev)) {
3020 			err = mlx4_bond(dev);
3021 			if (err)
3022 				en_err(bond->priv, "Fail to bond device\n");
3023 		}
3024 		if (!err) {
3025 			err = mlx4_port_map_set(dev, &bond->port_map);
3026 			if (err)
3027 				en_err(bond->priv, "Fail to set port map [%d][%d]: %d\n",
3028 				       bond->port_map.port1,
3029 				       bond->port_map.port2,
3030 				       err);
3031 		}
3032 	} else if (mlx4_is_bonded(dev)) {
3033 		err = mlx4_unbond(dev);
3034 		if (err)
3035 			en_err(bond->priv, "Fail to unbond device\n");
3036 	}
3037 	dev_put(bond->priv->dev);
3038 	kfree(bond);
3039 }
3040 
3041 static int mlx4_en_queue_bond_work(struct mlx4_en_priv *priv, int is_bonded,
3042 				   u8 v2p_p1, u8 v2p_p2)
3043 {
3044 	struct mlx4_en_bond *bond = NULL;
3045 
3046 	bond = kzalloc(sizeof(*bond), GFP_ATOMIC);
3047 	if (!bond)
3048 		return -ENOMEM;
3049 
3050 	INIT_WORK(&bond->work, mlx4_en_bond_work);
3051 	bond->priv = priv;
3052 	bond->is_bonded = is_bonded;
3053 	bond->port_map.port1 = v2p_p1;
3054 	bond->port_map.port2 = v2p_p2;
3055 	dev_hold(priv->dev);
3056 	queue_work(priv->mdev->workqueue, &bond->work);
3057 	return 0;
3058 }
3059 
3060 int mlx4_en_netdev_event(struct notifier_block *this,
3061 			 unsigned long event, void *ptr)
3062 {
3063 	struct net_device *ndev = netdev_notifier_info_to_dev(ptr);
3064 	u8 port = 0;
3065 	struct mlx4_en_dev *mdev;
3066 	struct mlx4_dev *dev;
3067 	int i, num_eth_ports = 0;
3068 	bool do_bond = true;
3069 	struct mlx4_en_priv *priv;
3070 	u8 v2p_port1 = 0;
3071 	u8 v2p_port2 = 0;
3072 
3073 	if (!net_eq(dev_net(ndev), &init_net))
3074 		return NOTIFY_DONE;
3075 
3076 	mdev = container_of(this, struct mlx4_en_dev, nb);
3077 	dev = mdev->dev;
3078 
3079 	/* Go into this mode only when two network devices set on two ports
3080 	 * of the same mlx4 device are slaves of the same bonding master
3081 	 */
3082 	mlx4_foreach_port(i, dev, MLX4_PORT_TYPE_ETH) {
3083 		++num_eth_ports;
3084 		if (!port && (mdev->pndev[i] == ndev))
3085 			port = i;
3086 		mdev->upper[i] = mdev->pndev[i] ?
3087 			netdev_master_upper_dev_get(mdev->pndev[i]) : NULL;
3088 		/* condition not met: network device is a slave */
3089 		if (!mdev->upper[i])
3090 			do_bond = false;
3091 		if (num_eth_ports < 2)
3092 			continue;
3093 		/* condition not met: same master */
3094 		if (mdev->upper[i] != mdev->upper[i-1])
3095 			do_bond = false;
3096 	}
3097 	/* condition not met: 2 salves */
3098 	do_bond = (num_eth_ports ==  2) ? do_bond : false;
3099 
3100 	/* handle only events that come with enough info */
3101 	if ((do_bond && (event != NETDEV_BONDING_INFO)) || !port)
3102 		return NOTIFY_DONE;
3103 
3104 	priv = netdev_priv(ndev);
3105 	if (do_bond) {
3106 		struct netdev_notifier_bonding_info *notifier_info = ptr;
3107 		struct netdev_bonding_info *bonding_info =
3108 			&notifier_info->bonding_info;
3109 
3110 		/* required mode 1, 2 or 4 */
3111 		if ((bonding_info->master.bond_mode != BOND_MODE_ACTIVEBACKUP) &&
3112 		    (bonding_info->master.bond_mode != BOND_MODE_XOR) &&
3113 		    (bonding_info->master.bond_mode != BOND_MODE_8023AD))
3114 			do_bond = false;
3115 
3116 		/* require exactly 2 slaves */
3117 		if (bonding_info->master.num_slaves != 2)
3118 			do_bond = false;
3119 
3120 		/* calc v2p */
3121 		if (do_bond) {
3122 			if (bonding_info->master.bond_mode ==
3123 			    BOND_MODE_ACTIVEBACKUP) {
3124 				/* in active-backup mode virtual ports are
3125 				 * mapped to the physical port of the active
3126 				 * slave */
3127 				if (bonding_info->slave.state ==
3128 				    BOND_STATE_BACKUP) {
3129 					if (port == 1) {
3130 						v2p_port1 = 2;
3131 						v2p_port2 = 2;
3132 					} else {
3133 						v2p_port1 = 1;
3134 						v2p_port2 = 1;
3135 					}
3136 				} else { /* BOND_STATE_ACTIVE */
3137 					if (port == 1) {
3138 						v2p_port1 = 1;
3139 						v2p_port2 = 1;
3140 					} else {
3141 						v2p_port1 = 2;
3142 						v2p_port2 = 2;
3143 					}
3144 				}
3145 			} else { /* Active-Active */
3146 				/* in active-active mode a virtual port is
3147 				 * mapped to the native physical port if and only
3148 				 * if the physical port is up */
3149 				__s8 link = bonding_info->slave.link;
3150 
3151 				if (port == 1)
3152 					v2p_port2 = 2;
3153 				else
3154 					v2p_port1 = 1;
3155 				if ((link == BOND_LINK_UP) ||
3156 				    (link == BOND_LINK_FAIL)) {
3157 					if (port == 1)
3158 						v2p_port1 = 1;
3159 					else
3160 						v2p_port2 = 2;
3161 				} else { /* BOND_LINK_DOWN || BOND_LINK_BACK */
3162 					if (port == 1)
3163 						v2p_port1 = 2;
3164 					else
3165 						v2p_port2 = 1;
3166 				}
3167 			}
3168 		}
3169 	}
3170 
3171 	mlx4_en_queue_bond_work(priv, do_bond,
3172 				v2p_port1, v2p_port2);
3173 
3174 	return NOTIFY_DONE;
3175 }
3176 
3177 void mlx4_en_update_pfc_stats_bitmap(struct mlx4_dev *dev,
3178 				     struct mlx4_en_stats_bitmap *stats_bitmap,
3179 				     u8 rx_ppp, u8 rx_pause,
3180 				     u8 tx_ppp, u8 tx_pause)
3181 {
3182 	int last_i = NUM_MAIN_STATS + NUM_PORT_STATS + NUM_PF_STATS;
3183 
3184 	if (!mlx4_is_slave(dev) &&
3185 	    (dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_FLOWSTATS_EN)) {
3186 		mutex_lock(&stats_bitmap->mutex);
3187 		bitmap_clear(stats_bitmap->bitmap, last_i, NUM_FLOW_STATS);
3188 
3189 		if (rx_ppp)
3190 			bitmap_set(stats_bitmap->bitmap, last_i,
3191 				   NUM_FLOW_PRIORITY_STATS_RX);
3192 		last_i += NUM_FLOW_PRIORITY_STATS_RX;
3193 
3194 		if (rx_pause && !(rx_ppp))
3195 			bitmap_set(stats_bitmap->bitmap, last_i,
3196 				   NUM_FLOW_STATS_RX);
3197 		last_i += NUM_FLOW_STATS_RX;
3198 
3199 		if (tx_ppp)
3200 			bitmap_set(stats_bitmap->bitmap, last_i,
3201 				   NUM_FLOW_PRIORITY_STATS_TX);
3202 		last_i += NUM_FLOW_PRIORITY_STATS_TX;
3203 
3204 		if (tx_pause && !(tx_ppp))
3205 			bitmap_set(stats_bitmap->bitmap, last_i,
3206 				   NUM_FLOW_STATS_TX);
3207 		last_i += NUM_FLOW_STATS_TX;
3208 
3209 		mutex_unlock(&stats_bitmap->mutex);
3210 	}
3211 }
3212 
3213 void mlx4_en_set_stats_bitmap(struct mlx4_dev *dev,
3214 			      struct mlx4_en_stats_bitmap *stats_bitmap,
3215 			      u8 rx_ppp, u8 rx_pause,
3216 			      u8 tx_ppp, u8 tx_pause)
3217 {
3218 	int last_i = 0;
3219 
3220 	mutex_init(&stats_bitmap->mutex);
3221 	bitmap_zero(stats_bitmap->bitmap, NUM_ALL_STATS);
3222 
3223 	if (mlx4_is_slave(dev)) {
3224 		bitmap_set(stats_bitmap->bitmap, last_i +
3225 					 MLX4_FIND_NETDEV_STAT(rx_packets), 1);
3226 		bitmap_set(stats_bitmap->bitmap, last_i +
3227 					 MLX4_FIND_NETDEV_STAT(tx_packets), 1);
3228 		bitmap_set(stats_bitmap->bitmap, last_i +
3229 					 MLX4_FIND_NETDEV_STAT(rx_bytes), 1);
3230 		bitmap_set(stats_bitmap->bitmap, last_i +
3231 					 MLX4_FIND_NETDEV_STAT(tx_bytes), 1);
3232 		bitmap_set(stats_bitmap->bitmap, last_i +
3233 					 MLX4_FIND_NETDEV_STAT(rx_dropped), 1);
3234 		bitmap_set(stats_bitmap->bitmap, last_i +
3235 					 MLX4_FIND_NETDEV_STAT(tx_dropped), 1);
3236 	} else {
3237 		bitmap_set(stats_bitmap->bitmap, last_i, NUM_MAIN_STATS);
3238 	}
3239 	last_i += NUM_MAIN_STATS;
3240 
3241 	bitmap_set(stats_bitmap->bitmap, last_i, NUM_PORT_STATS);
3242 	last_i += NUM_PORT_STATS;
3243 
3244 	if (mlx4_is_master(dev))
3245 		bitmap_set(stats_bitmap->bitmap, last_i,
3246 			   NUM_PF_STATS);
3247 	last_i += NUM_PF_STATS;
3248 
3249 	mlx4_en_update_pfc_stats_bitmap(dev, stats_bitmap,
3250 					rx_ppp, rx_pause,
3251 					tx_ppp, tx_pause);
3252 	last_i += NUM_FLOW_STATS;
3253 
3254 	if (!mlx4_is_slave(dev))
3255 		bitmap_set(stats_bitmap->bitmap, last_i, NUM_PKT_STATS);
3256 	last_i += NUM_PKT_STATS;
3257 
3258 	bitmap_set(stats_bitmap->bitmap, last_i, NUM_XDP_STATS);
3259 	last_i += NUM_XDP_STATS;
3260 
3261 	if (!mlx4_is_slave(dev))
3262 		bitmap_set(stats_bitmap->bitmap, last_i, NUM_PHY_STATS);
3263 	last_i += NUM_PHY_STATS;
3264 }
3265 
3266 int mlx4_en_init_netdev(struct mlx4_en_dev *mdev, int port,
3267 			struct mlx4_en_port_profile *prof)
3268 {
3269 	struct net_device *dev;
3270 	struct mlx4_en_priv *priv;
3271 	int i, t;
3272 	int err;
3273 
3274 	dev = alloc_etherdev_mqs(sizeof(struct mlx4_en_priv),
3275 				 MAX_TX_RINGS, MAX_RX_RINGS);
3276 	if (dev == NULL)
3277 		return -ENOMEM;
3278 
3279 	netif_set_real_num_tx_queues(dev, prof->tx_ring_num[TX]);
3280 	netif_set_real_num_rx_queues(dev, prof->rx_ring_num);
3281 
3282 	SET_NETDEV_DEV(dev, &mdev->dev->persist->pdev->dev);
3283 	dev->dev_port = port - 1;
3284 
3285 	/*
3286 	 * Initialize driver private data
3287 	 */
3288 
3289 	priv = netdev_priv(dev);
3290 	memset(priv, 0, sizeof(struct mlx4_en_priv));
3291 	priv->counter_index = MLX4_SINK_COUNTER_INDEX(mdev->dev);
3292 	spin_lock_init(&priv->stats_lock);
3293 	INIT_WORK(&priv->rx_mode_task, mlx4_en_do_set_rx_mode);
3294 	INIT_WORK(&priv->watchdog_task, mlx4_en_restart);
3295 	INIT_WORK(&priv->linkstate_task, mlx4_en_linkstate);
3296 	INIT_DELAYED_WORK(&priv->stats_task, mlx4_en_do_get_stats);
3297 	INIT_DELAYED_WORK(&priv->service_task, mlx4_en_service_task);
3298 	INIT_WORK(&priv->vxlan_add_task, mlx4_en_add_vxlan_offloads);
3299 	INIT_WORK(&priv->vxlan_del_task, mlx4_en_del_vxlan_offloads);
3300 #ifdef CONFIG_RFS_ACCEL
3301 	INIT_LIST_HEAD(&priv->filters);
3302 	spin_lock_init(&priv->filters_lock);
3303 #endif
3304 
3305 	priv->dev = dev;
3306 	priv->mdev = mdev;
3307 	priv->ddev = &mdev->pdev->dev;
3308 	priv->prof = prof;
3309 	priv->port = port;
3310 	priv->port_up = false;
3311 	priv->flags = prof->flags;
3312 	priv->pflags = MLX4_EN_PRIV_FLAGS_BLUEFLAME;
3313 	priv->ctrl_flags = cpu_to_be32(MLX4_WQE_CTRL_CQ_UPDATE |
3314 			MLX4_WQE_CTRL_SOLICITED);
3315 	priv->num_tx_rings_p_up = mdev->profile.max_num_tx_rings_p_up;
3316 	priv->tx_work_limit = MLX4_EN_DEFAULT_TX_WORK;
3317 	netdev_rss_key_fill(priv->rss_key, sizeof(priv->rss_key));
3318 
3319 	for (t = 0; t < MLX4_EN_NUM_TX_TYPES; t++) {
3320 		priv->tx_ring_num[t] = prof->tx_ring_num[t];
3321 		if (!priv->tx_ring_num[t])
3322 			continue;
3323 
3324 		priv->tx_ring[t] = kcalloc(MAX_TX_RINGS,
3325 					   sizeof(struct mlx4_en_tx_ring *),
3326 					   GFP_KERNEL);
3327 		if (!priv->tx_ring[t]) {
3328 			err = -ENOMEM;
3329 			goto out;
3330 		}
3331 		priv->tx_cq[t] = kcalloc(MAX_TX_RINGS,
3332 					 sizeof(struct mlx4_en_cq *),
3333 					 GFP_KERNEL);
3334 		if (!priv->tx_cq[t]) {
3335 			err = -ENOMEM;
3336 			goto out;
3337 		}
3338 	}
3339 	priv->rx_ring_num = prof->rx_ring_num;
3340 	priv->cqe_factor = (mdev->dev->caps.cqe_size == 64) ? 1 : 0;
3341 	priv->cqe_size = mdev->dev->caps.cqe_size;
3342 	priv->mac_index = -1;
3343 	priv->msg_enable = MLX4_EN_MSG_LEVEL;
3344 #ifdef CONFIG_MLX4_EN_DCB
3345 	if (!mlx4_is_slave(priv->mdev->dev)) {
3346 		u8 prio;
3347 
3348 		for (prio = 0; prio < IEEE_8021QAZ_MAX_TCS; ++prio) {
3349 			priv->ets.prio_tc[prio] = prio;
3350 			priv->ets.tc_tsa[prio]  = IEEE_8021QAZ_TSA_VENDOR;
3351 		}
3352 
3353 		priv->dcbx_cap = DCB_CAP_DCBX_VER_CEE | DCB_CAP_DCBX_HOST |
3354 			DCB_CAP_DCBX_VER_IEEE;
3355 		priv->flags |= MLX4_EN_DCB_ENABLED;
3356 		priv->cee_config.pfc_state = false;
3357 
3358 		for (i = 0; i < MLX4_EN_NUM_UP_HIGH; i++)
3359 			priv->cee_config.dcb_pfc[i] = pfc_disabled;
3360 
3361 		if (mdev->dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_ETS_CFG) {
3362 			dev->dcbnl_ops = &mlx4_en_dcbnl_ops;
3363 		} else {
3364 			en_info(priv, "enabling only PFC DCB ops\n");
3365 			dev->dcbnl_ops = &mlx4_en_dcbnl_pfc_ops;
3366 		}
3367 	}
3368 #endif
3369 
3370 	for (i = 0; i < MLX4_EN_MAC_HASH_SIZE; ++i)
3371 		INIT_HLIST_HEAD(&priv->mac_hash[i]);
3372 
3373 	/* Query for default mac and max mtu */
3374 	priv->max_mtu = mdev->dev->caps.eth_mtu_cap[priv->port];
3375 
3376 	if (mdev->dev->caps.rx_checksum_flags_port[priv->port] &
3377 	    MLX4_RX_CSUM_MODE_VAL_NON_TCP_UDP)
3378 		priv->flags |= MLX4_EN_FLAG_RX_CSUM_NON_TCP_UDP;
3379 
3380 	/* Set default MAC */
3381 	dev->addr_len = ETH_ALEN;
3382 	mlx4_en_u64_to_mac(dev->dev_addr, mdev->dev->caps.def_mac[priv->port]);
3383 	if (!is_valid_ether_addr(dev->dev_addr)) {
3384 		en_err(priv, "Port: %d, invalid mac burned: %pM, quiting\n",
3385 		       priv->port, dev->dev_addr);
3386 		err = -EINVAL;
3387 		goto out;
3388 	} else if (mlx4_is_slave(priv->mdev->dev) &&
3389 		   (priv->mdev->dev->port_random_macs & 1 << priv->port)) {
3390 		/* Random MAC was assigned in mlx4_slave_cap
3391 		 * in mlx4_core module
3392 		 */
3393 		dev->addr_assign_type |= NET_ADDR_RANDOM;
3394 		en_warn(priv, "Assigned random MAC address %pM\n", dev->dev_addr);
3395 	}
3396 
3397 	memcpy(priv->current_mac, dev->dev_addr, sizeof(priv->current_mac));
3398 
3399 	priv->stride = roundup_pow_of_two(sizeof(struct mlx4_en_rx_desc) +
3400 					  DS_SIZE * MLX4_EN_MAX_RX_FRAGS);
3401 	err = mlx4_en_alloc_resources(priv);
3402 	if (err)
3403 		goto out;
3404 
3405 	/* Initialize time stamping config */
3406 	priv->hwtstamp_config.flags = 0;
3407 	priv->hwtstamp_config.tx_type = HWTSTAMP_TX_OFF;
3408 	priv->hwtstamp_config.rx_filter = HWTSTAMP_FILTER_NONE;
3409 
3410 	/* Allocate page for receive rings */
3411 	err = mlx4_alloc_hwq_res(mdev->dev, &priv->res,
3412 				MLX4_EN_PAGE_SIZE);
3413 	if (err) {
3414 		en_err(priv, "Failed to allocate page for rx qps\n");
3415 		goto out;
3416 	}
3417 	priv->allocated = 1;
3418 
3419 	/*
3420 	 * Initialize netdev entry points
3421 	 */
3422 	if (mlx4_is_master(priv->mdev->dev))
3423 		dev->netdev_ops = &mlx4_netdev_ops_master;
3424 	else
3425 		dev->netdev_ops = &mlx4_netdev_ops;
3426 	dev->watchdog_timeo = MLX4_EN_WATCHDOG_TIMEOUT;
3427 	netif_set_real_num_tx_queues(dev, priv->tx_ring_num[TX]);
3428 	netif_set_real_num_rx_queues(dev, priv->rx_ring_num);
3429 
3430 	dev->ethtool_ops = &mlx4_en_ethtool_ops;
3431 
3432 	/*
3433 	 * Set driver features
3434 	 */
3435 	dev->hw_features = NETIF_F_SG | NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM;
3436 	if (mdev->LSO_support)
3437 		dev->hw_features |= NETIF_F_TSO | NETIF_F_TSO6;
3438 
3439 	dev->vlan_features = dev->hw_features;
3440 
3441 	dev->hw_features |= NETIF_F_RXCSUM | NETIF_F_RXHASH;
3442 	dev->features = dev->hw_features | NETIF_F_HIGHDMA |
3443 			NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_CTAG_RX |
3444 			NETIF_F_HW_VLAN_CTAG_FILTER;
3445 	dev->hw_features |= NETIF_F_LOOPBACK |
3446 			NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_CTAG_RX;
3447 
3448 	if (!(mdev->dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_SKIP_OUTER_VLAN)) {
3449 		dev->features |= NETIF_F_HW_VLAN_STAG_RX |
3450 			NETIF_F_HW_VLAN_STAG_FILTER;
3451 		dev->hw_features |= NETIF_F_HW_VLAN_STAG_RX;
3452 	}
3453 
3454 	if (mlx4_is_slave(mdev->dev)) {
3455 		bool vlan_offload_disabled;
3456 		int phv;
3457 
3458 		err = get_phv_bit(mdev->dev, port, &phv);
3459 		if (!err && phv) {
3460 			dev->hw_features |= NETIF_F_HW_VLAN_STAG_TX;
3461 			priv->pflags |= MLX4_EN_PRIV_FLAGS_PHV;
3462 		}
3463 		err = mlx4_get_is_vlan_offload_disabled(mdev->dev, port,
3464 							&vlan_offload_disabled);
3465 		if (!err && vlan_offload_disabled) {
3466 			dev->hw_features &= ~(NETIF_F_HW_VLAN_CTAG_TX |
3467 					      NETIF_F_HW_VLAN_CTAG_RX |
3468 					      NETIF_F_HW_VLAN_STAG_TX |
3469 					      NETIF_F_HW_VLAN_STAG_RX);
3470 			dev->features &= ~(NETIF_F_HW_VLAN_CTAG_TX |
3471 					   NETIF_F_HW_VLAN_CTAG_RX |
3472 					   NETIF_F_HW_VLAN_STAG_TX |
3473 					   NETIF_F_HW_VLAN_STAG_RX);
3474 		}
3475 	} else {
3476 		if (mdev->dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_PHV_EN &&
3477 		    !(mdev->dev->caps.flags2 &
3478 		      MLX4_DEV_CAP_FLAG2_SKIP_OUTER_VLAN))
3479 			dev->hw_features |= NETIF_F_HW_VLAN_STAG_TX;
3480 	}
3481 
3482 	if (mdev->dev->caps.flags & MLX4_DEV_CAP_FLAG_FCS_KEEP)
3483 		dev->hw_features |= NETIF_F_RXFCS;
3484 
3485 	if (mdev->dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_IGNORE_FCS)
3486 		dev->hw_features |= NETIF_F_RXALL;
3487 
3488 	if (mdev->dev->caps.steering_mode ==
3489 	    MLX4_STEERING_MODE_DEVICE_MANAGED &&
3490 	    mdev->dev->caps.dmfs_high_steer_mode != MLX4_STEERING_DMFS_A0_STATIC)
3491 		dev->hw_features |= NETIF_F_NTUPLE;
3492 
3493 	if (mdev->dev->caps.steering_mode != MLX4_STEERING_MODE_A0)
3494 		dev->priv_flags |= IFF_UNICAST_FLT;
3495 
3496 	/* Setting a default hash function value */
3497 	if (mdev->dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_RSS_TOP) {
3498 		priv->rss_hash_fn = ETH_RSS_HASH_TOP;
3499 	} else if (mdev->dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_RSS_XOR) {
3500 		priv->rss_hash_fn = ETH_RSS_HASH_XOR;
3501 	} else {
3502 		en_warn(priv,
3503 			"No RSS hash capabilities exposed, using Toeplitz\n");
3504 		priv->rss_hash_fn = ETH_RSS_HASH_TOP;
3505 	}
3506 
3507 	if (mdev->dev->caps.tunnel_offload_mode == MLX4_TUNNEL_OFFLOAD_MODE_VXLAN) {
3508 		dev->hw_features |= NETIF_F_GSO_UDP_TUNNEL |
3509 				    NETIF_F_GSO_UDP_TUNNEL_CSUM |
3510 				    NETIF_F_GSO_PARTIAL;
3511 		dev->features    |= NETIF_F_GSO_UDP_TUNNEL |
3512 				    NETIF_F_GSO_UDP_TUNNEL_CSUM |
3513 				    NETIF_F_GSO_PARTIAL;
3514 		dev->gso_partial_features = NETIF_F_GSO_UDP_TUNNEL_CSUM;
3515 	}
3516 
3517 	/* MTU range: 46 - hw-specific max */
3518 	dev->min_mtu = MLX4_EN_MIN_MTU;
3519 	dev->max_mtu = priv->max_mtu;
3520 
3521 	mdev->pndev[port] = dev;
3522 	mdev->upper[port] = NULL;
3523 
3524 	netif_carrier_off(dev);
3525 	mlx4_en_set_default_moderation(priv);
3526 
3527 	en_warn(priv, "Using %d TX rings\n", prof->tx_ring_num[TX]);
3528 	en_warn(priv, "Using %d RX rings\n", prof->rx_ring_num);
3529 
3530 	mlx4_en_update_loopback_state(priv->dev, priv->dev->features);
3531 
3532 	/* Configure port */
3533 	mlx4_en_calc_rx_buf(dev);
3534 	err = mlx4_SET_PORT_general(mdev->dev, priv->port,
3535 				    priv->rx_skb_size + ETH_FCS_LEN,
3536 				    prof->tx_pause, prof->tx_ppp,
3537 				    prof->rx_pause, prof->rx_ppp);
3538 	if (err) {
3539 		en_err(priv, "Failed setting port general configurations for port %d, with error %d\n",
3540 		       priv->port, err);
3541 		goto out;
3542 	}
3543 
3544 	if (mdev->dev->caps.tunnel_offload_mode == MLX4_TUNNEL_OFFLOAD_MODE_VXLAN) {
3545 		err = mlx4_SET_PORT_VXLAN(mdev->dev, priv->port, VXLAN_STEER_BY_OUTER_MAC, 1);
3546 		if (err) {
3547 			en_err(priv, "Failed setting port L2 tunnel configuration, err %d\n",
3548 			       err);
3549 			goto out;
3550 		}
3551 	}
3552 
3553 	/* Init port */
3554 	en_warn(priv, "Initializing port\n");
3555 	err = mlx4_INIT_PORT(mdev->dev, priv->port);
3556 	if (err) {
3557 		en_err(priv, "Failed Initializing port\n");
3558 		goto out;
3559 	}
3560 	queue_delayed_work(mdev->workqueue, &priv->stats_task, STATS_DELAY);
3561 
3562 	/* Initialize time stamp mechanism */
3563 	if (mdev->dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_TS)
3564 		mlx4_en_init_timestamp(mdev);
3565 
3566 	queue_delayed_work(mdev->workqueue, &priv->service_task,
3567 			   SERVICE_TASK_DELAY);
3568 
3569 	mlx4_en_set_stats_bitmap(mdev->dev, &priv->stats_bitmap,
3570 				 mdev->profile.prof[priv->port].rx_ppp,
3571 				 mdev->profile.prof[priv->port].rx_pause,
3572 				 mdev->profile.prof[priv->port].tx_ppp,
3573 				 mdev->profile.prof[priv->port].tx_pause);
3574 
3575 	err = register_netdev(dev);
3576 	if (err) {
3577 		en_err(priv, "Netdev registration failed for port %d\n", port);
3578 		goto out;
3579 	}
3580 
3581 	priv->registered = 1;
3582 	devlink_port_type_eth_set(mlx4_get_devlink_port(mdev->dev, priv->port),
3583 				  dev);
3584 
3585 	return 0;
3586 
3587 out:
3588 	mlx4_en_destroy_netdev(dev);
3589 	return err;
3590 }
3591 
3592 int mlx4_en_reset_config(struct net_device *dev,
3593 			 struct hwtstamp_config ts_config,
3594 			 netdev_features_t features)
3595 {
3596 	struct mlx4_en_priv *priv = netdev_priv(dev);
3597 	struct mlx4_en_dev *mdev = priv->mdev;
3598 	struct mlx4_en_port_profile new_prof;
3599 	struct mlx4_en_priv *tmp;
3600 	int port_up = 0;
3601 	int err = 0;
3602 
3603 	if (priv->hwtstamp_config.tx_type == ts_config.tx_type &&
3604 	    priv->hwtstamp_config.rx_filter == ts_config.rx_filter &&
3605 	    !DEV_FEATURE_CHANGED(dev, features, NETIF_F_HW_VLAN_CTAG_RX) &&
3606 	    !DEV_FEATURE_CHANGED(dev, features, NETIF_F_RXFCS))
3607 		return 0; /* Nothing to change */
3608 
3609 	if (DEV_FEATURE_CHANGED(dev, features, NETIF_F_HW_VLAN_CTAG_RX) &&
3610 	    (features & NETIF_F_HW_VLAN_CTAG_RX) &&
3611 	    (priv->hwtstamp_config.rx_filter != HWTSTAMP_FILTER_NONE)) {
3612 		en_warn(priv, "Can't turn ON rx vlan offload while time-stamping rx filter is ON\n");
3613 		return -EINVAL;
3614 	}
3615 
3616 	tmp = kzalloc(sizeof(*tmp), GFP_KERNEL);
3617 	if (!tmp)
3618 		return -ENOMEM;
3619 
3620 	mutex_lock(&mdev->state_lock);
3621 
3622 	memcpy(&new_prof, priv->prof, sizeof(struct mlx4_en_port_profile));
3623 	memcpy(&new_prof.hwtstamp_config, &ts_config, sizeof(ts_config));
3624 
3625 	err = mlx4_en_try_alloc_resources(priv, tmp, &new_prof, true);
3626 	if (err)
3627 		goto out;
3628 
3629 	if (priv->port_up) {
3630 		port_up = 1;
3631 		mlx4_en_stop_port(dev, 1);
3632 	}
3633 
3634 	mlx4_en_safe_replace_resources(priv, tmp);
3635 
3636 	if (DEV_FEATURE_CHANGED(dev, features, NETIF_F_HW_VLAN_CTAG_RX)) {
3637 		if (features & NETIF_F_HW_VLAN_CTAG_RX)
3638 			dev->features |= NETIF_F_HW_VLAN_CTAG_RX;
3639 		else
3640 			dev->features &= ~NETIF_F_HW_VLAN_CTAG_RX;
3641 	} else if (ts_config.rx_filter == HWTSTAMP_FILTER_NONE) {
3642 		/* RX time-stamping is OFF, update the RX vlan offload
3643 		 * to the latest wanted state
3644 		 */
3645 		if (dev->wanted_features & NETIF_F_HW_VLAN_CTAG_RX)
3646 			dev->features |= NETIF_F_HW_VLAN_CTAG_RX;
3647 		else
3648 			dev->features &= ~NETIF_F_HW_VLAN_CTAG_RX;
3649 	}
3650 
3651 	if (DEV_FEATURE_CHANGED(dev, features, NETIF_F_RXFCS)) {
3652 		if (features & NETIF_F_RXFCS)
3653 			dev->features |= NETIF_F_RXFCS;
3654 		else
3655 			dev->features &= ~NETIF_F_RXFCS;
3656 	}
3657 
3658 	/* RX vlan offload and RX time-stamping can't co-exist !
3659 	 * Regardless of the caller's choice,
3660 	 * Turn Off RX vlan offload in case of time-stamping is ON
3661 	 */
3662 	if (ts_config.rx_filter != HWTSTAMP_FILTER_NONE) {
3663 		if (dev->features & NETIF_F_HW_VLAN_CTAG_RX)
3664 			en_warn(priv, "Turning off RX vlan offload since RX time-stamping is ON\n");
3665 		dev->features &= ~NETIF_F_HW_VLAN_CTAG_RX;
3666 	}
3667 
3668 	if (port_up) {
3669 		err = mlx4_en_start_port(dev);
3670 		if (err)
3671 			en_err(priv, "Failed starting port\n");
3672 	}
3673 
3674 out:
3675 	mutex_unlock(&mdev->state_lock);
3676 	kfree(tmp);
3677 	if (!err)
3678 		netdev_features_change(dev);
3679 	return err;
3680 }
3681