xref: /freebsd/sys/dev/mlx4/mlx4_en/mlx4_en_netdev.c (revision f37852c17391fdf0e8309bcf684384dd0d854e43)
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 #ifdef CONFIG_NET_RX_BUSY_POLL
38 #include <net/busy_poll.h>
39 #endif
40 
41 #include <linux/list.h>
42 #include <linux/if_ether.h>
43 
44 #include <dev/mlx4/driver.h>
45 #include <dev/mlx4/device.h>
46 #include <dev/mlx4/cmd.h>
47 #include <dev/mlx4/cq.h>
48 
49 #include <sys/sockio.h>
50 #include <sys/sysctl.h>
51 
52 #include "en.h"
53 #include "en_port.h"
54 
55 static void mlx4_en_sysctl_stat(struct mlx4_en_priv *priv);
56 static void mlx4_en_sysctl_conf(struct mlx4_en_priv *priv);
57 
58 #ifdef CONFIG_NET_RX_BUSY_POLL
59 /* must be called with local_bh_disable()d */
60 static int mlx4_en_low_latency_recv(struct napi_struct *napi)
61 {
62 	struct mlx4_en_cq *cq = container_of(napi, struct mlx4_en_cq, napi);
63 	struct net_device *dev = cq->dev;
64 	struct mlx4_en_priv *priv = netdev_priv(dev);
65 	struct mlx4_en_rx_ring *rx_ring = priv->rx_ring[cq->ring];
66 	int done;
67 
68 	if (!priv->port_up)
69 		return LL_FLUSH_FAILED;
70 
71 	if (!mlx4_en_cq_lock_poll(cq))
72 		return LL_FLUSH_BUSY;
73 
74 	done = mlx4_en_process_rx_cq(dev, cq, 4);
75 #ifdef LL_EXTENDED_STATS
76 	if (done)
77 		rx_ring->cleaned += done;
78 	else
79 		rx_ring->misses++;
80 #endif
81 
82 	mlx4_en_cq_unlock_poll(cq);
83 
84 	return done;
85 }
86 #endif	/* CONFIG_NET_RX_BUSY_POLL */
87 
88 #ifdef CONFIG_RFS_ACCEL
89 
90 struct mlx4_en_filter {
91 	struct list_head next;
92 	struct work_struct work;
93 
94 	u8     ip_proto;
95 	__be32 src_ip;
96 	__be32 dst_ip;
97 	__be16 src_port;
98 	__be16 dst_port;
99 
100 	int rxq_index;
101 	struct mlx4_en_priv *priv;
102 	u32 flow_id;			/* RFS infrastructure id */
103 	int id;				/* mlx4_en driver id */
104 	u64 reg_id;			/* Flow steering API id */
105 	u8 activated;			/* Used to prevent expiry before filter
106 					 * is attached
107 					 */
108 	struct hlist_node filter_chain;
109 };
110 
111 static void mlx4_en_filter_rfs_expire(struct mlx4_en_priv *priv);
112 
113 static enum mlx4_net_trans_rule_id mlx4_ip_proto_to_trans_rule_id(u8 ip_proto)
114 {
115 	switch (ip_proto) {
116 	case IPPROTO_UDP:
117 		return MLX4_NET_TRANS_RULE_ID_UDP;
118 	case IPPROTO_TCP:
119 		return MLX4_NET_TRANS_RULE_ID_TCP;
120 	default:
121 		return -EPROTONOSUPPORT;
122 	}
123 };
124 
125 static void mlx4_en_filter_work(struct work_struct *work)
126 {
127 	struct mlx4_en_filter *filter = container_of(work,
128 						     struct mlx4_en_filter,
129 						     work);
130 	struct mlx4_en_priv *priv = filter->priv;
131 	struct mlx4_spec_list spec_tcp_udp = {
132 		.id = mlx4_ip_proto_to_trans_rule_id(filter->ip_proto),
133 		{
134 			.tcp_udp = {
135 				.dst_port = filter->dst_port,
136 				.dst_port_msk = (__force __be16)-1,
137 				.src_port = filter->src_port,
138 				.src_port_msk = (__force __be16)-1,
139 			},
140 		},
141 	};
142 	struct mlx4_spec_list spec_ip = {
143 		.id = MLX4_NET_TRANS_RULE_ID_IPV4,
144 		{
145 			.ipv4 = {
146 				.dst_ip = filter->dst_ip,
147 				.dst_ip_msk = (__force __be32)-1,
148 				.src_ip = filter->src_ip,
149 				.src_ip_msk = (__force __be32)-1,
150 			},
151 		},
152 	};
153 	struct mlx4_spec_list spec_eth = {
154 		.id = MLX4_NET_TRANS_RULE_ID_ETH,
155 	};
156 	struct mlx4_net_trans_rule rule = {
157 		.list = LIST_HEAD_INIT(rule.list),
158 		.queue_mode = MLX4_NET_TRANS_Q_LIFO,
159 		.exclusive = 1,
160 		.allow_loopback = 1,
161 		.promisc_mode = MLX4_FS_REGULAR,
162 		.port = priv->port,
163 		.priority = MLX4_DOMAIN_RFS,
164 	};
165 	int rc;
166 	__be64 mac_mask = cpu_to_be64(MLX4_MAC_MASK << 16);
167 
168 	if (spec_tcp_udp.id < 0) {
169 		en_warn(priv, "RFS: ignoring unsupported ip protocol (%d)\n",
170 			filter->ip_proto);
171 		goto ignore;
172 	}
173 	list_add_tail(&spec_eth.list, &rule.list);
174 	list_add_tail(&spec_ip.list, &rule.list);
175 	list_add_tail(&spec_tcp_udp.list, &rule.list);
176 
177 	rule.qpn = priv->rss_map.qps[filter->rxq_index].qpn;
178 	memcpy(spec_eth.eth.dst_mac, priv->dev->dev_addr, ETH_ALEN);
179 	memcpy(spec_eth.eth.dst_mac_msk, &mac_mask, ETH_ALEN);
180 
181 	filter->activated = 0;
182 
183 	if (filter->reg_id) {
184 		rc = mlx4_flow_detach(priv->mdev->dev, filter->reg_id);
185 		if (rc && rc != -ENOENT)
186 			en_err(priv, "Error detaching flow. rc = %d\n", rc);
187 	}
188 
189 	rc = mlx4_flow_attach(priv->mdev->dev, &rule, &filter->reg_id);
190 	if (rc)
191 		en_err(priv, "Error attaching flow. err = %d\n", rc);
192 
193 ignore:
194 	mlx4_en_filter_rfs_expire(priv);
195 
196 	filter->activated = 1;
197 }
198 
199 static inline struct hlist_head *
200 filter_hash_bucket(struct mlx4_en_priv *priv, __be32 src_ip, __be32 dst_ip,
201 		   __be16 src_port, __be16 dst_port)
202 {
203 	unsigned long l;
204 	int bucket_idx;
205 
206 	l = (__force unsigned long)src_port |
207 	    ((__force unsigned long)dst_port << 2);
208 	l ^= (__force unsigned long)(src_ip ^ dst_ip);
209 
210 	bucket_idx = hash_long(l, MLX4_EN_FILTER_HASH_SHIFT);
211 
212 	return &priv->filter_hash[bucket_idx];
213 }
214 
215 static struct mlx4_en_filter *
216 mlx4_en_filter_alloc(struct mlx4_en_priv *priv, int rxq_index, __be32 src_ip,
217 		     __be32 dst_ip, u8 ip_proto, __be16 src_port,
218 		     __be16 dst_port, u32 flow_id)
219 {
220 	struct mlx4_en_filter *filter = NULL;
221 
222 	filter = kzalloc(sizeof(struct mlx4_en_filter), GFP_ATOMIC);
223 	if (!filter)
224 		return NULL;
225 
226 	filter->priv = priv;
227 	filter->rxq_index = rxq_index;
228 	INIT_WORK(&filter->work, mlx4_en_filter_work);
229 
230 	filter->src_ip = src_ip;
231 	filter->dst_ip = dst_ip;
232 	filter->ip_proto = ip_proto;
233 	filter->src_port = src_port;
234 	filter->dst_port = dst_port;
235 
236 	filter->flow_id = flow_id;
237 
238 	filter->id = priv->last_filter_id++ % RPS_NO_FILTER;
239 
240 	list_add_tail(&filter->next, &priv->filters);
241 	hlist_add_head(&filter->filter_chain,
242 		       filter_hash_bucket(priv, src_ip, dst_ip, src_port,
243 					  dst_port));
244 
245 	return filter;
246 }
247 
248 static void mlx4_en_filter_free(struct mlx4_en_filter *filter)
249 {
250 	struct mlx4_en_priv *priv = filter->priv;
251 	int rc;
252 
253 	list_del(&filter->next);
254 
255 	rc = mlx4_flow_detach(priv->mdev->dev, filter->reg_id);
256 	if (rc && rc != -ENOENT)
257 		en_err(priv, "Error detaching flow. rc = %d\n", rc);
258 
259 	kfree(filter);
260 }
261 
262 static inline struct mlx4_en_filter *
263 mlx4_en_filter_find(struct mlx4_en_priv *priv, __be32 src_ip, __be32 dst_ip,
264 		    u8 ip_proto, __be16 src_port, __be16 dst_port)
265 {
266 	struct mlx4_en_filter *filter;
267 	struct mlx4_en_filter *ret = NULL;
268 
269 	hlist_for_each_entry(filter,
270 			     filter_hash_bucket(priv, src_ip, dst_ip,
271 						src_port, dst_port),
272 			     filter_chain) {
273 		if (filter->src_ip == src_ip &&
274 		    filter->dst_ip == dst_ip &&
275 		    filter->ip_proto == ip_proto &&
276 		    filter->src_port == src_port &&
277 		    filter->dst_port == dst_port) {
278 			ret = filter;
279 			break;
280 		}
281 	}
282 
283 	return ret;
284 }
285 
286 static int
287 mlx4_en_filter_rfs(struct net_device *net_dev, const struct sk_buff *skb,
288 		   u16 rxq_index, u32 flow_id)
289 {
290 	struct mlx4_en_priv *priv = netdev_priv(net_dev);
291 	struct mlx4_en_filter *filter;
292 	const struct iphdr *ip;
293 	const __be16 *ports;
294 	u8 ip_proto;
295 	__be32 src_ip;
296 	__be32 dst_ip;
297 	__be16 src_port;
298 	__be16 dst_port;
299 	int nhoff = skb_network_offset(skb);
300 	int ret = 0;
301 
302 	if (skb->protocol != htons(ETH_P_IP))
303 		return -EPROTONOSUPPORT;
304 
305 	ip = (const struct iphdr *)(skb->data + nhoff);
306 	if (ip_is_fragment(ip))
307 		return -EPROTONOSUPPORT;
308 
309 	if ((ip->protocol != IPPROTO_TCP) && (ip->protocol != IPPROTO_UDP))
310 		return -EPROTONOSUPPORT;
311 	ports = (const __be16 *)(skb->data + nhoff + 4 * ip->ihl);
312 
313 	ip_proto = ip->protocol;
314 	src_ip = ip->saddr;
315 	dst_ip = ip->daddr;
316 	src_port = ports[0];
317 	dst_port = ports[1];
318 
319 	spin_lock_bh(&priv->filters_lock);
320 	filter = mlx4_en_filter_find(priv, src_ip, dst_ip, ip_proto,
321 				     src_port, dst_port);
322 	if (filter) {
323 		if (filter->rxq_index == rxq_index)
324 			goto out;
325 
326 		filter->rxq_index = rxq_index;
327 	} else {
328 		filter = mlx4_en_filter_alloc(priv, rxq_index,
329 					      src_ip, dst_ip, ip_proto,
330 					      src_port, dst_port, flow_id);
331 		if (!filter) {
332 			ret = -ENOMEM;
333 			goto err;
334 		}
335 	}
336 
337 	queue_work(priv->mdev->workqueue, &filter->work);
338 
339 out:
340 	ret = filter->id;
341 err:
342 	spin_unlock_bh(&priv->filters_lock);
343 
344 	return ret;
345 }
346 
347 void mlx4_en_cleanup_filters(struct mlx4_en_priv *priv,
348 			     struct mlx4_en_rx_ring *rx_ring)
349 {
350 	struct mlx4_en_filter *filter, *tmp;
351 	LIST_HEAD(del_list);
352 
353 	spin_lock_bh(&priv->filters_lock);
354 	list_for_each_entry_safe(filter, tmp, &priv->filters, next) {
355 		list_move(&filter->next, &del_list);
356 		hlist_del(&filter->filter_chain);
357 	}
358 	spin_unlock_bh(&priv->filters_lock);
359 
360 	list_for_each_entry_safe(filter, tmp, &del_list, next) {
361 		cancel_work_sync(&filter->work);
362 		mlx4_en_filter_free(filter);
363 	}
364 }
365 
366 static void mlx4_en_filter_rfs_expire(struct mlx4_en_priv *priv)
367 {
368 	struct mlx4_en_filter *filter = NULL, *tmp, *last_filter = NULL;
369 	LIST_HEAD(del_list);
370 	int i = 0;
371 
372 	spin_lock_bh(&priv->filters_lock);
373 	list_for_each_entry_safe(filter, tmp, &priv->filters, next) {
374 		if (i > MLX4_EN_FILTER_EXPIRY_QUOTA)
375 			break;
376 
377 		if (filter->activated &&
378 		    !work_pending(&filter->work) &&
379 		    rps_may_expire_flow(priv->dev,
380 					filter->rxq_index, filter->flow_id,
381 					filter->id)) {
382 			list_move(&filter->next, &del_list);
383 			hlist_del(&filter->filter_chain);
384 		} else
385 			last_filter = filter;
386 
387 		i++;
388 	}
389 
390 	if (last_filter && (&last_filter->next != priv->filters.next))
391 		list_move(&priv->filters, &last_filter->next);
392 
393 	spin_unlock_bh(&priv->filters_lock);
394 
395 	list_for_each_entry_safe(filter, tmp, &del_list, next)
396 		mlx4_en_filter_free(filter);
397 }
398 #endif
399 
400 static void mlx4_en_vlan_rx_add_vid(void *arg, struct net_device *dev, u16 vid)
401 {
402 	struct mlx4_en_priv *priv = netdev_priv(dev);
403 	struct mlx4_en_dev *mdev = priv->mdev;
404 	int err;
405 	int idx;
406 
407 	if (arg != priv)
408 		return;
409 
410 	en_dbg(HW, priv, "adding VLAN:%d\n", vid);
411 
412 	set_bit(vid, priv->active_vlans);
413 
414 	/* Add VID to port VLAN filter */
415 	mutex_lock(&mdev->state_lock);
416 	if (mdev->device_up && priv->port_up) {
417 		err = mlx4_SET_VLAN_FLTR(mdev->dev, priv);
418 		if (err)
419 			en_err(priv, "Failed configuring VLAN filter\n");
420 	}
421 	if (mlx4_register_vlan(mdev->dev, priv->port, vid, &idx))
422 		en_dbg(HW, priv, "failed adding vlan %d\n", vid);
423 	mutex_unlock(&mdev->state_lock);
424 
425 }
426 
427 static void mlx4_en_vlan_rx_kill_vid(void *arg, struct net_device *dev, u16 vid)
428 {
429 	struct mlx4_en_priv *priv = netdev_priv(dev);
430 	struct mlx4_en_dev *mdev = priv->mdev;
431 	int err;
432 
433 	if (arg != priv)
434 		return;
435 
436 	en_dbg(HW, priv, "Killing VID:%d\n", vid);
437 
438 	clear_bit(vid, priv->active_vlans);
439 
440 	/* Remove VID from port VLAN filter */
441 	mutex_lock(&mdev->state_lock);
442 	mlx4_unregister_vlan(mdev->dev, priv->port, vid);
443 
444 	if (mdev->device_up && priv->port_up) {
445 		err = mlx4_SET_VLAN_FLTR(mdev->dev, priv);
446 		if (err)
447 			en_err(priv, "Failed configuring VLAN filter\n");
448 	}
449 	mutex_unlock(&mdev->state_lock);
450 
451 }
452 
453 static int mlx4_en_uc_steer_add(struct mlx4_en_priv *priv,
454 				unsigned char *mac, int *qpn, u64 *reg_id)
455 {
456 	struct mlx4_en_dev *mdev = priv->mdev;
457 	struct mlx4_dev *dev = mdev->dev;
458 	int err;
459 
460 	switch (dev->caps.steering_mode) {
461 	case MLX4_STEERING_MODE_B0: {
462 		struct mlx4_qp qp;
463 		u8 gid[16] = {0};
464 
465 		qp.qpn = *qpn;
466 		memcpy(&gid[10], mac, ETH_ALEN);
467 		gid[5] = priv->port;
468 
469 		err = mlx4_unicast_attach(dev, &qp, gid, 0, MLX4_PROT_ETH);
470 		break;
471 	}
472 	case MLX4_STEERING_MODE_DEVICE_MANAGED: {
473 		struct mlx4_spec_list spec_eth = { {NULL} };
474 		__be64 mac_mask = cpu_to_be64(MLX4_MAC_MASK << 16);
475 
476 		struct mlx4_net_trans_rule rule = {
477 			.queue_mode = MLX4_NET_TRANS_Q_FIFO,
478 			.exclusive = 0,
479 			.allow_loopback = 1,
480 			.promisc_mode = MLX4_FS_REGULAR,
481 			.priority = MLX4_DOMAIN_NIC,
482 		};
483 
484 		rule.port = priv->port;
485 		rule.qpn = *qpn;
486 		INIT_LIST_HEAD(&rule.list);
487 
488 		spec_eth.id = MLX4_NET_TRANS_RULE_ID_ETH;
489 		memcpy(spec_eth.eth.dst_mac, mac, ETH_ALEN);
490 		memcpy(spec_eth.eth.dst_mac_msk, &mac_mask, ETH_ALEN);
491 		list_add_tail(&spec_eth.list, &rule.list);
492 
493 		err = mlx4_flow_attach(dev, &rule, reg_id);
494 		break;
495 	}
496 	default:
497 		return -EINVAL;
498 	}
499 	if (err)
500 		en_warn(priv, "Failed Attaching Unicast\n");
501 
502 	return err;
503 }
504 
505 static void mlx4_en_uc_steer_release(struct mlx4_en_priv *priv,
506 				     unsigned char *mac, int qpn, u64 reg_id)
507 {
508 	struct mlx4_en_dev *mdev = priv->mdev;
509 	struct mlx4_dev *dev = mdev->dev;
510 
511 	switch (dev->caps.steering_mode) {
512 	case MLX4_STEERING_MODE_B0: {
513 		struct mlx4_qp qp;
514 		u8 gid[16] = {0};
515 
516 		qp.qpn = qpn;
517 		memcpy(&gid[10], mac, ETH_ALEN);
518 		gid[5] = priv->port;
519 
520 		mlx4_unicast_detach(dev, &qp, gid, MLX4_PROT_ETH);
521 		break;
522 	}
523 	case MLX4_STEERING_MODE_DEVICE_MANAGED: {
524 		mlx4_flow_detach(dev, reg_id);
525 		break;
526 	}
527 	default:
528 		en_err(priv, "Invalid steering mode.\n");
529 	}
530 }
531 
532 static int mlx4_en_get_qp(struct mlx4_en_priv *priv)
533 {
534 	struct mlx4_en_dev *mdev = priv->mdev;
535 	struct mlx4_dev *dev = mdev->dev;
536 	struct mlx4_mac_entry *entry;
537 	int index = 0;
538 	int err = 0;
539 	u64 reg_id;
540 	int *qpn = &priv->base_qpn;
541 	u64 mac = mlx4_mac_to_u64(IF_LLADDR(priv->dev));
542 
543 	en_dbg(DRV, priv, "Registering MAC: %pM for adding\n",
544 	       IF_LLADDR(priv->dev));
545 	index = mlx4_register_mac(dev, priv->port, mac);
546 	if (index < 0) {
547 		err = index;
548 		en_err(priv, "Failed adding MAC: %pM\n",
549 		       IF_LLADDR(priv->dev));
550 		return err;
551 	}
552 
553 	if (dev->caps.steering_mode == MLX4_STEERING_MODE_A0) {
554 		int base_qpn = mlx4_get_base_qpn(dev, priv->port);
555 		*qpn = base_qpn + index;
556 		return 0;
557 	}
558 
559 	err = mlx4_qp_reserve_range(dev, 1, 1, qpn, 0);
560 	en_dbg(DRV, priv, "Reserved qp %d\n", *qpn);
561 	if (err) {
562 		en_err(priv, "Failed to reserve qp for mac registration\n");
563 		goto qp_err;
564 	}
565 
566 	err = mlx4_en_uc_steer_add(priv, IF_LLADDR(priv->dev), qpn, &reg_id);
567 	if (err)
568 		goto steer_err;
569 
570 	entry = kmalloc(sizeof(*entry), GFP_KERNEL);
571 	if (!entry) {
572 		err = -ENOMEM;
573 		goto alloc_err;
574 	}
575 	memcpy(entry->mac, IF_LLADDR(priv->dev), sizeof(entry->mac));
576 	entry->reg_id = reg_id;
577 
578 	hlist_add_head(&entry->hlist,
579 			   &priv->mac_hash[entry->mac[MLX4_EN_MAC_HASH_IDX]]);
580 
581 	return 0;
582 
583 alloc_err:
584 	mlx4_en_uc_steer_release(priv, IF_LLADDR(priv->dev), *qpn, reg_id);
585 
586 steer_err:
587 	mlx4_qp_release_range(dev, *qpn, 1);
588 
589 qp_err:
590 	mlx4_unregister_mac(dev, priv->port, mac);
591 	return err;
592 }
593 
594 static void mlx4_en_put_qp(struct mlx4_en_priv *priv)
595 {
596 	struct mlx4_en_dev *mdev = priv->mdev;
597 	struct mlx4_dev *dev = mdev->dev;
598 	int qpn = priv->base_qpn;
599 	u64 mac;
600 
601 	if (dev->caps.steering_mode == MLX4_STEERING_MODE_A0) {
602 		mac = mlx4_mac_to_u64(IF_LLADDR(priv->dev));
603 		en_dbg(DRV, priv, "Registering MAC: %pM for deleting\n",
604 		       IF_LLADDR(priv->dev));
605 		mlx4_unregister_mac(dev, priv->port, mac);
606 	} else {
607 		struct mlx4_mac_entry *entry;
608 		struct hlist_node *tmp;
609 		struct hlist_head *bucket;
610 		unsigned int i;
611 
612 		for (i = 0; i < MLX4_EN_MAC_HASH_SIZE; ++i) {
613 			bucket = &priv->mac_hash[i];
614 			hlist_for_each_entry_safe(entry, tmp, bucket, hlist) {
615 				mac = mlx4_mac_to_u64(entry->mac);
616 				en_dbg(DRV, priv, "Registering MAC: %pM for deleting\n",
617 				       entry->mac);
618 				mlx4_en_uc_steer_release(priv, entry->mac,
619 							 qpn, entry->reg_id);
620 
621 				mlx4_unregister_mac(dev, priv->port, mac);
622 				hlist_del(&entry->hlist);
623 				kfree(entry);
624 			}
625 		}
626 
627 		en_dbg(DRV, priv, "Releasing qp: port %d, qpn %d\n",
628 		       priv->port, qpn);
629 		mlx4_qp_release_range(dev, qpn, 1);
630 		priv->flags &= ~MLX4_EN_FLAG_FORCE_PROMISC;
631 	}
632 }
633 
634 static void mlx4_en_clear_list(struct net_device *dev)
635 {
636 	struct mlx4_en_priv *priv = netdev_priv(dev);
637 	struct mlx4_en_mc_list *tmp, *mc_to_del;
638 
639 	list_for_each_entry_safe(mc_to_del, tmp, &priv->mc_list, list) {
640 		list_del(&mc_to_del->list);
641 		kfree(mc_to_del);
642 	}
643 }
644 
645 static void mlx4_en_cache_mclist(struct net_device *dev)
646 {
647         struct ifmultiaddr *ifma;
648 	struct mlx4_en_mc_list *tmp;
649 	struct mlx4_en_priv *priv = netdev_priv(dev);
650 
651         if_maddr_rlock(dev);
652         TAILQ_FOREACH(ifma, &dev->if_multiaddrs, ifma_link) {
653                 if (ifma->ifma_addr->sa_family != AF_LINK)
654                         continue;
655                 if (((struct sockaddr_dl *)ifma->ifma_addr)->sdl_alen !=
656                                 ETHER_ADDR_LEN)
657                         continue;
658                 /* Make sure the list didn't grow. */
659 		tmp = kzalloc(sizeof(struct mlx4_en_mc_list), GFP_ATOMIC);
660 		if (tmp == NULL) {
661 			en_err(priv, "Failed to allocate multicast list\n");
662 			break;
663 		}
664 		memcpy(tmp->addr,
665 			LLADDR((struct sockaddr_dl *)ifma->ifma_addr), ETH_ALEN);
666 		list_add_tail(&tmp->list, &priv->mc_list);
667         }
668         if_maddr_runlock(dev);
669 }
670 
671 static void update_mclist_flags(struct mlx4_en_priv *priv,
672 				struct list_head *dst,
673 				struct list_head *src)
674 {
675 	struct mlx4_en_mc_list *dst_tmp, *src_tmp, *new_mc;
676 	bool found;
677 
678 	/* Find all the entries that should be removed from dst,
679 	 * These are the entries that are not found in src
680 	 */
681 	list_for_each_entry(dst_tmp, dst, list) {
682 		found = false;
683 		list_for_each_entry(src_tmp, src, list) {
684 			if (!memcmp(dst_tmp->addr, src_tmp->addr, ETH_ALEN)) {
685 				found = true;
686 				break;
687 			}
688 		}
689 		if (!found)
690 			dst_tmp->action = MCLIST_REM;
691 	}
692 
693 	/* Add entries that exist in src but not in dst
694 	 * mark them as need to add
695 	 */
696 	list_for_each_entry(src_tmp, src, list) {
697 		found = false;
698 		list_for_each_entry(dst_tmp, dst, list) {
699 			if (!memcmp(dst_tmp->addr, src_tmp->addr, ETH_ALEN)) {
700 				dst_tmp->action = MCLIST_NONE;
701 				found = true;
702 				break;
703 			}
704 		}
705 		if (!found) {
706 			new_mc = kmalloc(sizeof(struct mlx4_en_mc_list),
707 					 GFP_KERNEL);
708 			if (!new_mc) {
709 				en_err(priv, "Failed to allocate current multicast list\n");
710 				return;
711 			}
712 			memcpy(new_mc, src_tmp,
713 			       sizeof(struct mlx4_en_mc_list));
714 			new_mc->action = MCLIST_ADD;
715 			list_add_tail(&new_mc->list, dst);
716 		}
717 	}
718 }
719 
720 static void mlx4_en_set_rx_mode(struct net_device *dev)
721 {
722 	struct mlx4_en_priv *priv = netdev_priv(dev);
723 
724 	if (!priv->port_up)
725 		return;
726 
727 	queue_work(priv->mdev->workqueue, &priv->rx_mode_task);
728 }
729 
730 static void mlx4_en_set_promisc_mode(struct mlx4_en_priv *priv,
731 				     struct mlx4_en_dev *mdev)
732 {
733 	int err = 0;
734 	if (!(priv->flags & MLX4_EN_FLAG_PROMISC)) {
735 		priv->flags |= MLX4_EN_FLAG_PROMISC;
736 
737 		/* Enable promiscouos mode */
738 		switch (mdev->dev->caps.steering_mode) {
739 		case MLX4_STEERING_MODE_DEVICE_MANAGED:
740 			err = mlx4_flow_steer_promisc_add(mdev->dev,
741 							  priv->port,
742 							  priv->base_qpn,
743 							  MLX4_FS_ALL_DEFAULT);
744 			if (err)
745 				en_err(priv, "Failed enabling promiscuous mode\n");
746 			priv->flags |= MLX4_EN_FLAG_MC_PROMISC;
747 			break;
748 
749 		case MLX4_STEERING_MODE_B0:
750 			err = mlx4_unicast_promisc_add(mdev->dev,
751 						       priv->base_qpn,
752 						       priv->port);
753 			if (err)
754 				en_err(priv, "Failed enabling unicast promiscuous mode\n");
755 
756 			/* Add the default qp number as multicast
757 			 * promisc
758 			 */
759 			if (!(priv->flags & MLX4_EN_FLAG_MC_PROMISC)) {
760 				err = mlx4_multicast_promisc_add(mdev->dev,
761 								 priv->base_qpn,
762 								 priv->port);
763 				if (err)
764 					en_err(priv, "Failed enabling multicast promiscuous mode\n");
765 				priv->flags |= MLX4_EN_FLAG_MC_PROMISC;
766 			}
767 			break;
768 
769 		case MLX4_STEERING_MODE_A0:
770 			err = mlx4_SET_PORT_qpn_calc(mdev->dev,
771 						     priv->port,
772 						     priv->base_qpn,
773 						     1);
774 			if (err)
775 				en_err(priv, "Failed enabling promiscuous mode\n");
776 			break;
777 		}
778 
779 		/* Disable port multicast filter (unconditionally) */
780 		err = mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, 0,
781 					  0, MLX4_MCAST_DISABLE);
782 		if (err)
783 			en_err(priv, "Failed disabling multicast filter\n");
784 	}
785 }
786 
787 static void mlx4_en_clear_promisc_mode(struct mlx4_en_priv *priv,
788 				       struct mlx4_en_dev *mdev)
789 {
790 	int err = 0;
791 
792 	priv->flags &= ~MLX4_EN_FLAG_PROMISC;
793 
794 	/* Disable promiscouos mode */
795 	switch (mdev->dev->caps.steering_mode) {
796 	case MLX4_STEERING_MODE_DEVICE_MANAGED:
797 		err = mlx4_flow_steer_promisc_remove(mdev->dev,
798 						     priv->port,
799 						     MLX4_FS_ALL_DEFAULT);
800 		if (err)
801 			en_err(priv, "Failed disabling promiscuous mode\n");
802 		priv->flags &= ~MLX4_EN_FLAG_MC_PROMISC;
803 		break;
804 
805 	case MLX4_STEERING_MODE_B0:
806 		err = mlx4_unicast_promisc_remove(mdev->dev,
807 						  priv->base_qpn,
808 						  priv->port);
809 		if (err)
810 			en_err(priv, "Failed disabling unicast promiscuous mode\n");
811 		/* Disable Multicast promisc */
812 		if (priv->flags & MLX4_EN_FLAG_MC_PROMISC) {
813 			err = mlx4_multicast_promisc_remove(mdev->dev,
814 							    priv->base_qpn,
815 							    priv->port);
816 			if (err)
817 				en_err(priv, "Failed disabling multicast promiscuous mode\n");
818 			priv->flags &= ~MLX4_EN_FLAG_MC_PROMISC;
819 		}
820 		break;
821 
822 	case MLX4_STEERING_MODE_A0:
823 		err = mlx4_SET_PORT_qpn_calc(mdev->dev,
824 					     priv->port,
825 					     priv->base_qpn, 0);
826 		if (err)
827 			en_err(priv, "Failed disabling promiscuous mode\n");
828 		break;
829 	}
830 }
831 
832 static void mlx4_en_do_multicast(struct mlx4_en_priv *priv,
833 				 struct net_device *dev,
834 				 struct mlx4_en_dev *mdev)
835 {
836 	struct mlx4_en_mc_list *mclist, *tmp;
837 	u8 mc_list[16] = {0};
838 	int err = 0;
839 	u64 mcast_addr = 0;
840 
841 
842 	/* Enable/disable the multicast filter according to IFF_ALLMULTI */
843 	if (dev->if_flags & IFF_ALLMULTI) {
844 		err = mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, 0,
845 					  0, MLX4_MCAST_DISABLE);
846 		if (err)
847 			en_err(priv, "Failed disabling multicast filter\n");
848 
849 		/* Add the default qp number as multicast promisc */
850 		if (!(priv->flags & MLX4_EN_FLAG_MC_PROMISC)) {
851 			switch (mdev->dev->caps.steering_mode) {
852 			case MLX4_STEERING_MODE_DEVICE_MANAGED:
853 				err = mlx4_flow_steer_promisc_add(mdev->dev,
854 								  priv->port,
855 								  priv->base_qpn,
856 								  MLX4_FS_MC_DEFAULT);
857 				break;
858 
859 			case MLX4_STEERING_MODE_B0:
860 				err = mlx4_multicast_promisc_add(mdev->dev,
861 								 priv->base_qpn,
862 								 priv->port);
863 				break;
864 
865 			case MLX4_STEERING_MODE_A0:
866 				break;
867 			}
868 			if (err)
869 				en_err(priv, "Failed entering multicast promisc mode\n");
870 			priv->flags |= MLX4_EN_FLAG_MC_PROMISC;
871 		}
872 	} else {
873 		/* Disable Multicast promisc */
874 		if (priv->flags & MLX4_EN_FLAG_MC_PROMISC) {
875 			switch (mdev->dev->caps.steering_mode) {
876 			case MLX4_STEERING_MODE_DEVICE_MANAGED:
877 				err = mlx4_flow_steer_promisc_remove(mdev->dev,
878 								     priv->port,
879 								     MLX4_FS_MC_DEFAULT);
880 				break;
881 
882 			case MLX4_STEERING_MODE_B0:
883 				err = mlx4_multicast_promisc_remove(mdev->dev,
884 								    priv->base_qpn,
885 								    priv->port);
886 				break;
887 
888 			case MLX4_STEERING_MODE_A0:
889 				break;
890 			}
891 			if (err)
892 				en_err(priv, "Failed disabling multicast promiscuous mode\n");
893 			priv->flags &= ~MLX4_EN_FLAG_MC_PROMISC;
894 		}
895 
896 		err = mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, 0,
897 					  0, MLX4_MCAST_DISABLE);
898 		if (err)
899 			en_err(priv, "Failed disabling multicast filter\n");
900 
901 		/* Flush mcast filter and init it with broadcast address */
902 		mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, ETH_BCAST,
903 				    1, MLX4_MCAST_CONFIG);
904 
905 		/* Update multicast list - we cache all addresses so they won't
906 		 * change while HW is updated holding the command semaphor */
907 		mlx4_en_cache_mclist(dev);
908 		list_for_each_entry(mclist, &priv->mc_list, list) {
909 			mcast_addr = mlx4_mac_to_u64(mclist->addr);
910 			mlx4_SET_MCAST_FLTR(mdev->dev, priv->port,
911 					mcast_addr, 0, MLX4_MCAST_CONFIG);
912 		}
913 		err = mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, 0,
914 					  0, MLX4_MCAST_ENABLE);
915 		if (err)
916 			en_err(priv, "Failed enabling multicast filter\n");
917 
918 		update_mclist_flags(priv, &priv->curr_list, &priv->mc_list);
919 		list_for_each_entry_safe(mclist, tmp, &priv->curr_list, list) {
920 			if (mclist->action == MCLIST_REM) {
921 				/* detach this address and delete from list */
922 				memcpy(&mc_list[10], mclist->addr, ETH_ALEN);
923 				mc_list[5] = priv->port;
924 				err = mlx4_multicast_detach(mdev->dev,
925 							    &priv->rss_map.indir_qp,
926 							    mc_list,
927 							    MLX4_PROT_ETH,
928 							    mclist->reg_id);
929 				if (err)
930 					en_err(priv, "Fail to detach multicast address\n");
931 
932 				/* remove from list */
933 				list_del(&mclist->list);
934 				kfree(mclist);
935 			} else if (mclist->action == MCLIST_ADD) {
936 				/* attach the address */
937 				memcpy(&mc_list[10], mclist->addr, ETH_ALEN);
938 				/* needed for B0 steering support */
939 				mc_list[5] = priv->port;
940 				err = mlx4_multicast_attach(mdev->dev,
941 							    &priv->rss_map.indir_qp,
942 							    mc_list,
943 							    priv->port, 0,
944 							    MLX4_PROT_ETH,
945 							    &mclist->reg_id);
946 				if (err)
947 					en_err(priv, "Fail to attach multicast address\n");
948 
949 			}
950 		}
951 	}
952 }
953 
954 static void mlx4_en_do_set_rx_mode(struct work_struct *work)
955 {
956 	struct mlx4_en_priv *priv = container_of(work, struct mlx4_en_priv,
957 						 rx_mode_task);
958 	struct mlx4_en_dev *mdev = priv->mdev;
959 	struct net_device *dev = priv->dev;
960 
961 
962 	mutex_lock(&mdev->state_lock);
963 	if (!mdev->device_up) {
964 		en_dbg(HW, priv, "Card is not up, ignoring rx mode change.\n");
965 		goto out;
966 	}
967 	if (!priv->port_up) {
968 		en_dbg(HW, priv, "Port is down, ignoring rx mode change.\n");
969 		goto out;
970 	}
971 	if (!mlx4_en_QUERY_PORT(mdev, priv->port)) {
972 		if (priv->port_state.link_state) {
973 			priv->last_link_state = MLX4_DEV_EVENT_PORT_UP;
974 			/* update netif baudrate */
975 			priv->dev->if_baudrate =
976 			    IF_Mbps(priv->port_state.link_speed);
977 			/* Important note: the following call for if_link_state_change
978 			 * is needed for interface up scenario (start port, link state
979 			 * change) */
980 			if_link_state_change(priv->dev, LINK_STATE_UP);
981 			en_dbg(HW, priv, "Link Up\n");
982 		}
983 	}
984 
985 	/* Promsicuous mode: disable all filters */
986 	if ((dev->if_flags & IFF_PROMISC) ||
987 	    (priv->flags & MLX4_EN_FLAG_FORCE_PROMISC)) {
988 		mlx4_en_set_promisc_mode(priv, mdev);
989 		goto out;
990 	}
991 
992 	/* Not in promiscuous mode */
993 	if (priv->flags & MLX4_EN_FLAG_PROMISC)
994 		mlx4_en_clear_promisc_mode(priv, mdev);
995 
996 	mlx4_en_do_multicast(priv, dev, mdev);
997 out:
998 	mutex_unlock(&mdev->state_lock);
999 }
1000 
1001 #ifdef CONFIG_NET_POLL_CONTROLLER
1002 static void mlx4_en_netpoll(struct net_device *dev)
1003 {
1004 	struct mlx4_en_priv *priv = netdev_priv(dev);
1005 	struct mlx4_en_cq *cq;
1006 	unsigned long flags;
1007 	int i;
1008 
1009 	for (i = 0; i < priv->rx_ring_num; i++) {
1010 		cq = priv->rx_cq[i];
1011 		spin_lock_irqsave(&cq->lock, flags);
1012 		napi_synchronize(&cq->napi);
1013 		mlx4_en_process_rx_cq(dev, cq, 0);
1014 		spin_unlock_irqrestore(&cq->lock, flags);
1015 	}
1016 }
1017 #endif
1018 
1019 static void mlx4_en_watchdog_timeout(void *arg)
1020 {
1021         struct mlx4_en_priv *priv = arg;
1022         struct mlx4_en_dev *mdev = priv->mdev;
1023 
1024         en_dbg(DRV, priv, "Scheduling watchdog\n");
1025         queue_work(mdev->workqueue, &priv->watchdog_task);
1026         if (priv->port_up)
1027                 callout_reset(&priv->watchdog_timer, MLX4_EN_WATCHDOG_TIMEOUT,
1028                                 mlx4_en_watchdog_timeout, priv);
1029 }
1030 
1031 
1032 
1033 static void mlx4_en_set_default_moderation(struct mlx4_en_priv *priv)
1034 {
1035 	struct mlx4_en_cq *cq;
1036 	int i;
1037 
1038 	/* If we haven't received a specific coalescing setting
1039 	 * (module param), we set the moderation parameters as follows:
1040 	 * - moder_cnt is set to the number of mtu sized packets to
1041 	 *   satisfy our coelsing target.
1042 	 * - moder_time is set to a fixed value.
1043 	 */
1044 	priv->rx_frames = MLX4_EN_RX_COAL_TARGET / priv->dev->if_mtu + 1;
1045 	priv->rx_usecs = MLX4_EN_RX_COAL_TIME;
1046 	priv->tx_frames = MLX4_EN_TX_COAL_PKTS;
1047 	priv->tx_usecs = MLX4_EN_TX_COAL_TIME;
1048 	en_dbg(INTR, priv, "Default coalesing params for mtu: %u - "
1049 	       "rx_frames:%d rx_usecs:%d\n",
1050 	       (unsigned)priv->dev->if_mtu, priv->rx_frames, priv->rx_usecs);
1051 
1052 	/* Setup cq moderation params */
1053 	for (i = 0; i < priv->rx_ring_num; i++) {
1054 		cq = priv->rx_cq[i];
1055 		cq->moder_cnt = priv->rx_frames;
1056 		cq->moder_time = priv->rx_usecs;
1057 		priv->last_moder_time[i] = MLX4_EN_AUTO_CONF;
1058 		priv->last_moder_packets[i] = 0;
1059 		priv->last_moder_bytes[i] = 0;
1060 	}
1061 
1062 	for (i = 0; i < priv->tx_ring_num; i++) {
1063 		cq = priv->tx_cq[i];
1064 		cq->moder_cnt = priv->tx_frames;
1065 		cq->moder_time = priv->tx_usecs;
1066 	}
1067 
1068 	/* Reset auto-moderation params */
1069 	priv->pkt_rate_low = MLX4_EN_RX_RATE_LOW;
1070 	priv->rx_usecs_low = MLX4_EN_RX_COAL_TIME_LOW;
1071 	priv->pkt_rate_high = MLX4_EN_RX_RATE_HIGH;
1072 	priv->rx_usecs_high = MLX4_EN_RX_COAL_TIME_HIGH;
1073 	priv->sample_interval = MLX4_EN_SAMPLE_INTERVAL;
1074 	priv->adaptive_rx_coal = 1;
1075 	priv->last_moder_jiffies = 0;
1076 	priv->last_moder_tx_packets = 0;
1077 }
1078 
1079 static void mlx4_en_auto_moderation(struct mlx4_en_priv *priv)
1080 {
1081 	unsigned long period = (unsigned long) (jiffies - priv->last_moder_jiffies);
1082 	struct mlx4_en_cq *cq;
1083 	unsigned long packets;
1084 	unsigned long rate;
1085 	unsigned long avg_pkt_size;
1086 	unsigned long rx_packets;
1087 	unsigned long rx_bytes;
1088 	unsigned long rx_pkt_diff;
1089 	int moder_time;
1090 	int ring, err;
1091 
1092 	if (!priv->adaptive_rx_coal || period < priv->sample_interval * HZ)
1093 		return;
1094 
1095 	for (ring = 0; ring < priv->rx_ring_num; ring++) {
1096                 spin_lock(&priv->stats_lock);
1097 		rx_packets = priv->rx_ring[ring]->packets;
1098 		rx_bytes = priv->rx_ring[ring]->bytes;
1099 		spin_unlock(&priv->stats_lock);
1100 
1101 		rx_pkt_diff = ((unsigned long) (rx_packets -
1102 				priv->last_moder_packets[ring]));
1103 		packets = rx_pkt_diff;
1104 		rate = packets * HZ / period;
1105 		avg_pkt_size = packets ? ((unsigned long) (rx_bytes -
1106 				priv->last_moder_bytes[ring])) / packets : 0;
1107 
1108 		/* Apply auto-moderation only when packet rate
1109 		 * exceeds a rate that it matters */
1110 		if (rate > (MLX4_EN_RX_RATE_THRESH / priv->rx_ring_num) &&
1111 		    avg_pkt_size > MLX4_EN_AVG_PKT_SMALL) {
1112 			if (rate < priv->pkt_rate_low)
1113 				moder_time = priv->rx_usecs_low;
1114 			else if (rate > priv->pkt_rate_high)
1115 				moder_time = priv->rx_usecs_high;
1116 			else
1117 				moder_time = (rate - priv->pkt_rate_low) *
1118 					(priv->rx_usecs_high - priv->rx_usecs_low) /
1119 					(priv->pkt_rate_high - priv->pkt_rate_low) +
1120 					priv->rx_usecs_low;
1121 		} else {
1122 			moder_time = priv->rx_usecs_low;
1123 		}
1124 
1125 		if (moder_time != priv->last_moder_time[ring]) {
1126 			priv->last_moder_time[ring] = moder_time;
1127 			cq = priv->rx_cq[ring];
1128 			cq->moder_time = moder_time;
1129 			err = mlx4_en_set_cq_moder(priv, cq);
1130 			if (err)
1131 				en_err(priv, "Failed modifying moderation for cq:%d\n",
1132 				       ring);
1133 		}
1134 		priv->last_moder_packets[ring] = rx_packets;
1135 		priv->last_moder_bytes[ring] = rx_bytes;
1136 	}
1137 
1138 	priv->last_moder_jiffies = jiffies;
1139 }
1140 
1141 static void mlx4_en_do_get_stats(struct work_struct *work)
1142 {
1143 	struct delayed_work *delay = to_delayed_work(work);
1144 	struct mlx4_en_priv *priv = container_of(delay, struct mlx4_en_priv,
1145 						 stats_task);
1146 	struct mlx4_en_dev *mdev = priv->mdev;
1147 	int err;
1148 
1149 	mutex_lock(&mdev->state_lock);
1150 	if (mdev->device_up) {
1151 		if (priv->port_up) {
1152                         err = mlx4_en_DUMP_ETH_STATS(mdev, priv->port, 0);
1153 			if (err)
1154 				en_dbg(HW, priv, "Could not update stats\n");
1155 
1156 			mlx4_en_auto_moderation(priv);
1157 		}
1158 
1159 		queue_delayed_work(mdev->workqueue, &priv->stats_task, STATS_DELAY);
1160 	}
1161 	mutex_unlock(&mdev->state_lock);
1162 }
1163 
1164 /* mlx4_en_service_task - Run service task for tasks that needed to be done
1165  * periodically
1166  */
1167 static void mlx4_en_service_task(struct work_struct *work)
1168 {
1169 	struct delayed_work *delay = to_delayed_work(work);
1170 	struct mlx4_en_priv *priv = container_of(delay, struct mlx4_en_priv,
1171 						 service_task);
1172 	struct mlx4_en_dev *mdev = priv->mdev;
1173 
1174 	mutex_lock(&mdev->state_lock);
1175 	if (mdev->device_up) {
1176 		queue_delayed_work(mdev->workqueue, &priv->service_task,
1177 				   SERVICE_TASK_DELAY);
1178 	}
1179 	mutex_unlock(&mdev->state_lock);
1180 }
1181 
1182 static void mlx4_en_linkstate(struct work_struct *work)
1183 {
1184 	struct mlx4_en_priv *priv = container_of(work, struct mlx4_en_priv,
1185 						 linkstate_task);
1186 	struct mlx4_en_dev *mdev = priv->mdev;
1187 	int linkstate = priv->link_state;
1188 
1189 	mutex_lock(&mdev->state_lock);
1190 	/* If observable port state changed set carrier state and
1191 	 * report to system log */
1192 	if (priv->last_link_state != linkstate) {
1193 		if (linkstate == MLX4_DEV_EVENT_PORT_DOWN) {
1194 			en_info(priv, "Link Down\n");
1195 			if_link_state_change(priv->dev, LINK_STATE_DOWN);
1196 			/* update netif baudrate */
1197 			priv->dev->if_baudrate = 0;
1198 
1199 		/* make sure the port is up before notifying the OS.
1200 		 * This is tricky since we get here on INIT_PORT and
1201 		 * in such case we can't tell the OS the port is up.
1202 		 * To solve this there is a call to if_link_state_change
1203 		 * in set_rx_mode.
1204 		 * */
1205 		} else if (priv->port_up && (linkstate == MLX4_DEV_EVENT_PORT_UP)){
1206 			if (mlx4_en_QUERY_PORT(priv->mdev, priv->port))
1207 				en_info(priv, "Query port failed\n");
1208 			priv->dev->if_baudrate =
1209 			    IF_Mbps(priv->port_state.link_speed);
1210 			en_info(priv, "Link Up\n");
1211 			if_link_state_change(priv->dev, LINK_STATE_UP);
1212 		}
1213 	}
1214 	priv->last_link_state = linkstate;
1215 	mutex_unlock(&mdev->state_lock);
1216 }
1217 
1218 
1219 int mlx4_en_start_port(struct net_device *dev)
1220 {
1221 	struct mlx4_en_priv *priv = netdev_priv(dev);
1222 	struct mlx4_en_dev *mdev = priv->mdev;
1223 	struct mlx4_en_cq *cq;
1224 	struct mlx4_en_tx_ring *tx_ring;
1225 	int rx_index = 0;
1226 	int tx_index = 0;
1227 	int err = 0;
1228 	int i;
1229 	int j;
1230 	u8 mc_list[16] = {0};
1231 
1232 
1233 	if (priv->port_up) {
1234 		en_dbg(DRV, priv, "start port called while port already up\n");
1235 		return 0;
1236 	}
1237 
1238 	INIT_LIST_HEAD(&priv->mc_list);
1239 	INIT_LIST_HEAD(&priv->curr_list);
1240 	INIT_LIST_HEAD(&priv->ethtool_list);
1241 
1242 	/* Calculate Rx buf size */
1243 	dev->if_mtu = min(dev->if_mtu, priv->max_mtu);
1244         mlx4_en_calc_rx_buf(dev);
1245 	en_dbg(DRV, priv, "Rx buf size:%d\n", priv->rx_mb_size);
1246 
1247 	/* Configure rx cq's and rings */
1248 	err = mlx4_en_activate_rx_rings(priv);
1249 	if (err) {
1250 		en_err(priv, "Failed to activate RX rings\n");
1251 		return err;
1252 	}
1253 	for (i = 0; i < priv->rx_ring_num; i++) {
1254 		cq = priv->rx_cq[i];
1255 
1256 		mlx4_en_cq_init_lock(cq);
1257 		err = mlx4_en_activate_cq(priv, cq, i);
1258 		if (err) {
1259 			en_err(priv, "Failed activating Rx CQ\n");
1260 			goto cq_err;
1261 		}
1262 		for (j = 0; j < cq->size; j++)
1263 			cq->buf[j].owner_sr_opcode = MLX4_CQE_OWNER_MASK;
1264 		err = mlx4_en_set_cq_moder(priv, cq);
1265 		if (err) {
1266 			en_err(priv, "Failed setting cq moderation parameters");
1267 			mlx4_en_deactivate_cq(priv, cq);
1268 			goto cq_err;
1269 		}
1270 		mlx4_en_arm_cq(priv, cq);
1271 		priv->rx_ring[i]->cqn = cq->mcq.cqn;
1272 		++rx_index;
1273 	}
1274 
1275 	/* Set qp number */
1276 	en_dbg(DRV, priv, "Getting qp number for port %d\n", priv->port);
1277 	err = mlx4_en_get_qp(priv);
1278 	if (err) {
1279 		en_err(priv, "Failed getting eth qp\n");
1280 		goto cq_err;
1281 	}
1282 	mdev->mac_removed[priv->port] = 0;
1283 
1284 	/* gets default allocated counter index from func cap */
1285 	/* or sink counter index if no resources */
1286 	priv->counter_index = mdev->dev->caps.def_counter_index[priv->port - 1];
1287 
1288 	en_dbg(DRV, priv, "%s: default counter index %d for port %d\n",
1289 	       __func__, priv->counter_index, priv->port);
1290 
1291 	err = mlx4_en_config_rss_steer(priv);
1292 	if (err) {
1293 		en_err(priv, "Failed configuring rss steering\n");
1294 		goto mac_err;
1295 	}
1296 
1297 	err = mlx4_en_create_drop_qp(priv);
1298 	if (err)
1299 		goto rss_err;
1300 
1301 	/* Configure tx cq's and rings */
1302 	for (i = 0; i < priv->tx_ring_num; i++) {
1303 		/* Configure cq */
1304 		cq = priv->tx_cq[i];
1305 		err = mlx4_en_activate_cq(priv, cq, i);
1306 		if (err) {
1307 			en_err(priv, "Failed activating Tx CQ\n");
1308 			goto tx_err;
1309 		}
1310 		err = mlx4_en_set_cq_moder(priv, cq);
1311 		if (err) {
1312 			en_err(priv, "Failed setting cq moderation parameters");
1313 			mlx4_en_deactivate_cq(priv, cq);
1314 			goto tx_err;
1315 		}
1316 		en_dbg(DRV, priv, "Resetting index of collapsed CQ:%d to -1\n", i);
1317 		cq->buf->wqe_index = cpu_to_be16(0xffff);
1318 
1319 		/* Configure ring */
1320 		tx_ring = priv->tx_ring[i];
1321 
1322 		err = mlx4_en_activate_tx_ring(priv, tx_ring, cq->mcq.cqn,
1323 					       i / priv->num_tx_rings_p_up);
1324 		if (err) {
1325 			en_err(priv, "Failed activating Tx ring %d\n", i);
1326 			mlx4_en_deactivate_cq(priv, cq);
1327 			goto tx_err;
1328 		}
1329 
1330 		/* Arm CQ for TX completions */
1331 		mlx4_en_arm_cq(priv, cq);
1332 
1333 		/* Set initial ownership of all Tx TXBBs to SW (1) */
1334 		for (j = 0; j < tx_ring->buf_size; j += STAMP_STRIDE)
1335 			*((u32 *) (tx_ring->buf + j)) = 0xffffffff;
1336 		++tx_index;
1337 	}
1338 
1339 	/* Configure port */
1340 	err = mlx4_SET_PORT_general(mdev->dev, priv->port,
1341 				    priv->rx_mb_size,
1342 				    priv->prof->tx_pause,
1343 				    priv->prof->tx_ppp,
1344 				    priv->prof->rx_pause,
1345 				    priv->prof->rx_ppp);
1346 	if (err) {
1347 		en_err(priv, "Failed setting port general configurations for port %d, with error %d\n",
1348 		       priv->port, err);
1349 		goto tx_err;
1350 	}
1351 	/* Set default qp number */
1352 	err = mlx4_SET_PORT_qpn_calc(mdev->dev, priv->port, priv->base_qpn, 0);
1353 	if (err) {
1354 		en_err(priv, "Failed setting default qp numbers\n");
1355 		goto tx_err;
1356 	}
1357 
1358 	/* Init port */
1359 	en_dbg(HW, priv, "Initializing port\n");
1360 	err = mlx4_INIT_PORT(mdev->dev, priv->port);
1361 	if (err) {
1362 		en_err(priv, "Failed Initializing port\n");
1363 		goto tx_err;
1364 	}
1365 
1366 	/* Attach rx QP to bradcast address */
1367 	memset(&mc_list[10], 0xff, ETH_ALEN);
1368 	mc_list[5] = priv->port; /* needed for B0 steering support */
1369 	if (mlx4_multicast_attach(mdev->dev, &priv->rss_map.indir_qp, mc_list,
1370 				  priv->port, 0, MLX4_PROT_ETH,
1371 				  &priv->broadcast_id))
1372 		mlx4_warn(mdev, "Failed Attaching Broadcast\n");
1373 
1374 	/* Must redo promiscuous mode setup. */
1375 	priv->flags &= ~(MLX4_EN_FLAG_PROMISC | MLX4_EN_FLAG_MC_PROMISC);
1376 
1377 	/* Schedule multicast task to populate multicast list */
1378 	queue_work(mdev->workqueue, &priv->rx_mode_task);
1379 
1380 	mlx4_set_stats_bitmap(mdev->dev, priv->stats_bitmap);
1381 
1382 	priv->port_up = true;
1383 
1384         /* Enable the queues. */
1385         dev->if_drv_flags &= ~IFF_DRV_OACTIVE;
1386         dev->if_drv_flags |= IFF_DRV_RUNNING;
1387 #ifdef CONFIG_DEBUG_FS
1388 	mlx4_en_create_debug_files(priv);
1389 #endif
1390         callout_reset(&priv->watchdog_timer, MLX4_EN_WATCHDOG_TIMEOUT,
1391                     mlx4_en_watchdog_timeout, priv);
1392 
1393 
1394 	return 0;
1395 
1396 tx_err:
1397 	while (tx_index--) {
1398 		mlx4_en_deactivate_tx_ring(priv, priv->tx_ring[tx_index]);
1399 		mlx4_en_deactivate_cq(priv, priv->tx_cq[tx_index]);
1400 	}
1401 	mlx4_en_destroy_drop_qp(priv);
1402 rss_err:
1403 	mlx4_en_release_rss_steer(priv);
1404 mac_err:
1405 	mlx4_en_put_qp(priv);
1406 cq_err:
1407 	while (rx_index--)
1408 		mlx4_en_deactivate_cq(priv, priv->rx_cq[rx_index]);
1409 	for (i = 0; i < priv->rx_ring_num; i++)
1410 		mlx4_en_deactivate_rx_ring(priv, priv->rx_ring[i]);
1411 
1412 	return err; /* need to close devices */
1413 }
1414 
1415 
1416 void mlx4_en_stop_port(struct net_device *dev)
1417 {
1418 	struct mlx4_en_priv *priv = netdev_priv(dev);
1419 	struct mlx4_en_dev *mdev = priv->mdev;
1420 	struct mlx4_en_mc_list *mclist, *tmp;
1421 	int i;
1422 	u8 mc_list[16] = {0};
1423 
1424 	if (!priv->port_up) {
1425 		en_dbg(DRV, priv, "stop port called while port already down\n");
1426 		return;
1427 	}
1428 
1429 #ifdef CONFIG_DEBUG_FS
1430 	mlx4_en_delete_debug_files(priv);
1431 #endif
1432 
1433 	/* close port*/
1434 	mlx4_CLOSE_PORT(mdev->dev, priv->port);
1435 
1436 	/* Set port as not active */
1437 	priv->port_up = false;
1438 	if (priv->counter_index != 0xff) {
1439 		mlx4_counter_free(mdev->dev, priv->port, priv->counter_index);
1440 		priv->counter_index = 0xff;
1441 	}
1442 
1443 	/* Promsicuous mode */
1444 	if (mdev->dev->caps.steering_mode ==
1445 	    MLX4_STEERING_MODE_DEVICE_MANAGED) {
1446 		priv->flags &= ~(MLX4_EN_FLAG_PROMISC |
1447 				 MLX4_EN_FLAG_MC_PROMISC);
1448 		mlx4_flow_steer_promisc_remove(mdev->dev,
1449 					       priv->port,
1450 					       MLX4_FS_ALL_DEFAULT);
1451 		mlx4_flow_steer_promisc_remove(mdev->dev,
1452 					       priv->port,
1453 					       MLX4_FS_MC_DEFAULT);
1454 	} else if (priv->flags & MLX4_EN_FLAG_PROMISC) {
1455 		priv->flags &= ~MLX4_EN_FLAG_PROMISC;
1456 
1457 		/* Disable promiscouos mode */
1458 		mlx4_unicast_promisc_remove(mdev->dev, priv->base_qpn,
1459 					    priv->port);
1460 
1461 		/* Disable Multicast promisc */
1462 		if (priv->flags & MLX4_EN_FLAG_MC_PROMISC) {
1463 			mlx4_multicast_promisc_remove(mdev->dev, priv->base_qpn,
1464 						      priv->port);
1465 			priv->flags &= ~MLX4_EN_FLAG_MC_PROMISC;
1466 		}
1467 	}
1468 
1469 	/* Detach All multicasts */
1470 	memset(&mc_list[10], 0xff, ETH_ALEN);
1471 	mc_list[5] = priv->port; /* needed for B0 steering support */
1472 	mlx4_multicast_detach(mdev->dev, &priv->rss_map.indir_qp, mc_list,
1473 			      MLX4_PROT_ETH, priv->broadcast_id);
1474 	list_for_each_entry(mclist, &priv->curr_list, list) {
1475 		memcpy(&mc_list[10], mclist->addr, ETH_ALEN);
1476 		mc_list[5] = priv->port;
1477 		mlx4_multicast_detach(mdev->dev, &priv->rss_map.indir_qp,
1478 				      mc_list, MLX4_PROT_ETH, mclist->reg_id);
1479 	}
1480 	mlx4_en_clear_list(dev);
1481 	list_for_each_entry_safe(mclist, tmp, &priv->curr_list, list) {
1482 		list_del(&mclist->list);
1483 		kfree(mclist);
1484 	}
1485 
1486 	/* Flush multicast filter */
1487 	mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, 0, 1, MLX4_MCAST_CONFIG);
1488 	mlx4_en_destroy_drop_qp(priv);
1489 
1490 	/* Free TX Rings */
1491 	for (i = 0; i < priv->tx_ring_num; i++) {
1492 		mlx4_en_deactivate_tx_ring(priv, priv->tx_ring[i]);
1493 		mlx4_en_deactivate_cq(priv, priv->tx_cq[i]);
1494 	}
1495 	msleep(10);
1496 
1497 	for (i = 0; i < priv->tx_ring_num; i++)
1498 		mlx4_en_free_tx_buf(dev, priv->tx_ring[i]);
1499 
1500 	/* Free RSS qps */
1501 	mlx4_en_release_rss_steer(priv);
1502 
1503 	/* Unregister Mac address for the port */
1504 	mlx4_en_put_qp(priv);
1505 	mdev->mac_removed[priv->port] = 1;
1506 
1507 	/* Free RX Rings */
1508 	for (i = 0; i < priv->rx_ring_num; i++) {
1509 		struct mlx4_en_cq *cq = priv->rx_cq[i];
1510 		mlx4_en_deactivate_rx_ring(priv, priv->rx_ring[i]);
1511 		mlx4_en_deactivate_cq(priv, cq);
1512 	}
1513 
1514         callout_stop(&priv->watchdog_timer);
1515 
1516         dev->if_drv_flags &= ~(IFF_DRV_RUNNING | IFF_DRV_OACTIVE);
1517 }
1518 
1519 static void mlx4_en_restart(struct work_struct *work)
1520 {
1521 	struct mlx4_en_priv *priv = container_of(work, struct mlx4_en_priv,
1522 						 watchdog_task);
1523 	struct mlx4_en_dev *mdev = priv->mdev;
1524 	struct net_device *dev = priv->dev;
1525 	struct mlx4_en_tx_ring *ring;
1526 	int i;
1527 
1528 
1529 	if (priv->blocked == 0 || priv->port_up == 0)
1530 		return;
1531 	for (i = 0; i < priv->tx_ring_num; i++) {
1532 		ring = priv->tx_ring[i];
1533 		if (ring->blocked &&
1534 				ring->watchdog_time + MLX4_EN_WATCHDOG_TIMEOUT < ticks)
1535 			goto reset;
1536 	}
1537 	return;
1538 
1539 reset:
1540 	priv->port_stats.tx_timeout++;
1541 	en_dbg(DRV, priv, "Watchdog task called for port %d\n", priv->port);
1542 
1543 	mutex_lock(&mdev->state_lock);
1544 	if (priv->port_up) {
1545 		mlx4_en_stop_port(dev);
1546                 //for (i = 0; i < priv->tx_ring_num; i++)
1547                 //        netdev_tx_reset_queue(priv->tx_ring[i]->tx_queue);
1548 		if (mlx4_en_start_port(dev))
1549 			en_err(priv, "Failed restarting port %d\n", priv->port);
1550 	}
1551 	mutex_unlock(&mdev->state_lock);
1552 }
1553 
1554 static void mlx4_en_clear_stats(struct net_device *dev)
1555 {
1556 	struct mlx4_en_priv *priv = netdev_priv(dev);
1557 	struct mlx4_en_dev *mdev = priv->mdev;
1558 	int i;
1559 
1560 	if (!mlx4_is_slave(mdev->dev))
1561 		if (mlx4_en_DUMP_ETH_STATS(mdev, priv->port, 1))
1562 			en_dbg(HW, priv, "Failed dumping statistics\n");
1563 
1564 	memset(&priv->pstats, 0, sizeof(priv->pstats));
1565 	memset(&priv->pkstats, 0, sizeof(priv->pkstats));
1566 	memset(&priv->port_stats, 0, sizeof(priv->port_stats));
1567 	memset(&priv->vport_stats, 0, sizeof(priv->vport_stats));
1568 
1569 	for (i = 0; i < priv->tx_ring_num; i++) {
1570 		priv->tx_ring[i]->bytes = 0;
1571 		priv->tx_ring[i]->packets = 0;
1572 		priv->tx_ring[i]->tx_csum = 0;
1573 		priv->tx_ring[i]->oversized_packets = 0;
1574 	}
1575 	for (i = 0; i < priv->rx_ring_num; i++) {
1576 		priv->rx_ring[i]->bytes = 0;
1577 		priv->rx_ring[i]->packets = 0;
1578 		priv->rx_ring[i]->csum_ok = 0;
1579 		priv->rx_ring[i]->csum_none = 0;
1580 	}
1581 }
1582 
1583 static void mlx4_en_open(void* arg)
1584 {
1585 
1586         struct mlx4_en_priv *priv;
1587         struct mlx4_en_dev *mdev;
1588         struct net_device *dev;
1589         int err = 0;
1590 
1591         priv = arg;
1592         mdev = priv->mdev;
1593         dev = priv->dev;
1594 
1595 
1596 	mutex_lock(&mdev->state_lock);
1597 
1598 	if (!mdev->device_up) {
1599 		en_err(priv, "Cannot open - device down/disabled\n");
1600 		goto out;
1601 	}
1602 
1603 	/* Reset HW statistics and SW counters */
1604 	mlx4_en_clear_stats(dev);
1605 
1606 	err = mlx4_en_start_port(dev);
1607 	if (err)
1608 		en_err(priv, "Failed starting port:%d\n", priv->port);
1609 
1610 out:
1611 	mutex_unlock(&mdev->state_lock);
1612 	return;
1613 }
1614 
1615 void mlx4_en_free_resources(struct mlx4_en_priv *priv)
1616 {
1617 	int i;
1618 
1619 #ifdef CONFIG_RFS_ACCEL
1620 	if (priv->dev->rx_cpu_rmap) {
1621 		free_irq_cpu_rmap(priv->dev->rx_cpu_rmap);
1622 		priv->dev->rx_cpu_rmap = NULL;
1623 	}
1624 #endif
1625 
1626 	for (i = 0; i < priv->tx_ring_num; i++) {
1627 		if (priv->tx_ring && priv->tx_ring[i])
1628 			mlx4_en_destroy_tx_ring(priv, &priv->tx_ring[i]);
1629 		if (priv->tx_cq && priv->tx_cq[i])
1630 			mlx4_en_destroy_cq(priv, &priv->tx_cq[i]);
1631 	}
1632 
1633 	for (i = 0; i < priv->rx_ring_num; i++) {
1634 		if (priv->rx_ring[i])
1635 			mlx4_en_destroy_rx_ring(priv, &priv->rx_ring[i],
1636 				priv->prof->rx_ring_size, priv->stride);
1637 		if (priv->rx_cq[i])
1638 			mlx4_en_destroy_cq(priv, &priv->rx_cq[i]);
1639 	}
1640 
1641 	if (priv->stat_sysctl != NULL)
1642 		sysctl_ctx_free(&priv->stat_ctx);
1643 }
1644 
1645 int mlx4_en_alloc_resources(struct mlx4_en_priv *priv)
1646 {
1647 	struct mlx4_en_port_profile *prof = priv->prof;
1648 	int i;
1649 	int node = 0;
1650 
1651 	/* Create rx Rings */
1652 	for (i = 0; i < priv->rx_ring_num; i++) {
1653 		if (mlx4_en_create_cq(priv, &priv->rx_cq[i],
1654 				      prof->rx_ring_size, i, RX, node))
1655 			goto err;
1656 
1657 		if (mlx4_en_create_rx_ring(priv, &priv->rx_ring[i],
1658 					   prof->rx_ring_size, node))
1659 			goto err;
1660 	}
1661 
1662 	/* Create tx Rings */
1663 	for (i = 0; i < priv->tx_ring_num; i++) {
1664 		if (mlx4_en_create_cq(priv, &priv->tx_cq[i],
1665 				      prof->tx_ring_size, i, TX, node))
1666 			goto err;
1667 
1668 		if (mlx4_en_create_tx_ring(priv, &priv->tx_ring[i],
1669 					   prof->tx_ring_size, TXBB_SIZE, node, i))
1670 			goto err;
1671 	}
1672 
1673 #ifdef CONFIG_RFS_ACCEL
1674 	priv->dev->rx_cpu_rmap = alloc_irq_cpu_rmap(priv->rx_ring_num);
1675 	if (!priv->dev->rx_cpu_rmap)
1676 		goto err;
1677 #endif
1678         /* Re-create stat sysctls in case the number of rings changed. */
1679 	mlx4_en_sysctl_stat(priv);
1680 	return 0;
1681 
1682 err:
1683 	en_err(priv, "Failed to allocate NIC resources\n");
1684 	for (i = 0; i < priv->rx_ring_num; i++) {
1685 		if (priv->rx_ring[i])
1686 			mlx4_en_destroy_rx_ring(priv, &priv->rx_ring[i],
1687 						prof->rx_ring_size,
1688 						priv->stride);
1689 		if (priv->rx_cq[i])
1690 			mlx4_en_destroy_cq(priv, &priv->rx_cq[i]);
1691 	}
1692 	for (i = 0; i < priv->tx_ring_num; i++) {
1693 		if (priv->tx_ring[i])
1694 			mlx4_en_destroy_tx_ring(priv, &priv->tx_ring[i]);
1695 		if (priv->tx_cq[i])
1696 			mlx4_en_destroy_cq(priv, &priv->tx_cq[i]);
1697 	}
1698 	priv->port_up = false;
1699 	return -ENOMEM;
1700 }
1701 
1702 struct en_port_attribute {
1703 	struct attribute attr;
1704 	ssize_t (*show)(struct en_port *, struct en_port_attribute *, char *buf);
1705 	ssize_t (*store)(struct en_port *, struct en_port_attribute *, char *buf, size_t count);
1706 };
1707 
1708 #define PORT_ATTR_RO(_name) \
1709 struct en_port_attribute en_port_attr_##_name = __ATTR_RO(_name)
1710 
1711 #define EN_PORT_ATTR(_name, _mode, _show, _store) \
1712 struct en_port_attribute en_port_attr_##_name = __ATTR(_name, _mode, _show, _store)
1713 
1714 void mlx4_en_destroy_netdev(struct net_device *dev)
1715 {
1716 	struct mlx4_en_priv *priv = netdev_priv(dev);
1717 	struct mlx4_en_dev *mdev = priv->mdev;
1718 
1719 	en_dbg(DRV, priv, "Destroying netdev on port:%d\n", priv->port);
1720 
1721         if (priv->vlan_attach != NULL)
1722                 EVENTHANDLER_DEREGISTER(vlan_config, priv->vlan_attach);
1723         if (priv->vlan_detach != NULL)
1724                 EVENTHANDLER_DEREGISTER(vlan_unconfig, priv->vlan_detach);
1725 
1726 	/* Unregister device - this will close the port if it was up */
1727 	if (priv->registered) {
1728 		mutex_lock(&mdev->state_lock);
1729 		ether_ifdetach(dev);
1730 		mutex_unlock(&mdev->state_lock);
1731 	}
1732 
1733 	mutex_lock(&mdev->state_lock);
1734 	mlx4_en_stop_port(dev);
1735 	mutex_unlock(&mdev->state_lock);
1736 
1737 	if (priv->allocated)
1738 		mlx4_free_hwq_res(mdev->dev, &priv->res, MLX4_EN_PAGE_SIZE);
1739 
1740 	cancel_delayed_work(&priv->stats_task);
1741 	cancel_delayed_work(&priv->service_task);
1742 	/* flush any pending task for this netdev */
1743 	flush_workqueue(mdev->workqueue);
1744         callout_drain(&priv->watchdog_timer);
1745 
1746 	/* Detach the netdev so tasks would not attempt to access it */
1747 	mutex_lock(&mdev->state_lock);
1748 	mdev->pndev[priv->port] = NULL;
1749 	mutex_unlock(&mdev->state_lock);
1750 
1751 
1752 	mlx4_en_free_resources(priv);
1753 
1754 	/* freeing the sysctl conf cannot be called from within mlx4_en_free_resources */
1755 	if (priv->conf_sysctl != NULL)
1756 		sysctl_ctx_free(&priv->conf_ctx);
1757 
1758 	kfree(priv->tx_ring);
1759 	kfree(priv->tx_cq);
1760 
1761         kfree(priv);
1762         if_free(dev);
1763 
1764 }
1765 
1766 static int mlx4_en_change_mtu(struct net_device *dev, int new_mtu)
1767 {
1768 	struct mlx4_en_priv *priv = netdev_priv(dev);
1769 	struct mlx4_en_dev *mdev = priv->mdev;
1770 	int err = 0;
1771 
1772 	en_dbg(DRV, priv, "Change MTU called - current:%u new:%u\n",
1773 	       (unsigned)dev->if_mtu, (unsigned)new_mtu);
1774 
1775 	if ((new_mtu < MLX4_EN_MIN_MTU) || (new_mtu > priv->max_mtu)) {
1776 		en_err(priv, "Bad MTU size:%d.\n", new_mtu);
1777 		return -EPERM;
1778 	}
1779 	mutex_lock(&mdev->state_lock);
1780 	dev->if_mtu = new_mtu;
1781 	if (dev->if_drv_flags & IFF_DRV_RUNNING) {
1782 		if (!mdev->device_up) {
1783 			/* NIC is probably restarting - let watchdog task reset
1784 			 *                          * the port */
1785 			en_dbg(DRV, priv, "Change MTU called with card down!?\n");
1786 		} else {
1787 			mlx4_en_stop_port(dev);
1788 			err = mlx4_en_start_port(dev);
1789 			if (err) {
1790 				en_err(priv, "Failed restarting port:%d\n",
1791 						priv->port);
1792 				queue_work(mdev->workqueue, &priv->watchdog_task);
1793 			}
1794 		}
1795 	}
1796 	mutex_unlock(&mdev->state_lock);
1797 	return 0;
1798 }
1799 
1800 static int mlx4_en_calc_media(struct mlx4_en_priv *priv)
1801 {
1802 	int trans_type;
1803 	int active;
1804 
1805 	active = IFM_ETHER;
1806 	if (priv->last_link_state == MLX4_DEV_EVENT_PORT_DOWN)
1807 		return (active);
1808 	active |= IFM_FDX;
1809 	trans_type = priv->port_state.transciver;
1810 	/* XXX I don't know all of the transceiver values. */
1811 	switch (priv->port_state.link_speed) {
1812 	case 1000:
1813 		active |= IFM_1000_T;
1814 		break;
1815 	case 10000:
1816 		if (trans_type > 0 && trans_type <= 0xC)
1817 			active |= IFM_10G_SR;
1818 		else if (trans_type == 0x80 || trans_type == 0)
1819 			active |= IFM_10G_CX4;
1820 		break;
1821 	case 40000:
1822 		active |= IFM_40G_CR4;
1823 		break;
1824 	}
1825 	if (priv->prof->tx_pause)
1826 		active |= IFM_ETH_TXPAUSE;
1827 	if (priv->prof->rx_pause)
1828 		active |= IFM_ETH_RXPAUSE;
1829 
1830 	return (active);
1831 }
1832 
1833 static void mlx4_en_media_status(struct ifnet *dev, struct ifmediareq *ifmr)
1834 {
1835 	struct mlx4_en_priv *priv;
1836 
1837 	priv = dev->if_softc;
1838 	ifmr->ifm_status = IFM_AVALID;
1839 	if (priv->last_link_state != MLX4_DEV_EVENT_PORT_DOWN)
1840 		ifmr->ifm_status |= IFM_ACTIVE;
1841 	ifmr->ifm_active = mlx4_en_calc_media(priv);
1842 
1843 	return;
1844 }
1845 
1846 static int mlx4_en_media_change(struct ifnet *dev)
1847 {
1848 	struct mlx4_en_priv *priv;
1849         struct ifmedia *ifm;
1850 	int rxpause;
1851 	int txpause;
1852 	int error;
1853 
1854 	priv = dev->if_softc;
1855 	ifm = &priv->media;
1856 	rxpause = txpause = 0;
1857 	error = 0;
1858 
1859 	if (IFM_TYPE(ifm->ifm_media) != IFM_ETHER)
1860 		return (EINVAL);
1861         switch (IFM_SUBTYPE(ifm->ifm_media)) {
1862         case IFM_AUTO:
1863 		break;
1864 	case IFM_10G_SR:
1865 	case IFM_10G_CX4:
1866 	case IFM_1000_T:
1867 	case IFM_40G_CR4:
1868 		if ((IFM_SUBTYPE(ifm->ifm_media)
1869 			== IFM_SUBTYPE(mlx4_en_calc_media(priv)))
1870 			&& (ifm->ifm_media & IFM_FDX))
1871 			break;
1872 		/* Fallthrough */
1873 	default:
1874                 printf("%s: Only auto media type\n", if_name(dev));
1875                 return (EINVAL);
1876 	}
1877 	/* Allow user to set/clear pause */
1878 	if (IFM_OPTIONS(ifm->ifm_media) & IFM_ETH_RXPAUSE)
1879 		rxpause = 1;
1880 	if (IFM_OPTIONS(ifm->ifm_media) & IFM_ETH_TXPAUSE)
1881 		txpause = 1;
1882 	if (priv->prof->tx_pause != txpause || priv->prof->rx_pause != rxpause) {
1883 		priv->prof->tx_pause = txpause;
1884 		priv->prof->rx_pause = rxpause;
1885 		error = -mlx4_SET_PORT_general(priv->mdev->dev, priv->port,
1886 		     priv->rx_mb_size + ETHER_CRC_LEN, priv->prof->tx_pause,
1887 		     priv->prof->tx_ppp, priv->prof->rx_pause,
1888 		     priv->prof->rx_ppp);
1889 	}
1890 	return (error);
1891 }
1892 
1893 static int mlx4_en_ioctl(struct ifnet *dev, u_long command, caddr_t data)
1894 {
1895 	struct mlx4_en_priv *priv;
1896 	struct mlx4_en_dev *mdev;
1897 	struct ifreq *ifr;
1898 	int error;
1899 	int mask;
1900 
1901 	error = 0;
1902 	mask = 0;
1903 	priv = dev->if_softc;
1904 	mdev = priv->mdev;
1905 	ifr = (struct ifreq *) data;
1906 	switch (command) {
1907 
1908 	case SIOCSIFMTU:
1909 		error = -mlx4_en_change_mtu(dev, ifr->ifr_mtu);
1910 		break;
1911 	case SIOCSIFFLAGS:
1912 		if (dev->if_flags & IFF_UP) {
1913 			if ((dev->if_drv_flags & IFF_DRV_RUNNING) == 0) {
1914 				mutex_lock(&mdev->state_lock);
1915 				mlx4_en_start_port(dev);
1916 				mutex_unlock(&mdev->state_lock);
1917 			} else {
1918 				mlx4_en_set_rx_mode(dev);
1919 			}
1920 		} else {
1921 			mutex_lock(&mdev->state_lock);
1922 			if (dev->if_drv_flags & IFF_DRV_RUNNING) {
1923 				mlx4_en_stop_port(dev);
1924 				if_link_state_change(dev, LINK_STATE_DOWN);
1925 			}
1926 			mutex_unlock(&mdev->state_lock);
1927 		}
1928 		break;
1929 	case SIOCADDMULTI:
1930 	case SIOCDELMULTI:
1931 		mlx4_en_set_rx_mode(dev);
1932 		break;
1933 	case SIOCSIFMEDIA:
1934 	case SIOCGIFMEDIA:
1935 		error = ifmedia_ioctl(dev, ifr, &priv->media, command);
1936 		break;
1937 	case SIOCSIFCAP:
1938 		mutex_lock(&mdev->state_lock);
1939 		mask = ifr->ifr_reqcap ^ dev->if_capenable;
1940 		if (mask & IFCAP_TXCSUM) {
1941 			dev->if_capenable ^= IFCAP_TXCSUM;
1942 			dev->if_hwassist ^= (CSUM_TCP | CSUM_UDP | CSUM_IP);
1943 
1944 			if (IFCAP_TSO4 & dev->if_capenable &&
1945 			    !(IFCAP_TXCSUM & dev->if_capenable)) {
1946 				dev->if_capenable &= ~IFCAP_TSO4;
1947 				dev->if_hwassist &= ~CSUM_IP_TSO;
1948 				if_printf(dev,
1949 				    "tso4 disabled due to -txcsum.\n");
1950 			}
1951 		}
1952 		if (mask & IFCAP_TXCSUM_IPV6) {
1953 			dev->if_capenable ^= IFCAP_TXCSUM_IPV6;
1954 			dev->if_hwassist ^= (CSUM_UDP_IPV6 | CSUM_TCP_IPV6);
1955 
1956 			if (IFCAP_TSO6 & dev->if_capenable &&
1957 			    !(IFCAP_TXCSUM_IPV6 & dev->if_capenable)) {
1958 				dev->if_capenable &= ~IFCAP_TSO6;
1959 				dev->if_hwassist &= ~CSUM_IP6_TSO;
1960 				if_printf(dev,
1961 				    "tso6 disabled due to -txcsum6.\n");
1962 			}
1963 		}
1964 		if (mask & IFCAP_RXCSUM)
1965 			dev->if_capenable ^= IFCAP_RXCSUM;
1966 		if (mask & IFCAP_RXCSUM_IPV6)
1967 			dev->if_capenable ^= IFCAP_RXCSUM_IPV6;
1968 
1969 		if (mask & IFCAP_TSO4) {
1970 			if (!(IFCAP_TSO4 & dev->if_capenable) &&
1971 			    !(IFCAP_TXCSUM & dev->if_capenable)) {
1972 				if_printf(dev, "enable txcsum first.\n");
1973 				error = EAGAIN;
1974 				goto out;
1975 			}
1976 			dev->if_capenable ^= IFCAP_TSO4;
1977 			dev->if_hwassist ^= CSUM_IP_TSO;
1978 		}
1979 		if (mask & IFCAP_TSO6) {
1980 			if (!(IFCAP_TSO6 & dev->if_capenable) &&
1981 			    !(IFCAP_TXCSUM_IPV6 & dev->if_capenable)) {
1982 				if_printf(dev, "enable txcsum6 first.\n");
1983 				error = EAGAIN;
1984 				goto out;
1985 			}
1986 			dev->if_capenable ^= IFCAP_TSO6;
1987 			dev->if_hwassist ^= CSUM_IP6_TSO;
1988 		}
1989 		if (mask & IFCAP_LRO)
1990 			dev->if_capenable ^= IFCAP_LRO;
1991 		if (mask & IFCAP_VLAN_HWTAGGING)
1992 			dev->if_capenable ^= IFCAP_VLAN_HWTAGGING;
1993 		if (mask & IFCAP_VLAN_HWFILTER)
1994 			dev->if_capenable ^= IFCAP_VLAN_HWFILTER;
1995 		if (mask & IFCAP_WOL_MAGIC)
1996 			dev->if_capenable ^= IFCAP_WOL_MAGIC;
1997 		if (dev->if_drv_flags & IFF_DRV_RUNNING)
1998 			mlx4_en_start_port(dev);
1999 out:
2000 		mutex_unlock(&mdev->state_lock);
2001 		VLAN_CAPABILITIES(dev);
2002 		break;
2003 #if __FreeBSD_version >= 1100036
2004 	case SIOCGI2C: {
2005 		struct ifi2creq i2c;
2006 
2007 		error = copyin(ifr->ifr_data, &i2c, sizeof(i2c));
2008 		if (error)
2009 			break;
2010 		if (i2c.len > sizeof(i2c.data)) {
2011 			error = EINVAL;
2012 			break;
2013 		}
2014 		/*
2015 		 * Note that we ignore i2c.addr here. The driver hardcodes
2016 		 * the address to 0x50, while standard expects it to be 0xA0.
2017 		 */
2018 		error = mlx4_get_module_info(mdev->dev, priv->port,
2019 		    i2c.offset, i2c.len, i2c.data);
2020 		if (error < 0) {
2021 			error = -error;
2022 			break;
2023 		}
2024 		error = copyout(&i2c, ifr->ifr_data, sizeof(i2c));
2025 		break;
2026 	}
2027 #endif
2028 	default:
2029 		error = ether_ioctl(dev, command, data);
2030 		break;
2031 	}
2032 
2033 	return (error);
2034 }
2035 
2036 
2037 int mlx4_en_init_netdev(struct mlx4_en_dev *mdev, int port,
2038 			struct mlx4_en_port_profile *prof)
2039 {
2040 	struct net_device *dev;
2041 	struct mlx4_en_priv *priv;
2042 	uint8_t dev_addr[ETHER_ADDR_LEN];
2043 	int err;
2044 	int i;
2045 
2046 	priv = kzalloc(sizeof(*priv), GFP_KERNEL);
2047 	dev = priv->dev = if_alloc(IFT_ETHER);
2048 	if (dev == NULL) {
2049 		en_err(priv, "Net device allocation failed\n");
2050 		kfree(priv);
2051 		return -ENOMEM;
2052 	}
2053 	dev->if_softc = priv;
2054 	if_initname(dev, "mlxen", (device_get_unit(
2055 	    mdev->pdev->dev.bsddev) * MLX4_MAX_PORTS) + port - 1);
2056 	dev->if_mtu = ETHERMTU;
2057 	dev->if_init = mlx4_en_open;
2058 	dev->if_flags = IFF_BROADCAST | IFF_SIMPLEX | IFF_MULTICAST;
2059 	dev->if_ioctl = mlx4_en_ioctl;
2060 	dev->if_transmit = mlx4_en_transmit;
2061 	dev->if_qflush = mlx4_en_qflush;
2062 	dev->if_snd.ifq_maxlen = prof->tx_ring_size;
2063 
2064 	/*
2065 	 * Initialize driver private data
2066 	 */
2067 	priv->counter_index = 0xff;
2068 	spin_lock_init(&priv->stats_lock);
2069 	INIT_WORK(&priv->rx_mode_task, mlx4_en_do_set_rx_mode);
2070 	INIT_WORK(&priv->watchdog_task, mlx4_en_restart);
2071 	INIT_WORK(&priv->linkstate_task, mlx4_en_linkstate);
2072 	INIT_DELAYED_WORK(&priv->stats_task, mlx4_en_do_get_stats);
2073 	INIT_DELAYED_WORK(&priv->service_task, mlx4_en_service_task);
2074 	callout_init(&priv->watchdog_timer, 1);
2075 #ifdef CONFIG_RFS_ACCEL
2076 	INIT_LIST_HEAD(&priv->filters);
2077 	spin_lock_init(&priv->filters_lock);
2078 #endif
2079 
2080 	priv->msg_enable = MLX4_EN_MSG_LEVEL;
2081 	priv->dev = dev;
2082 	priv->mdev = mdev;
2083 	priv->ddev = &mdev->pdev->dev;
2084 	priv->prof = prof;
2085 	priv->port = port;
2086 	priv->port_up = false;
2087 	priv->flags = prof->flags;
2088 
2089 	priv->num_tx_rings_p_up = mdev->profile.num_tx_rings_p_up;
2090 	priv->tx_ring_num = prof->tx_ring_num;
2091 	priv->tx_ring = kcalloc(MAX_TX_RINGS,
2092 				sizeof(struct mlx4_en_tx_ring *), GFP_KERNEL);
2093 	if (!priv->tx_ring) {
2094 		err = -ENOMEM;
2095 		goto out;
2096 	}
2097 	priv->tx_cq = kcalloc(sizeof(struct mlx4_en_cq *), MAX_TX_RINGS,
2098 			GFP_KERNEL);
2099 	if (!priv->tx_cq) {
2100 		err = -ENOMEM;
2101 		goto out;
2102 	}
2103 
2104 	priv->rx_ring_num = prof->rx_ring_num;
2105 	priv->cqe_factor = (mdev->dev->caps.cqe_size == 64) ? 1 : 0;
2106 	priv->mac_index = -1;
2107 	priv->last_ifq_jiffies = 0;
2108 	priv->if_counters_rx_errors = 0;
2109 	priv->if_counters_rx_no_buffer = 0;
2110 #ifdef CONFIG_MLX4_EN_DCB
2111 	if (!mlx4_is_slave(priv->mdev->dev)) {
2112 		priv->dcbx_cap = DCB_CAP_DCBX_HOST;
2113 		priv->flags |= MLX4_EN_FLAG_DCB_ENABLED;
2114 		if (mdev->dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_ETS_CFG) {
2115 			dev->dcbnl_ops = &mlx4_en_dcbnl_ops;
2116 		} else {
2117 			en_info(priv, "QoS disabled - no HW support\n");
2118 			dev->dcbnl_ops = &mlx4_en_dcbnl_pfc_ops;
2119 		}
2120 	}
2121 #endif
2122 
2123 	for (i = 0; i < MLX4_EN_MAC_HASH_SIZE; ++i)
2124 		INIT_HLIST_HEAD(&priv->mac_hash[i]);
2125 
2126 	/* Query for default mac and max mtu */
2127 	priv->max_mtu = mdev->dev->caps.eth_mtu_cap[priv->port];
2128         priv->mac = mdev->dev->caps.def_mac[priv->port];
2129         if (ILLEGAL_MAC(priv->mac)) {
2130 #if BITS_PER_LONG == 64
2131                 en_err(priv, "Port: %d, invalid mac burned: 0x%lx, quiting\n",
2132                                 priv->port, priv->mac);
2133 #elif BITS_PER_LONG == 32
2134                 en_err(priv, "Port: %d, invalid mac burned: 0x%llx, quiting\n",
2135                                 priv->port, priv->mac);
2136 #endif
2137                 err = -EINVAL;
2138                 goto out;
2139         }
2140 
2141 	priv->stride = roundup_pow_of_two(sizeof(struct mlx4_en_rx_desc) +
2142 					  DS_SIZE);
2143 
2144 	mlx4_en_sysctl_conf(priv);
2145 
2146 	err = mlx4_en_alloc_resources(priv);
2147 	if (err)
2148 		goto out;
2149 
2150 	/* Allocate page for receive rings */
2151 	err = mlx4_alloc_hwq_res(mdev->dev, &priv->res,
2152 				MLX4_EN_PAGE_SIZE, MLX4_EN_PAGE_SIZE);
2153 	if (err) {
2154 		en_err(priv, "Failed to allocate page for rx qps\n");
2155 		goto out;
2156 	}
2157 	priv->allocated = 1;
2158 
2159 	/*
2160 	 * Set driver features
2161 	 */
2162 	dev->if_capabilities |= IFCAP_HWCSUM | IFCAP_HWCSUM_IPV6;
2163 	dev->if_capabilities |= IFCAP_VLAN_MTU | IFCAP_VLAN_HWTAGGING;
2164 	dev->if_capabilities |= IFCAP_VLAN_HWCSUM | IFCAP_VLAN_HWFILTER;
2165 	dev->if_capabilities |= IFCAP_LINKSTATE | IFCAP_JUMBO_MTU;
2166 	dev->if_capabilities |= IFCAP_LRO;
2167 	dev->if_capabilities |= IFCAP_HWSTATS;
2168 
2169 	if (mdev->LSO_support)
2170 		dev->if_capabilities |= IFCAP_TSO4 | IFCAP_TSO6 | IFCAP_VLAN_HWTSO;
2171 
2172 #if __FreeBSD_version >= 1100000
2173 	/* set TSO limits so that we don't have to drop TX packets */
2174 	dev->if_hw_tsomax = MLX4_EN_TX_MAX_PAYLOAD_SIZE - (ETHER_HDR_LEN + ETHER_VLAN_ENCAP_LEN) /* hdr */;
2175 	dev->if_hw_tsomaxsegcount = MLX4_EN_TX_MAX_MBUF_FRAGS - 1 /* hdr */;
2176 	dev->if_hw_tsomaxsegsize = MLX4_EN_TX_MAX_MBUF_SIZE;
2177 #endif
2178 
2179 	dev->if_capenable = dev->if_capabilities;
2180 
2181 	dev->if_hwassist = 0;
2182 	if (dev->if_capenable & (IFCAP_TSO4 | IFCAP_TSO6))
2183 		dev->if_hwassist |= CSUM_TSO;
2184 	if (dev->if_capenable & IFCAP_TXCSUM)
2185 		dev->if_hwassist |= (CSUM_TCP | CSUM_UDP | CSUM_IP);
2186 	if (dev->if_capenable & IFCAP_TXCSUM_IPV6)
2187 		dev->if_hwassist |= (CSUM_UDP_IPV6 | CSUM_TCP_IPV6);
2188 
2189 
2190         /* Register for VLAN events */
2191 	priv->vlan_attach = EVENTHANDLER_REGISTER(vlan_config,
2192             mlx4_en_vlan_rx_add_vid, priv, EVENTHANDLER_PRI_FIRST);
2193 	priv->vlan_detach = EVENTHANDLER_REGISTER(vlan_unconfig,
2194             mlx4_en_vlan_rx_kill_vid, priv, EVENTHANDLER_PRI_FIRST);
2195 
2196 	mdev->pndev[priv->port] = dev;
2197 
2198 	priv->last_link_state = MLX4_DEV_EVENT_PORT_DOWN;
2199         mlx4_en_set_default_moderation(priv);
2200 
2201 	/* Set default MAC */
2202 	for (i = 0; i < ETHER_ADDR_LEN; i++)
2203 		dev_addr[ETHER_ADDR_LEN - 1 - i] = (u8) (priv->mac >> (8 * i));
2204 
2205 
2206 	ether_ifattach(dev, dev_addr);
2207 	if_link_state_change(dev, LINK_STATE_DOWN);
2208 	ifmedia_init(&priv->media, IFM_IMASK | IFM_ETH_FMASK,
2209 	    mlx4_en_media_change, mlx4_en_media_status);
2210 	ifmedia_add(&priv->media, IFM_ETHER | IFM_FDX | IFM_1000_T, 0, NULL);
2211 	ifmedia_add(&priv->media, IFM_ETHER | IFM_FDX | IFM_10G_SR, 0, NULL);
2212 	ifmedia_add(&priv->media, IFM_ETHER | IFM_FDX | IFM_10G_CX4, 0, NULL);
2213 	ifmedia_add(&priv->media, IFM_ETHER | IFM_FDX | IFM_40G_CR4, 0, NULL);
2214 	ifmedia_add(&priv->media, IFM_ETHER | IFM_AUTO, 0, NULL);
2215 	ifmedia_set(&priv->media, IFM_ETHER | IFM_AUTO);
2216 
2217 	en_warn(priv, "Using %d TX rings\n", prof->tx_ring_num);
2218 	en_warn(priv, "Using %d RX rings\n", prof->rx_ring_num);
2219 
2220 	priv->registered = 1;
2221 
2222         en_warn(priv, "Using %d TX rings\n", prof->tx_ring_num);
2223         en_warn(priv, "Using %d RX rings\n", prof->rx_ring_num);
2224 
2225 
2226 	priv->rx_mb_size = dev->if_mtu + ETH_HLEN + VLAN_HLEN + ETH_FCS_LEN;
2227 	err = mlx4_SET_PORT_general(mdev->dev, priv->port,
2228 				    priv->rx_mb_size,
2229 				    prof->tx_pause, prof->tx_ppp,
2230 				    prof->rx_pause, prof->rx_ppp);
2231 	if (err) {
2232 		en_err(priv, "Failed setting port general configurations "
2233 		       "for port %d, with error %d\n", priv->port, err);
2234 		goto out;
2235 	}
2236 
2237 	/* Init port */
2238 	en_warn(priv, "Initializing port\n");
2239 	err = mlx4_INIT_PORT(mdev->dev, priv->port);
2240 	if (err) {
2241 		en_err(priv, "Failed Initializing port\n");
2242 		goto out;
2243 	}
2244 
2245 	queue_delayed_work(mdev->workqueue, &priv->stats_task, STATS_DELAY);
2246 
2247         if (mdev->dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_TS)
2248                 queue_delayed_work(mdev->workqueue, &priv->service_task, SERVICE_TASK_DELAY);
2249 
2250 	return 0;
2251 
2252 out:
2253 	mlx4_en_destroy_netdev(dev);
2254 	return err;
2255 }
2256 
2257 static int mlx4_en_set_ring_size(struct net_device *dev,
2258     int rx_size, int tx_size)
2259 {
2260         struct mlx4_en_priv *priv = netdev_priv(dev);
2261         struct mlx4_en_dev *mdev = priv->mdev;
2262         int port_up = 0;
2263         int err = 0;
2264 
2265         rx_size = roundup_pow_of_two(rx_size);
2266         rx_size = max_t(u32, rx_size, MLX4_EN_MIN_RX_SIZE);
2267         rx_size = min_t(u32, rx_size, MLX4_EN_MAX_RX_SIZE);
2268         tx_size = roundup_pow_of_two(tx_size);
2269         tx_size = max_t(u32, tx_size, MLX4_EN_MIN_TX_SIZE);
2270         tx_size = min_t(u32, tx_size, MLX4_EN_MAX_TX_SIZE);
2271 
2272         if (rx_size == (priv->port_up ?
2273             priv->rx_ring[0]->actual_size : priv->rx_ring[0]->size) &&
2274             tx_size == priv->tx_ring[0]->size)
2275                 return 0;
2276         mutex_lock(&mdev->state_lock);
2277         if (priv->port_up) {
2278                 port_up = 1;
2279                 mlx4_en_stop_port(dev);
2280         }
2281         mlx4_en_free_resources(priv);
2282         priv->prof->tx_ring_size = tx_size;
2283         priv->prof->rx_ring_size = rx_size;
2284         err = mlx4_en_alloc_resources(priv);
2285         if (err) {
2286                 en_err(priv, "Failed reallocating port resources\n");
2287                 goto out;
2288         }
2289         if (port_up) {
2290                 err = mlx4_en_start_port(dev);
2291                 if (err)
2292                         en_err(priv, "Failed starting port\n");
2293         }
2294 out:
2295         mutex_unlock(&mdev->state_lock);
2296         return err;
2297 }
2298 static int mlx4_en_set_rx_ring_size(SYSCTL_HANDLER_ARGS)
2299 {
2300         struct mlx4_en_priv *priv;
2301         int size;
2302         int error;
2303 
2304         priv = arg1;
2305         size = priv->prof->rx_ring_size;
2306         error = sysctl_handle_int(oidp, &size, 0, req);
2307         if (error || !req->newptr)
2308                 return (error);
2309         error = -mlx4_en_set_ring_size(priv->dev, size,
2310             priv->prof->tx_ring_size);
2311         return (error);
2312 }
2313 
2314 static int mlx4_en_set_tx_ring_size(SYSCTL_HANDLER_ARGS)
2315 {
2316         struct mlx4_en_priv *priv;
2317         int size;
2318         int error;
2319 
2320         priv = arg1;
2321         size = priv->prof->tx_ring_size;
2322         error = sysctl_handle_int(oidp, &size, 0, req);
2323         if (error || !req->newptr)
2324                 return (error);
2325         error = -mlx4_en_set_ring_size(priv->dev, priv->prof->rx_ring_size,
2326             size);
2327 
2328         return (error);
2329 }
2330 
2331 static int mlx4_en_get_module_info(struct net_device *dev,
2332 				   struct ethtool_modinfo *modinfo)
2333 {
2334 	struct mlx4_en_priv *priv = netdev_priv(dev);
2335 	struct mlx4_en_dev *mdev = priv->mdev;
2336 	int ret;
2337 	u8 data[4];
2338 
2339 	/* Read first 2 bytes to get Module & REV ID */
2340 	ret = mlx4_get_module_info(mdev->dev, priv->port,
2341 				   0/*offset*/, 2/*size*/, data);
2342 
2343 	if (ret < 2) {
2344 		en_err(priv, "Failed to read eeprom module first two bytes, error: 0x%x\n", -ret);
2345 		return -EIO;
2346 	}
2347 
2348 	switch (data[0] /* identifier */) {
2349 	case MLX4_MODULE_ID_QSFP:
2350 		modinfo->type = ETH_MODULE_SFF_8436;
2351 		modinfo->eeprom_len = ETH_MODULE_SFF_8436_LEN;
2352 		break;
2353 	case MLX4_MODULE_ID_QSFP_PLUS:
2354 		if (data[1] >= 0x3) { /* revision id */
2355 			modinfo->type = ETH_MODULE_SFF_8636;
2356 			modinfo->eeprom_len = ETH_MODULE_SFF_8636_LEN;
2357 		} else {
2358 			modinfo->type = ETH_MODULE_SFF_8436;
2359 			modinfo->eeprom_len = ETH_MODULE_SFF_8436_LEN;
2360 		}
2361 		break;
2362 	case MLX4_MODULE_ID_QSFP28:
2363 		modinfo->type = ETH_MODULE_SFF_8636;
2364 		modinfo->eeprom_len = ETH_MODULE_SFF_8636_LEN;
2365 		break;
2366 	case MLX4_MODULE_ID_SFP:
2367 		modinfo->type = ETH_MODULE_SFF_8472;
2368 		modinfo->eeprom_len = ETH_MODULE_SFF_8472_LEN;
2369 		break;
2370 	default:
2371 		en_err(priv, "mlx4_en_get_module_info :  Not recognized cable type\n");
2372 		return -EINVAL;
2373 	}
2374 
2375 	return 0;
2376 }
2377 
2378 static int mlx4_en_get_module_eeprom(struct net_device *dev,
2379 				     struct ethtool_eeprom *ee,
2380 				     u8 *data)
2381 {
2382 	struct mlx4_en_priv *priv = netdev_priv(dev);
2383 	struct mlx4_en_dev *mdev = priv->mdev;
2384 	int offset = ee->offset;
2385 	int i = 0, ret;
2386 
2387 	if (ee->len == 0)
2388 		return -EINVAL;
2389 
2390 	memset(data, 0, ee->len);
2391 
2392 	while (i < ee->len) {
2393 		en_dbg(DRV, priv,
2394 		       "mlx4_get_module_info i(%d) offset(%d) len(%d)\n",
2395 		       i, offset, ee->len - i);
2396 
2397 		ret = mlx4_get_module_info(mdev->dev, priv->port,
2398 					   offset, ee->len - i, data + i);
2399 
2400 		if (!ret) /* Done reading */
2401 			return 0;
2402 
2403 		if (ret < 0) {
2404 			en_err(priv,
2405 			       "mlx4_get_module_info i(%d) offset(%d) bytes_to_read(%d) - FAILED (0x%x)\n",
2406 			       i, offset, ee->len - i, ret);
2407 			return -1;
2408 		}
2409 
2410 		i += ret;
2411 		offset += ret;
2412 	}
2413 	return 0;
2414 }
2415 
2416 static void mlx4_en_print_eeprom(u8 *data, __u32 len)
2417 {
2418 	int		i;
2419 	int		j = 0;
2420 	int		row = 0;
2421 	const int	NUM_OF_BYTES = 16;
2422 
2423 	printf("\nOffset\t\tValues\n");
2424 	printf("------\t\t------\n");
2425 	while(row < len){
2426 		printf("0x%04x\t\t",row);
2427 		for(i=0; i < NUM_OF_BYTES; i++){
2428 			printf("%02x ", data[j]);
2429 			row++;
2430 			j++;
2431 		}
2432 		printf("\n");
2433 	}
2434 }
2435 
2436 /* Read cable EEPROM module information by first inspecting the first
2437  * two bytes to get the length and then read the rest of the information.
2438  * The information is printed to dmesg. */
2439 static int mlx4_en_read_eeprom(SYSCTL_HANDLER_ARGS)
2440 {
2441 
2442 	u8*		data;
2443 	int		error;
2444 	int		result = 0;
2445 	struct		mlx4_en_priv *priv;
2446 	struct		net_device *dev;
2447 	struct		ethtool_modinfo modinfo;
2448 	struct		ethtool_eeprom ee;
2449 
2450 	error = sysctl_handle_int(oidp, &result, 0, req);
2451 	if (error || !req->newptr)
2452 		return (error);
2453 
2454 	if (result == 1) {
2455 		priv = arg1;
2456 		dev = priv->dev;
2457 		data = kmalloc(PAGE_SIZE, GFP_KERNEL);
2458 
2459 		error = mlx4_en_get_module_info(dev, &modinfo);
2460 		if (error) {
2461 			en_err(priv,
2462 			       "mlx4_en_get_module_info returned with error - FAILED (0x%x)\n",
2463 			       -error);
2464 			goto out;
2465 		}
2466 
2467 		ee.len = modinfo.eeprom_len;
2468 		ee.offset = 0;
2469 
2470 		error = mlx4_en_get_module_eeprom(dev, &ee, data);
2471 		if (error) {
2472 			en_err(priv,
2473 			       "mlx4_en_get_module_eeprom returned with error - FAILED (0x%x)\n",
2474 			       -error);
2475 			/* Continue printing partial information in case of an error */
2476 		}
2477 
2478 		/* EEPROM information will be printed in dmesg */
2479 		mlx4_en_print_eeprom(data, ee.len);
2480 out:
2481 		kfree(data);
2482 	}
2483 	/* Return zero to prevent sysctl failure. */
2484 	return (0);
2485 }
2486 
2487 static int mlx4_en_set_tx_ppp(SYSCTL_HANDLER_ARGS)
2488 {
2489         struct mlx4_en_priv *priv;
2490         int ppp;
2491         int error;
2492 
2493         priv = arg1;
2494         ppp = priv->prof->tx_ppp;
2495         error = sysctl_handle_int(oidp, &ppp, 0, req);
2496         if (error || !req->newptr)
2497                 return (error);
2498         if (ppp > 0xff || ppp < 0)
2499                 return (-EINVAL);
2500         priv->prof->tx_ppp = ppp;
2501         error = -mlx4_SET_PORT_general(priv->mdev->dev, priv->port,
2502                                        priv->rx_mb_size + ETHER_CRC_LEN,
2503                                        priv->prof->tx_pause,
2504                                        priv->prof->tx_ppp,
2505                                        priv->prof->rx_pause,
2506                                        priv->prof->rx_ppp);
2507 
2508         return (error);
2509 }
2510 
2511 static int mlx4_en_set_rx_ppp(SYSCTL_HANDLER_ARGS)
2512 {
2513         struct mlx4_en_priv *priv;
2514         struct mlx4_en_dev *mdev;
2515         int ppp;
2516         int error;
2517         int port_up;
2518 
2519         port_up = 0;
2520         priv = arg1;
2521         mdev = priv->mdev;
2522         ppp = priv->prof->rx_ppp;
2523         error = sysctl_handle_int(oidp, &ppp, 0, req);
2524         if (error || !req->newptr)
2525                 return (error);
2526         if (ppp > 0xff || ppp < 0)
2527                 return (-EINVAL);
2528         /* See if we have to change the number of tx queues. */
2529         if (!ppp != !priv->prof->rx_ppp) {
2530                 mutex_lock(&mdev->state_lock);
2531                 if (priv->port_up) {
2532                         port_up = 1;
2533                         mlx4_en_stop_port(priv->dev);
2534                 }
2535                 mlx4_en_free_resources(priv);
2536                 priv->prof->rx_ppp = ppp;
2537                 error = -mlx4_en_alloc_resources(priv);
2538                 if (error)
2539                         en_err(priv, "Failed reallocating port resources\n");
2540                 if (error == 0 && port_up) {
2541                         error = -mlx4_en_start_port(priv->dev);
2542                         if (error)
2543                                 en_err(priv, "Failed starting port\n");
2544                 }
2545                 mutex_unlock(&mdev->state_lock);
2546                 return (error);
2547 
2548         }
2549         priv->prof->rx_ppp = ppp;
2550         error = -mlx4_SET_PORT_general(priv->mdev->dev, priv->port,
2551                                        priv->rx_mb_size + ETHER_CRC_LEN,
2552                                        priv->prof->tx_pause,
2553                                        priv->prof->tx_ppp,
2554                                        priv->prof->rx_pause,
2555                                        priv->prof->rx_ppp);
2556 
2557         return (error);
2558 }
2559 
2560 static void mlx4_en_sysctl_conf(struct mlx4_en_priv *priv)
2561 {
2562         struct net_device *dev;
2563         struct sysctl_ctx_list *ctx;
2564         struct sysctl_oid *node;
2565         struct sysctl_oid_list *node_list;
2566         struct sysctl_oid *coal;
2567         struct sysctl_oid_list *coal_list;
2568 	const char *pnameunit;
2569 
2570         dev = priv->dev;
2571         ctx = &priv->conf_ctx;
2572 	pnameunit = device_get_nameunit(priv->mdev->pdev->dev.bsddev);
2573 
2574         sysctl_ctx_init(ctx);
2575         priv->conf_sysctl = SYSCTL_ADD_NODE(ctx, SYSCTL_STATIC_CHILDREN(_hw),
2576             OID_AUTO, dev->if_xname, CTLFLAG_RD, 0, "mlx4 10gig ethernet");
2577         node = SYSCTL_ADD_NODE(ctx, SYSCTL_CHILDREN(priv->conf_sysctl), OID_AUTO,
2578             "conf", CTLFLAG_RD, NULL, "Configuration");
2579         node_list = SYSCTL_CHILDREN(node);
2580 
2581         SYSCTL_ADD_UINT(ctx, node_list, OID_AUTO, "msg_enable",
2582             CTLFLAG_RW, &priv->msg_enable, 0,
2583             "Driver message enable bitfield");
2584         SYSCTL_ADD_UINT(ctx, node_list, OID_AUTO, "rx_rings",
2585             CTLFLAG_RD, &priv->rx_ring_num, 0,
2586             "Number of receive rings");
2587         SYSCTL_ADD_UINT(ctx, node_list, OID_AUTO, "tx_rings",
2588             CTLFLAG_RD, &priv->tx_ring_num, 0,
2589             "Number of transmit rings");
2590         SYSCTL_ADD_PROC(ctx, node_list, OID_AUTO, "rx_size",
2591             CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, priv, 0,
2592             mlx4_en_set_rx_ring_size, "I", "Receive ring size");
2593         SYSCTL_ADD_PROC(ctx, node_list, OID_AUTO, "tx_size",
2594             CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, priv, 0,
2595             mlx4_en_set_tx_ring_size, "I", "Transmit ring size");
2596         SYSCTL_ADD_PROC(ctx, node_list, OID_AUTO, "tx_ppp",
2597             CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, priv, 0,
2598             mlx4_en_set_tx_ppp, "I", "TX Per-priority pause");
2599         SYSCTL_ADD_PROC(ctx, node_list, OID_AUTO, "rx_ppp",
2600             CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, priv, 0,
2601             mlx4_en_set_rx_ppp, "I", "RX Per-priority pause");
2602         SYSCTL_ADD_UINT(ctx, node_list, OID_AUTO, "port_num",
2603             CTLFLAG_RD, &priv->port, 0,
2604             "Port Number");
2605         SYSCTL_ADD_STRING(ctx, node_list, OID_AUTO, "device_name",
2606 	    CTLFLAG_RD, __DECONST(void *, pnameunit), 0,
2607 	    "PCI device name");
2608 
2609         /* Add coalescer configuration. */
2610         coal = SYSCTL_ADD_NODE(ctx, node_list, OID_AUTO,
2611             "coalesce", CTLFLAG_RD, NULL, "Interrupt coalesce configuration");
2612         coal_list = SYSCTL_CHILDREN(coal);
2613         SYSCTL_ADD_UINT(ctx, coal_list, OID_AUTO, "pkt_rate_low",
2614             CTLFLAG_RW, &priv->pkt_rate_low, 0,
2615             "Packets per-second for minimum delay");
2616         SYSCTL_ADD_UINT(ctx, coal_list, OID_AUTO, "rx_usecs_low",
2617             CTLFLAG_RW, &priv->rx_usecs_low, 0,
2618             "Minimum RX delay in micro-seconds");
2619         SYSCTL_ADD_UINT(ctx, coal_list, OID_AUTO, "pkt_rate_high",
2620             CTLFLAG_RW, &priv->pkt_rate_high, 0,
2621             "Packets per-second for maximum delay");
2622         SYSCTL_ADD_UINT(ctx, coal_list, OID_AUTO, "rx_usecs_high",
2623             CTLFLAG_RW, &priv->rx_usecs_high, 0,
2624             "Maximum RX delay in micro-seconds");
2625         SYSCTL_ADD_UINT(ctx, coal_list, OID_AUTO, "sample_interval",
2626             CTLFLAG_RW, &priv->sample_interval, 0,
2627             "adaptive frequency in units of HZ ticks");
2628         SYSCTL_ADD_UINT(ctx, coal_list, OID_AUTO, "adaptive_rx_coal",
2629             CTLFLAG_RW, &priv->adaptive_rx_coal, 0,
2630             "Enable adaptive rx coalescing");
2631 	/* EEPROM support */
2632 	SYSCTL_ADD_PROC(ctx, node_list, OID_AUTO, "eeprom_info",
2633 	    CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, priv, 0,
2634 	    mlx4_en_read_eeprom, "I", "EEPROM information");
2635 }
2636 
2637 static void mlx4_en_sysctl_stat(struct mlx4_en_priv *priv)
2638 {
2639 	struct sysctl_ctx_list *ctx;
2640 	struct sysctl_oid_list *node_list;
2641 	struct sysctl_oid *ring_node;
2642 	struct sysctl_oid_list *ring_list;
2643 	struct mlx4_en_tx_ring *tx_ring;
2644 	struct mlx4_en_rx_ring *rx_ring;
2645 	char namebuf[128];
2646 	int i;
2647 
2648 	ctx = &priv->stat_ctx;
2649 	sysctl_ctx_init(ctx);
2650 	priv->stat_sysctl = SYSCTL_ADD_NODE(ctx, SYSCTL_CHILDREN(priv->conf_sysctl), OID_AUTO,
2651 	    "stat", CTLFLAG_RD, NULL, "Statistics");
2652 	node_list = SYSCTL_CHILDREN(priv->stat_sysctl);
2653 
2654 #ifdef MLX4_EN_PERF_STAT
2655 	SYSCTL_ADD_UINT(ctx, node_list, OID_AUTO, "tx_poll", CTLFLAG_RD,
2656 	    &priv->pstats.tx_poll, "TX Poll calls");
2657 	SYSCTL_ADD_QUAD(ctx, node_list, OID_AUTO, "tx_pktsz_avg", CTLFLAG_RD,
2658 	    &priv->pstats.tx_pktsz_avg, "TX average packet size");
2659 	SYSCTL_ADD_UINT(ctx, node_list, OID_AUTO, "inflight_avg", CTLFLAG_RD,
2660 	    &priv->pstats.inflight_avg, "TX average packets in-flight");
2661 	SYSCTL_ADD_UINT(ctx, node_list, OID_AUTO, "tx_coal_avg", CTLFLAG_RD,
2662 	    &priv->pstats.tx_coal_avg, "TX average coalesced completions");
2663 	SYSCTL_ADD_UINT(ctx, node_list, OID_AUTO, "rx_coal_avg", CTLFLAG_RD,
2664 	    &priv->pstats.rx_coal_avg, "RX average coalesced completions");
2665 #endif
2666 
2667 	SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "tso_packets", CTLFLAG_RD,
2668 	    &priv->port_stats.tso_packets, "TSO packets sent");
2669 	SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "queue_stopped", CTLFLAG_RD,
2670 	    &priv->port_stats.queue_stopped, "Queue full");
2671 	SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "wake_queue", CTLFLAG_RD,
2672 	    &priv->port_stats.wake_queue, "Queue resumed after full");
2673 	SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "tx_timeout", CTLFLAG_RD,
2674 	    &priv->port_stats.tx_timeout, "Transmit timeouts");
2675 	SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "tx_oversized_packets", CTLFLAG_RD,
2676 	    &priv->port_stats.oversized_packets, "TX oversized packets, m_defrag failed");
2677 	SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "rx_alloc_failed", CTLFLAG_RD,
2678 	    &priv->port_stats.rx_alloc_failed, "RX failed to allocate mbuf");
2679 	SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "rx_chksum_good", CTLFLAG_RD,
2680 	    &priv->port_stats.rx_chksum_good, "RX checksum offload success");
2681 	SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "rx_chksum_none", CTLFLAG_RD,
2682 	    &priv->port_stats.rx_chksum_none, "RX without checksum offload");
2683 	SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "tx_chksum_offload",
2684 	    CTLFLAG_RD, &priv->port_stats.tx_chksum_offload,
2685 	    "TX checksum offloads");
2686 
2687 	/* Could strdup the names and add in a loop.  This is simpler. */
2688 	SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "rx_bytes", CTLFLAG_RD,
2689 	    &priv->pkstats.rx_bytes, "RX Bytes");
2690 	SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "rx_packets", CTLFLAG_RD,
2691 	    &priv->pkstats.rx_packets, "RX packets");
2692 	SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "rx_multicast_packets", CTLFLAG_RD,
2693 	    &priv->pkstats.rx_multicast_packets, "RX Multicast Packets");
2694 	SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "rx_broadcast_packets", CTLFLAG_RD,
2695 	    &priv->pkstats.rx_broadcast_packets, "RX Broadcast Packets");
2696 	SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "rx_errors", CTLFLAG_RD,
2697 	    &priv->pkstats.rx_errors, "RX Errors");
2698 	SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "rx_dropped", CTLFLAG_RD,
2699 	    &priv->pkstats.rx_dropped, "RX Dropped");
2700 	SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "rx_length_errors", CTLFLAG_RD,
2701 	    &priv->pkstats.rx_length_errors, "RX Length Errors");
2702 	SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "rx_over_errors", CTLFLAG_RD,
2703 	    &priv->pkstats.rx_over_errors, "RX Over Errors");
2704 	SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "rx_crc_errors", CTLFLAG_RD,
2705 	    &priv->pkstats.rx_crc_errors, "RX CRC Errors");
2706 	SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "rx_jabbers", CTLFLAG_RD,
2707 	    &priv->pkstats.rx_jabbers, "RX Jabbers");
2708 
2709 
2710 	SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "rx_in_range_length_error", CTLFLAG_RD,
2711 	    &priv->pkstats.rx_in_range_length_error, "RX IN_Range Length Error");
2712 	SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "rx_out_range_length_error",
2713 		CTLFLAG_RD, &priv->pkstats.rx_out_range_length_error,
2714 		"RX Out Range Length Error");
2715 	SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "rx_lt_64_bytes_packets", CTLFLAG_RD,
2716 	    &priv->pkstats.rx_lt_64_bytes_packets, "RX Lt 64 Bytes Packets");
2717 	SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "rx_127_bytes_packets", CTLFLAG_RD,
2718 	    &priv->pkstats.rx_127_bytes_packets, "RX 127 bytes Packets");
2719 	SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "rx_255_bytes_packets", CTLFLAG_RD,
2720 	    &priv->pkstats.rx_255_bytes_packets, "RX 255 bytes Packets");
2721 	SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "rx_511_bytes_packets", CTLFLAG_RD,
2722 	    &priv->pkstats.rx_511_bytes_packets, "RX 511 bytes Packets");
2723 	SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "rx_1023_bytes_packets", CTLFLAG_RD,
2724 	    &priv->pkstats.rx_1023_bytes_packets, "RX 1023 bytes Packets");
2725 	SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "rx_1518_bytes_packets", CTLFLAG_RD,
2726 	    &priv->pkstats.rx_1518_bytes_packets, "RX 1518 bytes Packets");
2727 	SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "rx_1522_bytes_packets", CTLFLAG_RD,
2728 	    &priv->pkstats.rx_1522_bytes_packets, "RX 1522 bytes Packets");
2729 	SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "rx_1548_bytes_packets", CTLFLAG_RD,
2730 	    &priv->pkstats.rx_1548_bytes_packets, "RX 1548 bytes Packets");
2731 	SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "rx_gt_1548_bytes_packets", CTLFLAG_RD,
2732 	    &priv->pkstats.rx_gt_1548_bytes_packets,
2733 	    "RX Greater Then 1548 bytes Packets");
2734 
2735 struct mlx4_en_pkt_stats {
2736 	unsigned long tx_packets;
2737 	unsigned long tx_bytes;
2738 	unsigned long tx_multicast_packets;
2739 	unsigned long tx_broadcast_packets;
2740 	unsigned long tx_errors;
2741 	unsigned long tx_dropped;
2742 	unsigned long tx_lt_64_bytes_packets;
2743 	unsigned long tx_127_bytes_packets;
2744 	unsigned long tx_255_bytes_packets;
2745 	unsigned long tx_511_bytes_packets;
2746 	unsigned long tx_1023_bytes_packets;
2747 	unsigned long tx_1518_bytes_packets;
2748 	unsigned long tx_1522_bytes_packets;
2749 	unsigned long tx_1548_bytes_packets;
2750 	unsigned long tx_gt_1548_bytes_packets;
2751 	unsigned long rx_prio[NUM_PRIORITIES][NUM_PRIORITY_STATS];
2752 	unsigned long tx_prio[NUM_PRIORITIES][NUM_PRIORITY_STATS];
2753 #define NUM_PKT_STATS		72
2754 };
2755 
2756 
2757 	SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "tx_packets", CTLFLAG_RD,
2758 	    &priv->pkstats.tx_packets, "TX packets");
2759 	SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "tx_bytes", CTLFLAG_RD,
2760 	    &priv->pkstats.tx_bytes, "TX Bytes");
2761 	SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "tx_multicast_packets", CTLFLAG_RD,
2762 	    &priv->pkstats.tx_multicast_packets, "TX Multicast Packets");
2763 	SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "tx_broadcast_packets", CTLFLAG_RD,
2764 	    &priv->pkstats.tx_broadcast_packets, "TX Broadcast Packets");
2765 	SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "tx_errors", CTLFLAG_RD,
2766 	    &priv->pkstats.tx_errors, "TX Errors");
2767 	SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "tx_dropped", CTLFLAG_RD,
2768 	    &priv->pkstats.tx_dropped, "TX Dropped");
2769 	SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "tx_lt_64_bytes_packets", CTLFLAG_RD,
2770 	    &priv->pkstats.tx_lt_64_bytes_packets, "TX Less Then 64 Bytes Packets");
2771 	SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "tx_127_bytes_packets", CTLFLAG_RD,
2772 	    &priv->pkstats.tx_127_bytes_packets, "TX 127 Bytes Packets");
2773 	SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "tx_255_bytes_packets", CTLFLAG_RD,
2774 	    &priv->pkstats.tx_255_bytes_packets, "TX 255 Bytes Packets");
2775 	SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "tx_511_bytes_packets", CTLFLAG_RD,
2776 	    &priv->pkstats.tx_511_bytes_packets, "TX 511 Bytes Packets");
2777 	SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "tx_1023_bytes_packets", CTLFLAG_RD,
2778 	    &priv->pkstats.tx_1023_bytes_packets, "TX 1023 Bytes Packets");
2779 	SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "tx_1518_bytes_packets", CTLFLAG_RD,
2780 	    &priv->pkstats.tx_1518_bytes_packets, "TX 1518 Bytes Packets");
2781 	SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "tx_1522_bytes_packets", CTLFLAG_RD,
2782 	    &priv->pkstats.tx_1522_bytes_packets, "TX 1522 Bytes Packets");
2783 	SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "tx_1548_bytes_packets", CTLFLAG_RD,
2784 	    &priv->pkstats.tx_1548_bytes_packets, "TX 1548 Bytes Packets");
2785 	SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "tx_gt_1548_bytes_packets", CTLFLAG_RD,
2786 	    &priv->pkstats.tx_gt_1548_bytes_packets,
2787 	    "TX Greater Then 1548 Bytes Packets");
2788 
2789 
2790 
2791 	for (i = 0; i < priv->tx_ring_num; i++) {
2792 		tx_ring = priv->tx_ring[i];
2793 		snprintf(namebuf, sizeof(namebuf), "tx_ring%d", i);
2794 		ring_node = SYSCTL_ADD_NODE(ctx, node_list, OID_AUTO, namebuf,
2795 		    CTLFLAG_RD, NULL, "TX Ring");
2796 		ring_list = SYSCTL_CHILDREN(ring_node);
2797 		SYSCTL_ADD_ULONG(ctx, ring_list, OID_AUTO, "packets",
2798 		    CTLFLAG_RD, &tx_ring->packets, "TX packets");
2799 		SYSCTL_ADD_ULONG(ctx, ring_list, OID_AUTO, "bytes",
2800 		    CTLFLAG_RD, &tx_ring->bytes, "TX bytes");
2801 	}
2802 
2803 	for (i = 0; i < priv->rx_ring_num; i++) {
2804 		rx_ring = priv->rx_ring[i];
2805 		snprintf(namebuf, sizeof(namebuf), "rx_ring%d", i);
2806 		ring_node = SYSCTL_ADD_NODE(ctx, node_list, OID_AUTO, namebuf,
2807 		    CTLFLAG_RD, NULL, "RX Ring");
2808 		ring_list = SYSCTL_CHILDREN(ring_node);
2809 		SYSCTL_ADD_ULONG(ctx, ring_list, OID_AUTO, "packets",
2810 		    CTLFLAG_RD, &rx_ring->packets, "RX packets");
2811 		SYSCTL_ADD_ULONG(ctx, ring_list, OID_AUTO, "bytes",
2812 		    CTLFLAG_RD, &rx_ring->bytes, "RX bytes");
2813 		SYSCTL_ADD_ULONG(ctx, ring_list, OID_AUTO, "error",
2814 		    CTLFLAG_RD, &rx_ring->errors, "RX soft errors");
2815 	}
2816 }
2817