xref: /freebsd/sys/dev/mlx4/mlx4_en/mlx4_en_netdev.c (revision 52c2bb75163559a6e2866ad374a7de67a4ea1273)
1 /*
2  * Copyright (c) 2007, 2014 Mellanox Technologies. All rights reserved.
3  *
4  * This software is available to you under a choice of one of two
5  * licenses.  You may choose to be licensed under the terms of the GNU
6  * General Public License (GPL) Version 2, available from the file
7  * COPYING in the main directory of this source tree, or the
8  * OpenIB.org BSD license below:
9  *
10  *     Redistribution and use in source and binary forms, with or
11  *     without modification, are permitted provided that the following
12  *     conditions are met:
13  *
14  *      - Redistributions of source code must retain the above
15  *        copyright notice, this list of conditions and the following
16  *        disclaimer.
17  *
18  *      - Redistributions in binary form must reproduce the above
19  *        copyright notice, this list of conditions and the following
20  *        disclaimer in the documentation and/or other materials
21  *        provided with the distribution.
22  *
23  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
24  * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
25  * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
26  * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
27  * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
28  * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
29  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
30  * SOFTWARE.
31  *
32  */
33 
34 #include <linux/etherdevice.h>
35 #include <linux/delay.h>
36 #include <linux/slab.h>
37 #include <linux/compat.h>
38 #ifdef CONFIG_NET_RX_BUSY_POLL
39 #include <net/busy_poll.h>
40 #endif
41 
42 #include <linux/list.h>
43 #include <linux/if_ether.h>
44 
45 #include <dev/mlx4/driver.h>
46 #include <dev/mlx4/device.h>
47 #include <dev/mlx4/cmd.h>
48 #include <dev/mlx4/cq.h>
49 
50 #include <sys/sockio.h>
51 #include <sys/sysctl.h>
52 
53 #include "en.h"
54 #include "en_port.h"
55 
56 NETDUMP_DEFINE(mlx4_en);
57 
58 static void mlx4_en_sysctl_stat(struct mlx4_en_priv *priv);
59 static void mlx4_en_sysctl_conf(struct mlx4_en_priv *priv);
60 
61 #ifdef CONFIG_NET_RX_BUSY_POLL
62 /* must be called with local_bh_disable()d */
63 static int mlx4_en_low_latency_recv(struct napi_struct *napi)
64 {
65 	struct mlx4_en_cq *cq = container_of(napi, struct mlx4_en_cq, napi);
66 	struct net_device *dev = cq->dev;
67 	struct mlx4_en_priv *priv = netdev_priv(dev);
68 	struct mlx4_en_rx_ring *rx_ring = priv->rx_ring[cq->ring];
69 	int done;
70 
71 	if (!priv->port_up)
72 		return LL_FLUSH_FAILED;
73 
74 	if (!mlx4_en_cq_lock_poll(cq))
75 		return LL_FLUSH_BUSY;
76 
77 	done = mlx4_en_process_rx_cq(dev, cq, 4);
78 #ifdef LL_EXTENDED_STATS
79 	if (likely(done))
80 		rx_ring->cleaned += done;
81 	else
82 		rx_ring->misses++;
83 #endif
84 
85 	mlx4_en_cq_unlock_poll(cq);
86 
87 	return done;
88 }
89 #endif	/* CONFIG_NET_RX_BUSY_POLL */
90 
91 #ifdef CONFIG_RFS_ACCEL
92 
93 struct mlx4_en_filter {
94 	struct list_head next;
95 	struct work_struct work;
96 
97 	u8     ip_proto;
98 	__be32 src_ip;
99 	__be32 dst_ip;
100 	__be16 src_port;
101 	__be16 dst_port;
102 
103 	int rxq_index;
104 	struct mlx4_en_priv *priv;
105 	u32 flow_id;			/* RFS infrastructure id */
106 	int id;				/* mlx4_en driver id */
107 	u64 reg_id;			/* Flow steering API id */
108 	u8 activated;			/* Used to prevent expiry before filter
109 					 * is attached
110 					 */
111 	struct hlist_node filter_chain;
112 };
113 
114 static void mlx4_en_filter_rfs_expire(struct mlx4_en_priv *priv);
115 
116 static enum mlx4_net_trans_rule_id mlx4_ip_proto_to_trans_rule_id(u8 ip_proto)
117 {
118 	switch (ip_proto) {
119 	case IPPROTO_UDP:
120 		return MLX4_NET_TRANS_RULE_ID_UDP;
121 	case IPPROTO_TCP:
122 		return MLX4_NET_TRANS_RULE_ID_TCP;
123 	default:
124 		return MLX4_NET_TRANS_RULE_NUM;
125 	}
126 };
127 
128 static void mlx4_en_filter_work(struct work_struct *work)
129 {
130 	struct mlx4_en_filter *filter = container_of(work,
131 						     struct mlx4_en_filter,
132 						     work);
133 	struct mlx4_en_priv *priv = filter->priv;
134 	struct mlx4_spec_list spec_tcp_udp = {
135 		.id = mlx4_ip_proto_to_trans_rule_id(filter->ip_proto),
136 		{
137 			.tcp_udp = {
138 				.dst_port = filter->dst_port,
139 				.dst_port_msk = (__force __be16)-1,
140 				.src_port = filter->src_port,
141 				.src_port_msk = (__force __be16)-1,
142 			},
143 		},
144 	};
145 	struct mlx4_spec_list spec_ip = {
146 		.id = MLX4_NET_TRANS_RULE_ID_IPV4,
147 		{
148 			.ipv4 = {
149 				.dst_ip = filter->dst_ip,
150 				.dst_ip_msk = (__force __be32)-1,
151 				.src_ip = filter->src_ip,
152 				.src_ip_msk = (__force __be32)-1,
153 			},
154 		},
155 	};
156 	struct mlx4_spec_list spec_eth = {
157 		.id = MLX4_NET_TRANS_RULE_ID_ETH,
158 	};
159 	struct mlx4_net_trans_rule rule = {
160 		.list = LIST_HEAD_INIT(rule.list),
161 		.queue_mode = MLX4_NET_TRANS_Q_LIFO,
162 		.exclusive = 1,
163 		.allow_loopback = 1,
164 		.promisc_mode = MLX4_FS_REGULAR,
165 		.port = priv->port,
166 		.priority = MLX4_DOMAIN_RFS,
167 	};
168 	int rc;
169 	__be64 mac_mask = cpu_to_be64(MLX4_MAC_MASK << 16);
170 
171 	if (spec_tcp_udp.id >= MLX4_NET_TRANS_RULE_NUM) {
172 		en_warn(priv, "RFS: ignoring unsupported ip protocol (%d)\n",
173 			filter->ip_proto);
174 		goto ignore;
175 	}
176 	list_add_tail(&spec_eth.list, &rule.list);
177 	list_add_tail(&spec_ip.list, &rule.list);
178 	list_add_tail(&spec_tcp_udp.list, &rule.list);
179 
180 	rule.qpn = priv->rss_map.qps[filter->rxq_index].qpn;
181 	memcpy(spec_eth.eth.dst_mac, priv->dev->dev_addr, ETH_ALEN);
182 	memcpy(spec_eth.eth.dst_mac_msk, &mac_mask, ETH_ALEN);
183 
184 	filter->activated = 0;
185 
186 	if (filter->reg_id) {
187 		rc = mlx4_flow_detach(priv->mdev->dev, filter->reg_id);
188 		if (rc && rc != -ENOENT)
189 			en_err(priv, "Error detaching flow. rc = %d\n", rc);
190 	}
191 
192 	rc = mlx4_flow_attach(priv->mdev->dev, &rule, &filter->reg_id);
193 	if (rc)
194 		en_err(priv, "Error attaching flow. err = %d\n", rc);
195 
196 ignore:
197 	mlx4_en_filter_rfs_expire(priv);
198 
199 	filter->activated = 1;
200 }
201 
202 static inline struct hlist_head *
203 filter_hash_bucket(struct mlx4_en_priv *priv, __be32 src_ip, __be32 dst_ip,
204 		   __be16 src_port, __be16 dst_port)
205 {
206 	unsigned long l;
207 	int bucket_idx;
208 
209 	l = (__force unsigned long)src_port |
210 	    ((__force unsigned long)dst_port << 2);
211 	l ^= (__force unsigned long)(src_ip ^ dst_ip);
212 
213 	bucket_idx = hash_long(l, MLX4_EN_FILTER_HASH_SHIFT);
214 
215 	return &priv->filter_hash[bucket_idx];
216 }
217 
218 static struct mlx4_en_filter *
219 mlx4_en_filter_alloc(struct mlx4_en_priv *priv, int rxq_index, __be32 src_ip,
220 		     __be32 dst_ip, u8 ip_proto, __be16 src_port,
221 		     __be16 dst_port, u32 flow_id)
222 {
223 	struct mlx4_en_filter *filter = NULL;
224 
225 	filter = kzalloc(sizeof(struct mlx4_en_filter), GFP_ATOMIC);
226 	if (!filter)
227 		return NULL;
228 
229 	filter->priv = priv;
230 	filter->rxq_index = rxq_index;
231 	INIT_WORK(&filter->work, mlx4_en_filter_work);
232 
233 	filter->src_ip = src_ip;
234 	filter->dst_ip = dst_ip;
235 	filter->ip_proto = ip_proto;
236 	filter->src_port = src_port;
237 	filter->dst_port = dst_port;
238 
239 	filter->flow_id = flow_id;
240 
241 	filter->id = priv->last_filter_id++ % RPS_NO_FILTER;
242 
243 	list_add_tail(&filter->next, &priv->filters);
244 	hlist_add_head(&filter->filter_chain,
245 		       filter_hash_bucket(priv, src_ip, dst_ip, src_port,
246 					  dst_port));
247 
248 	return filter;
249 }
250 
251 static void mlx4_en_filter_free(struct mlx4_en_filter *filter)
252 {
253 	struct mlx4_en_priv *priv = filter->priv;
254 	int rc;
255 
256 	list_del(&filter->next);
257 
258 	rc = mlx4_flow_detach(priv->mdev->dev, filter->reg_id);
259 	if (rc && rc != -ENOENT)
260 		en_err(priv, "Error detaching flow. rc = %d\n", rc);
261 
262 	kfree(filter);
263 }
264 
265 static inline struct mlx4_en_filter *
266 mlx4_en_filter_find(struct mlx4_en_priv *priv, __be32 src_ip, __be32 dst_ip,
267 		    u8 ip_proto, __be16 src_port, __be16 dst_port)
268 {
269 	struct mlx4_en_filter *filter;
270 	struct mlx4_en_filter *ret = NULL;
271 
272 	hlist_for_each_entry(filter,
273 			     filter_hash_bucket(priv, src_ip, dst_ip,
274 						src_port, dst_port),
275 			     filter_chain) {
276 		if (filter->src_ip == src_ip &&
277 		    filter->dst_ip == dst_ip &&
278 		    filter->ip_proto == ip_proto &&
279 		    filter->src_port == src_port &&
280 		    filter->dst_port == dst_port) {
281 			ret = filter;
282 			break;
283 		}
284 	}
285 
286 	return ret;
287 }
288 
289 static int
290 mlx4_en_filter_rfs(struct net_device *net_dev, const struct sk_buff *skb,
291 		   u16 rxq_index, u32 flow_id)
292 {
293 	struct mlx4_en_priv *priv = netdev_priv(net_dev);
294 	struct mlx4_en_filter *filter;
295 	const struct iphdr *ip;
296 	const __be16 *ports;
297 	u8 ip_proto;
298 	__be32 src_ip;
299 	__be32 dst_ip;
300 	__be16 src_port;
301 	__be16 dst_port;
302 	int nhoff = skb_network_offset(skb);
303 	int ret = 0;
304 
305 	if (skb->protocol != htons(ETH_P_IP))
306 		return -EPROTONOSUPPORT;
307 
308 	ip = (const struct iphdr *)(skb->data + nhoff);
309 	if (ip_is_fragment(ip))
310 		return -EPROTONOSUPPORT;
311 
312 	if ((ip->protocol != IPPROTO_TCP) && (ip->protocol != IPPROTO_UDP))
313 		return -EPROTONOSUPPORT;
314 	ports = (const __be16 *)(skb->data + nhoff + 4 * ip->ihl);
315 
316 	ip_proto = ip->protocol;
317 	src_ip = ip->saddr;
318 	dst_ip = ip->daddr;
319 	src_port = ports[0];
320 	dst_port = ports[1];
321 
322 	spin_lock_bh(&priv->filters_lock);
323 	filter = mlx4_en_filter_find(priv, src_ip, dst_ip, ip_proto,
324 				     src_port, dst_port);
325 	if (filter) {
326 		if (filter->rxq_index == rxq_index)
327 			goto out;
328 
329 		filter->rxq_index = rxq_index;
330 	} else {
331 		filter = mlx4_en_filter_alloc(priv, rxq_index,
332 					      src_ip, dst_ip, ip_proto,
333 					      src_port, dst_port, flow_id);
334 		if (!filter) {
335 			ret = -ENOMEM;
336 			goto err;
337 		}
338 	}
339 
340 	queue_work(priv->mdev->workqueue, &filter->work);
341 
342 out:
343 	ret = filter->id;
344 err:
345 	spin_unlock_bh(&priv->filters_lock);
346 
347 	return ret;
348 }
349 
350 void mlx4_en_cleanup_filters(struct mlx4_en_priv *priv)
351 {
352 	struct mlx4_en_filter *filter, *tmp;
353 	LIST_HEAD(del_list);
354 
355 	spin_lock_bh(&priv->filters_lock);
356 	list_for_each_entry_safe(filter, tmp, &priv->filters, next) {
357 		list_move(&filter->next, &del_list);
358 		hlist_del(&filter->filter_chain);
359 	}
360 	spin_unlock_bh(&priv->filters_lock);
361 
362 	list_for_each_entry_safe(filter, tmp, &del_list, next) {
363 		cancel_work_sync(&filter->work);
364 		mlx4_en_filter_free(filter);
365 	}
366 }
367 
368 static void mlx4_en_filter_rfs_expire(struct mlx4_en_priv *priv)
369 {
370 	struct mlx4_en_filter *filter = NULL, *tmp, *last_filter = NULL;
371 	LIST_HEAD(del_list);
372 	int i = 0;
373 
374 	spin_lock_bh(&priv->filters_lock);
375 	list_for_each_entry_safe(filter, tmp, &priv->filters, next) {
376 		if (i > MLX4_EN_FILTER_EXPIRY_QUOTA)
377 			break;
378 
379 		if (filter->activated &&
380 		    !work_pending(&filter->work) &&
381 		    rps_may_expire_flow(priv->dev,
382 					filter->rxq_index, filter->flow_id,
383 					filter->id)) {
384 			list_move(&filter->next, &del_list);
385 			hlist_del(&filter->filter_chain);
386 		} else
387 			last_filter = filter;
388 
389 		i++;
390 	}
391 
392 	if (last_filter && (&last_filter->next != priv->filters.next))
393 		list_move(&priv->filters, &last_filter->next);
394 
395 	spin_unlock_bh(&priv->filters_lock);
396 
397 	list_for_each_entry_safe(filter, tmp, &del_list, next)
398 		mlx4_en_filter_free(filter);
399 }
400 #endif
401 
402 static void mlx4_en_vlan_rx_add_vid(void *arg, struct net_device *dev, u16 vid)
403 {
404 	struct mlx4_en_priv *priv = netdev_priv(dev);
405 	struct mlx4_en_dev *mdev = priv->mdev;
406 	int err;
407 	int idx;
408 
409 	if (arg != priv)
410 		return;
411 
412 	en_dbg(HW, priv, "adding VLAN:%d\n", vid);
413 
414 	set_bit(vid, priv->active_vlans);
415 
416 	/* Add VID to port VLAN filter */
417 	mutex_lock(&mdev->state_lock);
418 	if (mdev->device_up && priv->port_up) {
419 		err = mlx4_SET_VLAN_FLTR(mdev->dev, priv);
420 		if (err)
421 			en_err(priv, "Failed configuring VLAN filter\n");
422 	}
423 	if (mlx4_register_vlan(mdev->dev, priv->port, vid, &idx))
424 		en_dbg(HW, priv, "failed adding vlan %d\n", vid);
425 	mutex_unlock(&mdev->state_lock);
426 
427 }
428 
429 static void mlx4_en_vlan_rx_kill_vid(void *arg, struct net_device *dev, u16 vid)
430 {
431 	struct mlx4_en_priv *priv = netdev_priv(dev);
432 	struct mlx4_en_dev *mdev = priv->mdev;
433 	int err;
434 
435 	if (arg != priv)
436 		return;
437 
438 	en_dbg(HW, priv, "Killing VID:%d\n", vid);
439 
440 	clear_bit(vid, priv->active_vlans);
441 
442 	/* Remove VID from port VLAN filter */
443 	mutex_lock(&mdev->state_lock);
444 	mlx4_unregister_vlan(mdev->dev, priv->port, vid);
445 
446 	if (mdev->device_up && priv->port_up) {
447 		err = mlx4_SET_VLAN_FLTR(mdev->dev, priv);
448 		if (err)
449 			en_err(priv, "Failed configuring VLAN filter\n");
450 	}
451 	mutex_unlock(&mdev->state_lock);
452 
453 }
454 
455 static int mlx4_en_tunnel_steer_add(struct mlx4_en_priv *priv, unsigned char *addr,
456 				    int qpn, u64 *reg_id)
457 {
458 	int err;
459 
460 	if (priv->mdev->dev->caps.tunnel_offload_mode != MLX4_TUNNEL_OFFLOAD_MODE_VXLAN ||
461 	    priv->mdev->dev->caps.dmfs_high_steer_mode == MLX4_STEERING_DMFS_A0_STATIC)
462 		return 0; /* do nothing */
463 
464 	err = mlx4_tunnel_steer_add(priv->mdev->dev, addr, priv->port, qpn,
465 				    MLX4_DOMAIN_NIC, reg_id);
466 	if (err) {
467 		en_err(priv, "failed to add vxlan steering rule, err %d\n", err);
468 		return err;
469 	}
470 	en_dbg(DRV, priv, "added vxlan steering rule, mac %pM reg_id %llx\n", addr, (long long)*reg_id);
471 	return 0;
472 }
473 
474 static int mlx4_en_uc_steer_add(struct mlx4_en_priv *priv,
475 				unsigned char *mac, int *qpn, u64 *reg_id)
476 {
477 	struct mlx4_en_dev *mdev = priv->mdev;
478 	struct mlx4_dev *dev = mdev->dev;
479 	int err;
480 
481 	switch (dev->caps.steering_mode) {
482 	case MLX4_STEERING_MODE_B0: {
483 		struct mlx4_qp qp;
484 		u8 gid[16] = {0};
485 
486 		qp.qpn = *qpn;
487 		memcpy(&gid[10], mac, ETH_ALEN);
488 		gid[5] = priv->port;
489 
490 		err = mlx4_unicast_attach(dev, &qp, gid, 0, MLX4_PROT_ETH);
491 		break;
492 	}
493 	case MLX4_STEERING_MODE_DEVICE_MANAGED: {
494 		struct mlx4_spec_list spec_eth = { {NULL} };
495 		__be64 mac_mask = cpu_to_be64(MLX4_MAC_MASK << 16);
496 
497 		struct mlx4_net_trans_rule rule = {
498 			.queue_mode = MLX4_NET_TRANS_Q_FIFO,
499 			.exclusive = 0,
500 			.allow_loopback = 1,
501 			.promisc_mode = MLX4_FS_REGULAR,
502 			.priority = MLX4_DOMAIN_NIC,
503 		};
504 
505 		rule.port = priv->port;
506 		rule.qpn = *qpn;
507 		INIT_LIST_HEAD(&rule.list);
508 
509 		spec_eth.id = MLX4_NET_TRANS_RULE_ID_ETH;
510 		memcpy(spec_eth.eth.dst_mac, mac, ETH_ALEN);
511 		memcpy(spec_eth.eth.dst_mac_msk, &mac_mask, ETH_ALEN);
512 		list_add_tail(&spec_eth.list, &rule.list);
513 
514 		err = mlx4_flow_attach(dev, &rule, reg_id);
515 		break;
516 	}
517 	default:
518 		return -EINVAL;
519 	}
520 	if (err)
521 		en_warn(priv, "Failed Attaching Unicast\n");
522 
523 	return err;
524 }
525 
526 static void mlx4_en_uc_steer_release(struct mlx4_en_priv *priv,
527 				     unsigned char *mac, int qpn, u64 reg_id)
528 {
529 	struct mlx4_en_dev *mdev = priv->mdev;
530 	struct mlx4_dev *dev = mdev->dev;
531 
532 	switch (dev->caps.steering_mode) {
533 	case MLX4_STEERING_MODE_B0: {
534 		struct mlx4_qp qp;
535 		u8 gid[16] = {0};
536 
537 		qp.qpn = qpn;
538 		memcpy(&gid[10], mac, ETH_ALEN);
539 		gid[5] = priv->port;
540 
541 		mlx4_unicast_detach(dev, &qp, gid, MLX4_PROT_ETH);
542 		break;
543 	}
544 	case MLX4_STEERING_MODE_DEVICE_MANAGED: {
545 		mlx4_flow_detach(dev, reg_id);
546 		break;
547 	}
548 	default:
549 		en_err(priv, "Invalid steering mode.\n");
550 	}
551 }
552 
553 static int mlx4_en_get_qp(struct mlx4_en_priv *priv)
554 {
555 	struct mlx4_en_dev *mdev = priv->mdev;
556 	struct mlx4_dev *dev = mdev->dev;
557 	int index = 0;
558 	int err = 0;
559 	int *qpn = &priv->base_qpn;
560 	u64 mac = mlx4_mac_to_u64(IF_LLADDR(priv->dev));
561 
562 	en_dbg(DRV, priv, "Registering MAC: %pM for adding\n",
563 	       IF_LLADDR(priv->dev));
564 	index = mlx4_register_mac(dev, priv->port, mac);
565 	if (index < 0) {
566 		err = index;
567 		en_err(priv, "Failed adding MAC: %pM\n",
568 		       IF_LLADDR(priv->dev));
569 		return err;
570 	}
571 
572 	if (dev->caps.steering_mode == MLX4_STEERING_MODE_A0) {
573 		int base_qpn = mlx4_get_base_qpn(dev, priv->port);
574 		*qpn = base_qpn + index;
575 		return 0;
576 	}
577 
578 	err = mlx4_qp_reserve_range(dev, 1, 1, qpn, MLX4_RESERVE_A0_QP);
579 	en_dbg(DRV, priv, "Reserved qp %d\n", *qpn);
580 	if (err) {
581 		en_err(priv, "Failed to reserve qp for mac registration\n");
582 		mlx4_unregister_mac(dev, priv->port, mac);
583 		return err;
584 	}
585 
586 	return 0;
587 }
588 
589 static void mlx4_en_put_qp(struct mlx4_en_priv *priv)
590 {
591 	struct mlx4_en_dev *mdev = priv->mdev;
592 	struct mlx4_dev *dev = mdev->dev;
593 	int qpn = priv->base_qpn;
594 
595 	if (dev->caps.steering_mode == MLX4_STEERING_MODE_A0) {
596 		u64 mac = mlx4_mac_to_u64(IF_LLADDR(priv->dev));
597 		en_dbg(DRV, priv, "Registering MAC: %pM for deleting\n",
598 		       IF_LLADDR(priv->dev));
599 		mlx4_unregister_mac(dev, priv->port, mac);
600 	} else {
601 		en_dbg(DRV, priv, "Releasing qp: port %d, qpn %d\n",
602 		       priv->port, qpn);
603 		mlx4_qp_release_range(dev, qpn, 1);
604 		priv->flags &= ~MLX4_EN_FLAG_FORCE_PROMISC;
605 	}
606 }
607 
608 static void mlx4_en_clear_uclist(struct net_device *dev)
609 {
610 	struct mlx4_en_priv *priv = netdev_priv(dev);
611 	struct mlx4_en_addr_list *tmp, *uc_to_del;
612 
613 	list_for_each_entry_safe(uc_to_del, tmp, &priv->uc_list, list) {
614 		list_del(&uc_to_del->list);
615 		kfree(uc_to_del);
616 	}
617 }
618 
619 static void mlx4_en_cache_uclist(struct net_device *dev)
620 {
621 	struct mlx4_en_priv *priv = netdev_priv(dev);
622 	struct mlx4_en_addr_list *tmp;
623 	struct ifaddr *ifa;
624 
625 	mlx4_en_clear_uclist(dev);
626 
627 	if_addr_rlock(dev);
628 	CK_STAILQ_FOREACH(ifa, &dev->if_addrhead, ifa_link) {
629 		if (ifa->ifa_addr->sa_family != AF_LINK)
630 			continue;
631 		if (((struct sockaddr_dl *)ifa->ifa_addr)->sdl_alen !=
632 				ETHER_ADDR_LEN)
633 			continue;
634 		tmp = kzalloc(sizeof(struct mlx4_en_addr_list), GFP_ATOMIC);
635 		if (tmp == NULL) {
636 			en_err(priv, "Failed to allocate address list\n");
637 			break;
638 		}
639 		memcpy(tmp->addr,
640 			LLADDR((struct sockaddr_dl *)ifa->ifa_addr), ETH_ALEN);
641 		list_add_tail(&tmp->list, &priv->uc_list);
642 	}
643 	if_addr_runlock(dev);
644 }
645 
646 static void mlx4_en_clear_mclist(struct net_device *dev)
647 {
648 	struct mlx4_en_priv *priv = netdev_priv(dev);
649 	struct mlx4_en_addr_list *tmp, *mc_to_del;
650 
651 	list_for_each_entry_safe(mc_to_del, tmp, &priv->mc_list, list) {
652 		list_del(&mc_to_del->list);
653 		kfree(mc_to_del);
654 	}
655 }
656 
657 static void mlx4_en_cache_mclist(struct net_device *dev)
658 {
659 	struct mlx4_en_priv *priv = netdev_priv(dev);
660 	struct mlx4_en_addr_list *tmp;
661 	struct ifmultiaddr *ifma;
662 
663 	mlx4_en_clear_mclist(dev);
664 
665 	if_maddr_rlock(dev);
666 	CK_STAILQ_FOREACH(ifma, &dev->if_multiaddrs, ifma_link) {
667 		if (ifma->ifma_addr->sa_family != AF_LINK)
668 			continue;
669 		if (((struct sockaddr_dl *)ifma->ifma_addr)->sdl_alen !=
670 				ETHER_ADDR_LEN)
671 			continue;
672 		tmp = kzalloc(sizeof(struct mlx4_en_addr_list), GFP_ATOMIC);
673 		if (tmp == NULL) {
674 			en_err(priv, "Failed to allocate address list\n");
675 			break;
676 		}
677 		memcpy(tmp->addr,
678 			LLADDR((struct sockaddr_dl *)ifma->ifma_addr), ETH_ALEN);
679 		list_add_tail(&tmp->list, &priv->mc_list);
680 	}
681 	if_maddr_runlock(dev);
682 }
683 
684 static void update_addr_list_flags(struct mlx4_en_priv *priv,
685 				struct list_head *dst,
686 				struct list_head *src)
687 {
688 	struct mlx4_en_addr_list *dst_tmp, *src_tmp, *new_mc;
689 	bool found;
690 
691 	/* Find all the entries that should be removed from dst,
692 	 * These are the entries that are not found in src
693 	 */
694 	list_for_each_entry(dst_tmp, dst, list) {
695 		found = false;
696 		list_for_each_entry(src_tmp, src, list) {
697 			if (!memcmp(dst_tmp->addr, src_tmp->addr, ETH_ALEN)) {
698 				found = true;
699 				break;
700 			}
701 		}
702 		if (!found)
703 			dst_tmp->action = MLX4_ADDR_LIST_REM;
704 	}
705 
706 	/* Add entries that exist in src but not in dst
707 	 * mark them as need to add
708 	 */
709 	list_for_each_entry(src_tmp, src, list) {
710 		found = false;
711 		list_for_each_entry(dst_tmp, dst, list) {
712 			if (!memcmp(dst_tmp->addr, src_tmp->addr, ETH_ALEN)) {
713 				dst_tmp->action = MLX4_ADDR_LIST_NONE;
714 				found = true;
715 				break;
716 			}
717 		}
718 		if (!found) {
719 			new_mc = kmalloc(sizeof(struct mlx4_en_addr_list),
720 					 GFP_KERNEL);
721 			if (!new_mc) {
722 				en_err(priv, "Failed to allocate current multicast list\n");
723 				return;
724 			}
725 			memcpy(new_mc, src_tmp,
726 			       sizeof(struct mlx4_en_addr_list));
727 			new_mc->action = MLX4_ADDR_LIST_ADD;
728 			list_add_tail(&new_mc->list, dst);
729 		}
730 	}
731 }
732 
733 static void mlx4_en_set_rx_mode(struct net_device *dev)
734 {
735 	struct mlx4_en_priv *priv = netdev_priv(dev);
736 
737 	if (!priv->port_up)
738 		return;
739 
740 	queue_work(priv->mdev->workqueue, &priv->rx_mode_task);
741 }
742 
743 static void mlx4_en_set_promisc_mode(struct mlx4_en_priv *priv,
744 				     struct mlx4_en_dev *mdev)
745 {
746 	int err = 0;
747 
748 	if (!(priv->flags & MLX4_EN_FLAG_PROMISC)) {
749 		priv->flags |= MLX4_EN_FLAG_PROMISC;
750 
751 		/* Enable promiscouos mode */
752 		switch (mdev->dev->caps.steering_mode) {
753 		case MLX4_STEERING_MODE_DEVICE_MANAGED:
754 			err = mlx4_flow_steer_promisc_add(mdev->dev,
755 							  priv->port,
756 							  priv->base_qpn,
757 							  MLX4_FS_ALL_DEFAULT);
758 			if (err)
759 				en_err(priv, "Failed enabling promiscuous mode\n");
760 			priv->flags |= MLX4_EN_FLAG_MC_PROMISC;
761 			break;
762 
763 		case MLX4_STEERING_MODE_B0:
764 			err = mlx4_unicast_promisc_add(mdev->dev,
765 						       priv->base_qpn,
766 						       priv->port);
767 			if (err)
768 				en_err(priv, "Failed enabling unicast promiscuous mode\n");
769 
770 			/* Add the default qp number as multicast
771 			 * promisc
772 			 */
773 			if (!(priv->flags & MLX4_EN_FLAG_MC_PROMISC)) {
774 				err = mlx4_multicast_promisc_add(mdev->dev,
775 								 priv->base_qpn,
776 								 priv->port);
777 				if (err)
778 					en_err(priv, "Failed enabling multicast promiscuous mode\n");
779 				priv->flags |= MLX4_EN_FLAG_MC_PROMISC;
780 			}
781 			break;
782 
783 		case MLX4_STEERING_MODE_A0:
784 			err = mlx4_SET_PORT_qpn_calc(mdev->dev,
785 						     priv->port,
786 						     priv->base_qpn,
787 						     1);
788 			if (err)
789 				en_err(priv, "Failed enabling promiscuous mode\n");
790 			break;
791 		}
792 
793 		/* Disable port multicast filter (unconditionally) */
794 		err = mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, 0,
795 					  0, MLX4_MCAST_DISABLE);
796 		if (err)
797 			en_err(priv, "Failed disabling multicast filter\n");
798 	}
799 }
800 
801 static void mlx4_en_clear_promisc_mode(struct mlx4_en_priv *priv,
802 				       struct mlx4_en_dev *mdev)
803 {
804 	int err = 0;
805 
806 	priv->flags &= ~MLX4_EN_FLAG_PROMISC;
807 
808 	/* Disable promiscouos mode */
809 	switch (mdev->dev->caps.steering_mode) {
810 	case MLX4_STEERING_MODE_DEVICE_MANAGED:
811 		err = mlx4_flow_steer_promisc_remove(mdev->dev,
812 						     priv->port,
813 						     MLX4_FS_ALL_DEFAULT);
814 		if (err)
815 			en_err(priv, "Failed disabling promiscuous mode\n");
816 		priv->flags &= ~MLX4_EN_FLAG_MC_PROMISC;
817 		break;
818 
819 	case MLX4_STEERING_MODE_B0:
820 		err = mlx4_unicast_promisc_remove(mdev->dev,
821 						  priv->base_qpn,
822 						  priv->port);
823 		if (err)
824 			en_err(priv, "Failed disabling unicast promiscuous mode\n");
825 		/* Disable Multicast promisc */
826 		if (priv->flags & MLX4_EN_FLAG_MC_PROMISC) {
827 			err = mlx4_multicast_promisc_remove(mdev->dev,
828 							    priv->base_qpn,
829 							    priv->port);
830 			if (err)
831 				en_err(priv, "Failed disabling multicast promiscuous mode\n");
832 			priv->flags &= ~MLX4_EN_FLAG_MC_PROMISC;
833 		}
834 		break;
835 
836 	case MLX4_STEERING_MODE_A0:
837 		err = mlx4_SET_PORT_qpn_calc(mdev->dev,
838 					     priv->port,
839 					     priv->base_qpn, 0);
840 		if (err)
841 			en_err(priv, "Failed disabling promiscuous mode\n");
842 		break;
843 	}
844 }
845 
846 static void mlx4_en_do_multicast(struct mlx4_en_priv *priv,
847 				 struct net_device *dev,
848 				 struct mlx4_en_dev *mdev)
849 {
850 	struct mlx4_en_addr_list *addr_list, *tmp;
851 	u8 mc_list[16] = {0};
852 	int err = 0;
853 	u64 mcast_addr = 0;
854 
855 
856 	/* Enable/disable the multicast filter according to IFF_ALLMULTI */
857 	if (dev->if_flags & IFF_ALLMULTI) {
858 		err = mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, 0,
859 					  0, MLX4_MCAST_DISABLE);
860 		if (err)
861 			en_err(priv, "Failed disabling multicast filter\n");
862 
863 		/* Add the default qp number as multicast promisc */
864 		if (!(priv->flags & MLX4_EN_FLAG_MC_PROMISC)) {
865 			switch (mdev->dev->caps.steering_mode) {
866 			case MLX4_STEERING_MODE_DEVICE_MANAGED:
867 				err = mlx4_flow_steer_promisc_add(mdev->dev,
868 								  priv->port,
869 								  priv->base_qpn,
870 								  MLX4_FS_MC_DEFAULT);
871 				break;
872 
873 			case MLX4_STEERING_MODE_B0:
874 				err = mlx4_multicast_promisc_add(mdev->dev,
875 								 priv->base_qpn,
876 								 priv->port);
877 				break;
878 
879 			case MLX4_STEERING_MODE_A0:
880 				break;
881 			}
882 			if (err)
883 				en_err(priv, "Failed entering multicast promisc mode\n");
884 			priv->flags |= MLX4_EN_FLAG_MC_PROMISC;
885 		}
886 	} else {
887 		/* Disable Multicast promisc */
888 		if (priv->flags & MLX4_EN_FLAG_MC_PROMISC) {
889 			switch (mdev->dev->caps.steering_mode) {
890 			case MLX4_STEERING_MODE_DEVICE_MANAGED:
891 				err = mlx4_flow_steer_promisc_remove(mdev->dev,
892 								     priv->port,
893 								     MLX4_FS_MC_DEFAULT);
894 				break;
895 
896 			case MLX4_STEERING_MODE_B0:
897 				err = mlx4_multicast_promisc_remove(mdev->dev,
898 								    priv->base_qpn,
899 								    priv->port);
900 				break;
901 
902 			case MLX4_STEERING_MODE_A0:
903 				break;
904 			}
905 			if (err)
906 				en_err(priv, "Failed disabling multicast promiscuous mode\n");
907 			priv->flags &= ~MLX4_EN_FLAG_MC_PROMISC;
908 		}
909 
910 		err = mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, 0,
911 					  0, MLX4_MCAST_DISABLE);
912 		if (err)
913 			en_err(priv, "Failed disabling multicast filter\n");
914 
915 		/* Flush mcast filter and init it with broadcast address */
916 		mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, ETH_BCAST,
917 				    1, MLX4_MCAST_CONFIG);
918 
919 		/* Update multicast list - we cache all addresses so they won't
920 		 * change while HW is updated holding the command semaphor */
921 		mlx4_en_cache_mclist(dev);
922 		list_for_each_entry(addr_list, &priv->mc_list, list) {
923 			mcast_addr = mlx4_mac_to_u64(addr_list->addr);
924 			mlx4_SET_MCAST_FLTR(mdev->dev, priv->port,
925 					mcast_addr, 0, MLX4_MCAST_CONFIG);
926 		}
927 		err = mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, 0,
928 					  0, MLX4_MCAST_ENABLE);
929 		if (err)
930 			en_err(priv, "Failed enabling multicast filter\n");
931 
932 		update_addr_list_flags(priv, &priv->curr_mc_list, &priv->mc_list);
933 
934 		list_for_each_entry_safe(addr_list, tmp, &priv->curr_mc_list, list) {
935 			if (addr_list->action == MLX4_ADDR_LIST_REM) {
936 				/* detach this address and delete from list */
937 				memcpy(&mc_list[10], addr_list->addr, ETH_ALEN);
938 				mc_list[5] = priv->port;
939 				err = mlx4_multicast_detach(mdev->dev,
940 							    &priv->rss_map.indir_qp,
941 							    mc_list,
942 							    MLX4_PROT_ETH,
943 							    addr_list->reg_id);
944 				if (err)
945 					en_err(priv, "Fail to detach multicast address\n");
946 
947 				if (addr_list->tunnel_reg_id) {
948 					err = mlx4_flow_detach(priv->mdev->dev, addr_list->tunnel_reg_id);
949 					if (err)
950 						en_err(priv, "Failed to detach multicast address\n");
951 				}
952 
953 				/* remove from list */
954 				list_del(&addr_list->list);
955 				kfree(addr_list);
956 			} else if (addr_list->action == MLX4_ADDR_LIST_ADD) {
957 				/* attach the address */
958 				memcpy(&mc_list[10], addr_list->addr, ETH_ALEN);
959 				/* needed for B0 steering support */
960 				mc_list[5] = priv->port;
961 				err = mlx4_multicast_attach(mdev->dev,
962 							    &priv->rss_map.indir_qp,
963 							    mc_list,
964 							    priv->port, 0,
965 							    MLX4_PROT_ETH,
966 							    &addr_list->reg_id);
967 				if (err)
968 					en_err(priv, "Fail to attach multicast address\n");
969 
970 				err = mlx4_en_tunnel_steer_add(priv, &mc_list[10], priv->base_qpn,
971 							       &addr_list->tunnel_reg_id);
972 				if (err)
973 					en_err(priv, "Failed to attach multicast address\n");
974 			}
975 		}
976 	}
977 }
978 
979 static void mlx4_en_do_unicast(struct mlx4_en_priv *priv,
980 			       struct net_device *dev,
981 			       struct mlx4_en_dev *mdev)
982 {
983 	struct mlx4_en_addr_list *addr_list, *tmp;
984 	int err;
985 
986 	/* Update unicast list */
987 	mlx4_en_cache_uclist(dev);
988 
989 	update_addr_list_flags(priv, &priv->curr_uc_list, &priv->uc_list);
990 
991 	list_for_each_entry_safe(addr_list, tmp, &priv->curr_uc_list, list) {
992 		if (addr_list->action == MLX4_ADDR_LIST_REM) {
993 			mlx4_en_uc_steer_release(priv, addr_list->addr,
994 						 priv->rss_map.indir_qp.qpn,
995 						 addr_list->reg_id);
996 			/* remove from list */
997 			list_del(&addr_list->list);
998 			kfree(addr_list);
999 		} else if (addr_list->action == MLX4_ADDR_LIST_ADD) {
1000 			err = mlx4_en_uc_steer_add(priv, addr_list->addr,
1001 						   &priv->rss_map.indir_qp.qpn,
1002 						   &addr_list->reg_id);
1003 			if (err)
1004 				en_err(priv, "Fail to add unicast address\n");
1005 		}
1006 	}
1007 }
1008 
1009 static void mlx4_en_do_set_rx_mode(struct work_struct *work)
1010 {
1011 	struct mlx4_en_priv *priv = container_of(work, struct mlx4_en_priv,
1012 						 rx_mode_task);
1013 	struct mlx4_en_dev *mdev = priv->mdev;
1014 	struct net_device *dev = priv->dev;
1015 
1016 	mutex_lock(&mdev->state_lock);
1017 	if (!mdev->device_up) {
1018 		en_dbg(HW, priv, "Card is not up, ignoring rx mode change.\n");
1019 		goto out;
1020 	}
1021 	if (!priv->port_up) {
1022 		en_dbg(HW, priv, "Port is down, ignoring rx mode change.\n");
1023 		goto out;
1024 	}
1025 	if (!mlx4_en_QUERY_PORT(mdev, priv->port)) {
1026 		if (priv->port_state.link_state) {
1027 			priv->last_link_state = MLX4_DEV_EVENT_PORT_UP;
1028 			/* update netif baudrate */
1029 			priv->dev->if_baudrate =
1030 			    IF_Mbps(priv->port_state.link_speed);
1031 			/* Important note: the following call for if_link_state_change
1032 			 * is needed for interface up scenario (start port, link state
1033 			 * change) */
1034 			if_link_state_change(priv->dev, LINK_STATE_UP);
1035 			en_dbg(HW, priv, "Link Up\n");
1036 		}
1037 	}
1038 
1039 	/* Set unicast rules */
1040 	mlx4_en_do_unicast(priv, dev, mdev);
1041 
1042 	/* Promsicuous mode: disable all filters */
1043 	if ((dev->if_flags & IFF_PROMISC) ||
1044 	    (priv->flags & MLX4_EN_FLAG_FORCE_PROMISC)) {
1045 		mlx4_en_set_promisc_mode(priv, mdev);
1046 	} else if (priv->flags & MLX4_EN_FLAG_PROMISC) {
1047 		/* Not in promiscuous mode */
1048 		mlx4_en_clear_promisc_mode(priv, mdev);
1049 	}
1050 
1051 	/* Set multicast rules */
1052 	mlx4_en_do_multicast(priv, dev, mdev);
1053 out:
1054 	mutex_unlock(&mdev->state_lock);
1055 }
1056 
1057 static void mlx4_en_watchdog_timeout(void *arg)
1058 {
1059         struct mlx4_en_priv *priv = arg;
1060         struct mlx4_en_dev *mdev = priv->mdev;
1061 
1062         en_dbg(DRV, priv, "Scheduling watchdog\n");
1063         queue_work(mdev->workqueue, &priv->watchdog_task);
1064         if (priv->port_up)
1065                 callout_reset(&priv->watchdog_timer, MLX4_EN_WATCHDOG_TIMEOUT,
1066                                 mlx4_en_watchdog_timeout, priv);
1067 }
1068 
1069 
1070 
1071 static void mlx4_en_set_default_moderation(struct mlx4_en_priv *priv)
1072 {
1073 	struct mlx4_en_cq *cq;
1074 	int i;
1075 
1076 	/* If we haven't received a specific coalescing setting
1077 	 * (module param), we set the moderation parameters as follows:
1078 	 * - moder_cnt is set to the number of mtu sized packets to
1079 	 *   satisfy our coalescing target.
1080 	 * - moder_time is set to a fixed value.
1081 	 */
1082 	priv->rx_frames = MLX4_EN_RX_COAL_TARGET;
1083 	priv->rx_usecs = MLX4_EN_RX_COAL_TIME;
1084 	priv->tx_frames = MLX4_EN_TX_COAL_PKTS;
1085 	priv->tx_usecs = MLX4_EN_TX_COAL_TIME;
1086 	en_dbg(INTR, priv, "Default coalesing params for mtu: %u - "
1087 	       "rx_frames:%d rx_usecs:%d\n",
1088 	       (unsigned)priv->dev->if_mtu, priv->rx_frames, priv->rx_usecs);
1089 
1090 	/* Setup cq moderation params */
1091 	for (i = 0; i < priv->rx_ring_num; i++) {
1092 		cq = priv->rx_cq[i];
1093 		cq->moder_cnt = priv->rx_frames;
1094 		cq->moder_time = priv->rx_usecs;
1095 		priv->last_moder_time[i] = MLX4_EN_AUTO_CONF;
1096 		priv->last_moder_packets[i] = 0;
1097 		priv->last_moder_bytes[i] = 0;
1098 	}
1099 
1100 	for (i = 0; i < priv->tx_ring_num; i++) {
1101 		cq = priv->tx_cq[i];
1102 		cq->moder_cnt = priv->tx_frames;
1103 		cq->moder_time = priv->tx_usecs;
1104 	}
1105 
1106 	/* Reset auto-moderation params */
1107 	priv->pkt_rate_low = MLX4_EN_RX_RATE_LOW;
1108 	priv->rx_usecs_low = MLX4_EN_RX_COAL_TIME_LOW;
1109 	priv->pkt_rate_high = MLX4_EN_RX_RATE_HIGH;
1110 	priv->rx_usecs_high = MLX4_EN_RX_COAL_TIME_HIGH;
1111 	priv->sample_interval = MLX4_EN_SAMPLE_INTERVAL;
1112 	priv->adaptive_rx_coal = 1;
1113 	priv->last_moder_jiffies = 0;
1114 	priv->last_moder_tx_packets = 0;
1115 }
1116 
1117 static void mlx4_en_auto_moderation(struct mlx4_en_priv *priv)
1118 {
1119 	unsigned long period = (unsigned long) (jiffies - priv->last_moder_jiffies);
1120 	struct mlx4_en_cq *cq;
1121 	unsigned long packets;
1122 	unsigned long rate;
1123 	unsigned long avg_pkt_size;
1124 	unsigned long rx_packets;
1125 	unsigned long rx_bytes;
1126 	unsigned long rx_pkt_diff;
1127 	int moder_time;
1128 	int ring, err;
1129 
1130 	if (!priv->adaptive_rx_coal || period < priv->sample_interval * HZ)
1131 		return;
1132 
1133 	for (ring = 0; ring < priv->rx_ring_num; ring++) {
1134                 spin_lock(&priv->stats_lock);
1135 		rx_packets = priv->rx_ring[ring]->packets;
1136 		rx_bytes = priv->rx_ring[ring]->bytes;
1137 		spin_unlock(&priv->stats_lock);
1138 
1139 		rx_pkt_diff = ((unsigned long) (rx_packets -
1140 				priv->last_moder_packets[ring]));
1141 		packets = rx_pkt_diff;
1142 		rate = packets * HZ / period;
1143 		avg_pkt_size = packets ? ((unsigned long) (rx_bytes -
1144 				priv->last_moder_bytes[ring])) / packets : 0;
1145 
1146 		/* Apply auto-moderation only when packet rate
1147 		 * exceeds a rate that it matters */
1148 		if (rate > (MLX4_EN_RX_RATE_THRESH / priv->rx_ring_num) &&
1149 		    avg_pkt_size > MLX4_EN_AVG_PKT_SMALL) {
1150 			if (rate < priv->pkt_rate_low)
1151 				moder_time = priv->rx_usecs_low;
1152 			else if (rate > priv->pkt_rate_high)
1153 				moder_time = priv->rx_usecs_high;
1154 			else
1155 				moder_time = (rate - priv->pkt_rate_low) *
1156 					(priv->rx_usecs_high - priv->rx_usecs_low) /
1157 					(priv->pkt_rate_high - priv->pkt_rate_low) +
1158 					priv->rx_usecs_low;
1159 		} else {
1160 			moder_time = priv->rx_usecs_low;
1161 		}
1162 
1163 		if (moder_time != priv->last_moder_time[ring]) {
1164 			priv->last_moder_time[ring] = moder_time;
1165 			cq = priv->rx_cq[ring];
1166 			cq->moder_time = moder_time;
1167 			cq->moder_cnt = priv->rx_frames;
1168 			err = mlx4_en_set_cq_moder(priv, cq);
1169 			if (err)
1170 				en_err(priv, "Failed modifying moderation for cq:%d\n",
1171 				       ring);
1172 		}
1173 		priv->last_moder_packets[ring] = rx_packets;
1174 		priv->last_moder_bytes[ring] = rx_bytes;
1175 	}
1176 
1177 	priv->last_moder_jiffies = jiffies;
1178 }
1179 
1180 static void mlx4_en_do_get_stats(struct work_struct *work)
1181 {
1182 	struct delayed_work *delay = to_delayed_work(work);
1183 	struct mlx4_en_priv *priv = container_of(delay, struct mlx4_en_priv,
1184 						 stats_task);
1185 	struct mlx4_en_dev *mdev = priv->mdev;
1186 	int err;
1187 
1188 	mutex_lock(&mdev->state_lock);
1189 	if (mdev->device_up) {
1190 		if (priv->port_up) {
1191 			if (mlx4_is_slave(mdev->dev))
1192 				err = mlx4_en_get_vport_stats(mdev, priv->port);
1193 			else
1194 				err = mlx4_en_DUMP_ETH_STATS(mdev, priv->port, 0);
1195 			if (err)
1196 				en_dbg(HW, priv, "Could not update stats\n");
1197 
1198 			mlx4_en_auto_moderation(priv);
1199 		}
1200 
1201 		queue_delayed_work(mdev->workqueue, &priv->stats_task, STATS_DELAY);
1202 	}
1203 	mutex_unlock(&mdev->state_lock);
1204 }
1205 
1206 /* mlx4_en_service_task - Run service task for tasks that needed to be done
1207  * periodically
1208  */
1209 static void mlx4_en_service_task(struct work_struct *work)
1210 {
1211 	struct delayed_work *delay = to_delayed_work(work);
1212 	struct mlx4_en_priv *priv = container_of(delay, struct mlx4_en_priv,
1213 						 service_task);
1214 	struct mlx4_en_dev *mdev = priv->mdev;
1215 
1216 	mutex_lock(&mdev->state_lock);
1217 	if (mdev->device_up) {
1218 		queue_delayed_work(mdev->workqueue, &priv->service_task,
1219 				   SERVICE_TASK_DELAY);
1220 	}
1221 	mutex_unlock(&mdev->state_lock);
1222 }
1223 
1224 static void mlx4_en_linkstate(struct work_struct *work)
1225 {
1226 	struct mlx4_en_priv *priv = container_of(work, struct mlx4_en_priv,
1227 						 linkstate_task);
1228 	struct mlx4_en_dev *mdev = priv->mdev;
1229 	int linkstate = priv->link_state;
1230 
1231 	mutex_lock(&mdev->state_lock);
1232 	/* If observable port state changed set carrier state and
1233 	 * report to system log */
1234 	if (priv->last_link_state != linkstate) {
1235 		if (linkstate == MLX4_DEV_EVENT_PORT_DOWN) {
1236 			en_info(priv, "Link Down\n");
1237 			if_link_state_change(priv->dev, LINK_STATE_DOWN);
1238 			/* update netif baudrate */
1239 			priv->dev->if_baudrate = 0;
1240 
1241 		/* make sure the port is up before notifying the OS.
1242 		 * This is tricky since we get here on INIT_PORT and
1243 		 * in such case we can't tell the OS the port is up.
1244 		 * To solve this there is a call to if_link_state_change
1245 		 * in set_rx_mode.
1246 		 * */
1247 		} else if (priv->port_up && (linkstate == MLX4_DEV_EVENT_PORT_UP)){
1248 			if (mlx4_en_QUERY_PORT(priv->mdev, priv->port))
1249 				en_info(priv, "Query port failed\n");
1250 			priv->dev->if_baudrate =
1251 			    IF_Mbps(priv->port_state.link_speed);
1252 			en_info(priv, "Link Up\n");
1253 			if_link_state_change(priv->dev, LINK_STATE_UP);
1254 		}
1255 	}
1256 	priv->last_link_state = linkstate;
1257 	mutex_unlock(&mdev->state_lock);
1258 }
1259 
1260 
1261 int mlx4_en_start_port(struct net_device *dev)
1262 {
1263 	struct mlx4_en_priv *priv = netdev_priv(dev);
1264 	struct mlx4_en_dev *mdev = priv->mdev;
1265 	struct mlx4_en_cq *cq;
1266 	struct mlx4_en_tx_ring *tx_ring;
1267 	int rx_index = 0;
1268 	int tx_index = 0;
1269 	int err = 0;
1270 	int i;
1271 	int j;
1272 	u8 mc_list[16] = {0};
1273 
1274 
1275 	if (priv->port_up) {
1276 		en_dbg(DRV, priv, "start port called while port already up\n");
1277 		return 0;
1278 	}
1279 
1280 	INIT_LIST_HEAD(&priv->mc_list);
1281 	INIT_LIST_HEAD(&priv->uc_list);
1282 	INIT_LIST_HEAD(&priv->curr_mc_list);
1283 	INIT_LIST_HEAD(&priv->curr_uc_list);
1284 	INIT_LIST_HEAD(&priv->ethtool_list);
1285 
1286 	/* Calculate Rx buf size */
1287 	dev->if_mtu = min(dev->if_mtu, priv->max_mtu);
1288         mlx4_en_calc_rx_buf(dev);
1289 	en_dbg(DRV, priv, "Rx buf size:%d\n", priv->rx_mb_size);
1290 
1291 	/* Configure rx cq's and rings */
1292 	err = mlx4_en_activate_rx_rings(priv);
1293 	if (err) {
1294 		en_err(priv, "Failed to activate RX rings\n");
1295 		return err;
1296 	}
1297 	for (i = 0; i < priv->rx_ring_num; i++) {
1298 		cq = priv->rx_cq[i];
1299 
1300 		mlx4_en_cq_init_lock(cq);
1301 		err = mlx4_en_activate_cq(priv, cq, i);
1302 		if (err) {
1303 			en_err(priv, "Failed activating Rx CQ\n");
1304 			goto cq_err;
1305 		}
1306 		for (j = 0; j < cq->size; j++)
1307 			cq->buf[j].owner_sr_opcode = MLX4_CQE_OWNER_MASK;
1308 		err = mlx4_en_set_cq_moder(priv, cq);
1309 		if (err) {
1310 			en_err(priv, "Failed setting cq moderation parameters");
1311 			mlx4_en_deactivate_cq(priv, cq);
1312 			goto cq_err;
1313 		}
1314 		mlx4_en_arm_cq(priv, cq);
1315 		priv->rx_ring[i]->cqn = cq->mcq.cqn;
1316 		++rx_index;
1317 	}
1318 
1319 	/* Set qp number */
1320 	en_dbg(DRV, priv, "Getting qp number for port %d\n", priv->port);
1321 	err = mlx4_en_get_qp(priv);
1322 	if (err) {
1323 		en_err(priv, "Failed getting eth qp\n");
1324 		goto cq_err;
1325 	}
1326 	mdev->mac_removed[priv->port] = 0;
1327 
1328 	priv->counter_index =
1329 			mlx4_get_default_counter_index(mdev->dev, priv->port);
1330 
1331 	err = mlx4_en_config_rss_steer(priv);
1332 	if (err) {
1333 		en_err(priv, "Failed configuring rss steering\n");
1334 		goto mac_err;
1335 	}
1336 
1337 	err = mlx4_en_create_drop_qp(priv);
1338 	if (err)
1339 		goto rss_err;
1340 
1341 	/* Configure tx cq's and rings */
1342 	for (i = 0; i < priv->tx_ring_num; i++) {
1343 		/* Configure cq */
1344 		cq = priv->tx_cq[i];
1345 		err = mlx4_en_activate_cq(priv, cq, i);
1346 		if (err) {
1347 			en_err(priv, "Failed activating Tx CQ\n");
1348 			goto tx_err;
1349 		}
1350 		err = mlx4_en_set_cq_moder(priv, cq);
1351 		if (err) {
1352 			en_err(priv, "Failed setting cq moderation parameters");
1353 			mlx4_en_deactivate_cq(priv, cq);
1354 			goto tx_err;
1355 		}
1356 		en_dbg(DRV, priv, "Resetting index of collapsed CQ:%d to -1\n", i);
1357 		cq->buf->wqe_index = cpu_to_be16(0xffff);
1358 
1359 		/* Configure ring */
1360 		tx_ring = priv->tx_ring[i];
1361 
1362 		err = mlx4_en_activate_tx_ring(priv, tx_ring, cq->mcq.cqn,
1363 					       i / priv->num_tx_rings_p_up);
1364 		if (err) {
1365 			en_err(priv, "Failed activating Tx ring %d\n", i);
1366 			mlx4_en_deactivate_cq(priv, cq);
1367 			goto tx_err;
1368 		}
1369 
1370 		/* Arm CQ for TX completions */
1371 		mlx4_en_arm_cq(priv, cq);
1372 
1373 		/* Set initial ownership of all Tx TXBBs to SW (1) */
1374 		for (j = 0; j < tx_ring->buf_size; j += STAMP_STRIDE)
1375 			*((u32 *) (tx_ring->buf + j)) = INIT_OWNER_BIT;
1376 		++tx_index;
1377 	}
1378 
1379 	/* Configure port */
1380 	err = mlx4_SET_PORT_general(mdev->dev, priv->port,
1381 				    priv->rx_mb_size,
1382 				    priv->prof->tx_pause,
1383 				    priv->prof->tx_ppp,
1384 				    priv->prof->rx_pause,
1385 				    priv->prof->rx_ppp);
1386 	if (err) {
1387 		en_err(priv, "Failed setting port general configurations for port %d, with error %d\n",
1388 		       priv->port, err);
1389 		goto tx_err;
1390 	}
1391 	/* Set default qp number */
1392 	err = mlx4_SET_PORT_qpn_calc(mdev->dev, priv->port, priv->base_qpn, 0);
1393 	if (err) {
1394 		en_err(priv, "Failed setting default qp numbers\n");
1395 		goto tx_err;
1396 	}
1397 
1398 	/* Init port */
1399 	en_dbg(HW, priv, "Initializing port\n");
1400 	err = mlx4_INIT_PORT(mdev->dev, priv->port);
1401 	if (err) {
1402 		en_err(priv, "Failed Initializing port\n");
1403 		goto tx_err;
1404 	}
1405 
1406 	/* Attach rx QP to bradcast address */
1407 	memset(&mc_list[10], 0xff, ETH_ALEN);
1408 	mc_list[5] = priv->port; /* needed for B0 steering support */
1409 	if (mlx4_multicast_attach(mdev->dev, &priv->rss_map.indir_qp, mc_list,
1410 				  priv->port, 0, MLX4_PROT_ETH,
1411 				  &priv->broadcast_id))
1412 		mlx4_warn(mdev, "Failed Attaching Broadcast\n");
1413 
1414 	/* Must redo promiscuous mode setup. */
1415 	priv->flags &= ~(MLX4_EN_FLAG_PROMISC | MLX4_EN_FLAG_MC_PROMISC);
1416 
1417 	/* Schedule multicast task to populate multicast list */
1418 	queue_work(mdev->workqueue, &priv->rx_mode_task);
1419 
1420 	priv->port_up = true;
1421 
1422         /* Enable the queues. */
1423         dev->if_drv_flags &= ~IFF_DRV_OACTIVE;
1424         dev->if_drv_flags |= IFF_DRV_RUNNING;
1425 #ifdef CONFIG_DEBUG_FS
1426 	mlx4_en_create_debug_files(priv);
1427 #endif
1428         callout_reset(&priv->watchdog_timer, MLX4_EN_WATCHDOG_TIMEOUT,
1429                     mlx4_en_watchdog_timeout, priv);
1430 
1431 
1432 	return 0;
1433 
1434 tx_err:
1435 	while (tx_index--) {
1436 		mlx4_en_deactivate_tx_ring(priv, priv->tx_ring[tx_index]);
1437 		mlx4_en_deactivate_cq(priv, priv->tx_cq[tx_index]);
1438 	}
1439 	mlx4_en_destroy_drop_qp(priv);
1440 rss_err:
1441 	mlx4_en_release_rss_steer(priv);
1442 mac_err:
1443 	mlx4_en_put_qp(priv);
1444 cq_err:
1445 	while (rx_index--)
1446 		mlx4_en_deactivate_cq(priv, priv->rx_cq[rx_index]);
1447 	for (i = 0; i < priv->rx_ring_num; i++)
1448 		mlx4_en_deactivate_rx_ring(priv, priv->rx_ring[i]);
1449 
1450 	return err; /* need to close devices */
1451 }
1452 
1453 
1454 void mlx4_en_stop_port(struct net_device *dev)
1455 {
1456 	struct mlx4_en_priv *priv = netdev_priv(dev);
1457 	struct mlx4_en_dev *mdev = priv->mdev;
1458 	struct mlx4_en_addr_list *addr_list, *tmp;
1459 	int i;
1460 	u8 mc_list[16] = {0};
1461 
1462 	if (!priv->port_up) {
1463 		en_dbg(DRV, priv, "stop port called while port already down\n");
1464 		return;
1465 	}
1466 
1467 #ifdef CONFIG_DEBUG_FS
1468 	mlx4_en_delete_debug_files(priv);
1469 #endif
1470 
1471 	/* close port*/
1472 	mlx4_CLOSE_PORT(mdev->dev, priv->port);
1473 
1474 	/* Set port as not active */
1475 	priv->port_up = false;
1476 	priv->counter_index = MLX4_SINK_COUNTER_INDEX(mdev->dev);
1477 
1478 	/* Promsicuous mode */
1479 	if (mdev->dev->caps.steering_mode ==
1480 	    MLX4_STEERING_MODE_DEVICE_MANAGED) {
1481 		priv->flags &= ~(MLX4_EN_FLAG_PROMISC |
1482 				 MLX4_EN_FLAG_MC_PROMISC);
1483 		mlx4_flow_steer_promisc_remove(mdev->dev,
1484 					       priv->port,
1485 					       MLX4_FS_ALL_DEFAULT);
1486 		mlx4_flow_steer_promisc_remove(mdev->dev,
1487 					       priv->port,
1488 					       MLX4_FS_MC_DEFAULT);
1489 	} else if (priv->flags & MLX4_EN_FLAG_PROMISC) {
1490 		priv->flags &= ~MLX4_EN_FLAG_PROMISC;
1491 
1492 		/* Disable promiscouos mode */
1493 		mlx4_unicast_promisc_remove(mdev->dev, priv->base_qpn,
1494 					    priv->port);
1495 
1496 		/* Disable Multicast promisc */
1497 		if (priv->flags & MLX4_EN_FLAG_MC_PROMISC) {
1498 			mlx4_multicast_promisc_remove(mdev->dev, priv->base_qpn,
1499 						      priv->port);
1500 			priv->flags &= ~MLX4_EN_FLAG_MC_PROMISC;
1501 		}
1502 	}
1503 
1504 	/* Detach All unicasts */
1505 	list_for_each_entry(addr_list, &priv->curr_uc_list, list) {
1506 		mlx4_en_uc_steer_release(priv, addr_list->addr,
1507 					 priv->rss_map.indir_qp.qpn,
1508 					 addr_list->reg_id);
1509 	}
1510 	mlx4_en_clear_uclist(dev);
1511 	list_for_each_entry_safe(addr_list, tmp, &priv->curr_uc_list, list) {
1512 		list_del(&addr_list->list);
1513 		kfree(addr_list);
1514 	}
1515 
1516 	/* Detach All multicasts */
1517 	memset(&mc_list[10], 0xff, ETH_ALEN);
1518 	mc_list[5] = priv->port; /* needed for B0 steering support */
1519 	mlx4_multicast_detach(mdev->dev, &priv->rss_map.indir_qp, mc_list,
1520 			      MLX4_PROT_ETH, priv->broadcast_id);
1521 	list_for_each_entry(addr_list, &priv->curr_mc_list, list) {
1522 		memcpy(&mc_list[10], addr_list->addr, ETH_ALEN);
1523 		mc_list[5] = priv->port;
1524 		mlx4_multicast_detach(mdev->dev, &priv->rss_map.indir_qp,
1525 				      mc_list, MLX4_PROT_ETH, addr_list->reg_id);
1526 	}
1527 	mlx4_en_clear_mclist(dev);
1528 	list_for_each_entry_safe(addr_list, tmp, &priv->curr_mc_list, list) {
1529 		list_del(&addr_list->list);
1530 		kfree(addr_list);
1531 	}
1532 
1533 	/* Flush multicast filter */
1534 	mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, 0, 1, MLX4_MCAST_CONFIG);
1535 	mlx4_en_destroy_drop_qp(priv);
1536 
1537 	/* Free TX Rings */
1538 	for (i = 0; i < priv->tx_ring_num; i++) {
1539 		mlx4_en_deactivate_tx_ring(priv, priv->tx_ring[i]);
1540 		mlx4_en_deactivate_cq(priv, priv->tx_cq[i]);
1541 	}
1542 	msleep(10);
1543 
1544 	for (i = 0; i < priv->tx_ring_num; i++)
1545 		mlx4_en_free_tx_buf(dev, priv->tx_ring[i]);
1546 
1547 	/* Free RSS qps */
1548 	mlx4_en_release_rss_steer(priv);
1549 
1550 	/* Unregister Mac address for the port */
1551 	mlx4_en_put_qp(priv);
1552 	mdev->mac_removed[priv->port] = 1;
1553 
1554 	/* Free RX Rings */
1555 	for (i = 0; i < priv->rx_ring_num; i++) {
1556 		struct mlx4_en_cq *cq = priv->rx_cq[i];
1557 		mlx4_en_deactivate_rx_ring(priv, priv->rx_ring[i]);
1558 		mlx4_en_deactivate_cq(priv, cq);
1559 	}
1560 
1561         callout_stop(&priv->watchdog_timer);
1562 
1563         dev->if_drv_flags &= ~(IFF_DRV_RUNNING | IFF_DRV_OACTIVE);
1564 }
1565 
1566 static void mlx4_en_restart(struct work_struct *work)
1567 {
1568 	struct mlx4_en_priv *priv = container_of(work, struct mlx4_en_priv,
1569 						 watchdog_task);
1570 	struct mlx4_en_dev *mdev = priv->mdev;
1571 	struct net_device *dev = priv->dev;
1572 	struct mlx4_en_tx_ring *ring;
1573 	int i;
1574 
1575 
1576 	if (priv->blocked == 0 || priv->port_up == 0)
1577 		return;
1578 	for (i = 0; i < priv->tx_ring_num; i++) {
1579 		int watchdog_time;
1580 
1581 		ring = priv->tx_ring[i];
1582 		watchdog_time = READ_ONCE(ring->watchdog_time);
1583 		if (watchdog_time != 0 &&
1584 		    time_after(ticks, ring->watchdog_time))
1585 			goto reset;
1586 	}
1587 	return;
1588 
1589 reset:
1590 	priv->port_stats.tx_timeout++;
1591 	en_dbg(DRV, priv, "Watchdog task called for port %d\n", priv->port);
1592 
1593 	mutex_lock(&mdev->state_lock);
1594 	if (priv->port_up) {
1595 		mlx4_en_stop_port(dev);
1596                 //for (i = 0; i < priv->tx_ring_num; i++)
1597                 //        netdev_tx_reset_queue(priv->tx_ring[i]->tx_queue);
1598 		if (mlx4_en_start_port(dev))
1599 			en_err(priv, "Failed restarting port %d\n", priv->port);
1600 	}
1601 	mutex_unlock(&mdev->state_lock);
1602 }
1603 
1604 static void mlx4_en_clear_stats(struct net_device *dev)
1605 {
1606 	struct mlx4_en_priv *priv = netdev_priv(dev);
1607 	struct mlx4_en_dev *mdev = priv->mdev;
1608 	int i;
1609 
1610 	if (!mlx4_is_slave(mdev->dev))
1611 		if (mlx4_en_DUMP_ETH_STATS(mdev, priv->port, 1))
1612 			en_dbg(HW, priv, "Failed dumping statistics\n");
1613 
1614 	memset(&priv->pstats, 0, sizeof(priv->pstats));
1615 	memset(&priv->pkstats, 0, sizeof(priv->pkstats));
1616 	memset(&priv->port_stats, 0, sizeof(priv->port_stats));
1617 	memset(&priv->vport_stats, 0, sizeof(priv->vport_stats));
1618 
1619 	for (i = 0; i < priv->tx_ring_num; i++) {
1620 		priv->tx_ring[i]->bytes = 0;
1621 		priv->tx_ring[i]->packets = 0;
1622 		priv->tx_ring[i]->tx_csum = 0;
1623 		priv->tx_ring[i]->oversized_packets = 0;
1624 	}
1625 	for (i = 0; i < priv->rx_ring_num; i++) {
1626 		priv->rx_ring[i]->bytes = 0;
1627 		priv->rx_ring[i]->packets = 0;
1628 		priv->rx_ring[i]->csum_ok = 0;
1629 		priv->rx_ring[i]->csum_none = 0;
1630 	}
1631 }
1632 
1633 static void mlx4_en_open(void* arg)
1634 {
1635 
1636         struct mlx4_en_priv *priv;
1637         struct mlx4_en_dev *mdev;
1638         struct net_device *dev;
1639         int err = 0;
1640 
1641         priv = arg;
1642         mdev = priv->mdev;
1643         dev = priv->dev;
1644 
1645 
1646 	mutex_lock(&mdev->state_lock);
1647 
1648 	if (!mdev->device_up) {
1649 		en_err(priv, "Cannot open - device down/disabled\n");
1650 		goto out;
1651 	}
1652 
1653 	/* Reset HW statistics and SW counters */
1654 	mlx4_en_clear_stats(dev);
1655 
1656 	err = mlx4_en_start_port(dev);
1657 	if (err)
1658 		en_err(priv, "Failed starting port:%d\n", priv->port);
1659 
1660 out:
1661 	mutex_unlock(&mdev->state_lock);
1662 	return;
1663 }
1664 
1665 void mlx4_en_free_resources(struct mlx4_en_priv *priv)
1666 {
1667 	int i;
1668 
1669 #ifdef CONFIG_RFS_ACCEL
1670 	if (priv->dev->rx_cpu_rmap) {
1671 		free_irq_cpu_rmap(priv->dev->rx_cpu_rmap);
1672 		priv->dev->rx_cpu_rmap = NULL;
1673 	}
1674 #endif
1675 
1676 	for (i = 0; i < priv->tx_ring_num; i++) {
1677 		if (priv->tx_ring && priv->tx_ring[i])
1678 			mlx4_en_destroy_tx_ring(priv, &priv->tx_ring[i]);
1679 		if (priv->tx_cq && priv->tx_cq[i])
1680 			mlx4_en_destroy_cq(priv, &priv->tx_cq[i]);
1681 	}
1682 
1683 	for (i = 0; i < priv->rx_ring_num; i++) {
1684 		if (priv->rx_ring[i])
1685 			mlx4_en_destroy_rx_ring(priv, &priv->rx_ring[i],
1686 				priv->prof->rx_ring_size);
1687 		if (priv->rx_cq[i])
1688 			mlx4_en_destroy_cq(priv, &priv->rx_cq[i]);
1689 	}
1690 
1691 	if (priv->stat_sysctl != NULL)
1692 		sysctl_ctx_free(&priv->stat_ctx);
1693 }
1694 
1695 int mlx4_en_alloc_resources(struct mlx4_en_priv *priv)
1696 {
1697 	struct mlx4_en_port_profile *prof = priv->prof;
1698 	int i;
1699 	int node = 0;
1700 
1701 	/* Create rx Rings */
1702 	for (i = 0; i < priv->rx_ring_num; i++) {
1703 		if (mlx4_en_create_cq(priv, &priv->rx_cq[i],
1704 				      prof->rx_ring_size, i, RX, node))
1705 			goto err;
1706 
1707 		if (mlx4_en_create_rx_ring(priv, &priv->rx_ring[i],
1708 					   prof->rx_ring_size, node))
1709 			goto err;
1710 	}
1711 
1712 	/* Create tx Rings */
1713 	for (i = 0; i < priv->tx_ring_num; i++) {
1714 		if (mlx4_en_create_cq(priv, &priv->tx_cq[i],
1715 				      prof->tx_ring_size, i, TX, node))
1716 			goto err;
1717 
1718 		if (mlx4_en_create_tx_ring(priv, &priv->tx_ring[i],
1719 					   prof->tx_ring_size, TXBB_SIZE, node, i))
1720 			goto err;
1721 	}
1722 
1723 #ifdef CONFIG_RFS_ACCEL
1724 	priv->dev->rx_cpu_rmap = alloc_irq_cpu_rmap(priv->rx_ring_num);
1725 	if (!priv->dev->rx_cpu_rmap)
1726 		goto err;
1727 #endif
1728         /* Re-create stat sysctls in case the number of rings changed. */
1729 	mlx4_en_sysctl_stat(priv);
1730 	return 0;
1731 
1732 err:
1733 	en_err(priv, "Failed to allocate NIC resources\n");
1734 	for (i = 0; i < priv->rx_ring_num; i++) {
1735 		if (priv->rx_ring[i])
1736 			mlx4_en_destroy_rx_ring(priv, &priv->rx_ring[i],
1737 						prof->rx_ring_size);
1738 		if (priv->rx_cq[i])
1739 			mlx4_en_destroy_cq(priv, &priv->rx_cq[i]);
1740 	}
1741 	for (i = 0; i < priv->tx_ring_num; i++) {
1742 		if (priv->tx_ring[i])
1743 			mlx4_en_destroy_tx_ring(priv, &priv->tx_ring[i]);
1744 		if (priv->tx_cq[i])
1745 			mlx4_en_destroy_cq(priv, &priv->tx_cq[i]);
1746 	}
1747 	priv->port_up = false;
1748 	return -ENOMEM;
1749 }
1750 
1751 struct en_port_attribute {
1752 	struct attribute attr;
1753 	ssize_t (*show)(struct en_port *, struct en_port_attribute *, char *buf);
1754 	ssize_t (*store)(struct en_port *, struct en_port_attribute *, char *buf, size_t count);
1755 };
1756 
1757 #define PORT_ATTR_RO(_name) \
1758 struct en_port_attribute en_port_attr_##_name = __ATTR_RO(_name)
1759 
1760 #define EN_PORT_ATTR(_name, _mode, _show, _store) \
1761 struct en_port_attribute en_port_attr_##_name = __ATTR(_name, _mode, _show, _store)
1762 
1763 void mlx4_en_destroy_netdev(struct net_device *dev)
1764 {
1765 	struct mlx4_en_priv *priv = netdev_priv(dev);
1766 	struct mlx4_en_dev *mdev = priv->mdev;
1767 
1768 	en_dbg(DRV, priv, "Destroying netdev on port:%d\n", priv->port);
1769 
1770 	/* don't allow more IOCTLs */
1771 	priv->gone = 1;
1772 
1773 	/* XXX wait a bit to allow IOCTL handlers to complete */
1774 	pause("W", hz);
1775 
1776 	if (priv->vlan_attach != NULL)
1777 		EVENTHANDLER_DEREGISTER(vlan_config, priv->vlan_attach);
1778 	if (priv->vlan_detach != NULL)
1779 		EVENTHANDLER_DEREGISTER(vlan_unconfig, priv->vlan_detach);
1780 
1781 	mutex_lock(&mdev->state_lock);
1782 	mlx4_en_stop_port(dev);
1783 	mutex_unlock(&mdev->state_lock);
1784 
1785 	/* Unregister device - this will close the port if it was up */
1786 	if (priv->registered)
1787 		ether_ifdetach(dev);
1788 
1789 	if (priv->allocated)
1790 		mlx4_free_hwq_res(mdev->dev, &priv->res, MLX4_EN_PAGE_SIZE);
1791 
1792 	cancel_delayed_work(&priv->stats_task);
1793 	cancel_delayed_work(&priv->service_task);
1794 	/* flush any pending task for this netdev */
1795 	flush_workqueue(mdev->workqueue);
1796         callout_drain(&priv->watchdog_timer);
1797 
1798 	/* Detach the netdev so tasks would not attempt to access it */
1799 	mutex_lock(&mdev->state_lock);
1800 	mdev->pndev[priv->port] = NULL;
1801 	mutex_unlock(&mdev->state_lock);
1802 
1803 
1804 	mlx4_en_free_resources(priv);
1805 
1806 	/* freeing the sysctl conf cannot be called from within mlx4_en_free_resources */
1807 	if (priv->conf_sysctl != NULL)
1808 		sysctl_ctx_free(&priv->conf_ctx);
1809 
1810 	kfree(priv->tx_ring);
1811 	kfree(priv->tx_cq);
1812 
1813         kfree(priv);
1814         if_free(dev);
1815 
1816 }
1817 
1818 static int mlx4_en_change_mtu(struct net_device *dev, int new_mtu)
1819 {
1820 	struct mlx4_en_priv *priv = netdev_priv(dev);
1821 	struct mlx4_en_dev *mdev = priv->mdev;
1822 	int err = 0;
1823 
1824 	en_dbg(DRV, priv, "Change MTU called - current:%u new:%u\n",
1825 	       (unsigned)dev->if_mtu, (unsigned)new_mtu);
1826 
1827 	if ((new_mtu < MLX4_EN_MIN_MTU) || (new_mtu > priv->max_mtu)) {
1828 		en_err(priv, "Bad MTU size:%d, max %u.\n", new_mtu,
1829 		    priv->max_mtu);
1830 		return -EPERM;
1831 	}
1832 	mutex_lock(&mdev->state_lock);
1833 	dev->if_mtu = new_mtu;
1834 	if (dev->if_drv_flags & IFF_DRV_RUNNING) {
1835 		if (!mdev->device_up) {
1836 			/* NIC is probably restarting - let watchdog task reset
1837 			 *                          * the port */
1838 			en_dbg(DRV, priv, "Change MTU called with card down!?\n");
1839 		} else {
1840 			mlx4_en_stop_port(dev);
1841 			err = mlx4_en_start_port(dev);
1842 			if (err) {
1843 				en_err(priv, "Failed restarting port:%d\n",
1844 						priv->port);
1845 				queue_work(mdev->workqueue, &priv->watchdog_task);
1846 			}
1847 		}
1848 	}
1849 	mutex_unlock(&mdev->state_lock);
1850 	return 0;
1851 }
1852 
1853 static int mlx4_en_calc_media(struct mlx4_en_priv *priv)
1854 {
1855 	int trans_type;
1856 	int active;
1857 
1858 	active = IFM_ETHER;
1859 	if (priv->last_link_state == MLX4_DEV_EVENT_PORT_DOWN)
1860 		return (active);
1861 	active |= IFM_FDX;
1862 	trans_type = priv->port_state.transceiver;
1863 	/* XXX I don't know all of the transceiver values. */
1864 	switch (priv->port_state.link_speed) {
1865 	case 100:
1866 		active |= IFM_100_T;
1867 		break;
1868 	case 1000:
1869 		active |= IFM_1000_T;
1870 		break;
1871 	case 10000:
1872 		if (trans_type > 0 && trans_type <= 0xC)
1873 			active |= IFM_10G_SR;
1874 		else if (trans_type == 0x80 || trans_type == 0)
1875 			active |= IFM_10G_CX4;
1876 		break;
1877 	case 40000:
1878 		active |= IFM_40G_CR4;
1879 		break;
1880 	}
1881 	if (priv->prof->tx_pause)
1882 		active |= IFM_ETH_TXPAUSE;
1883 	if (priv->prof->rx_pause)
1884 		active |= IFM_ETH_RXPAUSE;
1885 
1886 	return (active);
1887 }
1888 
1889 static void mlx4_en_media_status(struct ifnet *dev, struct ifmediareq *ifmr)
1890 {
1891 	struct mlx4_en_priv *priv;
1892 
1893 	priv = dev->if_softc;
1894 	ifmr->ifm_status = IFM_AVALID;
1895 	if (priv->last_link_state != MLX4_DEV_EVENT_PORT_DOWN)
1896 		ifmr->ifm_status |= IFM_ACTIVE;
1897 	ifmr->ifm_active = mlx4_en_calc_media(priv);
1898 
1899 	return;
1900 }
1901 
1902 static int mlx4_en_media_change(struct ifnet *dev)
1903 {
1904 	struct mlx4_en_priv *priv;
1905         struct ifmedia *ifm;
1906 	int rxpause;
1907 	int txpause;
1908 	int error;
1909 
1910 	priv = dev->if_softc;
1911 	ifm = &priv->media;
1912 	rxpause = txpause = 0;
1913 	error = 0;
1914 
1915 	if (IFM_TYPE(ifm->ifm_media) != IFM_ETHER)
1916 		return (EINVAL);
1917         switch (IFM_SUBTYPE(ifm->ifm_media)) {
1918         case IFM_AUTO:
1919 		break;
1920 	case IFM_10G_SR:
1921 	case IFM_10G_CX4:
1922 	case IFM_1000_T:
1923 	case IFM_40G_CR4:
1924 		if ((IFM_SUBTYPE(ifm->ifm_media)
1925 			== IFM_SUBTYPE(mlx4_en_calc_media(priv)))
1926 			&& (ifm->ifm_media & IFM_FDX))
1927 			break;
1928 		/* Fallthrough */
1929 	default:
1930                 printf("%s: Only auto media type\n", if_name(dev));
1931                 return (EINVAL);
1932 	}
1933 	/* Allow user to set/clear pause */
1934 	if (IFM_OPTIONS(ifm->ifm_media) & IFM_ETH_RXPAUSE)
1935 		rxpause = 1;
1936 	if (IFM_OPTIONS(ifm->ifm_media) & IFM_ETH_TXPAUSE)
1937 		txpause = 1;
1938 	if (priv->prof->tx_pause != txpause || priv->prof->rx_pause != rxpause) {
1939 		priv->prof->tx_pause = txpause;
1940 		priv->prof->rx_pause = rxpause;
1941 		error = -mlx4_SET_PORT_general(priv->mdev->dev, priv->port,
1942 		     priv->rx_mb_size + ETHER_CRC_LEN, priv->prof->tx_pause,
1943 		     priv->prof->tx_ppp, priv->prof->rx_pause,
1944 		     priv->prof->rx_ppp);
1945 	}
1946 	return (error);
1947 }
1948 
1949 static int mlx4_en_ioctl(struct ifnet *dev, u_long command, caddr_t data)
1950 {
1951 	struct mlx4_en_priv *priv;
1952 	struct mlx4_en_dev *mdev;
1953 	struct ifreq *ifr;
1954 	int error;
1955 	int mask;
1956 	struct ifrsskey *ifrk;
1957 	const u32 *key;
1958 	struct ifrsshash *ifrh;
1959 	u8 rss_mask;
1960 
1961 	error = 0;
1962 	mask = 0;
1963 	priv = dev->if_softc;
1964 
1965 	/* check if detaching */
1966 	if (priv == NULL || priv->gone != 0)
1967 		return (ENXIO);
1968 
1969 	mdev = priv->mdev;
1970 	ifr = (struct ifreq *) data;
1971 
1972 	switch (command) {
1973 	case SIOCSIFMTU:
1974 		error = -mlx4_en_change_mtu(dev, ifr->ifr_mtu);
1975 		break;
1976 	case SIOCSIFFLAGS:
1977 		if (dev->if_flags & IFF_UP) {
1978 			if ((dev->if_drv_flags & IFF_DRV_RUNNING) == 0) {
1979 				mutex_lock(&mdev->state_lock);
1980 				mlx4_en_start_port(dev);
1981 				mutex_unlock(&mdev->state_lock);
1982 			} else {
1983 				mlx4_en_set_rx_mode(dev);
1984 			}
1985 		} else {
1986 			mutex_lock(&mdev->state_lock);
1987 			if (dev->if_drv_flags & IFF_DRV_RUNNING) {
1988 				mlx4_en_stop_port(dev);
1989 				if_link_state_change(dev, LINK_STATE_DOWN);
1990 			}
1991 			mutex_unlock(&mdev->state_lock);
1992 		}
1993 		break;
1994 	case SIOCADDMULTI:
1995 	case SIOCDELMULTI:
1996 		mlx4_en_set_rx_mode(dev);
1997 		break;
1998 	case SIOCSIFMEDIA:
1999 	case SIOCGIFMEDIA:
2000 		error = ifmedia_ioctl(dev, ifr, &priv->media, command);
2001 		break;
2002 	case SIOCSIFCAP:
2003 		mutex_lock(&mdev->state_lock);
2004 		mask = ifr->ifr_reqcap ^ dev->if_capenable;
2005 		if (mask & IFCAP_TXCSUM) {
2006 			dev->if_capenable ^= IFCAP_TXCSUM;
2007 			dev->if_hwassist ^= (CSUM_TCP | CSUM_UDP | CSUM_IP);
2008 
2009 			if (IFCAP_TSO4 & dev->if_capenable &&
2010 			    !(IFCAP_TXCSUM & dev->if_capenable)) {
2011 				dev->if_capenable &= ~IFCAP_TSO4;
2012 				dev->if_hwassist &= ~CSUM_IP_TSO;
2013 				if_printf(dev,
2014 				    "tso4 disabled due to -txcsum.\n");
2015 			}
2016 		}
2017 		if (mask & IFCAP_TXCSUM_IPV6) {
2018 			dev->if_capenable ^= IFCAP_TXCSUM_IPV6;
2019 			dev->if_hwassist ^= (CSUM_UDP_IPV6 | CSUM_TCP_IPV6);
2020 
2021 			if (IFCAP_TSO6 & dev->if_capenable &&
2022 			    !(IFCAP_TXCSUM_IPV6 & dev->if_capenable)) {
2023 				dev->if_capenable &= ~IFCAP_TSO6;
2024 				dev->if_hwassist &= ~CSUM_IP6_TSO;
2025 				if_printf(dev,
2026 				    "tso6 disabled due to -txcsum6.\n");
2027 			}
2028 		}
2029 		if (mask & IFCAP_RXCSUM)
2030 			dev->if_capenable ^= IFCAP_RXCSUM;
2031 		if (mask & IFCAP_RXCSUM_IPV6)
2032 			dev->if_capenable ^= IFCAP_RXCSUM_IPV6;
2033 
2034 		if (mask & IFCAP_TSO4) {
2035 			if (!(IFCAP_TSO4 & dev->if_capenable) &&
2036 			    !(IFCAP_TXCSUM & dev->if_capenable)) {
2037 				if_printf(dev, "enable txcsum first.\n");
2038 				error = EAGAIN;
2039 				goto out;
2040 			}
2041 			dev->if_capenable ^= IFCAP_TSO4;
2042 			dev->if_hwassist ^= CSUM_IP_TSO;
2043 		}
2044 		if (mask & IFCAP_TSO6) {
2045 			if (!(IFCAP_TSO6 & dev->if_capenable) &&
2046 			    !(IFCAP_TXCSUM_IPV6 & dev->if_capenable)) {
2047 				if_printf(dev, "enable txcsum6 first.\n");
2048 				error = EAGAIN;
2049 				goto out;
2050 			}
2051 			dev->if_capenable ^= IFCAP_TSO6;
2052 			dev->if_hwassist ^= CSUM_IP6_TSO;
2053 		}
2054 		if (mask & IFCAP_LRO)
2055 			dev->if_capenable ^= IFCAP_LRO;
2056 		if (mask & IFCAP_VLAN_HWTAGGING)
2057 			dev->if_capenable ^= IFCAP_VLAN_HWTAGGING;
2058 		if (mask & IFCAP_VLAN_HWFILTER)
2059 			dev->if_capenable ^= IFCAP_VLAN_HWFILTER;
2060 		if (mask & IFCAP_WOL_MAGIC)
2061 			dev->if_capenable ^= IFCAP_WOL_MAGIC;
2062 		if (dev->if_drv_flags & IFF_DRV_RUNNING)
2063 			mlx4_en_start_port(dev);
2064 out:
2065 		mutex_unlock(&mdev->state_lock);
2066 		VLAN_CAPABILITIES(dev);
2067 		break;
2068 #if __FreeBSD_version >= 1100036
2069 	case SIOCGI2C: {
2070 		struct ifi2creq i2c;
2071 
2072 		error = copyin(ifr_data_get_ptr(ifr), &i2c, sizeof(i2c));
2073 		if (error)
2074 			break;
2075 		if (i2c.len > sizeof(i2c.data)) {
2076 			error = EINVAL;
2077 			break;
2078 		}
2079 		/*
2080 		 * Note that we ignore i2c.addr here. The driver hardcodes
2081 		 * the address to 0x50, while standard expects it to be 0xA0.
2082 		 */
2083 		error = mlx4_get_module_info(mdev->dev, priv->port,
2084 		    i2c.offset, i2c.len, i2c.data);
2085 		if (error < 0) {
2086 			error = -error;
2087 			break;
2088 		}
2089 		error = copyout(&i2c, ifr_data_get_ptr(ifr), sizeof(i2c));
2090 		break;
2091 	}
2092 #endif
2093 	case SIOCGIFRSSKEY:
2094 		ifrk = (struct ifrsskey *)data;
2095 		ifrk->ifrk_func = RSS_FUNC_TOEPLITZ;
2096 		mutex_lock(&mdev->state_lock);
2097 		key = mlx4_en_get_rss_key(priv, &ifrk->ifrk_keylen);
2098 		if (ifrk->ifrk_keylen > RSS_KEYLEN)
2099 			error = EINVAL;
2100 		else
2101 			memcpy(ifrk->ifrk_key, key, ifrk->ifrk_keylen);
2102 		mutex_unlock(&mdev->state_lock);
2103 		break;
2104 
2105 	case SIOCGIFRSSHASH:
2106 		mutex_lock(&mdev->state_lock);
2107 		rss_mask = mlx4_en_get_rss_mask(priv);
2108 		mutex_unlock(&mdev->state_lock);
2109 		ifrh = (struct ifrsshash *)data;
2110 		ifrh->ifrh_func = RSS_FUNC_TOEPLITZ;
2111 		ifrh->ifrh_types = 0;
2112 		if (rss_mask & MLX4_RSS_IPV4)
2113 			ifrh->ifrh_types |= RSS_TYPE_IPV4;
2114 		if (rss_mask & MLX4_RSS_TCP_IPV4)
2115 			ifrh->ifrh_types |= RSS_TYPE_TCP_IPV4;
2116 		if (rss_mask & MLX4_RSS_IPV6)
2117 			ifrh->ifrh_types |= RSS_TYPE_IPV6;
2118 		if (rss_mask & MLX4_RSS_TCP_IPV6)
2119 			ifrh->ifrh_types |= RSS_TYPE_TCP_IPV6;
2120 		if (rss_mask & MLX4_RSS_UDP_IPV4)
2121 			ifrh->ifrh_types |= RSS_TYPE_UDP_IPV4;
2122 		if (rss_mask & MLX4_RSS_UDP_IPV6)
2123 			ifrh->ifrh_types |= RSS_TYPE_UDP_IPV6;
2124 		break;
2125 
2126 	default:
2127 		error = ether_ioctl(dev, command, data);
2128 		break;
2129 	}
2130 
2131 	return (error);
2132 }
2133 
2134 
2135 int mlx4_en_init_netdev(struct mlx4_en_dev *mdev, int port,
2136 			struct mlx4_en_port_profile *prof)
2137 {
2138 	struct net_device *dev;
2139 	struct mlx4_en_priv *priv;
2140 	uint8_t dev_addr[ETHER_ADDR_LEN];
2141 	int err;
2142 	int i;
2143 
2144 	priv = kzalloc(sizeof(*priv), GFP_KERNEL);
2145 	dev = priv->dev = if_alloc(IFT_ETHER);
2146 	if (dev == NULL) {
2147 		en_err(priv, "Net device allocation failed\n");
2148 		kfree(priv);
2149 		return -ENOMEM;
2150 	}
2151 	dev->if_softc = priv;
2152 	if_initname(dev, "mlxen", (device_get_unit(
2153 	    mdev->pdev->dev.bsddev) * MLX4_MAX_PORTS) + port - 1);
2154 	dev->if_mtu = ETHERMTU;
2155 	dev->if_init = mlx4_en_open;
2156 	dev->if_flags = IFF_BROADCAST | IFF_SIMPLEX | IFF_MULTICAST;
2157 	dev->if_ioctl = mlx4_en_ioctl;
2158 	dev->if_transmit = mlx4_en_transmit;
2159 	dev->if_qflush = mlx4_en_qflush;
2160 	dev->if_snd.ifq_maxlen = prof->tx_ring_size;
2161 
2162 	/*
2163 	 * Initialize driver private data
2164 	 */
2165 	priv->counter_index = 0xff;
2166 	spin_lock_init(&priv->stats_lock);
2167 	INIT_WORK(&priv->rx_mode_task, mlx4_en_do_set_rx_mode);
2168 	INIT_WORK(&priv->watchdog_task, mlx4_en_restart);
2169 	INIT_WORK(&priv->linkstate_task, mlx4_en_linkstate);
2170 	INIT_DELAYED_WORK(&priv->stats_task, mlx4_en_do_get_stats);
2171 	INIT_DELAYED_WORK(&priv->service_task, mlx4_en_service_task);
2172 	callout_init(&priv->watchdog_timer, 1);
2173 #ifdef CONFIG_RFS_ACCEL
2174 	INIT_LIST_HEAD(&priv->filters);
2175 	spin_lock_init(&priv->filters_lock);
2176 #endif
2177 
2178 	priv->msg_enable = MLX4_EN_MSG_LEVEL;
2179 	priv->dev = dev;
2180 	priv->mdev = mdev;
2181 	priv->ddev = &mdev->pdev->dev;
2182 	priv->prof = prof;
2183 	priv->port = port;
2184 	priv->port_up = false;
2185 	priv->flags = prof->flags;
2186 
2187 	priv->num_tx_rings_p_up = mdev->profile.num_tx_rings_p_up;
2188 	priv->tx_ring_num = prof->tx_ring_num;
2189 	priv->tx_ring = kcalloc(MAX_TX_RINGS,
2190 				sizeof(struct mlx4_en_tx_ring *), GFP_KERNEL);
2191 	if (!priv->tx_ring) {
2192 		err = -ENOMEM;
2193 		goto out;
2194 	}
2195 	priv->tx_cq = kcalloc(sizeof(struct mlx4_en_cq *), MAX_TX_RINGS,
2196 			GFP_KERNEL);
2197 	if (!priv->tx_cq) {
2198 		err = -ENOMEM;
2199 		goto out;
2200 	}
2201 
2202 	priv->rx_ring_num = prof->rx_ring_num;
2203 	priv->cqe_factor = (mdev->dev->caps.cqe_size == 64) ? 1 : 0;
2204 	priv->mac_index = -1;
2205 	priv->last_ifq_jiffies = 0;
2206 	priv->if_counters_rx_errors = 0;
2207 	priv->if_counters_rx_no_buffer = 0;
2208 #ifdef CONFIG_MLX4_EN_DCB
2209 	if (!mlx4_is_slave(priv->mdev->dev)) {
2210 		priv->dcbx_cap = DCB_CAP_DCBX_HOST;
2211 		priv->flags |= MLX4_EN_FLAG_DCB_ENABLED;
2212 		if (mdev->dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_ETS_CFG) {
2213 			dev->dcbnl_ops = &mlx4_en_dcbnl_ops;
2214 		} else {
2215 			en_info(priv, "QoS disabled - no HW support\n");
2216 			dev->dcbnl_ops = &mlx4_en_dcbnl_pfc_ops;
2217 		}
2218 	}
2219 #endif
2220 
2221 	/* Query for default mac and max mtu */
2222 	priv->max_mtu = mdev->dev->caps.eth_mtu_cap[priv->port];
2223         priv->mac = mdev->dev->caps.def_mac[priv->port];
2224         if (ILLEGAL_MAC(priv->mac)) {
2225 #if BITS_PER_LONG == 64
2226                 en_err(priv, "Port: %d, invalid mac burned: 0x%lx, quiting\n",
2227                                 priv->port, priv->mac);
2228 #elif BITS_PER_LONG == 32
2229                 en_err(priv, "Port: %d, invalid mac burned: 0x%llx, quiting\n",
2230                                 priv->port, priv->mac);
2231 #endif
2232                 err = -EINVAL;
2233                 goto out;
2234         }
2235 
2236 	mlx4_en_sysctl_conf(priv);
2237 
2238 	err = mlx4_en_alloc_resources(priv);
2239 	if (err)
2240 		goto out;
2241 
2242 	/* Allocate page for receive rings */
2243 	err = mlx4_alloc_hwq_res(mdev->dev, &priv->res,
2244 				MLX4_EN_PAGE_SIZE, MLX4_EN_PAGE_SIZE);
2245 	if (err) {
2246 		en_err(priv, "Failed to allocate page for rx qps\n");
2247 		goto out;
2248 	}
2249 	priv->allocated = 1;
2250 
2251 	/*
2252 	 * Set driver features
2253 	 */
2254 	dev->if_capabilities |= IFCAP_HWCSUM | IFCAP_HWCSUM_IPV6;
2255 	dev->if_capabilities |= IFCAP_VLAN_MTU | IFCAP_VLAN_HWTAGGING;
2256 	dev->if_capabilities |= IFCAP_VLAN_HWCSUM | IFCAP_VLAN_HWFILTER;
2257 	dev->if_capabilities |= IFCAP_LINKSTATE | IFCAP_JUMBO_MTU;
2258 	dev->if_capabilities |= IFCAP_LRO;
2259 	dev->if_capabilities |= IFCAP_HWSTATS;
2260 
2261 	if (mdev->LSO_support)
2262 		dev->if_capabilities |= IFCAP_TSO4 | IFCAP_TSO6 | IFCAP_VLAN_HWTSO;
2263 
2264 #if __FreeBSD_version >= 1100000
2265 	/* set TSO limits so that we don't have to drop TX packets */
2266 	dev->if_hw_tsomax = MLX4_EN_TX_MAX_PAYLOAD_SIZE - (ETHER_HDR_LEN + ETHER_VLAN_ENCAP_LEN) /* hdr */;
2267 	dev->if_hw_tsomaxsegcount = MLX4_EN_TX_MAX_MBUF_FRAGS - 1 /* hdr */;
2268 	dev->if_hw_tsomaxsegsize = MLX4_EN_TX_MAX_MBUF_SIZE;
2269 #endif
2270 
2271 	dev->if_capenable = dev->if_capabilities;
2272 
2273 	dev->if_hwassist = 0;
2274 	if (dev->if_capenable & (IFCAP_TSO4 | IFCAP_TSO6))
2275 		dev->if_hwassist |= CSUM_TSO;
2276 	if (dev->if_capenable & IFCAP_TXCSUM)
2277 		dev->if_hwassist |= (CSUM_TCP | CSUM_UDP | CSUM_IP);
2278 	if (dev->if_capenable & IFCAP_TXCSUM_IPV6)
2279 		dev->if_hwassist |= (CSUM_UDP_IPV6 | CSUM_TCP_IPV6);
2280 
2281 
2282         /* Register for VLAN events */
2283 	priv->vlan_attach = EVENTHANDLER_REGISTER(vlan_config,
2284             mlx4_en_vlan_rx_add_vid, priv, EVENTHANDLER_PRI_FIRST);
2285 	priv->vlan_detach = EVENTHANDLER_REGISTER(vlan_unconfig,
2286             mlx4_en_vlan_rx_kill_vid, priv, EVENTHANDLER_PRI_FIRST);
2287 
2288 	mdev->pndev[priv->port] = dev;
2289 
2290 	priv->last_link_state = MLX4_DEV_EVENT_PORT_DOWN;
2291         mlx4_en_set_default_moderation(priv);
2292 
2293 	/* Set default MAC */
2294 	for (i = 0; i < ETHER_ADDR_LEN; i++)
2295 		dev_addr[ETHER_ADDR_LEN - 1 - i] = (u8) (priv->mac >> (8 * i));
2296 
2297 
2298 	ether_ifattach(dev, dev_addr);
2299 	if_link_state_change(dev, LINK_STATE_DOWN);
2300 	ifmedia_init(&priv->media, IFM_IMASK | IFM_ETH_FMASK,
2301 	    mlx4_en_media_change, mlx4_en_media_status);
2302 	ifmedia_add(&priv->media, IFM_ETHER | IFM_FDX | IFM_1000_T, 0, NULL);
2303 	ifmedia_add(&priv->media, IFM_ETHER | IFM_FDX | IFM_10G_SR, 0, NULL);
2304 	ifmedia_add(&priv->media, IFM_ETHER | IFM_FDX | IFM_10G_CX4, 0, NULL);
2305 	ifmedia_add(&priv->media, IFM_ETHER | IFM_FDX | IFM_40G_CR4, 0, NULL);
2306 	ifmedia_add(&priv->media, IFM_ETHER | IFM_AUTO, 0, NULL);
2307 	ifmedia_set(&priv->media, IFM_ETHER | IFM_AUTO);
2308 
2309 	NETDUMP_SET(dev, mlx4_en);
2310 
2311 	en_warn(priv, "Using %d TX rings\n", prof->tx_ring_num);
2312 	en_warn(priv, "Using %d RX rings\n", prof->rx_ring_num);
2313 
2314 	priv->registered = 1;
2315 
2316         en_warn(priv, "Using %d TX rings\n", prof->tx_ring_num);
2317         en_warn(priv, "Using %d RX rings\n", prof->rx_ring_num);
2318 
2319 
2320 	priv->rx_mb_size = dev->if_mtu + ETH_HLEN + VLAN_HLEN + ETH_FCS_LEN;
2321 	err = mlx4_SET_PORT_general(mdev->dev, priv->port,
2322 				    priv->rx_mb_size,
2323 				    prof->tx_pause, prof->tx_ppp,
2324 				    prof->rx_pause, prof->rx_ppp);
2325 	if (err) {
2326 		en_err(priv, "Failed setting port general configurations "
2327 		       "for port %d, with error %d\n", priv->port, err);
2328 		goto out;
2329 	}
2330 
2331 	/* Init port */
2332 	en_warn(priv, "Initializing port\n");
2333 	err = mlx4_INIT_PORT(mdev->dev, priv->port);
2334 	if (err) {
2335 		en_err(priv, "Failed Initializing port\n");
2336 		goto out;
2337 	}
2338 
2339 	queue_delayed_work(mdev->workqueue, &priv->stats_task, STATS_DELAY);
2340 
2341         if (mdev->dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_TS)
2342                 queue_delayed_work(mdev->workqueue, &priv->service_task, SERVICE_TASK_DELAY);
2343 
2344 	return 0;
2345 
2346 out:
2347 	mlx4_en_destroy_netdev(dev);
2348 	return err;
2349 }
2350 
2351 static int mlx4_en_set_ring_size(struct net_device *dev,
2352     int rx_size, int tx_size)
2353 {
2354         struct mlx4_en_priv *priv = netdev_priv(dev);
2355         struct mlx4_en_dev *mdev = priv->mdev;
2356         int port_up = 0;
2357         int err = 0;
2358 
2359         rx_size = roundup_pow_of_two(rx_size);
2360         rx_size = max_t(u32, rx_size, MLX4_EN_MIN_RX_SIZE);
2361         rx_size = min_t(u32, rx_size, MLX4_EN_MAX_RX_SIZE);
2362         tx_size = roundup_pow_of_two(tx_size);
2363         tx_size = max_t(u32, tx_size, MLX4_EN_MIN_TX_SIZE);
2364         tx_size = min_t(u32, tx_size, MLX4_EN_MAX_TX_SIZE);
2365 
2366         if (rx_size == (priv->port_up ?
2367             priv->rx_ring[0]->actual_size : priv->rx_ring[0]->size) &&
2368             tx_size == priv->tx_ring[0]->size)
2369                 return 0;
2370         mutex_lock(&mdev->state_lock);
2371         if (priv->port_up) {
2372                 port_up = 1;
2373                 mlx4_en_stop_port(dev);
2374         }
2375         mlx4_en_free_resources(priv);
2376         priv->prof->tx_ring_size = tx_size;
2377         priv->prof->rx_ring_size = rx_size;
2378         err = mlx4_en_alloc_resources(priv);
2379         if (err) {
2380                 en_err(priv, "Failed reallocating port resources\n");
2381                 goto out;
2382         }
2383         if (port_up) {
2384                 err = mlx4_en_start_port(dev);
2385                 if (err)
2386                         en_err(priv, "Failed starting port\n");
2387         }
2388 out:
2389         mutex_unlock(&mdev->state_lock);
2390         return err;
2391 }
2392 static int mlx4_en_set_rx_ring_size(SYSCTL_HANDLER_ARGS)
2393 {
2394         struct mlx4_en_priv *priv;
2395         int size;
2396         int error;
2397 
2398         priv = arg1;
2399         size = priv->prof->rx_ring_size;
2400         error = sysctl_handle_int(oidp, &size, 0, req);
2401         if (error || !req->newptr)
2402                 return (error);
2403         error = -mlx4_en_set_ring_size(priv->dev, size,
2404             priv->prof->tx_ring_size);
2405         return (error);
2406 }
2407 
2408 static int mlx4_en_set_tx_ring_size(SYSCTL_HANDLER_ARGS)
2409 {
2410         struct mlx4_en_priv *priv;
2411         int size;
2412         int error;
2413 
2414         priv = arg1;
2415         size = priv->prof->tx_ring_size;
2416         error = sysctl_handle_int(oidp, &size, 0, req);
2417         if (error || !req->newptr)
2418                 return (error);
2419         error = -mlx4_en_set_ring_size(priv->dev, priv->prof->rx_ring_size,
2420             size);
2421 
2422         return (error);
2423 }
2424 
2425 static int mlx4_en_get_module_info(struct net_device *dev,
2426 				   struct ethtool_modinfo *modinfo)
2427 {
2428 	struct mlx4_en_priv *priv = netdev_priv(dev);
2429 	struct mlx4_en_dev *mdev = priv->mdev;
2430 	int ret;
2431 	u8 data[4];
2432 
2433 	/* Read first 2 bytes to get Module & REV ID */
2434 	ret = mlx4_get_module_info(mdev->dev, priv->port,
2435 				   0/*offset*/, 2/*size*/, data);
2436 
2437 	if (ret < 2) {
2438 		en_err(priv, "Failed to read eeprom module first two bytes, error: 0x%x\n", -ret);
2439 		return -EIO;
2440 	}
2441 
2442 	switch (data[0] /* identifier */) {
2443 	case MLX4_MODULE_ID_QSFP:
2444 		modinfo->type = ETH_MODULE_SFF_8436;
2445 		modinfo->eeprom_len = ETH_MODULE_SFF_8436_LEN;
2446 		break;
2447 	case MLX4_MODULE_ID_QSFP_PLUS:
2448 		if (data[1] >= 0x3) { /* revision id */
2449 			modinfo->type = ETH_MODULE_SFF_8636;
2450 			modinfo->eeprom_len = ETH_MODULE_SFF_8636_LEN;
2451 		} else {
2452 			modinfo->type = ETH_MODULE_SFF_8436;
2453 			modinfo->eeprom_len = ETH_MODULE_SFF_8436_LEN;
2454 		}
2455 		break;
2456 	case MLX4_MODULE_ID_QSFP28:
2457 		modinfo->type = ETH_MODULE_SFF_8636;
2458 		modinfo->eeprom_len = ETH_MODULE_SFF_8636_LEN;
2459 		break;
2460 	case MLX4_MODULE_ID_SFP:
2461 		modinfo->type = ETH_MODULE_SFF_8472;
2462 		modinfo->eeprom_len = ETH_MODULE_SFF_8472_LEN;
2463 		break;
2464 	default:
2465 		en_err(priv, "mlx4_en_get_module_info :  Not recognized cable type\n");
2466 		return -EINVAL;
2467 	}
2468 
2469 	return 0;
2470 }
2471 
2472 static int mlx4_en_get_module_eeprom(struct net_device *dev,
2473 				     struct ethtool_eeprom *ee,
2474 				     u8 *data)
2475 {
2476 	struct mlx4_en_priv *priv = netdev_priv(dev);
2477 	struct mlx4_en_dev *mdev = priv->mdev;
2478 	int offset = ee->offset;
2479 	int i = 0, ret;
2480 
2481 	if (ee->len == 0)
2482 		return -EINVAL;
2483 
2484 	memset(data, 0, ee->len);
2485 
2486 	while (i < ee->len) {
2487 		en_dbg(DRV, priv,
2488 		       "mlx4_get_module_info i(%d) offset(%d) len(%d)\n",
2489 		       i, offset, ee->len - i);
2490 
2491 		ret = mlx4_get_module_info(mdev->dev, priv->port,
2492 					   offset, ee->len - i, data + i);
2493 
2494 		if (!ret) /* Done reading */
2495 			return 0;
2496 
2497 		if (ret < 0) {
2498 			en_err(priv,
2499 			       "mlx4_get_module_info i(%d) offset(%d) bytes_to_read(%d) - FAILED (0x%x)\n",
2500 			       i, offset, ee->len - i, ret);
2501 			return -1;
2502 		}
2503 
2504 		i += ret;
2505 		offset += ret;
2506 	}
2507 	return 0;
2508 }
2509 
2510 static void mlx4_en_print_eeprom(u8 *data, __u32 len)
2511 {
2512 	int		i;
2513 	int		j = 0;
2514 	int		row = 0;
2515 	const int	NUM_OF_BYTES = 16;
2516 
2517 	printf("\nOffset\t\tValues\n");
2518 	printf("------\t\t------\n");
2519 	while(row < len){
2520 		printf("0x%04x\t\t",row);
2521 		for(i=0; i < NUM_OF_BYTES; i++){
2522 			printf("%02x ", data[j]);
2523 			row++;
2524 			j++;
2525 		}
2526 		printf("\n");
2527 	}
2528 }
2529 
2530 /* Read cable EEPROM module information by first inspecting the first
2531  * two bytes to get the length and then read the rest of the information.
2532  * The information is printed to dmesg. */
2533 static int mlx4_en_read_eeprom(SYSCTL_HANDLER_ARGS)
2534 {
2535 
2536 	u8*		data;
2537 	int		error;
2538 	int		result = 0;
2539 	struct		mlx4_en_priv *priv;
2540 	struct		net_device *dev;
2541 	struct		ethtool_modinfo modinfo;
2542 	struct		ethtool_eeprom ee;
2543 
2544 	error = sysctl_handle_int(oidp, &result, 0, req);
2545 	if (error || !req->newptr)
2546 		return (error);
2547 
2548 	if (result == 1) {
2549 		priv = arg1;
2550 		dev = priv->dev;
2551 		data = kmalloc(PAGE_SIZE, GFP_KERNEL);
2552 
2553 		error = mlx4_en_get_module_info(dev, &modinfo);
2554 		if (error) {
2555 			en_err(priv,
2556 			       "mlx4_en_get_module_info returned with error - FAILED (0x%x)\n",
2557 			       -error);
2558 			goto out;
2559 		}
2560 
2561 		ee.len = modinfo.eeprom_len;
2562 		ee.offset = 0;
2563 
2564 		error = mlx4_en_get_module_eeprom(dev, &ee, data);
2565 		if (error) {
2566 			en_err(priv,
2567 			       "mlx4_en_get_module_eeprom returned with error - FAILED (0x%x)\n",
2568 			       -error);
2569 			/* Continue printing partial information in case of an error */
2570 		}
2571 
2572 		/* EEPROM information will be printed in dmesg */
2573 		mlx4_en_print_eeprom(data, ee.len);
2574 out:
2575 		kfree(data);
2576 	}
2577 	/* Return zero to prevent sysctl failure. */
2578 	return (0);
2579 }
2580 
2581 static int mlx4_en_set_tx_ppp(SYSCTL_HANDLER_ARGS)
2582 {
2583         struct mlx4_en_priv *priv;
2584         int ppp;
2585         int error;
2586 
2587         priv = arg1;
2588         ppp = priv->prof->tx_ppp;
2589         error = sysctl_handle_int(oidp, &ppp, 0, req);
2590         if (error || !req->newptr)
2591                 return (error);
2592         if (ppp > 0xff || ppp < 0)
2593                 return (-EINVAL);
2594         priv->prof->tx_ppp = ppp;
2595         error = -mlx4_SET_PORT_general(priv->mdev->dev, priv->port,
2596                                        priv->rx_mb_size + ETHER_CRC_LEN,
2597                                        priv->prof->tx_pause,
2598                                        priv->prof->tx_ppp,
2599                                        priv->prof->rx_pause,
2600                                        priv->prof->rx_ppp);
2601 
2602         return (error);
2603 }
2604 
2605 static int mlx4_en_set_rx_ppp(SYSCTL_HANDLER_ARGS)
2606 {
2607         struct mlx4_en_priv *priv;
2608         struct mlx4_en_dev *mdev;
2609         int ppp;
2610         int error;
2611         int port_up;
2612 
2613         port_up = 0;
2614         priv = arg1;
2615         mdev = priv->mdev;
2616         ppp = priv->prof->rx_ppp;
2617         error = sysctl_handle_int(oidp, &ppp, 0, req);
2618         if (error || !req->newptr)
2619                 return (error);
2620         if (ppp > 0xff || ppp < 0)
2621                 return (-EINVAL);
2622         /* See if we have to change the number of tx queues. */
2623         if (!ppp != !priv->prof->rx_ppp) {
2624                 mutex_lock(&mdev->state_lock);
2625                 if (priv->port_up) {
2626                         port_up = 1;
2627                         mlx4_en_stop_port(priv->dev);
2628                 }
2629                 mlx4_en_free_resources(priv);
2630                 priv->prof->rx_ppp = ppp;
2631                 error = -mlx4_en_alloc_resources(priv);
2632                 if (error)
2633                         en_err(priv, "Failed reallocating port resources\n");
2634                 if (error == 0 && port_up) {
2635                         error = -mlx4_en_start_port(priv->dev);
2636                         if (error)
2637                                 en_err(priv, "Failed starting port\n");
2638                 }
2639                 mutex_unlock(&mdev->state_lock);
2640                 return (error);
2641 
2642         }
2643         priv->prof->rx_ppp = ppp;
2644         error = -mlx4_SET_PORT_general(priv->mdev->dev, priv->port,
2645                                        priv->rx_mb_size + ETHER_CRC_LEN,
2646                                        priv->prof->tx_pause,
2647                                        priv->prof->tx_ppp,
2648                                        priv->prof->rx_pause,
2649                                        priv->prof->rx_ppp);
2650 
2651         return (error);
2652 }
2653 
2654 static void mlx4_en_sysctl_conf(struct mlx4_en_priv *priv)
2655 {
2656         struct net_device *dev;
2657         struct sysctl_ctx_list *ctx;
2658         struct sysctl_oid *node;
2659         struct sysctl_oid_list *node_list;
2660         struct sysctl_oid *coal;
2661         struct sysctl_oid_list *coal_list;
2662 	const char *pnameunit;
2663         dev = priv->dev;
2664         ctx = &priv->conf_ctx;
2665 	pnameunit = device_get_nameunit(priv->mdev->pdev->dev.bsddev);
2666 
2667         sysctl_ctx_init(ctx);
2668         priv->conf_sysctl = SYSCTL_ADD_NODE(ctx, SYSCTL_STATIC_CHILDREN(_hw),
2669             OID_AUTO, dev->if_xname, CTLFLAG_RD, 0, "mlx4 10gig ethernet");
2670         node = SYSCTL_ADD_NODE(ctx, SYSCTL_CHILDREN(priv->conf_sysctl), OID_AUTO,
2671             "conf", CTLFLAG_RD, NULL, "Configuration");
2672         node_list = SYSCTL_CHILDREN(node);
2673 
2674         SYSCTL_ADD_UINT(ctx, node_list, OID_AUTO, "msg_enable",
2675             CTLFLAG_RW, &priv->msg_enable, 0,
2676             "Driver message enable bitfield");
2677         SYSCTL_ADD_UINT(ctx, node_list, OID_AUTO, "rx_rings",
2678             CTLFLAG_RD, &priv->rx_ring_num, 0,
2679             "Number of receive rings");
2680         SYSCTL_ADD_UINT(ctx, node_list, OID_AUTO, "tx_rings",
2681             CTLFLAG_RD, &priv->tx_ring_num, 0,
2682             "Number of transmit rings");
2683         SYSCTL_ADD_PROC(ctx, node_list, OID_AUTO, "rx_size",
2684             CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, priv, 0,
2685             mlx4_en_set_rx_ring_size, "I", "Receive ring size");
2686         SYSCTL_ADD_PROC(ctx, node_list, OID_AUTO, "tx_size",
2687             CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, priv, 0,
2688             mlx4_en_set_tx_ring_size, "I", "Transmit ring size");
2689         SYSCTL_ADD_PROC(ctx, node_list, OID_AUTO, "tx_ppp",
2690             CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, priv, 0,
2691             mlx4_en_set_tx_ppp, "I", "TX Per-priority pause");
2692         SYSCTL_ADD_PROC(ctx, node_list, OID_AUTO, "rx_ppp",
2693             CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, priv, 0,
2694             mlx4_en_set_rx_ppp, "I", "RX Per-priority pause");
2695         SYSCTL_ADD_UINT(ctx, node_list, OID_AUTO, "port_num",
2696             CTLFLAG_RD, &priv->port, 0,
2697             "Port Number");
2698         SYSCTL_ADD_STRING(ctx, node_list, OID_AUTO, "device_name",
2699 	    CTLFLAG_RD, __DECONST(void *, pnameunit), 0,
2700 	    "PCI device name");
2701         /* Add coalescer configuration. */
2702         coal = SYSCTL_ADD_NODE(ctx, node_list, OID_AUTO,
2703             "coalesce", CTLFLAG_RD, NULL, "Interrupt coalesce configuration");
2704         coal_list = SYSCTL_CHILDREN(coal);
2705         SYSCTL_ADD_UINT(ctx, coal_list, OID_AUTO, "pkt_rate_low",
2706             CTLFLAG_RW, &priv->pkt_rate_low, 0,
2707             "Packets per-second for minimum delay");
2708         SYSCTL_ADD_UINT(ctx, coal_list, OID_AUTO, "rx_usecs_low",
2709             CTLFLAG_RW, &priv->rx_usecs_low, 0,
2710             "Minimum RX delay in micro-seconds");
2711         SYSCTL_ADD_UINT(ctx, coal_list, OID_AUTO, "pkt_rate_high",
2712             CTLFLAG_RW, &priv->pkt_rate_high, 0,
2713             "Packets per-second for maximum delay");
2714         SYSCTL_ADD_UINT(ctx, coal_list, OID_AUTO, "rx_usecs_high",
2715             CTLFLAG_RW, &priv->rx_usecs_high, 0,
2716             "Maximum RX delay in micro-seconds");
2717         SYSCTL_ADD_UINT(ctx, coal_list, OID_AUTO, "sample_interval",
2718             CTLFLAG_RW, &priv->sample_interval, 0,
2719             "adaptive frequency in units of HZ ticks");
2720         SYSCTL_ADD_UINT(ctx, coal_list, OID_AUTO, "adaptive_rx_coal",
2721             CTLFLAG_RW, &priv->adaptive_rx_coal, 0,
2722             "Enable adaptive rx coalescing");
2723 	/* EEPROM support */
2724 	SYSCTL_ADD_PROC(ctx, node_list, OID_AUTO, "eeprom_info",
2725 	    CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, priv, 0,
2726 	    mlx4_en_read_eeprom, "I", "EEPROM information");
2727 }
2728 
2729 static void mlx4_en_sysctl_stat(struct mlx4_en_priv *priv)
2730 {
2731 	struct sysctl_ctx_list *ctx;
2732 	struct sysctl_oid_list *node_list;
2733 	struct sysctl_oid *ring_node;
2734 	struct sysctl_oid_list *ring_list;
2735 	struct mlx4_en_tx_ring *tx_ring;
2736 	struct mlx4_en_rx_ring *rx_ring;
2737 	char namebuf[128];
2738 	int i;
2739 
2740 	ctx = &priv->stat_ctx;
2741 	sysctl_ctx_init(ctx);
2742 	priv->stat_sysctl = SYSCTL_ADD_NODE(ctx, SYSCTL_CHILDREN(priv->conf_sysctl), OID_AUTO,
2743 	    "stat", CTLFLAG_RD, NULL, "Statistics");
2744 	node_list = SYSCTL_CHILDREN(priv->stat_sysctl);
2745 
2746 #ifdef MLX4_EN_PERF_STAT
2747 	SYSCTL_ADD_UINT(ctx, node_list, OID_AUTO, "tx_poll", CTLFLAG_RD,
2748 	    &priv->pstats.tx_poll, "TX Poll calls");
2749 	SYSCTL_ADD_QUAD(ctx, node_list, OID_AUTO, "tx_pktsz_avg", CTLFLAG_RD,
2750 	    &priv->pstats.tx_pktsz_avg, "TX average packet size");
2751 	SYSCTL_ADD_UINT(ctx, node_list, OID_AUTO, "inflight_avg", CTLFLAG_RD,
2752 	    &priv->pstats.inflight_avg, "TX average packets in-flight");
2753 	SYSCTL_ADD_UINT(ctx, node_list, OID_AUTO, "tx_coal_avg", CTLFLAG_RD,
2754 	    &priv->pstats.tx_coal_avg, "TX average coalesced completions");
2755 	SYSCTL_ADD_UINT(ctx, node_list, OID_AUTO, "rx_coal_avg", CTLFLAG_RD,
2756 	    &priv->pstats.rx_coal_avg, "RX average coalesced completions");
2757 #endif
2758 
2759 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "tso_packets", CTLFLAG_RD,
2760 	    &priv->port_stats.tso_packets, 0, "TSO packets sent");
2761 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "queue_stopped", CTLFLAG_RD,
2762 	    &priv->port_stats.queue_stopped, 0, "Queue full");
2763 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "wake_queue", CTLFLAG_RD,
2764 	    &priv->port_stats.wake_queue, 0, "Queue resumed after full");
2765 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "tx_timeout", CTLFLAG_RD,
2766 	    &priv->port_stats.tx_timeout, 0, "Transmit timeouts");
2767 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "tx_oversized_packets", CTLFLAG_RD,
2768 	    &priv->port_stats.oversized_packets, 0, "TX oversized packets, m_defrag failed");
2769 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_alloc_failed", CTLFLAG_RD,
2770 	    &priv->port_stats.rx_alloc_failed, 0, "RX failed to allocate mbuf");
2771 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_chksum_good", CTLFLAG_RD,
2772             &priv->port_stats.rx_chksum_good, 0, "RX checksum offload success");
2773 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_chksum_none", CTLFLAG_RD,
2774 	    &priv->port_stats.rx_chksum_none, 0, "RX without checksum offload");
2775 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "tx_chksum_offload",
2776 	    CTLFLAG_RD, &priv->port_stats.tx_chksum_offload, 0,
2777 	    "TX checksum offloads");
2778 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "defrag_attempts",
2779 	    CTLFLAG_RD, &priv->port_stats.defrag_attempts, 0,
2780 	    "Oversized chains defragged");
2781 
2782 	/* Could strdup the names and add in a loop.  This is simpler. */
2783 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_bytes", CTLFLAG_RD,
2784 	    &priv->pkstats.rx_bytes, 0, "RX Bytes");
2785 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_packets", CTLFLAG_RD,
2786 	    &priv->pkstats.rx_packets, 0, "RX packets");
2787 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_multicast_packets", CTLFLAG_RD,
2788 	    &priv->pkstats.rx_multicast_packets, 0, "RX Multicast Packets");
2789 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_broadcast_packets", CTLFLAG_RD,
2790 	    &priv->pkstats.rx_broadcast_packets, 0, "RX Broadcast Packets");
2791 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_errors", CTLFLAG_RD,
2792 	    &priv->pkstats.rx_errors, 0, "RX Errors");
2793 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_dropped", CTLFLAG_RD,
2794 	    &priv->pkstats.rx_dropped, 0, "RX Dropped");
2795 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_length_errors", CTLFLAG_RD,
2796 	    &priv->pkstats.rx_length_errors, 0, "RX Length Errors");
2797 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_over_errors", CTLFLAG_RD,
2798 	    &priv->pkstats.rx_over_errors, 0, "RX Over Errors");
2799 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_crc_errors", CTLFLAG_RD,
2800 	    &priv->pkstats.rx_crc_errors, 0, "RX CRC Errors");
2801 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_jabbers", CTLFLAG_RD,
2802 	    &priv->pkstats.rx_jabbers, 0, "RX Jabbers");
2803 
2804 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_in_range_length_error", CTLFLAG_RD,
2805 	    &priv->pkstats.rx_in_range_length_error, 0, "RX IN_Range Length Error");
2806 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_out_range_length_error",
2807 	    CTLFLAG_RD, &priv->pkstats.rx_out_range_length_error, 0,
2808 	    "RX Out Range Length Error");
2809 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_lt_64_bytes_packets", CTLFLAG_RD,
2810 	    &priv->pkstats.rx_lt_64_bytes_packets, 0, "RX Lt 64 Bytes Packets");
2811 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_127_bytes_packets", CTLFLAG_RD,
2812 	    &priv->pkstats.rx_127_bytes_packets, 0, "RX 127 bytes Packets");
2813 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_255_bytes_packets", CTLFLAG_RD,
2814 	    &priv->pkstats.rx_255_bytes_packets, 0, "RX 255 bytes Packets");
2815 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_511_bytes_packets", CTLFLAG_RD,
2816 	    &priv->pkstats.rx_511_bytes_packets, 0, "RX 511 bytes Packets");
2817 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_1023_bytes_packets", CTLFLAG_RD,
2818 	    &priv->pkstats.rx_1023_bytes_packets, 0, "RX 1023 bytes Packets");
2819 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_1518_bytes_packets", CTLFLAG_RD,
2820 	    &priv->pkstats.rx_1518_bytes_packets, 0, "RX 1518 bytes Packets");
2821 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_1522_bytes_packets", CTLFLAG_RD,
2822 	    &priv->pkstats.rx_1522_bytes_packets, 0, "RX 1522 bytes Packets");
2823 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_1548_bytes_packets", CTLFLAG_RD,
2824 	    &priv->pkstats.rx_1548_bytes_packets, 0, "RX 1548 bytes Packets");
2825 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_gt_1548_bytes_packets", CTLFLAG_RD,
2826 	    &priv->pkstats.rx_gt_1548_bytes_packets, 0,
2827 	    "RX Greater Then 1548 bytes Packets");
2828 
2829 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "tx_packets", CTLFLAG_RD,
2830 	    &priv->pkstats.tx_packets, 0, "TX packets");
2831 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "tx_bytes", CTLFLAG_RD,
2832 	    &priv->pkstats.tx_bytes, 0, "TX Bytes");
2833 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "tx_multicast_packets", CTLFLAG_RD,
2834 	    &priv->pkstats.tx_multicast_packets, 0, "TX Multicast Packets");
2835 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "tx_broadcast_packets", CTLFLAG_RD,
2836 	    &priv->pkstats.tx_broadcast_packets, 0, "TX Broadcast Packets");
2837 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "tx_errors", CTLFLAG_RD,
2838 	    &priv->pkstats.tx_errors, 0, "TX Errors");
2839 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "tx_dropped", CTLFLAG_RD,
2840 	    &priv->pkstats.tx_dropped, 0, "TX Dropped");
2841 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "tx_lt_64_bytes_packets", CTLFLAG_RD,
2842 	    &priv->pkstats.tx_lt_64_bytes_packets, 0, "TX Less Then 64 Bytes Packets");
2843 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "tx_127_bytes_packets", CTLFLAG_RD,
2844 	    &priv->pkstats.tx_127_bytes_packets, 0, "TX 127 Bytes Packets");
2845 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "tx_255_bytes_packets", CTLFLAG_RD,
2846 	    &priv->pkstats.tx_255_bytes_packets, 0, "TX 255 Bytes Packets");
2847 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "tx_511_bytes_packets", CTLFLAG_RD,
2848 	    &priv->pkstats.tx_511_bytes_packets, 0, "TX 511 Bytes Packets");
2849 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "tx_1023_bytes_packets", CTLFLAG_RD,
2850 	    &priv->pkstats.tx_1023_bytes_packets, 0, "TX 1023 Bytes Packets");
2851 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "tx_1518_bytes_packets", CTLFLAG_RD,
2852 	    &priv->pkstats.tx_1518_bytes_packets, 0, "TX 1518 Bytes Packets");
2853 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "tx_1522_bytes_packets", CTLFLAG_RD,
2854 	    &priv->pkstats.tx_1522_bytes_packets, 0, "TX 1522 Bytes Packets");
2855 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "tx_1548_bytes_packets", CTLFLAG_RD,
2856 	    &priv->pkstats.tx_1548_bytes_packets, 0, "TX 1548 Bytes Packets");
2857 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "tx_gt_1548_bytes_packets", CTLFLAG_RD,
2858 	    &priv->pkstats.tx_gt_1548_bytes_packets, 0,
2859 	    "TX Greater Then 1548 Bytes Packets");
2860 
2861 	for (i = 0; i < priv->tx_ring_num; i++) {
2862 		tx_ring = priv->tx_ring[i];
2863 		snprintf(namebuf, sizeof(namebuf), "tx_ring%d", i);
2864 		ring_node = SYSCTL_ADD_NODE(ctx, node_list, OID_AUTO, namebuf,
2865 		    CTLFLAG_RD, NULL, "TX Ring");
2866 		ring_list = SYSCTL_CHILDREN(ring_node);
2867 		SYSCTL_ADD_U64(ctx, ring_list, OID_AUTO, "packets",
2868 		    CTLFLAG_RD, &tx_ring->packets, 0, "TX packets");
2869 		SYSCTL_ADD_U64(ctx, ring_list, OID_AUTO, "bytes",
2870 		    CTLFLAG_RD, &tx_ring->bytes, 0, "TX bytes");
2871 		SYSCTL_ADD_U64(ctx, ring_list, OID_AUTO, "tso_packets",
2872 		    CTLFLAG_RD, &tx_ring->tso_packets, 0, "TSO packets");
2873 		SYSCTL_ADD_U64(ctx, ring_list, OID_AUTO, "defrag_attempts",
2874 		    CTLFLAG_RD, &tx_ring->defrag_attempts, 0,
2875 		    "Oversized chains defragged");
2876 	}
2877 
2878 	for (i = 0; i < priv->rx_ring_num; i++) {
2879 		rx_ring = priv->rx_ring[i];
2880 		snprintf(namebuf, sizeof(namebuf), "rx_ring%d", i);
2881 		ring_node = SYSCTL_ADD_NODE(ctx, node_list, OID_AUTO, namebuf,
2882 		    CTLFLAG_RD, NULL, "RX Ring");
2883 		ring_list = SYSCTL_CHILDREN(ring_node);
2884 		SYSCTL_ADD_U64(ctx, ring_list, OID_AUTO, "packets",
2885 		    CTLFLAG_RD, &rx_ring->packets, 0, "RX packets");
2886 		SYSCTL_ADD_U64(ctx, ring_list, OID_AUTO, "bytes",
2887 		    CTLFLAG_RD, &rx_ring->bytes, 0, "RX bytes");
2888 		SYSCTL_ADD_U64(ctx, ring_list, OID_AUTO, "error",
2889 		    CTLFLAG_RD, &rx_ring->errors, 0, "RX soft errors");
2890 	}
2891 }
2892 
2893 #ifdef NETDUMP
2894 static void
2895 mlx4_en_netdump_init(struct ifnet *dev, int *nrxr, int *ncl, int *clsize)
2896 {
2897 	struct mlx4_en_priv *priv;
2898 
2899 	priv = if_getsoftc(dev);
2900 	mutex_lock(&priv->mdev->state_lock);
2901 	*nrxr = priv->rx_ring_num;
2902 	*ncl = NETDUMP_MAX_IN_FLIGHT;
2903 	*clsize = priv->rx_mb_size;
2904 	mutex_unlock(&priv->mdev->state_lock);
2905 }
2906 
2907 static void
2908 mlx4_en_netdump_event(struct ifnet *dev, enum netdump_ev event)
2909 {
2910 }
2911 
2912 static int
2913 mlx4_en_netdump_transmit(struct ifnet *dev, struct mbuf *m)
2914 {
2915 	struct mlx4_en_priv *priv;
2916 	int err;
2917 
2918 	priv = if_getsoftc(dev);
2919 	if ((if_getdrvflags(dev) & (IFF_DRV_RUNNING | IFF_DRV_OACTIVE)) !=
2920 	    IFF_DRV_RUNNING || !priv->link_state)
2921 		return (ENOENT);
2922 
2923 	err = mlx4_en_xmit(priv, 0, &m);
2924 	if (err != 0 && m != NULL)
2925 		m_freem(m);
2926 	return (err);
2927 }
2928 
2929 static int
2930 mlx4_en_netdump_poll(struct ifnet *dev, int count)
2931 {
2932 	struct mlx4_en_priv *priv;
2933 
2934 	priv = if_getsoftc(dev);
2935 	if ((if_getdrvflags(dev) & IFF_DRV_RUNNING) == 0 || !priv->link_state)
2936 		return (ENOENT);
2937 
2938 	mlx4_poll_interrupts(priv->mdev->dev);
2939 
2940 	return (0);
2941 }
2942 #endif /* NETDUMP */
2943