xref: /freebsd/sys/dev/mlx4/mlx4_en/mlx4_en_netdev.c (revision 2e3f49888ec8851bafb22011533217487764fdb0)
1 /*
2  * Copyright (c) 2007, 2014 Mellanox Technologies. All rights reserved.
3  *
4  * This software is available to you under a choice of one of two
5  * licenses.  You may choose to be licensed under the terms of the GNU
6  * General Public License (GPL) Version 2, available from the file
7  * COPYING in the main directory of this source tree, or the
8  * OpenIB.org BSD license below:
9  *
10  *     Redistribution and use in source and binary forms, with or
11  *     without modification, are permitted provided that the following
12  *     conditions are met:
13  *
14  *      - Redistributions of source code must retain the above
15  *        copyright notice, this list of conditions and the following
16  *        disclaimer.
17  *
18  *      - Redistributions in binary form must reproduce the above
19  *        copyright notice, this list of conditions and the following
20  *        disclaimer in the documentation and/or other materials
21  *        provided with the distribution.
22  *
23  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
24  * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
25  * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
26  * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
27  * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
28  * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
29  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
30  * SOFTWARE.
31  *
32  */
33 
34 #include <linux/etherdevice.h>
35 #include <linux/delay.h>
36 #include <linux/slab.h>
37 #include <linux/compat.h>
38 #ifdef CONFIG_NET_RX_BUSY_POLL
39 #include <net/busy_poll.h>
40 #endif
41 
42 #include <linux/list.h>
43 #include <linux/if_ether.h>
44 
45 #include <dev/mlx4/driver.h>
46 #include <dev/mlx4/device.h>
47 #include <dev/mlx4/cmd.h>
48 #include <dev/mlx4/cq.h>
49 
50 #include <sys/eventhandler.h>
51 #include <sys/sockio.h>
52 #include <sys/sysctl.h>
53 
54 #include "en.h"
55 #include "en_port.h"
56 
57 DEBUGNET_DEFINE(mlx4_en);
58 
59 static void mlx4_en_sysctl_stat(struct mlx4_en_priv *priv);
60 static void mlx4_en_sysctl_conf(struct mlx4_en_priv *priv);
61 
62 #ifdef CONFIG_NET_RX_BUSY_POLL
63 /* must be called with local_bh_disable()d */
64 static int mlx4_en_low_latency_recv(struct napi_struct *napi)
65 {
66 	struct mlx4_en_cq *cq = container_of(napi, struct mlx4_en_cq, napi);
67 	if_t dev = cq->dev;
68 	struct mlx4_en_priv *priv = mlx4_netdev_priv(dev);
69 	struct mlx4_en_rx_ring *rx_ring = priv->rx_ring[cq->ring];
70 	int done;
71 
72 	if (!priv->port_up)
73 		return LL_FLUSH_FAILED;
74 
75 	if (!mlx4_en_cq_lock_poll(cq))
76 		return LL_FLUSH_BUSY;
77 
78 	done = mlx4_en_process_rx_cq(dev, cq, 4);
79 #ifdef LL_EXTENDED_STATS
80 	if (likely(done))
81 		rx_ring->cleaned += done;
82 	else
83 		rx_ring->misses++;
84 #endif
85 
86 	mlx4_en_cq_unlock_poll(cq);
87 
88 	return done;
89 }
90 #endif	/* CONFIG_NET_RX_BUSY_POLL */
91 
92 #ifdef CONFIG_RFS_ACCEL
93 
94 struct mlx4_en_filter {
95 	struct list_head next;
96 	struct work_struct work;
97 
98 	u8     ip_proto;
99 	__be32 src_ip;
100 	__be32 dst_ip;
101 	__be16 src_port;
102 	__be16 dst_port;
103 
104 	int rxq_index;
105 	struct mlx4_en_priv *priv;
106 	u32 flow_id;			/* RFS infrastructure id */
107 	int id;				/* mlx4_en driver id */
108 	u64 reg_id;			/* Flow steering API id */
109 	u8 activated;			/* Used to prevent expiry before filter
110 					 * is attached
111 					 */
112 	struct hlist_node filter_chain;
113 };
114 
115 static void mlx4_en_filter_rfs_expire(struct mlx4_en_priv *priv);
116 
117 static enum mlx4_net_trans_rule_id mlx4_ip_proto_to_trans_rule_id(u8 ip_proto)
118 {
119 	switch (ip_proto) {
120 	case IPPROTO_UDP:
121 		return MLX4_NET_TRANS_RULE_ID_UDP;
122 	case IPPROTO_TCP:
123 		return MLX4_NET_TRANS_RULE_ID_TCP;
124 	default:
125 		return MLX4_NET_TRANS_RULE_NUM;
126 	}
127 };
128 
129 static void mlx4_en_filter_work(struct work_struct *work)
130 {
131 	struct mlx4_en_filter *filter = container_of(work,
132 						     struct mlx4_en_filter,
133 						     work);
134 	struct mlx4_en_priv *priv = filter->priv;
135 	struct mlx4_spec_list spec_tcp_udp = {
136 		.id = mlx4_ip_proto_to_trans_rule_id(filter->ip_proto),
137 		{
138 			.tcp_udp = {
139 				.dst_port = filter->dst_port,
140 				.dst_port_msk = (__force __be16)-1,
141 				.src_port = filter->src_port,
142 				.src_port_msk = (__force __be16)-1,
143 			},
144 		},
145 	};
146 	struct mlx4_spec_list spec_ip = {
147 		.id = MLX4_NET_TRANS_RULE_ID_IPV4,
148 		{
149 			.ipv4 = {
150 				.dst_ip = filter->dst_ip,
151 				.dst_ip_msk = (__force __be32)-1,
152 				.src_ip = filter->src_ip,
153 				.src_ip_msk = (__force __be32)-1,
154 			},
155 		},
156 	};
157 	struct mlx4_spec_list spec_eth = {
158 		.id = MLX4_NET_TRANS_RULE_ID_ETH,
159 	};
160 	struct mlx4_net_trans_rule rule = {
161 		.list = LIST_HEAD_INIT(rule.list),
162 		.queue_mode = MLX4_NET_TRANS_Q_LIFO,
163 		.exclusive = 1,
164 		.allow_loopback = 1,
165 		.promisc_mode = MLX4_FS_REGULAR,
166 		.port = priv->port,
167 		.priority = MLX4_DOMAIN_RFS,
168 	};
169 	int rc;
170 	__be64 mac_mask = cpu_to_be64(MLX4_MAC_MASK << 16);
171 
172 	if (spec_tcp_udp.id >= MLX4_NET_TRANS_RULE_NUM) {
173 		en_warn(priv, "RFS: ignoring unsupported ip protocol (%d)\n",
174 			filter->ip_proto);
175 		goto ignore;
176 	}
177 	list_add_tail(&spec_eth.list, &rule.list);
178 	list_add_tail(&spec_ip.list, &rule.list);
179 	list_add_tail(&spec_tcp_udp.list, &rule.list);
180 
181 	rule.qpn = priv->rss_map.qps[filter->rxq_index].qpn;
182 	memcpy(spec_eth.eth.dst_mac, priv->dev->dev_addr, ETH_ALEN);
183 	memcpy(spec_eth.eth.dst_mac_msk, &mac_mask, ETH_ALEN);
184 
185 	filter->activated = 0;
186 
187 	if (filter->reg_id) {
188 		rc = mlx4_flow_detach(priv->mdev->dev, filter->reg_id);
189 		if (rc && rc != -ENOENT)
190 			en_err(priv, "Error detaching flow. rc = %d\n", rc);
191 	}
192 
193 	rc = mlx4_flow_attach(priv->mdev->dev, &rule, &filter->reg_id);
194 	if (rc)
195 		en_err(priv, "Error attaching flow. err = %d\n", rc);
196 
197 ignore:
198 	mlx4_en_filter_rfs_expire(priv);
199 
200 	filter->activated = 1;
201 }
202 
203 static inline struct hlist_head *
204 filter_hash_bucket(struct mlx4_en_priv *priv, __be32 src_ip, __be32 dst_ip,
205 		   __be16 src_port, __be16 dst_port)
206 {
207 	unsigned long l;
208 	int bucket_idx;
209 
210 	l = (__force unsigned long)src_port |
211 	    ((__force unsigned long)dst_port << 2);
212 	l ^= (__force unsigned long)(src_ip ^ dst_ip);
213 
214 	bucket_idx = hash_long(l, MLX4_EN_FILTER_HASH_SHIFT);
215 
216 	return &priv->filter_hash[bucket_idx];
217 }
218 
219 static struct mlx4_en_filter *
220 mlx4_en_filter_alloc(struct mlx4_en_priv *priv, int rxq_index, __be32 src_ip,
221 		     __be32 dst_ip, u8 ip_proto, __be16 src_port,
222 		     __be16 dst_port, u32 flow_id)
223 {
224 	struct mlx4_en_filter *filter = NULL;
225 
226 	filter = kzalloc(sizeof(struct mlx4_en_filter), GFP_ATOMIC);
227 	if (!filter)
228 		return NULL;
229 
230 	filter->priv = priv;
231 	filter->rxq_index = rxq_index;
232 	INIT_WORK(&filter->work, mlx4_en_filter_work);
233 
234 	filter->src_ip = src_ip;
235 	filter->dst_ip = dst_ip;
236 	filter->ip_proto = ip_proto;
237 	filter->src_port = src_port;
238 	filter->dst_port = dst_port;
239 
240 	filter->flow_id = flow_id;
241 
242 	filter->id = priv->last_filter_id++ % RPS_NO_FILTER;
243 
244 	list_add_tail(&filter->next, &priv->filters);
245 	hlist_add_head(&filter->filter_chain,
246 		       filter_hash_bucket(priv, src_ip, dst_ip, src_port,
247 					  dst_port));
248 
249 	return filter;
250 }
251 
252 static void mlx4_en_filter_free(struct mlx4_en_filter *filter)
253 {
254 	struct mlx4_en_priv *priv = filter->priv;
255 	int rc;
256 
257 	list_del(&filter->next);
258 
259 	rc = mlx4_flow_detach(priv->mdev->dev, filter->reg_id);
260 	if (rc && rc != -ENOENT)
261 		en_err(priv, "Error detaching flow. rc = %d\n", rc);
262 
263 	kfree(filter);
264 }
265 
266 static inline struct mlx4_en_filter *
267 mlx4_en_filter_find(struct mlx4_en_priv *priv, __be32 src_ip, __be32 dst_ip,
268 		    u8 ip_proto, __be16 src_port, __be16 dst_port)
269 {
270 	struct mlx4_en_filter *filter;
271 	struct mlx4_en_filter *ret = NULL;
272 
273 	hlist_for_each_entry(filter,
274 			     filter_hash_bucket(priv, src_ip, dst_ip,
275 						src_port, dst_port),
276 			     filter_chain) {
277 		if (filter->src_ip == src_ip &&
278 		    filter->dst_ip == dst_ip &&
279 		    filter->ip_proto == ip_proto &&
280 		    filter->src_port == src_port &&
281 		    filter->dst_port == dst_port) {
282 			ret = filter;
283 			break;
284 		}
285 	}
286 
287 	return ret;
288 }
289 
290 static int
291 mlx4_en_filter_rfs(if_t net_dev, const struct sk_buff *skb,
292 		   u16 rxq_index, u32 flow_id)
293 {
294 	struct mlx4_en_priv *priv = mlx4_netdev_priv(net_dev);
295 	struct mlx4_en_filter *filter;
296 	const struct iphdr *ip;
297 	const __be16 *ports;
298 	u8 ip_proto;
299 	__be32 src_ip;
300 	__be32 dst_ip;
301 	__be16 src_port;
302 	__be16 dst_port;
303 	int nhoff = skb_network_offset(skb);
304 	int ret = 0;
305 
306 	if (skb->protocol != htons(ETH_P_IP))
307 		return -EPROTONOSUPPORT;
308 
309 	ip = (const struct iphdr *)(skb->data + nhoff);
310 	if (ip_is_fragment(ip))
311 		return -EPROTONOSUPPORT;
312 
313 	if ((ip->protocol != IPPROTO_TCP) && (ip->protocol != IPPROTO_UDP))
314 		return -EPROTONOSUPPORT;
315 	ports = (const __be16 *)(skb->data + nhoff + 4 * ip->ihl);
316 
317 	ip_proto = ip->protocol;
318 	src_ip = ip->saddr;
319 	dst_ip = ip->daddr;
320 	src_port = ports[0];
321 	dst_port = ports[1];
322 
323 	spin_lock_bh(&priv->filters_lock);
324 	filter = mlx4_en_filter_find(priv, src_ip, dst_ip, ip_proto,
325 				     src_port, dst_port);
326 	if (filter) {
327 		if (filter->rxq_index == rxq_index)
328 			goto out;
329 
330 		filter->rxq_index = rxq_index;
331 	} else {
332 		filter = mlx4_en_filter_alloc(priv, rxq_index,
333 					      src_ip, dst_ip, ip_proto,
334 					      src_port, dst_port, flow_id);
335 		if (!filter) {
336 			ret = -ENOMEM;
337 			goto err;
338 		}
339 	}
340 
341 	queue_work(priv->mdev->workqueue, &filter->work);
342 
343 out:
344 	ret = filter->id;
345 err:
346 	spin_unlock_bh(&priv->filters_lock);
347 
348 	return ret;
349 }
350 
351 void mlx4_en_cleanup_filters(struct mlx4_en_priv *priv)
352 {
353 	struct mlx4_en_filter *filter, *tmp;
354 	LIST_HEAD(del_list);
355 
356 	spin_lock_bh(&priv->filters_lock);
357 	list_for_each_entry_safe(filter, tmp, &priv->filters, next) {
358 		list_move(&filter->next, &del_list);
359 		hlist_del(&filter->filter_chain);
360 	}
361 	spin_unlock_bh(&priv->filters_lock);
362 
363 	list_for_each_entry_safe(filter, tmp, &del_list, next) {
364 		cancel_work_sync(&filter->work);
365 		mlx4_en_filter_free(filter);
366 	}
367 }
368 
369 static void mlx4_en_filter_rfs_expire(struct mlx4_en_priv *priv)
370 {
371 	struct mlx4_en_filter *filter = NULL, *tmp, *last_filter = NULL;
372 	LIST_HEAD(del_list);
373 	int i = 0;
374 
375 	spin_lock_bh(&priv->filters_lock);
376 	list_for_each_entry_safe(filter, tmp, &priv->filters, next) {
377 		if (i > MLX4_EN_FILTER_EXPIRY_QUOTA)
378 			break;
379 
380 		if (filter->activated &&
381 		    !work_pending(&filter->work) &&
382 		    rps_may_expire_flow(priv->dev,
383 					filter->rxq_index, filter->flow_id,
384 					filter->id)) {
385 			list_move(&filter->next, &del_list);
386 			hlist_del(&filter->filter_chain);
387 		} else
388 			last_filter = filter;
389 
390 		i++;
391 	}
392 
393 	if (last_filter && (&last_filter->next != priv->filters.next))
394 		list_move(&priv->filters, &last_filter->next);
395 
396 	spin_unlock_bh(&priv->filters_lock);
397 
398 	list_for_each_entry_safe(filter, tmp, &del_list, next)
399 		mlx4_en_filter_free(filter);
400 }
401 #endif
402 
403 static void mlx4_en_vlan_rx_add_vid(void *arg, if_t dev, u16 vid)
404 {
405 	struct mlx4_en_priv *priv = mlx4_netdev_priv(dev);
406 	struct mlx4_en_dev *mdev = priv->mdev;
407 	int err;
408 	int idx;
409 
410 	if (arg != priv)
411 		return;
412 
413 	en_dbg(HW, priv, "adding VLAN:%d\n", vid);
414 
415 	set_bit(vid, priv->active_vlans);
416 
417 	/* Add VID to port VLAN filter */
418 	mutex_lock(&mdev->state_lock);
419 	if (mdev->device_up && priv->port_up) {
420 		err = mlx4_SET_VLAN_FLTR(mdev->dev, priv);
421 		if (err)
422 			en_err(priv, "Failed configuring VLAN filter\n");
423 	}
424 	if (mlx4_register_vlan(mdev->dev, priv->port, vid, &idx))
425 		en_dbg(HW, priv, "failed adding vlan %d\n", vid);
426 	mutex_unlock(&mdev->state_lock);
427 
428 }
429 
430 static void mlx4_en_vlan_rx_kill_vid(void *arg, if_t dev, u16 vid)
431 {
432 	struct mlx4_en_priv *priv = mlx4_netdev_priv(dev);
433 	struct mlx4_en_dev *mdev = priv->mdev;
434 	int err;
435 
436 	if (arg != priv)
437 		return;
438 
439 	en_dbg(HW, priv, "Killing VID:%d\n", vid);
440 
441 	clear_bit(vid, priv->active_vlans);
442 
443 	/* Remove VID from port VLAN filter */
444 	mutex_lock(&mdev->state_lock);
445 	mlx4_unregister_vlan(mdev->dev, priv->port, vid);
446 
447 	if (mdev->device_up && priv->port_up) {
448 		err = mlx4_SET_VLAN_FLTR(mdev->dev, priv);
449 		if (err)
450 			en_err(priv, "Failed configuring VLAN filter\n");
451 	}
452 	mutex_unlock(&mdev->state_lock);
453 
454 }
455 
456 static int mlx4_en_tunnel_steer_add(struct mlx4_en_priv *priv, unsigned char *addr,
457 				    int qpn, u64 *reg_id)
458 {
459 	int err;
460 
461 	if (priv->mdev->dev->caps.tunnel_offload_mode != MLX4_TUNNEL_OFFLOAD_MODE_VXLAN ||
462 	    priv->mdev->dev->caps.dmfs_high_steer_mode == MLX4_STEERING_DMFS_A0_STATIC)
463 		return 0; /* do nothing */
464 
465 	err = mlx4_tunnel_steer_add(priv->mdev->dev, addr, priv->port, qpn,
466 				    MLX4_DOMAIN_NIC, reg_id);
467 	if (err) {
468 		en_err(priv, "failed to add vxlan steering rule, err %d\n", err);
469 		return err;
470 	}
471 	en_dbg(DRV, priv, "added vxlan steering rule, mac %pM reg_id %llx\n", addr, (long long)*reg_id);
472 	return 0;
473 }
474 
475 static int mlx4_en_uc_steer_add(struct mlx4_en_priv *priv,
476 				unsigned char *mac, int *qpn, u64 *reg_id)
477 {
478 	struct mlx4_en_dev *mdev = priv->mdev;
479 	struct mlx4_dev *dev = mdev->dev;
480 	int err;
481 
482 	switch (dev->caps.steering_mode) {
483 	case MLX4_STEERING_MODE_B0: {
484 		struct mlx4_qp qp;
485 		u8 gid[16] = {0};
486 
487 		qp.qpn = *qpn;
488 		memcpy(&gid[10], mac, ETH_ALEN);
489 		gid[5] = priv->port;
490 
491 		err = mlx4_unicast_attach(dev, &qp, gid, 0, MLX4_PROT_ETH);
492 		break;
493 	}
494 	case MLX4_STEERING_MODE_DEVICE_MANAGED: {
495 		struct mlx4_spec_list spec_eth = { {NULL} };
496 		__be64 mac_mask = cpu_to_be64(MLX4_MAC_MASK << 16);
497 
498 		struct mlx4_net_trans_rule rule = {
499 			.queue_mode = MLX4_NET_TRANS_Q_FIFO,
500 			.exclusive = 0,
501 			.allow_loopback = 1,
502 			.promisc_mode = MLX4_FS_REGULAR,
503 			.priority = MLX4_DOMAIN_NIC,
504 		};
505 
506 		rule.port = priv->port;
507 		rule.qpn = *qpn;
508 		INIT_LIST_HEAD(&rule.list);
509 
510 		spec_eth.id = MLX4_NET_TRANS_RULE_ID_ETH;
511 		memcpy(spec_eth.eth.dst_mac, mac, ETH_ALEN);
512 		memcpy(spec_eth.eth.dst_mac_msk, &mac_mask, ETH_ALEN);
513 		list_add_tail(&spec_eth.list, &rule.list);
514 
515 		err = mlx4_flow_attach(dev, &rule, reg_id);
516 		break;
517 	}
518 	default:
519 		return -EINVAL;
520 	}
521 	if (err)
522 		en_warn(priv, "Failed Attaching Unicast\n");
523 
524 	return err;
525 }
526 
527 static void mlx4_en_uc_steer_release(struct mlx4_en_priv *priv,
528 				     unsigned char *mac, int qpn, u64 reg_id)
529 {
530 	struct mlx4_en_dev *mdev = priv->mdev;
531 	struct mlx4_dev *dev = mdev->dev;
532 
533 	switch (dev->caps.steering_mode) {
534 	case MLX4_STEERING_MODE_B0: {
535 		struct mlx4_qp qp;
536 		u8 gid[16] = {0};
537 
538 		qp.qpn = qpn;
539 		memcpy(&gid[10], mac, ETH_ALEN);
540 		gid[5] = priv->port;
541 
542 		mlx4_unicast_detach(dev, &qp, gid, MLX4_PROT_ETH);
543 		break;
544 	}
545 	case MLX4_STEERING_MODE_DEVICE_MANAGED: {
546 		mlx4_flow_detach(dev, reg_id);
547 		break;
548 	}
549 	default:
550 		en_err(priv, "Invalid steering mode.\n");
551 	}
552 }
553 
554 static int mlx4_en_get_qp(struct mlx4_en_priv *priv)
555 {
556 	struct mlx4_en_dev *mdev = priv->mdev;
557 	struct mlx4_dev *dev = mdev->dev;
558 	int index = 0;
559 	int err = 0;
560 	int *qpn = &priv->base_qpn;
561 	u64 mac = mlx4_mac_to_u64(if_getlladdr(priv->dev));
562 
563 	en_dbg(DRV, priv, "Registering MAC: %pM for adding\n",
564 	       if_getlladdr(priv->dev));
565 	index = mlx4_register_mac(dev, priv->port, mac);
566 	if (index < 0) {
567 		err = index;
568 		en_err(priv, "Failed adding MAC: %pM\n",
569 		       if_getlladdr(priv->dev));
570 		return err;
571 	}
572 
573 	if (dev->caps.steering_mode == MLX4_STEERING_MODE_A0) {
574 		int base_qpn = mlx4_get_base_qpn(dev, priv->port);
575 		*qpn = base_qpn + index;
576 		return 0;
577 	}
578 
579 	err = mlx4_qp_reserve_range(dev, 1, 1, qpn, MLX4_RESERVE_A0_QP);
580 	en_dbg(DRV, priv, "Reserved qp %d\n", *qpn);
581 	if (err) {
582 		en_err(priv, "Failed to reserve qp for mac registration\n");
583 		mlx4_unregister_mac(dev, priv->port, mac);
584 		return err;
585 	}
586 
587 	return 0;
588 }
589 
590 static void mlx4_en_put_qp(struct mlx4_en_priv *priv)
591 {
592 	struct mlx4_en_dev *mdev = priv->mdev;
593 	struct mlx4_dev *dev = mdev->dev;
594 	int qpn = priv->base_qpn;
595 
596 	if (dev->caps.steering_mode == MLX4_STEERING_MODE_A0) {
597 		u64 mac = mlx4_mac_to_u64(if_getlladdr(priv->dev));
598 		en_dbg(DRV, priv, "Registering MAC: %pM for deleting\n",
599 		       if_getlladdr(priv->dev));
600 		mlx4_unregister_mac(dev, priv->port, mac);
601 	} else {
602 		en_dbg(DRV, priv, "Releasing qp: port %d, qpn %d\n",
603 		       priv->port, qpn);
604 		mlx4_qp_release_range(dev, qpn, 1);
605 		priv->flags &= ~MLX4_EN_FLAG_FORCE_PROMISC;
606 	}
607 }
608 
609 static void mlx4_en_clear_uclist(if_t dev)
610 {
611 	struct mlx4_en_priv *priv = mlx4_netdev_priv(dev);
612 	struct mlx4_en_addr_list *tmp, *uc_to_del;
613 
614 	list_for_each_entry_safe(uc_to_del, tmp, &priv->uc_list, list) {
615 		list_del(&uc_to_del->list);
616 		kfree(uc_to_del);
617 	}
618 }
619 
620 static u_int mlx4_copy_addr(void *arg, struct sockaddr_dl *sdl, u_int cnt)
621 {
622 	struct mlx4_en_priv *priv = arg;
623 	struct mlx4_en_addr_list *tmp;
624 
625 	if (sdl->sdl_alen != ETHER_ADDR_LEN)	/* XXXGL: can that happen? */
626 		return (0);
627 	tmp = kzalloc(sizeof(struct mlx4_en_addr_list), GFP_ATOMIC);
628 	if (tmp == NULL) {
629 		en_err(priv, "Failed to allocate address list\n");
630 		return (0);
631 	}
632 	memcpy(tmp->addr, LLADDR(sdl), ETH_ALEN);
633 	list_add_tail(&tmp->list, &priv->uc_list);
634 
635 	return (1);
636 }
637 
638 static void mlx4_en_cache_uclist(if_t dev)
639 {
640 	struct mlx4_en_priv *priv = mlx4_netdev_priv(dev);
641 
642 	mlx4_en_clear_uclist(dev);
643 	if_foreach_lladdr(dev, mlx4_copy_addr, priv);
644 }
645 
646 static void mlx4_en_clear_mclist(if_t dev)
647 {
648 	struct mlx4_en_priv *priv = mlx4_netdev_priv(dev);
649 	struct mlx4_en_addr_list *tmp, *mc_to_del;
650 
651 	list_for_each_entry_safe(mc_to_del, tmp, &priv->mc_list, list) {
652 		list_del(&mc_to_del->list);
653 		kfree(mc_to_del);
654 	}
655 }
656 
657 static u_int mlx4_copy_maddr(void *arg, struct sockaddr_dl *sdl, u_int count)
658 {
659 	struct mlx4_en_priv *priv = arg;
660 	struct mlx4_en_addr_list *tmp;
661 
662 	if (sdl->sdl_alen != ETHER_ADDR_LEN)	/* XXXGL: can that happen? */
663 		return (0);
664 	tmp = kzalloc(sizeof(struct mlx4_en_addr_list), GFP_ATOMIC);
665 	if (tmp == NULL) {
666 		en_err(priv, "Failed to allocate address list\n");
667 		return (0);
668 	}
669 	memcpy(tmp->addr, LLADDR(sdl), ETH_ALEN);
670 	list_add_tail(&tmp->list, &priv->mc_list);
671 	return (1);
672 }
673 
674 static void mlx4_en_cache_mclist(if_t dev)
675 {
676 	struct mlx4_en_priv *priv = mlx4_netdev_priv(dev);
677 
678 	mlx4_en_clear_mclist(dev);
679 	if_foreach_llmaddr(dev, mlx4_copy_maddr, priv);
680 }
681 
682 static void update_addr_list_flags(struct mlx4_en_priv *priv,
683 				struct list_head *dst,
684 				struct list_head *src)
685 {
686 	struct mlx4_en_addr_list *dst_tmp, *src_tmp, *new_mc;
687 	bool found;
688 
689 	/* Find all the entries that should be removed from dst,
690 	 * These are the entries that are not found in src
691 	 */
692 	list_for_each_entry(dst_tmp, dst, list) {
693 		found = false;
694 		list_for_each_entry(src_tmp, src, list) {
695 			if (!memcmp(dst_tmp->addr, src_tmp->addr, ETH_ALEN)) {
696 				found = true;
697 				break;
698 			}
699 		}
700 		if (!found)
701 			dst_tmp->action = MLX4_ADDR_LIST_REM;
702 	}
703 
704 	/* Add entries that exist in src but not in dst
705 	 * mark them as need to add
706 	 */
707 	list_for_each_entry(src_tmp, src, list) {
708 		found = false;
709 		list_for_each_entry(dst_tmp, dst, list) {
710 			if (!memcmp(dst_tmp->addr, src_tmp->addr, ETH_ALEN)) {
711 				dst_tmp->action = MLX4_ADDR_LIST_NONE;
712 				found = true;
713 				break;
714 			}
715 		}
716 		if (!found) {
717 			new_mc = kmalloc(sizeof(struct mlx4_en_addr_list),
718 					 GFP_KERNEL);
719 			if (!new_mc) {
720 				en_err(priv, "Failed to allocate current multicast list\n");
721 				return;
722 			}
723 			memcpy(new_mc, src_tmp,
724 			       sizeof(struct mlx4_en_addr_list));
725 			new_mc->action = MLX4_ADDR_LIST_ADD;
726 			list_add_tail(&new_mc->list, dst);
727 		}
728 	}
729 }
730 
731 static void mlx4_en_set_rx_mode(if_t dev)
732 {
733 	struct mlx4_en_priv *priv = mlx4_netdev_priv(dev);
734 
735 	if (!priv->port_up)
736 		return;
737 
738 	queue_work(priv->mdev->workqueue, &priv->rx_mode_task);
739 }
740 
741 static void mlx4_en_set_promisc_mode(struct mlx4_en_priv *priv,
742 				     struct mlx4_en_dev *mdev)
743 {
744 	int err = 0;
745 
746 	if (!(priv->flags & MLX4_EN_FLAG_PROMISC)) {
747 		priv->flags |= MLX4_EN_FLAG_PROMISC;
748 
749 		/* Enable promiscouos mode */
750 		switch (mdev->dev->caps.steering_mode) {
751 		case MLX4_STEERING_MODE_DEVICE_MANAGED:
752 			err = mlx4_flow_steer_promisc_add(mdev->dev,
753 							  priv->port,
754 							  priv->base_qpn,
755 							  MLX4_FS_ALL_DEFAULT);
756 			if (err)
757 				en_err(priv, "Failed enabling promiscuous mode\n");
758 			priv->flags |= MLX4_EN_FLAG_MC_PROMISC;
759 			break;
760 
761 		case MLX4_STEERING_MODE_B0:
762 			err = mlx4_unicast_promisc_add(mdev->dev,
763 						       priv->base_qpn,
764 						       priv->port);
765 			if (err)
766 				en_err(priv, "Failed enabling unicast promiscuous mode\n");
767 
768 			/* Add the default qp number as multicast
769 			 * promisc
770 			 */
771 			if (!(priv->flags & MLX4_EN_FLAG_MC_PROMISC)) {
772 				err = mlx4_multicast_promisc_add(mdev->dev,
773 								 priv->base_qpn,
774 								 priv->port);
775 				if (err)
776 					en_err(priv, "Failed enabling multicast promiscuous mode\n");
777 				priv->flags |= MLX4_EN_FLAG_MC_PROMISC;
778 			}
779 			break;
780 
781 		case MLX4_STEERING_MODE_A0:
782 			err = mlx4_SET_PORT_qpn_calc(mdev->dev,
783 						     priv->port,
784 						     priv->base_qpn,
785 						     1);
786 			if (err)
787 				en_err(priv, "Failed enabling promiscuous mode\n");
788 			break;
789 		}
790 
791 		/* Disable port multicast filter (unconditionally) */
792 		err = mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, 0,
793 					  0, MLX4_MCAST_DISABLE);
794 		if (err)
795 			en_err(priv, "Failed disabling multicast filter\n");
796 	}
797 }
798 
799 static void mlx4_en_clear_promisc_mode(struct mlx4_en_priv *priv,
800 				       struct mlx4_en_dev *mdev)
801 {
802 	int err = 0;
803 
804 	priv->flags &= ~MLX4_EN_FLAG_PROMISC;
805 
806 	/* Disable promiscouos mode */
807 	switch (mdev->dev->caps.steering_mode) {
808 	case MLX4_STEERING_MODE_DEVICE_MANAGED:
809 		err = mlx4_flow_steer_promisc_remove(mdev->dev,
810 						     priv->port,
811 						     MLX4_FS_ALL_DEFAULT);
812 		if (err)
813 			en_err(priv, "Failed disabling promiscuous mode\n");
814 		priv->flags &= ~MLX4_EN_FLAG_MC_PROMISC;
815 		break;
816 
817 	case MLX4_STEERING_MODE_B0:
818 		err = mlx4_unicast_promisc_remove(mdev->dev,
819 						  priv->base_qpn,
820 						  priv->port);
821 		if (err)
822 			en_err(priv, "Failed disabling unicast promiscuous mode\n");
823 		/* Disable Multicast promisc */
824 		if (priv->flags & MLX4_EN_FLAG_MC_PROMISC) {
825 			err = mlx4_multicast_promisc_remove(mdev->dev,
826 							    priv->base_qpn,
827 							    priv->port);
828 			if (err)
829 				en_err(priv, "Failed disabling multicast promiscuous mode\n");
830 			priv->flags &= ~MLX4_EN_FLAG_MC_PROMISC;
831 		}
832 		break;
833 
834 	case MLX4_STEERING_MODE_A0:
835 		err = mlx4_SET_PORT_qpn_calc(mdev->dev,
836 					     priv->port,
837 					     priv->base_qpn, 0);
838 		if (err)
839 			en_err(priv, "Failed disabling promiscuous mode\n");
840 		break;
841 	}
842 }
843 
844 static void mlx4_en_do_multicast(struct mlx4_en_priv *priv,
845 				 if_t dev,
846 				 struct mlx4_en_dev *mdev)
847 {
848 	struct mlx4_en_addr_list *addr_list, *tmp;
849 	u8 mc_list[16] = {0};
850 	int err = 0;
851 	u64 mcast_addr = 0;
852 
853 	/*
854 	 * Enable/disable the multicast filter according to
855 	 * IFF_ALLMULTI and IFF_PROMISC:
856 	 */
857 	if (if_getflags(dev) & (IFF_ALLMULTI | IFF_PROMISC)) {
858 		err = mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, 0,
859 					  0, MLX4_MCAST_DISABLE);
860 		if (err)
861 			en_err(priv, "Failed disabling multicast filter\n");
862 
863 		/* Add the default qp number as multicast promisc */
864 		if (!(priv->flags & MLX4_EN_FLAG_MC_PROMISC)) {
865 			switch (mdev->dev->caps.steering_mode) {
866 			case MLX4_STEERING_MODE_DEVICE_MANAGED:
867 				err = mlx4_flow_steer_promisc_add(mdev->dev,
868 								  priv->port,
869 								  priv->base_qpn,
870 								  MLX4_FS_MC_DEFAULT);
871 				break;
872 
873 			case MLX4_STEERING_MODE_B0:
874 				err = mlx4_multicast_promisc_add(mdev->dev,
875 								 priv->base_qpn,
876 								 priv->port);
877 				break;
878 
879 			case MLX4_STEERING_MODE_A0:
880 				break;
881 			}
882 			if (err)
883 				en_err(priv, "Failed entering multicast promisc mode\n");
884 			priv->flags |= MLX4_EN_FLAG_MC_PROMISC;
885 		}
886 	} else {
887 		/* Disable Multicast promisc */
888 		if (priv->flags & MLX4_EN_FLAG_MC_PROMISC) {
889 			switch (mdev->dev->caps.steering_mode) {
890 			case MLX4_STEERING_MODE_DEVICE_MANAGED:
891 				err = mlx4_flow_steer_promisc_remove(mdev->dev,
892 								     priv->port,
893 								     MLX4_FS_MC_DEFAULT);
894 				break;
895 
896 			case MLX4_STEERING_MODE_B0:
897 				err = mlx4_multicast_promisc_remove(mdev->dev,
898 								    priv->base_qpn,
899 								    priv->port);
900 				break;
901 
902 			case MLX4_STEERING_MODE_A0:
903 				break;
904 			}
905 			if (err)
906 				en_err(priv, "Failed disabling multicast promiscuous mode\n");
907 			priv->flags &= ~MLX4_EN_FLAG_MC_PROMISC;
908 		}
909 
910 		err = mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, 0,
911 					  0, MLX4_MCAST_DISABLE);
912 		if (err)
913 			en_err(priv, "Failed disabling multicast filter\n");
914 
915 		/* Flush mcast filter and init it with broadcast address */
916 		mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, ETH_BCAST,
917 				    1, MLX4_MCAST_CONFIG);
918 
919 		/* Update multicast list - we cache all addresses so they won't
920 		 * change while HW is updated holding the command semaphor */
921 		mlx4_en_cache_mclist(dev);
922 		list_for_each_entry(addr_list, &priv->mc_list, list) {
923 			mcast_addr = mlx4_mac_to_u64(addr_list->addr);
924 			mlx4_SET_MCAST_FLTR(mdev->dev, priv->port,
925 					mcast_addr, 0, MLX4_MCAST_CONFIG);
926 		}
927 		err = mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, 0,
928 					  0, MLX4_MCAST_ENABLE);
929 		if (err)
930 			en_err(priv, "Failed enabling multicast filter\n");
931 
932 		update_addr_list_flags(priv, &priv->curr_mc_list, &priv->mc_list);
933 
934 		list_for_each_entry_safe(addr_list, tmp, &priv->curr_mc_list, list) {
935 			if (addr_list->action == MLX4_ADDR_LIST_REM) {
936 				/* detach this address and delete from list */
937 				memcpy(&mc_list[10], addr_list->addr, ETH_ALEN);
938 				mc_list[5] = priv->port;
939 				err = mlx4_multicast_detach(mdev->dev,
940 							    &priv->rss_map.indir_qp,
941 							    mc_list,
942 							    MLX4_PROT_ETH,
943 							    addr_list->reg_id);
944 				if (err)
945 					en_err(priv, "Fail to detach multicast address\n");
946 
947 				if (addr_list->tunnel_reg_id) {
948 					err = mlx4_flow_detach(priv->mdev->dev, addr_list->tunnel_reg_id);
949 					if (err)
950 						en_err(priv, "Failed to detach multicast address\n");
951 				}
952 
953 				/* remove from list */
954 				list_del(&addr_list->list);
955 				kfree(addr_list);
956 			} else if (addr_list->action == MLX4_ADDR_LIST_ADD) {
957 				/* attach the address */
958 				memcpy(&mc_list[10], addr_list->addr, ETH_ALEN);
959 				/* needed for B0 steering support */
960 				mc_list[5] = priv->port;
961 				err = mlx4_multicast_attach(mdev->dev,
962 							    &priv->rss_map.indir_qp,
963 							    mc_list,
964 							    priv->port, 0,
965 							    MLX4_PROT_ETH,
966 							    &addr_list->reg_id);
967 				if (err)
968 					en_err(priv, "Fail to attach multicast address\n");
969 
970 				err = mlx4_en_tunnel_steer_add(priv, &mc_list[10], priv->base_qpn,
971 							       &addr_list->tunnel_reg_id);
972 				if (err)
973 					en_err(priv, "Failed to attach multicast address\n");
974 			}
975 		}
976 	}
977 }
978 
979 static void mlx4_en_do_unicast(struct mlx4_en_priv *priv,
980 			       if_t 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 	if_t 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 			if_setbaudrate(priv->dev,
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 ((if_getflags(dev) & 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)if_getmtu(priv->dev), 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 			if_setbaudrate(priv->dev, 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 			if_setbaudrate(priv->dev,
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(if_t dev)
1262 {
1263 	struct mlx4_en_priv *priv = mlx4_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 	if_setmtu(dev, min(if_getmtu(dev), 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         if_setdrvflagbits(dev, IFF_DRV_RUNNING, IFF_DRV_OACTIVE);
1424 #ifdef CONFIG_DEBUG_FS
1425 	mlx4_en_create_debug_files(priv);
1426 #endif
1427         callout_reset(&priv->watchdog_timer, MLX4_EN_WATCHDOG_TIMEOUT,
1428                     mlx4_en_watchdog_timeout, priv);
1429 
1430 
1431 	return 0;
1432 
1433 tx_err:
1434 	while (tx_index--) {
1435 		mlx4_en_deactivate_tx_ring(priv, priv->tx_ring[tx_index]);
1436 		mlx4_en_deactivate_cq(priv, priv->tx_cq[tx_index]);
1437 	}
1438 	mlx4_en_destroy_drop_qp(priv);
1439 rss_err:
1440 	mlx4_en_release_rss_steer(priv);
1441 mac_err:
1442 	mlx4_en_put_qp(priv);
1443 cq_err:
1444 	while (rx_index--)
1445 		mlx4_en_deactivate_cq(priv, priv->rx_cq[rx_index]);
1446 	for (i = 0; i < priv->rx_ring_num; i++)
1447 		mlx4_en_deactivate_rx_ring(priv, priv->rx_ring[i]);
1448 
1449 	return err; /* need to close devices */
1450 }
1451 
1452 
1453 void mlx4_en_stop_port(if_t dev)
1454 {
1455 	struct mlx4_en_priv *priv = mlx4_netdev_priv(dev);
1456 	struct mlx4_en_dev *mdev = priv->mdev;
1457 	struct mlx4_en_addr_list *addr_list, *tmp;
1458 	int i;
1459 	u8 mc_list[16] = {0};
1460 
1461 	if (!priv->port_up) {
1462 		en_dbg(DRV, priv, "stop port called while port already down\n");
1463 		return;
1464 	}
1465 
1466 #ifdef CONFIG_DEBUG_FS
1467 	mlx4_en_delete_debug_files(priv);
1468 #endif
1469 
1470 	/* close port*/
1471 	mlx4_CLOSE_PORT(mdev->dev, priv->port);
1472 
1473 	/* Set port as not active */
1474 	priv->port_up = false;
1475 	priv->counter_index = MLX4_SINK_COUNTER_INDEX(mdev->dev);
1476 
1477 	/* Promsicuous mode */
1478 	if (mdev->dev->caps.steering_mode ==
1479 	    MLX4_STEERING_MODE_DEVICE_MANAGED) {
1480 		priv->flags &= ~(MLX4_EN_FLAG_PROMISC |
1481 				 MLX4_EN_FLAG_MC_PROMISC);
1482 		mlx4_flow_steer_promisc_remove(mdev->dev,
1483 					       priv->port,
1484 					       MLX4_FS_ALL_DEFAULT);
1485 		mlx4_flow_steer_promisc_remove(mdev->dev,
1486 					       priv->port,
1487 					       MLX4_FS_MC_DEFAULT);
1488 	} else if (priv->flags & MLX4_EN_FLAG_PROMISC) {
1489 		priv->flags &= ~MLX4_EN_FLAG_PROMISC;
1490 
1491 		/* Disable promiscouos mode */
1492 		mlx4_unicast_promisc_remove(mdev->dev, priv->base_qpn,
1493 					    priv->port);
1494 
1495 		/* Disable Multicast promisc */
1496 		if (priv->flags & MLX4_EN_FLAG_MC_PROMISC) {
1497 			mlx4_multicast_promisc_remove(mdev->dev, priv->base_qpn,
1498 						      priv->port);
1499 			priv->flags &= ~MLX4_EN_FLAG_MC_PROMISC;
1500 		}
1501 	}
1502 
1503 	/* Detach All unicasts */
1504 	list_for_each_entry(addr_list, &priv->curr_uc_list, list) {
1505 		mlx4_en_uc_steer_release(priv, addr_list->addr,
1506 					 priv->rss_map.indir_qp.qpn,
1507 					 addr_list->reg_id);
1508 	}
1509 	mlx4_en_clear_uclist(dev);
1510 	list_for_each_entry_safe(addr_list, tmp, &priv->curr_uc_list, list) {
1511 		list_del(&addr_list->list);
1512 		kfree(addr_list);
1513 	}
1514 
1515 	/* Detach All multicasts */
1516 	memset(&mc_list[10], 0xff, ETH_ALEN);
1517 	mc_list[5] = priv->port; /* needed for B0 steering support */
1518 	mlx4_multicast_detach(mdev->dev, &priv->rss_map.indir_qp, mc_list,
1519 			      MLX4_PROT_ETH, priv->broadcast_id);
1520 	list_for_each_entry(addr_list, &priv->curr_mc_list, list) {
1521 		memcpy(&mc_list[10], addr_list->addr, ETH_ALEN);
1522 		mc_list[5] = priv->port;
1523 		mlx4_multicast_detach(mdev->dev, &priv->rss_map.indir_qp,
1524 				      mc_list, MLX4_PROT_ETH, addr_list->reg_id);
1525 	}
1526 	mlx4_en_clear_mclist(dev);
1527 	list_for_each_entry_safe(addr_list, tmp, &priv->curr_mc_list, list) {
1528 		list_del(&addr_list->list);
1529 		kfree(addr_list);
1530 	}
1531 
1532 	/* Flush multicast filter */
1533 	mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, 0, 1, MLX4_MCAST_CONFIG);
1534 	mlx4_en_destroy_drop_qp(priv);
1535 
1536 	/* Free TX Rings */
1537 	for (i = 0; i < priv->tx_ring_num; i++) {
1538 		mlx4_en_deactivate_tx_ring(priv, priv->tx_ring[i]);
1539 		mlx4_en_deactivate_cq(priv, priv->tx_cq[i]);
1540 	}
1541 	msleep(10);
1542 
1543 	for (i = 0; i < priv->tx_ring_num; i++)
1544 		mlx4_en_free_tx_buf(dev, priv->tx_ring[i]);
1545 
1546 	/* Free RSS qps */
1547 	mlx4_en_release_rss_steer(priv);
1548 
1549 	/* Unregister Mac address for the port */
1550 	mlx4_en_put_qp(priv);
1551 	mdev->mac_removed[priv->port] = 1;
1552 
1553 	/* Free RX Rings */
1554 	for (i = 0; i < priv->rx_ring_num; i++) {
1555 		struct mlx4_en_cq *cq = priv->rx_cq[i];
1556 		mlx4_en_deactivate_rx_ring(priv, priv->rx_ring[i]);
1557 		mlx4_en_deactivate_cq(priv, cq);
1558 	}
1559 
1560         callout_stop(&priv->watchdog_timer);
1561 
1562         if_setdrvflagbits(dev, 0, IFF_DRV_RUNNING | IFF_DRV_OACTIVE);
1563 }
1564 
1565 static void mlx4_en_restart(struct work_struct *work)
1566 {
1567 	struct mlx4_en_priv *priv = container_of(work, struct mlx4_en_priv,
1568 						 watchdog_task);
1569 	struct mlx4_en_dev *mdev = priv->mdev;
1570 	if_t dev = priv->dev;
1571 	struct mlx4_en_tx_ring *ring;
1572 	int i;
1573 
1574 
1575 	if (priv->blocked == 0 || priv->port_up == 0)
1576 		return;
1577 	for (i = 0; i < priv->tx_ring_num; i++) {
1578 		int watchdog_time;
1579 
1580 		ring = priv->tx_ring[i];
1581 		watchdog_time = READ_ONCE(ring->watchdog_time);
1582 		if (watchdog_time != 0 &&
1583 		    time_after(ticks, ring->watchdog_time))
1584 			goto reset;
1585 	}
1586 	return;
1587 
1588 reset:
1589 	priv->port_stats.tx_timeout++;
1590 	en_dbg(DRV, priv, "Watchdog task called for port %d\n", priv->port);
1591 
1592 	mutex_lock(&mdev->state_lock);
1593 	if (priv->port_up) {
1594 		mlx4_en_stop_port(dev);
1595                 //for (i = 0; i < priv->tx_ring_num; i++)
1596                 //        netdev_tx_reset_queue(priv->tx_ring[i]->tx_queue);
1597 		if (mlx4_en_start_port(dev))
1598 			en_err(priv, "Failed restarting port %d\n", priv->port);
1599 	}
1600 	mutex_unlock(&mdev->state_lock);
1601 }
1602 
1603 static void mlx4_en_clear_stats(if_t dev)
1604 {
1605 	struct mlx4_en_priv *priv = mlx4_netdev_priv(dev);
1606 	struct mlx4_en_dev *mdev = priv->mdev;
1607 	int i;
1608 
1609 	if (!mlx4_is_slave(mdev->dev))
1610 		if (mlx4_en_DUMP_ETH_STATS(mdev, priv->port, 1))
1611 			en_dbg(HW, priv, "Failed dumping statistics\n");
1612 
1613 	memset(&priv->pstats, 0, sizeof(priv->pstats));
1614 	memset(&priv->pkstats, 0, sizeof(priv->pkstats));
1615 	memset(&priv->port_stats, 0, sizeof(priv->port_stats));
1616 	memset(&priv->vport_stats, 0, sizeof(priv->vport_stats));
1617 
1618 	for (i = 0; i < priv->tx_ring_num; i++) {
1619 		priv->tx_ring[i]->bytes = 0;
1620 		priv->tx_ring[i]->packets = 0;
1621 		priv->tx_ring[i]->tx_csum = 0;
1622 		priv->tx_ring[i]->oversized_packets = 0;
1623 	}
1624 	for (i = 0; i < priv->rx_ring_num; i++) {
1625 		priv->rx_ring[i]->bytes = 0;
1626 		priv->rx_ring[i]->packets = 0;
1627 		priv->rx_ring[i]->csum_ok = 0;
1628 		priv->rx_ring[i]->csum_none = 0;
1629 	}
1630 }
1631 
1632 static void mlx4_en_open(void* arg)
1633 {
1634 
1635         struct mlx4_en_priv *priv;
1636         struct mlx4_en_dev *mdev;
1637         if_t dev;
1638         int err = 0;
1639 
1640         priv = arg;
1641         mdev = priv->mdev;
1642         dev = priv->dev;
1643 
1644 
1645 	mutex_lock(&mdev->state_lock);
1646 
1647 	if (!mdev->device_up) {
1648 		en_err(priv, "Cannot open - device down/disabled\n");
1649 		goto out;
1650 	}
1651 
1652 	/* Reset HW statistics and SW counters */
1653 	mlx4_en_clear_stats(dev);
1654 
1655 	err = mlx4_en_start_port(dev);
1656 	if (err)
1657 		en_err(priv, "Failed starting port:%d\n", priv->port);
1658 
1659 out:
1660 	mutex_unlock(&mdev->state_lock);
1661 	return;
1662 }
1663 
1664 void mlx4_en_free_resources(struct mlx4_en_priv *priv)
1665 {
1666 	int i;
1667 
1668 #ifdef CONFIG_RFS_ACCEL
1669 	if (priv->dev->rx_cpu_rmap) {
1670 		free_irq_cpu_rmap(priv->dev->rx_cpu_rmap);
1671 		priv->dev->rx_cpu_rmap = NULL;
1672 	}
1673 #endif
1674 
1675 	for (i = 0; i < priv->tx_ring_num; i++) {
1676 		if (priv->tx_ring && priv->tx_ring[i])
1677 			mlx4_en_destroy_tx_ring(priv, &priv->tx_ring[i]);
1678 		if (priv->tx_cq && priv->tx_cq[i])
1679 			mlx4_en_destroy_cq(priv, &priv->tx_cq[i]);
1680 	}
1681 
1682 	for (i = 0; i < priv->rx_ring_num; i++) {
1683 		if (priv->rx_ring[i])
1684 			mlx4_en_destroy_rx_ring(priv, &priv->rx_ring[i],
1685 				priv->prof->rx_ring_size);
1686 		if (priv->rx_cq[i])
1687 			mlx4_en_destroy_cq(priv, &priv->rx_cq[i]);
1688 	}
1689 
1690 	if (priv->stat_sysctl != NULL)
1691 		sysctl_ctx_free(&priv->stat_ctx);
1692 }
1693 
1694 int mlx4_en_alloc_resources(struct mlx4_en_priv *priv)
1695 {
1696 	struct mlx4_en_port_profile *prof = priv->prof;
1697 	int i;
1698 	int node = 0;
1699 
1700 	/* Create rx Rings */
1701 	for (i = 0; i < priv->rx_ring_num; i++) {
1702 		if (mlx4_en_create_cq(priv, &priv->rx_cq[i],
1703 				      prof->rx_ring_size, i, RX, node))
1704 			goto err;
1705 
1706 		if (mlx4_en_create_rx_ring(priv, &priv->rx_ring[i],
1707 					   prof->rx_ring_size, node))
1708 			goto err;
1709 	}
1710 
1711 	/* Create tx Rings */
1712 	for (i = 0; i < priv->tx_ring_num; i++) {
1713 		if (mlx4_en_create_cq(priv, &priv->tx_cq[i],
1714 				      prof->tx_ring_size, i, TX, node))
1715 			goto err;
1716 
1717 		if (mlx4_en_create_tx_ring(priv, &priv->tx_ring[i],
1718 					   prof->tx_ring_size, TXBB_SIZE, node, i))
1719 			goto err;
1720 	}
1721 
1722 #ifdef CONFIG_RFS_ACCEL
1723 	priv->dev->rx_cpu_rmap = alloc_irq_cpu_rmap(priv->rx_ring_num);
1724 	if (!priv->dev->rx_cpu_rmap)
1725 		goto err;
1726 #endif
1727         /* Re-create stat sysctls in case the number of rings changed. */
1728 	mlx4_en_sysctl_stat(priv);
1729 	return 0;
1730 
1731 err:
1732 	en_err(priv, "Failed to allocate NIC resources\n");
1733 	for (i = 0; i < priv->rx_ring_num; i++) {
1734 		if (priv->rx_ring[i])
1735 			mlx4_en_destroy_rx_ring(priv, &priv->rx_ring[i],
1736 						prof->rx_ring_size);
1737 		if (priv->rx_cq[i])
1738 			mlx4_en_destroy_cq(priv, &priv->rx_cq[i]);
1739 	}
1740 	for (i = 0; i < priv->tx_ring_num; i++) {
1741 		if (priv->tx_ring[i])
1742 			mlx4_en_destroy_tx_ring(priv, &priv->tx_ring[i]);
1743 		if (priv->tx_cq[i])
1744 			mlx4_en_destroy_cq(priv, &priv->tx_cq[i]);
1745 	}
1746 	priv->port_up = false;
1747 	return -ENOMEM;
1748 }
1749 
1750 struct en_port_attribute {
1751 	struct attribute attr;
1752 	ssize_t (*show)(struct en_port *, struct en_port_attribute *, char *buf);
1753 	ssize_t (*store)(struct en_port *, struct en_port_attribute *, char *buf, size_t count);
1754 };
1755 
1756 #define PORT_ATTR_RO(_name) \
1757 struct en_port_attribute en_port_attr_##_name = __ATTR_RO(_name)
1758 
1759 #define EN_PORT_ATTR(_name, _mode, _show, _store) \
1760 struct en_port_attribute en_port_attr_##_name = __ATTR(_name, _mode, _show, _store)
1761 
1762 void mlx4_en_destroy_netdev(if_t dev)
1763 {
1764 	struct mlx4_en_priv *priv = mlx4_netdev_priv(dev);
1765 	struct mlx4_en_dev *mdev = priv->mdev;
1766 
1767 	en_dbg(DRV, priv, "Destroying netdev on port:%d\n", priv->port);
1768 
1769 	/* don't allow more IOCTLs */
1770 	priv->gone = 1;
1771 
1772 	/* XXX wait a bit to allow IOCTL handlers to complete */
1773 	pause("W", hz);
1774 
1775 	if (priv->vlan_attach != NULL)
1776 		EVENTHANDLER_DEREGISTER(vlan_config, priv->vlan_attach);
1777 	if (priv->vlan_detach != NULL)
1778 		EVENTHANDLER_DEREGISTER(vlan_unconfig, priv->vlan_detach);
1779 
1780 	mutex_lock(&mdev->state_lock);
1781 	mlx4_en_stop_port(dev);
1782 	mutex_unlock(&mdev->state_lock);
1783 
1784 	/* Unregister device - this will close the port if it was up */
1785 	if (priv->registered)
1786 		ether_ifdetach(dev);
1787 
1788 	if (priv->allocated)
1789 		mlx4_free_hwq_res(mdev->dev, &priv->res, MLX4_EN_PAGE_SIZE);
1790 
1791 	cancel_delayed_work(&priv->stats_task);
1792 	cancel_delayed_work(&priv->service_task);
1793 	/* flush any pending task for this netdev */
1794 	flush_workqueue(mdev->workqueue);
1795         callout_drain(&priv->watchdog_timer);
1796 
1797 	/* Detach the netdev so tasks would not attempt to access it */
1798 	mutex_lock(&mdev->state_lock);
1799 	mdev->pndev[priv->port] = NULL;
1800 	mutex_unlock(&mdev->state_lock);
1801 
1802 
1803 	mlx4_en_free_resources(priv);
1804 
1805 	/* freeing the sysctl conf cannot be called from within mlx4_en_free_resources */
1806 	if (priv->conf_sysctl != NULL)
1807 		sysctl_ctx_free(&priv->conf_ctx);
1808 
1809 	kfree(priv->tx_ring);
1810 	kfree(priv->tx_cq);
1811 
1812         kfree(priv);
1813         if_free(dev);
1814 
1815 }
1816 
1817 static int mlx4_en_change_mtu(if_t dev, int new_mtu)
1818 {
1819 	struct mlx4_en_priv *priv = mlx4_netdev_priv(dev);
1820 	struct mlx4_en_dev *mdev = priv->mdev;
1821 	int err = 0;
1822 
1823 	en_dbg(DRV, priv, "Change MTU called - current:%u new:%u\n",
1824 	       (unsigned)if_getmtu(dev), (unsigned)new_mtu);
1825 
1826 	if ((new_mtu < MLX4_EN_MIN_MTU) || (new_mtu > priv->max_mtu)) {
1827 		en_err(priv, "Bad MTU size:%d, max %u.\n", new_mtu,
1828 		    priv->max_mtu);
1829 		return -EPERM;
1830 	}
1831 	mutex_lock(&mdev->state_lock);
1832 	if_setmtu(dev, new_mtu);
1833 	if (if_getdrvflags(dev) & IFF_DRV_RUNNING) {
1834 		if (!mdev->device_up) {
1835 			/* NIC is probably restarting - let watchdog task reset
1836 			 *                          * the port */
1837 			en_dbg(DRV, priv, "Change MTU called with card down!?\n");
1838 		} else {
1839 			mlx4_en_stop_port(dev);
1840 			err = mlx4_en_start_port(dev);
1841 			if (err) {
1842 				en_err(priv, "Failed restarting port:%d\n",
1843 						priv->port);
1844 				queue_work(mdev->workqueue, &priv->watchdog_task);
1845 			}
1846 		}
1847 	}
1848 	mutex_unlock(&mdev->state_lock);
1849 	return 0;
1850 }
1851 
1852 static int mlx4_en_calc_media(struct mlx4_en_priv *priv)
1853 {
1854 	int trans_type;
1855 	int active;
1856 
1857 	active = IFM_ETHER;
1858 	if (priv->last_link_state == MLX4_DEV_EVENT_PORT_DOWN)
1859 		return (active);
1860 	active |= IFM_FDX;
1861 	trans_type = priv->port_state.transceiver;
1862 	/* XXX I don't know all of the transceiver values. */
1863 	switch (priv->port_state.link_speed) {
1864 	case 100:
1865 		active |= IFM_100_T;
1866 		break;
1867 	case 1000:
1868 		active |= IFM_1000_T;
1869 		break;
1870 	case 10000:
1871 		if (trans_type > 0 && trans_type <= 0xC)
1872 			active |= IFM_10G_SR;
1873 		else if (trans_type == 0x80 || trans_type == 0)
1874 			active |= IFM_10G_CX4;
1875 		break;
1876 	case 40000:
1877 		active |= IFM_40G_CR4;
1878 		break;
1879 	}
1880 	if (priv->prof->tx_pause)
1881 		active |= IFM_ETH_TXPAUSE;
1882 	if (priv->prof->rx_pause)
1883 		active |= IFM_ETH_RXPAUSE;
1884 
1885 	return (active);
1886 }
1887 
1888 static void mlx4_en_media_status(if_t dev, struct ifmediareq *ifmr)
1889 {
1890 	struct mlx4_en_priv *priv;
1891 
1892 	priv = if_getsoftc(dev);
1893 	ifmr->ifm_status = IFM_AVALID;
1894 	if (priv->last_link_state != MLX4_DEV_EVENT_PORT_DOWN)
1895 		ifmr->ifm_status |= IFM_ACTIVE;
1896 	ifmr->ifm_active = mlx4_en_calc_media(priv);
1897 
1898 	return;
1899 }
1900 
1901 static int mlx4_en_media_change(if_t dev)
1902 {
1903 	struct mlx4_en_priv *priv;
1904         struct ifmedia *ifm;
1905 	int rxpause;
1906 	int txpause;
1907 	int error;
1908 
1909 	priv = if_getsoftc(dev);
1910 	ifm = &priv->media;
1911 	rxpause = txpause = 0;
1912 	error = 0;
1913 
1914 	if (IFM_TYPE(ifm->ifm_media) != IFM_ETHER)
1915 		return (EINVAL);
1916         switch (IFM_SUBTYPE(ifm->ifm_media)) {
1917         case IFM_AUTO:
1918 		break;
1919 	case IFM_10G_SR:
1920 	case IFM_10G_CX4:
1921 	case IFM_1000_T:
1922 	case IFM_40G_CR4:
1923 		if ((IFM_SUBTYPE(ifm->ifm_media)
1924 			== IFM_SUBTYPE(mlx4_en_calc_media(priv)))
1925 			&& (ifm->ifm_media & IFM_FDX))
1926 			break;
1927 		/* Fallthrough */
1928 	default:
1929                 printf("%s: Only auto media type\n", if_name(dev));
1930                 return (EINVAL);
1931 	}
1932 	/* Allow user to set/clear pause */
1933 	if (IFM_OPTIONS(ifm->ifm_media) & IFM_ETH_RXPAUSE)
1934 		rxpause = 1;
1935 	if (IFM_OPTIONS(ifm->ifm_media) & IFM_ETH_TXPAUSE)
1936 		txpause = 1;
1937 	if (priv->prof->tx_pause != txpause || priv->prof->rx_pause != rxpause) {
1938 		priv->prof->tx_pause = txpause;
1939 		priv->prof->rx_pause = rxpause;
1940 		error = -mlx4_SET_PORT_general(priv->mdev->dev, priv->port,
1941 		     priv->rx_mb_size + ETHER_CRC_LEN, priv->prof->tx_pause,
1942 		     priv->prof->tx_ppp, priv->prof->rx_pause,
1943 		     priv->prof->rx_ppp);
1944 	}
1945 	return (error);
1946 }
1947 
1948 static int mlx4_en_ioctl(if_t dev, u_long command, caddr_t data)
1949 {
1950 	struct mlx4_en_priv *priv;
1951 	struct mlx4_en_dev *mdev;
1952 	struct ifreq *ifr;
1953 	int error;
1954 	int mask;
1955 	struct ifrsskey *ifrk;
1956 	const u32 *key;
1957 	struct ifrsshash *ifrh;
1958 	u8 rss_mask;
1959 
1960 	error = 0;
1961 	mask = 0;
1962 	priv = if_getsoftc(dev);
1963 
1964 	/* check if detaching */
1965 	if (priv == NULL || priv->gone != 0)
1966 		return (ENXIO);
1967 
1968 	mdev = priv->mdev;
1969 	ifr = (struct ifreq *) data;
1970 
1971 	switch (command) {
1972 	case SIOCSIFMTU:
1973 		error = -mlx4_en_change_mtu(dev, ifr->ifr_mtu);
1974 		break;
1975 	case SIOCSIFFLAGS:
1976 		if (if_getflags(dev) & IFF_UP) {
1977 			if ((if_getdrvflags(dev) & IFF_DRV_RUNNING) == 0) {
1978 				mutex_lock(&mdev->state_lock);
1979 				mlx4_en_start_port(dev);
1980 				mutex_unlock(&mdev->state_lock);
1981 			} else {
1982 				mlx4_en_set_rx_mode(dev);
1983 			}
1984 		} else {
1985 			mutex_lock(&mdev->state_lock);
1986 			if (if_getdrvflags(dev) & IFF_DRV_RUNNING) {
1987 				mlx4_en_stop_port(dev);
1988 				if_link_state_change(dev, LINK_STATE_DOWN);
1989 			}
1990 			mutex_unlock(&mdev->state_lock);
1991 		}
1992 		break;
1993 	case SIOCADDMULTI:
1994 	case SIOCDELMULTI:
1995 		mlx4_en_set_rx_mode(dev);
1996 		break;
1997 	case SIOCSIFMEDIA:
1998 	case SIOCGIFMEDIA:
1999 		error = ifmedia_ioctl(dev, ifr, &priv->media, command);
2000 		break;
2001 	case SIOCSIFCAP:
2002 		mutex_lock(&mdev->state_lock);
2003 		mask = ifr->ifr_reqcap ^ if_getcapenable(dev);
2004 		if (mask & IFCAP_TXCSUM) {
2005 			if_togglecapenable(dev, IFCAP_TXCSUM);
2006 			if_togglehwassist(dev, CSUM_TCP | CSUM_UDP | CSUM_IP);
2007 
2008 			if (IFCAP_TSO4 & if_getcapenable(dev) &&
2009 			    !(IFCAP_TXCSUM & if_getcapenable(dev))) {
2010 				mask &= ~IFCAP_TSO4;
2011 				if_setcapenablebit(dev, 0, IFCAP_TSO4);
2012 				if_sethwassistbits(dev, 0, CSUM_IP_TSO);
2013 				if_printf(dev,
2014 				    "tso4 disabled due to -txcsum.\n");
2015 			}
2016 		}
2017 		if (mask & IFCAP_TXCSUM_IPV6) {
2018 			if_togglecapenable(dev, IFCAP_TXCSUM_IPV6);
2019 			if_togglehwassist(dev, (CSUM_UDP_IPV6 | CSUM_TCP_IPV6));
2020 
2021 			if (IFCAP_TSO6 & if_getcapenable(dev) &&
2022 			    !(IFCAP_TXCSUM_IPV6 & if_getcapenable(dev))) {
2023 				mask &= ~IFCAP_TSO6;
2024 				if_setcapenablebit(dev, 0, IFCAP_TSO6);
2025 				if_sethwassistbits(dev, 0, CSUM_IP6_TSO);
2026 				if_printf(dev,
2027 				    "tso6 disabled due to -txcsum6.\n");
2028 			}
2029 		}
2030 		if (mask & IFCAP_RXCSUM)
2031 			if_togglecapenable(dev, IFCAP_RXCSUM);
2032 		if (mask & IFCAP_RXCSUM_IPV6)
2033 			if_togglecapenable(dev, IFCAP_RXCSUM_IPV6);
2034 
2035 		if (mask & IFCAP_TSO4) {
2036 			if (!(IFCAP_TSO4 & if_getcapenable(dev)) &&
2037 			    !(IFCAP_TXCSUM & if_getcapenable(dev))) {
2038 				if_printf(dev, "enable txcsum first.\n");
2039 				error = EAGAIN;
2040 				goto out;
2041 			}
2042 			if_togglecapenable(dev, IFCAP_TSO4);
2043 			if_togglehwassist(dev, CSUM_IP_TSO);
2044 		}
2045 		if (mask & IFCAP_TSO6) {
2046 			if (!(IFCAP_TSO6 & if_getcapenable(dev)) &&
2047 			    !(IFCAP_TXCSUM_IPV6 & if_getcapenable(dev))) {
2048 				if_printf(dev, "enable txcsum6 first.\n");
2049 				error = EAGAIN;
2050 				goto out;
2051 			}
2052 			if_togglecapenable(dev, IFCAP_TSO6);
2053 			if_togglehwassist(dev, CSUM_IP6_TSO);
2054 		}
2055 		if (mask & IFCAP_LRO)
2056 			if_togglecapenable(dev, IFCAP_LRO);
2057 		if (mask & IFCAP_VLAN_HWTAGGING)
2058 			if_togglecapenable(dev, IFCAP_VLAN_HWTAGGING);
2059 		if (mask & IFCAP_VLAN_HWFILTER)
2060 			if_togglecapenable(dev, IFCAP_VLAN_HWFILTER);
2061 		if (mask & IFCAP_WOL_MAGIC)
2062 			if_togglecapenable(dev, IFCAP_WOL_MAGIC);
2063 		if (if_getdrvflags(dev) & IFF_DRV_RUNNING)
2064 			mlx4_en_start_port(dev);
2065 out:
2066 		mutex_unlock(&mdev->state_lock);
2067 		VLAN_CAPABILITIES(dev);
2068 		break;
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 	case SIOCGIFRSSKEY:
2093 		ifrk = (struct ifrsskey *)data;
2094 		ifrk->ifrk_func = RSS_FUNC_TOEPLITZ;
2095 		mutex_lock(&mdev->state_lock);
2096 		key = mlx4_en_get_rss_key(priv, &ifrk->ifrk_keylen);
2097 		if (ifrk->ifrk_keylen > RSS_KEYLEN)
2098 			error = EINVAL;
2099 		else
2100 			memcpy(ifrk->ifrk_key, key, ifrk->ifrk_keylen);
2101 		mutex_unlock(&mdev->state_lock);
2102 		break;
2103 
2104 	case SIOCGIFRSSHASH:
2105 		mutex_lock(&mdev->state_lock);
2106 		rss_mask = mlx4_en_get_rss_mask(priv);
2107 		mutex_unlock(&mdev->state_lock);
2108 		ifrh = (struct ifrsshash *)data;
2109 		ifrh->ifrh_func = RSS_FUNC_TOEPLITZ;
2110 		ifrh->ifrh_types = 0;
2111 		if (rss_mask & MLX4_RSS_IPV4)
2112 			ifrh->ifrh_types |= RSS_TYPE_IPV4;
2113 		if (rss_mask & MLX4_RSS_TCP_IPV4)
2114 			ifrh->ifrh_types |= RSS_TYPE_TCP_IPV4;
2115 		if (rss_mask & MLX4_RSS_IPV6)
2116 			ifrh->ifrh_types |= RSS_TYPE_IPV6;
2117 		if (rss_mask & MLX4_RSS_TCP_IPV6)
2118 			ifrh->ifrh_types |= RSS_TYPE_TCP_IPV6;
2119 		if (rss_mask & MLX4_RSS_UDP_IPV4)
2120 			ifrh->ifrh_types |= RSS_TYPE_UDP_IPV4;
2121 		if (rss_mask & MLX4_RSS_UDP_IPV6)
2122 			ifrh->ifrh_types |= RSS_TYPE_UDP_IPV6;
2123 		break;
2124 
2125 	default:
2126 		error = ether_ioctl(dev, command, data);
2127 		break;
2128 	}
2129 
2130 	return (error);
2131 }
2132 
2133 
2134 int mlx4_en_init_netdev(struct mlx4_en_dev *mdev, int port,
2135 			struct mlx4_en_port_profile *prof)
2136 {
2137 	if_t dev;
2138 	struct mlx4_en_priv *priv;
2139 	uint32_t hwassist;
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 	if_setsoftc(dev, priv);
2152 	if_initname(dev, "mlxen", (device_get_unit(
2153 	    mdev->pdev->dev.bsddev) * MLX4_MAX_PORTS) + port - 1);
2154 	if_setmtu(dev, ETHERMTU);
2155 	if_setinitfn(dev, mlx4_en_open);
2156 	if_setflags(dev, IFF_BROADCAST | IFF_SIMPLEX | IFF_MULTICAST);
2157 	if_setioctlfn(dev, mlx4_en_ioctl);
2158 	if_settransmitfn(dev, mlx4_en_transmit);
2159 	if_setqflushfn(dev, mlx4_en_qflush);
2160 	if_setsendqlen(dev, 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 	if_setcapabilitiesbit(dev, IFCAP_HWCSUM | IFCAP_HWCSUM_IPV6 |
2255 	    IFCAP_VLAN_MTU | IFCAP_VLAN_HWTAGGING |
2256 	    IFCAP_VLAN_HWCSUM | IFCAP_VLAN_HWFILTER |
2257 	    IFCAP_LINKSTATE | IFCAP_JUMBO_MTU |
2258 	    IFCAP_LRO | IFCAP_HWSTATS, 0);
2259 
2260 	if (mdev->LSO_support)
2261 		if_setcapabilitiesbit(dev, IFCAP_TSO4 | IFCAP_TSO6 | IFCAP_VLAN_HWTSO, 0);
2262 
2263 	/* set TSO limits so that we don't have to drop TX packets */
2264 	if_sethwtsomax(dev, MLX4_EN_TX_MAX_PAYLOAD_SIZE - (ETHER_HDR_LEN + ETHER_VLAN_ENCAP_LEN) /* hdr */);
2265 	if_sethwtsomaxsegcount(dev, MLX4_EN_TX_MAX_MBUF_FRAGS - 1 /* hdr */);
2266 	if_sethwtsomaxsegsize(dev, MLX4_EN_TX_MAX_MBUF_SIZE);
2267 
2268 	if_setcapenable(dev, if_getcapabilities(dev));
2269 
2270 	hwassist = 0;
2271 	if (if_getcapenable(dev) & (IFCAP_TSO4 | IFCAP_TSO6))
2272 		hwassist |= CSUM_TSO;
2273 	if (if_getcapenable(dev) & IFCAP_TXCSUM)
2274 		hwassist |= (CSUM_TCP | CSUM_UDP | CSUM_IP);
2275 	if (if_getcapenable(dev) & IFCAP_TXCSUM_IPV6)
2276 		hwassist |= (CSUM_UDP_IPV6 | CSUM_TCP_IPV6);
2277 	if_sethwassist(dev, hwassist);
2278 
2279 
2280         /* Register for VLAN events */
2281 	priv->vlan_attach = EVENTHANDLER_REGISTER(vlan_config,
2282             mlx4_en_vlan_rx_add_vid, priv, EVENTHANDLER_PRI_FIRST);
2283 	priv->vlan_detach = EVENTHANDLER_REGISTER(vlan_unconfig,
2284             mlx4_en_vlan_rx_kill_vid, priv, EVENTHANDLER_PRI_FIRST);
2285 
2286 	mdev->pndev[priv->port] = dev;
2287 
2288 	priv->last_link_state = MLX4_DEV_EVENT_PORT_DOWN;
2289         mlx4_en_set_default_moderation(priv);
2290 
2291 	/* Set default MAC */
2292 	for (i = 0; i < ETHER_ADDR_LEN; i++)
2293 		dev_addr[ETHER_ADDR_LEN - 1 - i] = (u8) (priv->mac >> (8 * i));
2294 
2295 
2296 	if_link_state_change(dev, LINK_STATE_DOWN);
2297 	ifmedia_init(&priv->media, IFM_IMASK | IFM_ETH_FMASK,
2298 	    mlx4_en_media_change, mlx4_en_media_status);
2299 	ifmedia_add(&priv->media, IFM_ETHER | IFM_FDX | IFM_1000_T, 0, NULL);
2300 	ifmedia_add(&priv->media, IFM_ETHER | IFM_FDX | IFM_10G_SR, 0, NULL);
2301 	ifmedia_add(&priv->media, IFM_ETHER | IFM_FDX | IFM_10G_CX4, 0, NULL);
2302 	ifmedia_add(&priv->media, IFM_ETHER | IFM_FDX | IFM_40G_CR4, 0, NULL);
2303 	ifmedia_add(&priv->media, IFM_ETHER | IFM_AUTO, 0, NULL);
2304 	ifmedia_set(&priv->media, IFM_ETHER | IFM_AUTO);
2305 
2306 	DEBUGNET_SET(dev, mlx4_en);
2307 
2308 	ether_ifattach(dev, dev_addr);
2309 
2310 	en_warn(priv, "Using %d TX rings\n", prof->tx_ring_num);
2311 	en_warn(priv, "Using %d RX rings\n", prof->rx_ring_num);
2312 
2313 	priv->registered = 1;
2314 
2315         en_warn(priv, "Using %d TX rings\n", prof->tx_ring_num);
2316         en_warn(priv, "Using %d RX rings\n", prof->rx_ring_num);
2317 
2318 
2319 	priv->rx_mb_size = if_getmtu(dev) + ETH_HLEN + VLAN_HLEN + ETH_FCS_LEN;
2320 	err = mlx4_SET_PORT_general(mdev->dev, priv->port,
2321 				    priv->rx_mb_size,
2322 				    prof->tx_pause, prof->tx_ppp,
2323 				    prof->rx_pause, prof->rx_ppp);
2324 	if (err) {
2325 		en_err(priv, "Failed setting port general configurations "
2326 		       "for port %d, with error %d\n", priv->port, err);
2327 		goto out;
2328 	}
2329 
2330 	/* Init port */
2331 	en_warn(priv, "Initializing port\n");
2332 	err = mlx4_INIT_PORT(mdev->dev, priv->port);
2333 	if (err) {
2334 		en_err(priv, "Failed Initializing port\n");
2335 		goto out;
2336 	}
2337 
2338 	queue_delayed_work(mdev->workqueue, &priv->stats_task, STATS_DELAY);
2339 
2340         if (mdev->dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_TS)
2341                 queue_delayed_work(mdev->workqueue, &priv->service_task, SERVICE_TASK_DELAY);
2342 
2343 	return 0;
2344 
2345 out:
2346 	mlx4_en_destroy_netdev(dev);
2347 	return err;
2348 }
2349 
2350 static int mlx4_en_set_ring_size(if_t dev,
2351     int rx_size, int tx_size)
2352 {
2353         struct mlx4_en_priv *priv = mlx4_netdev_priv(dev);
2354         struct mlx4_en_dev *mdev = priv->mdev;
2355         int port_up = 0;
2356         int err = 0;
2357 
2358         rx_size = roundup_pow_of_two(rx_size);
2359         rx_size = max_t(u32, rx_size, MLX4_EN_MIN_RX_SIZE);
2360         rx_size = min_t(u32, rx_size, MLX4_EN_MAX_RX_SIZE);
2361         tx_size = roundup_pow_of_two(tx_size);
2362         tx_size = max_t(u32, tx_size, MLX4_EN_MIN_TX_SIZE);
2363         tx_size = min_t(u32, tx_size, MLX4_EN_MAX_TX_SIZE);
2364 
2365         if (rx_size == (priv->port_up ?
2366             priv->rx_ring[0]->actual_size : priv->rx_ring[0]->size) &&
2367             tx_size == priv->tx_ring[0]->size)
2368                 return 0;
2369         mutex_lock(&mdev->state_lock);
2370         if (priv->port_up) {
2371                 port_up = 1;
2372                 mlx4_en_stop_port(dev);
2373         }
2374         mlx4_en_free_resources(priv);
2375         priv->prof->tx_ring_size = tx_size;
2376         priv->prof->rx_ring_size = rx_size;
2377         err = mlx4_en_alloc_resources(priv);
2378         if (err) {
2379                 en_err(priv, "Failed reallocating port resources\n");
2380                 goto out;
2381         }
2382         if (port_up) {
2383                 err = mlx4_en_start_port(dev);
2384                 if (err)
2385                         en_err(priv, "Failed starting port\n");
2386         }
2387 out:
2388         mutex_unlock(&mdev->state_lock);
2389         return err;
2390 }
2391 static int mlx4_en_set_rx_ring_size(SYSCTL_HANDLER_ARGS)
2392 {
2393         struct mlx4_en_priv *priv;
2394         int size;
2395         int error;
2396 
2397         priv = arg1;
2398         size = priv->prof->rx_ring_size;
2399         error = sysctl_handle_int(oidp, &size, 0, req);
2400         if (error || !req->newptr)
2401                 return (error);
2402         error = -mlx4_en_set_ring_size(priv->dev, size,
2403             priv->prof->tx_ring_size);
2404         return (error);
2405 }
2406 
2407 static int mlx4_en_set_tx_ring_size(SYSCTL_HANDLER_ARGS)
2408 {
2409         struct mlx4_en_priv *priv;
2410         int size;
2411         int error;
2412 
2413         priv = arg1;
2414         size = priv->prof->tx_ring_size;
2415         error = sysctl_handle_int(oidp, &size, 0, req);
2416         if (error || !req->newptr)
2417                 return (error);
2418         error = -mlx4_en_set_ring_size(priv->dev, priv->prof->rx_ring_size,
2419             size);
2420 
2421         return (error);
2422 }
2423 
2424 static int mlx4_en_get_module_info(if_t dev,
2425 				   struct ethtool_modinfo *modinfo)
2426 {
2427 	struct mlx4_en_priv *priv = mlx4_netdev_priv(dev);
2428 	struct mlx4_en_dev *mdev = priv->mdev;
2429 	int ret;
2430 	u8 data[4];
2431 
2432 	/* Read first 2 bytes to get Module & REV ID */
2433 	ret = mlx4_get_module_info(mdev->dev, priv->port,
2434 				   0/*offset*/, 2/*size*/, data);
2435 
2436 	if (ret < 2) {
2437 		en_err(priv, "Failed to read eeprom module first two bytes, error: 0x%x\n", -ret);
2438 		return -EIO;
2439 	}
2440 
2441 	switch (data[0] /* identifier */) {
2442 	case MLX4_MODULE_ID_QSFP:
2443 		modinfo->type = ETH_MODULE_SFF_8436;
2444 		modinfo->eeprom_len = ETH_MODULE_SFF_8436_LEN;
2445 		break;
2446 	case MLX4_MODULE_ID_QSFP_PLUS:
2447 		if (data[1] >= 0x3) { /* revision id */
2448 			modinfo->type = ETH_MODULE_SFF_8636;
2449 			modinfo->eeprom_len = ETH_MODULE_SFF_8636_LEN;
2450 		} else {
2451 			modinfo->type = ETH_MODULE_SFF_8436;
2452 			modinfo->eeprom_len = ETH_MODULE_SFF_8436_LEN;
2453 		}
2454 		break;
2455 	case MLX4_MODULE_ID_QSFP28:
2456 		modinfo->type = ETH_MODULE_SFF_8636;
2457 		modinfo->eeprom_len = ETH_MODULE_SFF_8636_LEN;
2458 		break;
2459 	case MLX4_MODULE_ID_SFP:
2460 		modinfo->type = ETH_MODULE_SFF_8472;
2461 		modinfo->eeprom_len = ETH_MODULE_SFF_8472_LEN;
2462 		break;
2463 	default:
2464 		en_err(priv, "mlx4_en_get_module_info :  Not recognized cable type\n");
2465 		return -EINVAL;
2466 	}
2467 
2468 	return 0;
2469 }
2470 
2471 static int mlx4_en_get_module_eeprom(if_t dev,
2472 				     struct ethtool_eeprom *ee,
2473 				     u8 *data)
2474 {
2475 	struct mlx4_en_priv *priv = mlx4_netdev_priv(dev);
2476 	struct mlx4_en_dev *mdev = priv->mdev;
2477 	int offset = ee->offset;
2478 	int i = 0, ret;
2479 
2480 	if (ee->len == 0)
2481 		return -EINVAL;
2482 
2483 	memset(data, 0, ee->len);
2484 
2485 	while (i < ee->len) {
2486 		en_dbg(DRV, priv,
2487 		       "mlx4_get_module_info i(%d) offset(%d) len(%d)\n",
2488 		       i, offset, ee->len - i);
2489 
2490 		ret = mlx4_get_module_info(mdev->dev, priv->port,
2491 					   offset, ee->len - i, data + i);
2492 
2493 		if (!ret) /* Done reading */
2494 			return 0;
2495 
2496 		if (ret < 0) {
2497 			en_err(priv,
2498 			       "mlx4_get_module_info i(%d) offset(%d) bytes_to_read(%d) - FAILED (0x%x)\n",
2499 			       i, offset, ee->len - i, ret);
2500 			return -1;
2501 		}
2502 
2503 		i += ret;
2504 		offset += ret;
2505 	}
2506 	return 0;
2507 }
2508 
2509 static void mlx4_en_print_eeprom(u8 *data, __u32 len)
2510 {
2511 	int		i;
2512 	int		j = 0;
2513 	int		row = 0;
2514 	const int	NUM_OF_BYTES = 16;
2515 
2516 	printf("\nOffset\t\tValues\n");
2517 	printf("------\t\t------\n");
2518 	while(row < len){
2519 		printf("0x%04x\t\t",row);
2520 		for(i=0; i < NUM_OF_BYTES; i++){
2521 			printf("%02x ", data[j]);
2522 			row++;
2523 			j++;
2524 		}
2525 		printf("\n");
2526 	}
2527 }
2528 
2529 /* Read cable EEPROM module information by first inspecting the first
2530  * two bytes to get the length and then read the rest of the information.
2531  * The information is printed to dmesg. */
2532 static int mlx4_en_read_eeprom(SYSCTL_HANDLER_ARGS)
2533 {
2534 
2535 	u8*		data;
2536 	int		error;
2537 	int		result = 0;
2538 	struct		mlx4_en_priv *priv;
2539 	struct		ifnet *dev;
2540 	struct		ethtool_modinfo modinfo;
2541 	struct		ethtool_eeprom ee;
2542 
2543 	error = sysctl_handle_int(oidp, &result, 0, req);
2544 	if (error || !req->newptr)
2545 		return (error);
2546 
2547 	if (result == 1) {
2548 		priv = arg1;
2549 		dev = priv->dev;
2550 		data = kmalloc(PAGE_SIZE, GFP_KERNEL);
2551 
2552 		error = mlx4_en_get_module_info(dev, &modinfo);
2553 		if (error) {
2554 			en_err(priv,
2555 			       "mlx4_en_get_module_info returned with error - FAILED (0x%x)\n",
2556 			       -error);
2557 			goto out;
2558 		}
2559 
2560 		ee.len = modinfo.eeprom_len;
2561 		ee.offset = 0;
2562 
2563 		error = mlx4_en_get_module_eeprom(dev, &ee, data);
2564 		if (error) {
2565 			en_err(priv,
2566 			       "mlx4_en_get_module_eeprom returned with error - FAILED (0x%x)\n",
2567 			       -error);
2568 			/* Continue printing partial information in case of an error */
2569 		}
2570 
2571 		/* EEPROM information will be printed in dmesg */
2572 		mlx4_en_print_eeprom(data, ee.len);
2573 out:
2574 		kfree(data);
2575 	}
2576 	/* Return zero to prevent sysctl failure. */
2577 	return (0);
2578 }
2579 
2580 static int mlx4_en_set_tx_ppp(SYSCTL_HANDLER_ARGS)
2581 {
2582         struct mlx4_en_priv *priv;
2583         int ppp;
2584         int error;
2585 
2586         priv = arg1;
2587         ppp = priv->prof->tx_ppp;
2588         error = sysctl_handle_int(oidp, &ppp, 0, req);
2589         if (error || !req->newptr)
2590                 return (error);
2591         if (ppp > 0xff || ppp < 0)
2592                 return (-EINVAL);
2593         priv->prof->tx_ppp = ppp;
2594         error = -mlx4_SET_PORT_general(priv->mdev->dev, priv->port,
2595                                        priv->rx_mb_size + ETHER_CRC_LEN,
2596                                        priv->prof->tx_pause,
2597                                        priv->prof->tx_ppp,
2598                                        priv->prof->rx_pause,
2599                                        priv->prof->rx_ppp);
2600 
2601         return (error);
2602 }
2603 
2604 static int mlx4_en_set_rx_ppp(SYSCTL_HANDLER_ARGS)
2605 {
2606         struct mlx4_en_priv *priv;
2607         struct mlx4_en_dev *mdev;
2608         int ppp;
2609         int error;
2610         int port_up;
2611 
2612         port_up = 0;
2613         priv = arg1;
2614         mdev = priv->mdev;
2615         ppp = priv->prof->rx_ppp;
2616         error = sysctl_handle_int(oidp, &ppp, 0, req);
2617         if (error || !req->newptr)
2618                 return (error);
2619         if (ppp > 0xff || ppp < 0)
2620                 return (-EINVAL);
2621         /* See if we have to change the number of tx queues. */
2622         if (!ppp != !priv->prof->rx_ppp) {
2623                 mutex_lock(&mdev->state_lock);
2624                 if (priv->port_up) {
2625                         port_up = 1;
2626                         mlx4_en_stop_port(priv->dev);
2627                 }
2628                 mlx4_en_free_resources(priv);
2629                 priv->prof->rx_ppp = ppp;
2630                 error = -mlx4_en_alloc_resources(priv);
2631                 if (error)
2632                         en_err(priv, "Failed reallocating port resources\n");
2633                 if (error == 0 && port_up) {
2634                         error = -mlx4_en_start_port(priv->dev);
2635                         if (error)
2636                                 en_err(priv, "Failed starting port\n");
2637                 }
2638                 mutex_unlock(&mdev->state_lock);
2639                 return (error);
2640 
2641         }
2642         priv->prof->rx_ppp = ppp;
2643         error = -mlx4_SET_PORT_general(priv->mdev->dev, priv->port,
2644                                        priv->rx_mb_size + ETHER_CRC_LEN,
2645                                        priv->prof->tx_pause,
2646                                        priv->prof->tx_ppp,
2647                                        priv->prof->rx_pause,
2648                                        priv->prof->rx_ppp);
2649 
2650         return (error);
2651 }
2652 
2653 static void mlx4_en_sysctl_conf(struct mlx4_en_priv *priv)
2654 {
2655         if_t dev;
2656         struct sysctl_ctx_list *ctx;
2657         struct sysctl_oid *node;
2658         struct sysctl_oid_list *node_list;
2659         struct sysctl_oid *coal;
2660         struct sysctl_oid_list *coal_list;
2661 	const char *pnameunit;
2662         dev = priv->dev;
2663         ctx = &priv->conf_ctx;
2664 	pnameunit = device_get_nameunit(priv->mdev->pdev->dev.bsddev);
2665 
2666         sysctl_ctx_init(ctx);
2667         priv->conf_sysctl = SYSCTL_ADD_NODE(ctx, SYSCTL_STATIC_CHILDREN(_hw),
2668             OID_AUTO, if_name(dev), CTLFLAG_RD | CTLFLAG_MPSAFE, 0,
2669 	    "mlx4 10gig ethernet");
2670         node = SYSCTL_ADD_NODE(ctx, SYSCTL_CHILDREN(priv->conf_sysctl), OID_AUTO,
2671             "conf", CTLFLAG_RD | CTLFLAG_MPSAFE, 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 | CTLFLAG_MPSAFE, NULL,
2704 	    "Interrupt coalesce configuration");
2705         coal_list = SYSCTL_CHILDREN(coal);
2706         SYSCTL_ADD_UINT(ctx, coal_list, OID_AUTO, "pkt_rate_low",
2707             CTLFLAG_RW, &priv->pkt_rate_low, 0,
2708             "Packets per-second for minimum delay");
2709         SYSCTL_ADD_UINT(ctx, coal_list, OID_AUTO, "rx_usecs_low",
2710             CTLFLAG_RW, &priv->rx_usecs_low, 0,
2711             "Minimum RX delay in micro-seconds");
2712         SYSCTL_ADD_UINT(ctx, coal_list, OID_AUTO, "pkt_rate_high",
2713             CTLFLAG_RW, &priv->pkt_rate_high, 0,
2714             "Packets per-second for maximum delay");
2715         SYSCTL_ADD_UINT(ctx, coal_list, OID_AUTO, "rx_usecs_high",
2716             CTLFLAG_RW, &priv->rx_usecs_high, 0,
2717             "Maximum RX delay in micro-seconds");
2718         SYSCTL_ADD_UINT(ctx, coal_list, OID_AUTO, "sample_interval",
2719             CTLFLAG_RW, &priv->sample_interval, 0,
2720             "adaptive frequency in units of HZ ticks");
2721         SYSCTL_ADD_UINT(ctx, coal_list, OID_AUTO, "adaptive_rx_coal",
2722             CTLFLAG_RW, &priv->adaptive_rx_coal, 0,
2723             "Enable adaptive rx coalescing");
2724 	/* EEPROM support */
2725 	SYSCTL_ADD_PROC(ctx, node_list, OID_AUTO, "eeprom_info",
2726 	    CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, priv, 0,
2727 	    mlx4_en_read_eeprom, "I", "EEPROM information");
2728 }
2729 
2730 static void mlx4_en_sysctl_stat(struct mlx4_en_priv *priv)
2731 {
2732 	struct sysctl_ctx_list *ctx;
2733 	struct sysctl_oid_list *node_list;
2734 	struct sysctl_oid *ring_node;
2735 	struct sysctl_oid_list *ring_list;
2736 	struct mlx4_en_tx_ring *tx_ring;
2737 	struct mlx4_en_rx_ring *rx_ring;
2738 	char namebuf[128];
2739 	int i;
2740 
2741 	ctx = &priv->stat_ctx;
2742 	sysctl_ctx_init(ctx);
2743 	priv->stat_sysctl = SYSCTL_ADD_NODE(ctx, SYSCTL_CHILDREN(priv->conf_sysctl), OID_AUTO,
2744 	    "stat", CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, "Statistics");
2745 	node_list = SYSCTL_CHILDREN(priv->stat_sysctl);
2746 
2747 #ifdef MLX4_EN_PERF_STAT
2748 	SYSCTL_ADD_UINT(ctx, node_list, OID_AUTO, "tx_poll", CTLFLAG_RD,
2749 	    &priv->pstats.tx_poll, "TX Poll calls");
2750 	SYSCTL_ADD_QUAD(ctx, node_list, OID_AUTO, "tx_pktsz_avg", CTLFLAG_RD,
2751 	    &priv->pstats.tx_pktsz_avg, "TX average packet size");
2752 	SYSCTL_ADD_UINT(ctx, node_list, OID_AUTO, "inflight_avg", CTLFLAG_RD,
2753 	    &priv->pstats.inflight_avg, "TX average packets in-flight");
2754 	SYSCTL_ADD_UINT(ctx, node_list, OID_AUTO, "tx_coal_avg", CTLFLAG_RD,
2755 	    &priv->pstats.tx_coal_avg, "TX average coalesced completions");
2756 	SYSCTL_ADD_UINT(ctx, node_list, OID_AUTO, "rx_coal_avg", CTLFLAG_RD,
2757 	    &priv->pstats.rx_coal_avg, "RX average coalesced completions");
2758 #endif
2759 
2760 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "tso_packets", CTLFLAG_RD,
2761 	    &priv->port_stats.tso_packets, 0, "TSO packets sent");
2762 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "queue_stopped", CTLFLAG_RD,
2763 	    &priv->port_stats.queue_stopped, 0, "Queue full");
2764 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "wake_queue", CTLFLAG_RD,
2765 	    &priv->port_stats.wake_queue, 0, "Queue resumed after full");
2766 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "tx_timeout", CTLFLAG_RD,
2767 	    &priv->port_stats.tx_timeout, 0, "Transmit timeouts");
2768 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "tx_oversized_packets", CTLFLAG_RD,
2769 	    &priv->port_stats.oversized_packets, 0, "TX oversized packets, m_defrag failed");
2770 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_alloc_failed", CTLFLAG_RD,
2771 	    &priv->port_stats.rx_alloc_failed, 0, "RX failed to allocate mbuf");
2772 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_chksum_good", CTLFLAG_RD,
2773             &priv->port_stats.rx_chksum_good, 0, "RX checksum offload success");
2774 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_chksum_none", CTLFLAG_RD,
2775 	    &priv->port_stats.rx_chksum_none, 0, "RX without checksum offload");
2776 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "tx_chksum_offload",
2777 	    CTLFLAG_RD, &priv->port_stats.tx_chksum_offload, 0,
2778 	    "TX checksum offloads");
2779 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "defrag_attempts",
2780 	    CTLFLAG_RD, &priv->port_stats.defrag_attempts, 0,
2781 	    "Oversized chains defragged");
2782 
2783 	/* Could strdup the names and add in a loop.  This is simpler. */
2784 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_bytes", CTLFLAG_RD,
2785 	    &priv->pkstats.rx_bytes, 0, "RX Bytes");
2786 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_packets", CTLFLAG_RD,
2787 	    &priv->pkstats.rx_packets, 0, "RX packets");
2788 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_multicast_packets", CTLFLAG_RD,
2789 	    &priv->pkstats.rx_multicast_packets, 0, "RX Multicast Packets");
2790 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_broadcast_packets", CTLFLAG_RD,
2791 	    &priv->pkstats.rx_broadcast_packets, 0, "RX Broadcast Packets");
2792 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_errors", CTLFLAG_RD,
2793 	    &priv->pkstats.rx_errors, 0, "RX Errors");
2794 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_dropped", CTLFLAG_RD,
2795 	    &priv->pkstats.rx_dropped, 0, "RX Dropped");
2796 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_length_errors", CTLFLAG_RD,
2797 	    &priv->pkstats.rx_length_errors, 0, "RX Length Errors");
2798 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_over_errors", CTLFLAG_RD,
2799 	    &priv->pkstats.rx_over_errors, 0, "RX Over Errors");
2800 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_crc_errors", CTLFLAG_RD,
2801 	    &priv->pkstats.rx_crc_errors, 0, "RX CRC Errors");
2802 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_jabbers", CTLFLAG_RD,
2803 	    &priv->pkstats.rx_jabbers, 0, "RX Jabbers");
2804 
2805 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_in_range_length_error", CTLFLAG_RD,
2806 	    &priv->pkstats.rx_in_range_length_error, 0, "RX IN_Range Length Error");
2807 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_out_range_length_error",
2808 	    CTLFLAG_RD, &priv->pkstats.rx_out_range_length_error, 0,
2809 	    "RX Out Range Length Error");
2810 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_lt_64_bytes_packets", CTLFLAG_RD,
2811 	    &priv->pkstats.rx_lt_64_bytes_packets, 0, "RX Lt 64 Bytes Packets");
2812 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_127_bytes_packets", CTLFLAG_RD,
2813 	    &priv->pkstats.rx_127_bytes_packets, 0, "RX 127 bytes Packets");
2814 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_255_bytes_packets", CTLFLAG_RD,
2815 	    &priv->pkstats.rx_255_bytes_packets, 0, "RX 255 bytes Packets");
2816 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_511_bytes_packets", CTLFLAG_RD,
2817 	    &priv->pkstats.rx_511_bytes_packets, 0, "RX 511 bytes Packets");
2818 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_1023_bytes_packets", CTLFLAG_RD,
2819 	    &priv->pkstats.rx_1023_bytes_packets, 0, "RX 1023 bytes Packets");
2820 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_1518_bytes_packets", CTLFLAG_RD,
2821 	    &priv->pkstats.rx_1518_bytes_packets, 0, "RX 1518 bytes Packets");
2822 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_1522_bytes_packets", CTLFLAG_RD,
2823 	    &priv->pkstats.rx_1522_bytes_packets, 0, "RX 1522 bytes Packets");
2824 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_1548_bytes_packets", CTLFLAG_RD,
2825 	    &priv->pkstats.rx_1548_bytes_packets, 0, "RX 1548 bytes Packets");
2826 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_gt_1548_bytes_packets", CTLFLAG_RD,
2827 	    &priv->pkstats.rx_gt_1548_bytes_packets, 0,
2828 	    "RX Greater Then 1548 bytes Packets");
2829 
2830 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "tx_packets", CTLFLAG_RD,
2831 	    &priv->pkstats.tx_packets, 0, "TX packets");
2832 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "tx_bytes", CTLFLAG_RD,
2833 	    &priv->pkstats.tx_bytes, 0, "TX Bytes");
2834 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "tx_multicast_packets", CTLFLAG_RD,
2835 	    &priv->pkstats.tx_multicast_packets, 0, "TX Multicast Packets");
2836 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "tx_broadcast_packets", CTLFLAG_RD,
2837 	    &priv->pkstats.tx_broadcast_packets, 0, "TX Broadcast Packets");
2838 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "tx_errors", CTLFLAG_RD,
2839 	    &priv->pkstats.tx_errors, 0, "TX Errors");
2840 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "tx_dropped", CTLFLAG_RD,
2841 	    &priv->pkstats.tx_dropped, 0, "TX Dropped");
2842 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "tx_lt_64_bytes_packets", CTLFLAG_RD,
2843 	    &priv->pkstats.tx_lt_64_bytes_packets, 0, "TX Less Then 64 Bytes Packets");
2844 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "tx_127_bytes_packets", CTLFLAG_RD,
2845 	    &priv->pkstats.tx_127_bytes_packets, 0, "TX 127 Bytes Packets");
2846 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "tx_255_bytes_packets", CTLFLAG_RD,
2847 	    &priv->pkstats.tx_255_bytes_packets, 0, "TX 255 Bytes Packets");
2848 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "tx_511_bytes_packets", CTLFLAG_RD,
2849 	    &priv->pkstats.tx_511_bytes_packets, 0, "TX 511 Bytes Packets");
2850 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "tx_1023_bytes_packets", CTLFLAG_RD,
2851 	    &priv->pkstats.tx_1023_bytes_packets, 0, "TX 1023 Bytes Packets");
2852 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "tx_1518_bytes_packets", CTLFLAG_RD,
2853 	    &priv->pkstats.tx_1518_bytes_packets, 0, "TX 1518 Bytes Packets");
2854 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "tx_1522_bytes_packets", CTLFLAG_RD,
2855 	    &priv->pkstats.tx_1522_bytes_packets, 0, "TX 1522 Bytes Packets");
2856 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "tx_1548_bytes_packets", CTLFLAG_RD,
2857 	    &priv->pkstats.tx_1548_bytes_packets, 0, "TX 1548 Bytes Packets");
2858 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "tx_gt_1548_bytes_packets", CTLFLAG_RD,
2859 	    &priv->pkstats.tx_gt_1548_bytes_packets, 0,
2860 	    "TX Greater Then 1548 Bytes Packets");
2861 
2862 	for (i = 0; i < priv->tx_ring_num; i++) {
2863 		tx_ring = priv->tx_ring[i];
2864 		snprintf(namebuf, sizeof(namebuf), "tx_ring%d", i);
2865 		ring_node = SYSCTL_ADD_NODE(ctx, node_list, OID_AUTO, namebuf,
2866 		    CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, "TX Ring");
2867 		ring_list = SYSCTL_CHILDREN(ring_node);
2868 		SYSCTL_ADD_U64(ctx, ring_list, OID_AUTO, "packets",
2869 		    CTLFLAG_RD, &tx_ring->packets, 0, "TX packets");
2870 		SYSCTL_ADD_U64(ctx, ring_list, OID_AUTO, "bytes",
2871 		    CTLFLAG_RD, &tx_ring->bytes, 0, "TX bytes");
2872 		SYSCTL_ADD_U64(ctx, ring_list, OID_AUTO, "tso_packets",
2873 		    CTLFLAG_RD, &tx_ring->tso_packets, 0, "TSO packets");
2874 		SYSCTL_ADD_U64(ctx, ring_list, OID_AUTO, "defrag_attempts",
2875 		    CTLFLAG_RD, &tx_ring->defrag_attempts, 0,
2876 		    "Oversized chains defragged");
2877 	}
2878 
2879 	for (i = 0; i < priv->rx_ring_num; i++) {
2880 		rx_ring = priv->rx_ring[i];
2881 		snprintf(namebuf, sizeof(namebuf), "rx_ring%d", i);
2882 		ring_node = SYSCTL_ADD_NODE(ctx, node_list, OID_AUTO, namebuf,
2883 		    CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, "RX Ring");
2884 		ring_list = SYSCTL_CHILDREN(ring_node);
2885 		SYSCTL_ADD_U64(ctx, ring_list, OID_AUTO, "packets",
2886 		    CTLFLAG_RD, &rx_ring->packets, 0, "RX packets");
2887 		SYSCTL_ADD_U64(ctx, ring_list, OID_AUTO, "bytes",
2888 		    CTLFLAG_RD, &rx_ring->bytes, 0, "RX bytes");
2889 		SYSCTL_ADD_U64(ctx, ring_list, OID_AUTO, "error",
2890 		    CTLFLAG_RD, &rx_ring->errors, 0, "RX soft errors");
2891 	}
2892 }
2893 
2894 #ifdef DEBUGNET
2895 static void
2896 mlx4_en_debugnet_init(if_t dev, int *nrxr, int *ncl, int *clsize)
2897 {
2898 	struct mlx4_en_priv *priv;
2899 
2900 	priv = if_getsoftc(dev);
2901 	mutex_lock(&priv->mdev->state_lock);
2902 	*nrxr = priv->rx_ring_num;
2903 	*ncl = DEBUGNET_MAX_IN_FLIGHT;
2904 	*clsize = MCLBYTES;
2905 	mutex_unlock(&priv->mdev->state_lock);
2906 }
2907 
2908 static void
2909 mlx4_en_debugnet_event(if_t dev, enum debugnet_ev event)
2910 {
2911 }
2912 
2913 static int
2914 mlx4_en_debugnet_transmit(if_t dev, struct mbuf *m)
2915 {
2916 	struct mlx4_en_priv *priv;
2917 	int err;
2918 
2919 	priv = if_getsoftc(dev);
2920 	if ((if_getdrvflags(dev) & (IFF_DRV_RUNNING | IFF_DRV_OACTIVE)) !=
2921 	    IFF_DRV_RUNNING || !priv->link_state)
2922 		return (ENOENT);
2923 
2924 	err = mlx4_en_xmit(priv, 0, &m);
2925 	if (err != 0 && m != NULL)
2926 		m_freem(m);
2927 	return (err);
2928 }
2929 
2930 static int
2931 mlx4_en_debugnet_poll(if_t dev, int count)
2932 {
2933 	struct mlx4_en_priv *priv;
2934 
2935 	priv = if_getsoftc(dev);
2936 	if ((if_getdrvflags(dev) & IFF_DRV_RUNNING) == 0 || !priv->link_state)
2937 		return (ENOENT);
2938 
2939 	mlx4_poll_interrupts(priv->mdev->dev);
2940 
2941 	return (0);
2942 }
2943 #endif /* DEBUGNET */
2944