xref: /freebsd/sys/dev/mlx4/mlx4_en/mlx4_en_netdev.c (revision 0629b152762b06325dd75a41bcb0a2789514141b)
1 /*
2  * Copyright (c) 2007, 2014 Mellanox Technologies. All rights reserved.
3  *
4  * This software is available to you under a choice of one of two
5  * licenses.  You may choose to be licensed under the terms of the GNU
6  * General Public License (GPL) Version 2, available from the file
7  * COPYING in the main directory of this source tree, or the
8  * OpenIB.org BSD license below:
9  *
10  *     Redistribution and use in source and binary forms, with or
11  *     without modification, are permitted provided that the following
12  *     conditions are met:
13  *
14  *      - Redistributions of source code must retain the above
15  *        copyright notice, this list of conditions and the following
16  *        disclaimer.
17  *
18  *      - Redistributions in binary form must reproduce the above
19  *        copyright notice, this list of conditions and the following
20  *        disclaimer in the documentation and/or other materials
21  *        provided with the distribution.
22  *
23  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
24  * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
25  * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
26  * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
27  * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
28  * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
29  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
30  * SOFTWARE.
31  *
32  */
33 
34 #include <linux/etherdevice.h>
35 #include <linux/delay.h>
36 #include <linux/slab.h>
37 #include <linux/compat.h>
38 #ifdef CONFIG_NET_RX_BUSY_POLL
39 #include <net/busy_poll.h>
40 #endif
41 
42 #include <linux/list.h>
43 #include <linux/if_ether.h>
44 
45 #include <dev/mlx4/driver.h>
46 #include <dev/mlx4/device.h>
47 #include <dev/mlx4/cmd.h>
48 #include <dev/mlx4/cq.h>
49 
50 #include <sys/sockio.h>
51 #include <sys/sysctl.h>
52 
53 #include "en.h"
54 #include "en_port.h"
55 
56 static void mlx4_en_sysctl_stat(struct mlx4_en_priv *priv);
57 static void mlx4_en_sysctl_conf(struct mlx4_en_priv *priv);
58 
59 #ifdef CONFIG_NET_RX_BUSY_POLL
60 /* must be called with local_bh_disable()d */
61 static int mlx4_en_low_latency_recv(struct napi_struct *napi)
62 {
63 	struct mlx4_en_cq *cq = container_of(napi, struct mlx4_en_cq, napi);
64 	struct net_device *dev = cq->dev;
65 	struct mlx4_en_priv *priv = netdev_priv(dev);
66 	struct mlx4_en_rx_ring *rx_ring = priv->rx_ring[cq->ring];
67 	int done;
68 
69 	if (!priv->port_up)
70 		return LL_FLUSH_FAILED;
71 
72 	if (!mlx4_en_cq_lock_poll(cq))
73 		return LL_FLUSH_BUSY;
74 
75 	done = mlx4_en_process_rx_cq(dev, cq, 4);
76 #ifdef LL_EXTENDED_STATS
77 	if (likely(done))
78 		rx_ring->cleaned += done;
79 	else
80 		rx_ring->misses++;
81 #endif
82 
83 	mlx4_en_cq_unlock_poll(cq);
84 
85 	return done;
86 }
87 #endif	/* CONFIG_NET_RX_BUSY_POLL */
88 
89 #ifdef CONFIG_RFS_ACCEL
90 
91 struct mlx4_en_filter {
92 	struct list_head next;
93 	struct work_struct work;
94 
95 	u8     ip_proto;
96 	__be32 src_ip;
97 	__be32 dst_ip;
98 	__be16 src_port;
99 	__be16 dst_port;
100 
101 	int rxq_index;
102 	struct mlx4_en_priv *priv;
103 	u32 flow_id;			/* RFS infrastructure id */
104 	int id;				/* mlx4_en driver id */
105 	u64 reg_id;			/* Flow steering API id */
106 	u8 activated;			/* Used to prevent expiry before filter
107 					 * is attached
108 					 */
109 	struct hlist_node filter_chain;
110 };
111 
112 static void mlx4_en_filter_rfs_expire(struct mlx4_en_priv *priv);
113 
114 static enum mlx4_net_trans_rule_id mlx4_ip_proto_to_trans_rule_id(u8 ip_proto)
115 {
116 	switch (ip_proto) {
117 	case IPPROTO_UDP:
118 		return MLX4_NET_TRANS_RULE_ID_UDP;
119 	case IPPROTO_TCP:
120 		return MLX4_NET_TRANS_RULE_ID_TCP;
121 	default:
122 		return MLX4_NET_TRANS_RULE_NUM;
123 	}
124 };
125 
126 static void mlx4_en_filter_work(struct work_struct *work)
127 {
128 	struct mlx4_en_filter *filter = container_of(work,
129 						     struct mlx4_en_filter,
130 						     work);
131 	struct mlx4_en_priv *priv = filter->priv;
132 	struct mlx4_spec_list spec_tcp_udp = {
133 		.id = mlx4_ip_proto_to_trans_rule_id(filter->ip_proto),
134 		{
135 			.tcp_udp = {
136 				.dst_port = filter->dst_port,
137 				.dst_port_msk = (__force __be16)-1,
138 				.src_port = filter->src_port,
139 				.src_port_msk = (__force __be16)-1,
140 			},
141 		},
142 	};
143 	struct mlx4_spec_list spec_ip = {
144 		.id = MLX4_NET_TRANS_RULE_ID_IPV4,
145 		{
146 			.ipv4 = {
147 				.dst_ip = filter->dst_ip,
148 				.dst_ip_msk = (__force __be32)-1,
149 				.src_ip = filter->src_ip,
150 				.src_ip_msk = (__force __be32)-1,
151 			},
152 		},
153 	};
154 	struct mlx4_spec_list spec_eth = {
155 		.id = MLX4_NET_TRANS_RULE_ID_ETH,
156 	};
157 	struct mlx4_net_trans_rule rule = {
158 		.list = LIST_HEAD_INIT(rule.list),
159 		.queue_mode = MLX4_NET_TRANS_Q_LIFO,
160 		.exclusive = 1,
161 		.allow_loopback = 1,
162 		.promisc_mode = MLX4_FS_REGULAR,
163 		.port = priv->port,
164 		.priority = MLX4_DOMAIN_RFS,
165 	};
166 	int rc;
167 	__be64 mac_mask = cpu_to_be64(MLX4_MAC_MASK << 16);
168 
169 	if (spec_tcp_udp.id >= MLX4_NET_TRANS_RULE_NUM) {
170 		en_warn(priv, "RFS: ignoring unsupported ip protocol (%d)\n",
171 			filter->ip_proto);
172 		goto ignore;
173 	}
174 	list_add_tail(&spec_eth.list, &rule.list);
175 	list_add_tail(&spec_ip.list, &rule.list);
176 	list_add_tail(&spec_tcp_udp.list, &rule.list);
177 
178 	rule.qpn = priv->rss_map.qps[filter->rxq_index].qpn;
179 	memcpy(spec_eth.eth.dst_mac, priv->dev->dev_addr, ETH_ALEN);
180 	memcpy(spec_eth.eth.dst_mac_msk, &mac_mask, ETH_ALEN);
181 
182 	filter->activated = 0;
183 
184 	if (filter->reg_id) {
185 		rc = mlx4_flow_detach(priv->mdev->dev, filter->reg_id);
186 		if (rc && rc != -ENOENT)
187 			en_err(priv, "Error detaching flow. rc = %d\n", rc);
188 	}
189 
190 	rc = mlx4_flow_attach(priv->mdev->dev, &rule, &filter->reg_id);
191 	if (rc)
192 		en_err(priv, "Error attaching flow. err = %d\n", rc);
193 
194 ignore:
195 	mlx4_en_filter_rfs_expire(priv);
196 
197 	filter->activated = 1;
198 }
199 
200 static inline struct hlist_head *
201 filter_hash_bucket(struct mlx4_en_priv *priv, __be32 src_ip, __be32 dst_ip,
202 		   __be16 src_port, __be16 dst_port)
203 {
204 	unsigned long l;
205 	int bucket_idx;
206 
207 	l = (__force unsigned long)src_port |
208 	    ((__force unsigned long)dst_port << 2);
209 	l ^= (__force unsigned long)(src_ip ^ dst_ip);
210 
211 	bucket_idx = hash_long(l, MLX4_EN_FILTER_HASH_SHIFT);
212 
213 	return &priv->filter_hash[bucket_idx];
214 }
215 
216 static struct mlx4_en_filter *
217 mlx4_en_filter_alloc(struct mlx4_en_priv *priv, int rxq_index, __be32 src_ip,
218 		     __be32 dst_ip, u8 ip_proto, __be16 src_port,
219 		     __be16 dst_port, u32 flow_id)
220 {
221 	struct mlx4_en_filter *filter = NULL;
222 
223 	filter = kzalloc(sizeof(struct mlx4_en_filter), GFP_ATOMIC);
224 	if (!filter)
225 		return NULL;
226 
227 	filter->priv = priv;
228 	filter->rxq_index = rxq_index;
229 	INIT_WORK(&filter->work, mlx4_en_filter_work);
230 
231 	filter->src_ip = src_ip;
232 	filter->dst_ip = dst_ip;
233 	filter->ip_proto = ip_proto;
234 	filter->src_port = src_port;
235 	filter->dst_port = dst_port;
236 
237 	filter->flow_id = flow_id;
238 
239 	filter->id = priv->last_filter_id++ % RPS_NO_FILTER;
240 
241 	list_add_tail(&filter->next, &priv->filters);
242 	hlist_add_head(&filter->filter_chain,
243 		       filter_hash_bucket(priv, src_ip, dst_ip, src_port,
244 					  dst_port));
245 
246 	return filter;
247 }
248 
249 static void mlx4_en_filter_free(struct mlx4_en_filter *filter)
250 {
251 	struct mlx4_en_priv *priv = filter->priv;
252 	int rc;
253 
254 	list_del(&filter->next);
255 
256 	rc = mlx4_flow_detach(priv->mdev->dev, filter->reg_id);
257 	if (rc && rc != -ENOENT)
258 		en_err(priv, "Error detaching flow. rc = %d\n", rc);
259 
260 	kfree(filter);
261 }
262 
263 static inline struct mlx4_en_filter *
264 mlx4_en_filter_find(struct mlx4_en_priv *priv, __be32 src_ip, __be32 dst_ip,
265 		    u8 ip_proto, __be16 src_port, __be16 dst_port)
266 {
267 	struct mlx4_en_filter *filter;
268 	struct mlx4_en_filter *ret = NULL;
269 
270 	hlist_for_each_entry(filter,
271 			     filter_hash_bucket(priv, src_ip, dst_ip,
272 						src_port, dst_port),
273 			     filter_chain) {
274 		if (filter->src_ip == src_ip &&
275 		    filter->dst_ip == dst_ip &&
276 		    filter->ip_proto == ip_proto &&
277 		    filter->src_port == src_port &&
278 		    filter->dst_port == dst_port) {
279 			ret = filter;
280 			break;
281 		}
282 	}
283 
284 	return ret;
285 }
286 
287 static int
288 mlx4_en_filter_rfs(struct net_device *net_dev, const struct sk_buff *skb,
289 		   u16 rxq_index, u32 flow_id)
290 {
291 	struct mlx4_en_priv *priv = netdev_priv(net_dev);
292 	struct mlx4_en_filter *filter;
293 	const struct iphdr *ip;
294 	const __be16 *ports;
295 	u8 ip_proto;
296 	__be32 src_ip;
297 	__be32 dst_ip;
298 	__be16 src_port;
299 	__be16 dst_port;
300 	int nhoff = skb_network_offset(skb);
301 	int ret = 0;
302 
303 	if (skb->protocol != htons(ETH_P_IP))
304 		return -EPROTONOSUPPORT;
305 
306 	ip = (const struct iphdr *)(skb->data + nhoff);
307 	if (ip_is_fragment(ip))
308 		return -EPROTONOSUPPORT;
309 
310 	if ((ip->protocol != IPPROTO_TCP) && (ip->protocol != IPPROTO_UDP))
311 		return -EPROTONOSUPPORT;
312 	ports = (const __be16 *)(skb->data + nhoff + 4 * ip->ihl);
313 
314 	ip_proto = ip->protocol;
315 	src_ip = ip->saddr;
316 	dst_ip = ip->daddr;
317 	src_port = ports[0];
318 	dst_port = ports[1];
319 
320 	spin_lock_bh(&priv->filters_lock);
321 	filter = mlx4_en_filter_find(priv, src_ip, dst_ip, ip_proto,
322 				     src_port, dst_port);
323 	if (filter) {
324 		if (filter->rxq_index == rxq_index)
325 			goto out;
326 
327 		filter->rxq_index = rxq_index;
328 	} else {
329 		filter = mlx4_en_filter_alloc(priv, rxq_index,
330 					      src_ip, dst_ip, ip_proto,
331 					      src_port, dst_port, flow_id);
332 		if (!filter) {
333 			ret = -ENOMEM;
334 			goto err;
335 		}
336 	}
337 
338 	queue_work(priv->mdev->workqueue, &filter->work);
339 
340 out:
341 	ret = filter->id;
342 err:
343 	spin_unlock_bh(&priv->filters_lock);
344 
345 	return ret;
346 }
347 
348 void mlx4_en_cleanup_filters(struct mlx4_en_priv *priv)
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_tunnel_steer_add(struct mlx4_en_priv *priv, unsigned char *addr,
454 				    int qpn, u64 *reg_id)
455 {
456 	int err;
457 
458 	if (priv->mdev->dev->caps.tunnel_offload_mode != MLX4_TUNNEL_OFFLOAD_MODE_VXLAN ||
459 	    priv->mdev->dev->caps.dmfs_high_steer_mode == MLX4_STEERING_DMFS_A0_STATIC)
460 		return 0; /* do nothing */
461 
462 	err = mlx4_tunnel_steer_add(priv->mdev->dev, addr, priv->port, qpn,
463 				    MLX4_DOMAIN_NIC, reg_id);
464 	if (err) {
465 		en_err(priv, "failed to add vxlan steering rule, err %d\n", err);
466 		return err;
467 	}
468 	en_dbg(DRV, priv, "added vxlan steering rule, mac %pM reg_id %llx\n", addr, (long long)*reg_id);
469 	return 0;
470 }
471 
472 static int mlx4_en_uc_steer_add(struct mlx4_en_priv *priv,
473 				unsigned char *mac, int *qpn, u64 *reg_id)
474 {
475 	struct mlx4_en_dev *mdev = priv->mdev;
476 	struct mlx4_dev *dev = mdev->dev;
477 	int err;
478 
479 	switch (dev->caps.steering_mode) {
480 	case MLX4_STEERING_MODE_B0: {
481 		struct mlx4_qp qp;
482 		u8 gid[16] = {0};
483 
484 		qp.qpn = *qpn;
485 		memcpy(&gid[10], mac, ETH_ALEN);
486 		gid[5] = priv->port;
487 
488 		err = mlx4_unicast_attach(dev, &qp, gid, 0, MLX4_PROT_ETH);
489 		break;
490 	}
491 	case MLX4_STEERING_MODE_DEVICE_MANAGED: {
492 		struct mlx4_spec_list spec_eth = { {NULL} };
493 		__be64 mac_mask = cpu_to_be64(MLX4_MAC_MASK << 16);
494 
495 		struct mlx4_net_trans_rule rule = {
496 			.queue_mode = MLX4_NET_TRANS_Q_FIFO,
497 			.exclusive = 0,
498 			.allow_loopback = 1,
499 			.promisc_mode = MLX4_FS_REGULAR,
500 			.priority = MLX4_DOMAIN_NIC,
501 		};
502 
503 		rule.port = priv->port;
504 		rule.qpn = *qpn;
505 		INIT_LIST_HEAD(&rule.list);
506 
507 		spec_eth.id = MLX4_NET_TRANS_RULE_ID_ETH;
508 		memcpy(spec_eth.eth.dst_mac, mac, ETH_ALEN);
509 		memcpy(spec_eth.eth.dst_mac_msk, &mac_mask, ETH_ALEN);
510 		list_add_tail(&spec_eth.list, &rule.list);
511 
512 		err = mlx4_flow_attach(dev, &rule, reg_id);
513 		break;
514 	}
515 	default:
516 		return -EINVAL;
517 	}
518 	if (err)
519 		en_warn(priv, "Failed Attaching Unicast\n");
520 
521 	return err;
522 }
523 
524 static void mlx4_en_uc_steer_release(struct mlx4_en_priv *priv,
525 				     unsigned char *mac, int qpn, u64 reg_id)
526 {
527 	struct mlx4_en_dev *mdev = priv->mdev;
528 	struct mlx4_dev *dev = mdev->dev;
529 
530 	switch (dev->caps.steering_mode) {
531 	case MLX4_STEERING_MODE_B0: {
532 		struct mlx4_qp qp;
533 		u8 gid[16] = {0};
534 
535 		qp.qpn = qpn;
536 		memcpy(&gid[10], mac, ETH_ALEN);
537 		gid[5] = priv->port;
538 
539 		mlx4_unicast_detach(dev, &qp, gid, MLX4_PROT_ETH);
540 		break;
541 	}
542 	case MLX4_STEERING_MODE_DEVICE_MANAGED: {
543 		mlx4_flow_detach(dev, reg_id);
544 		break;
545 	}
546 	default:
547 		en_err(priv, "Invalid steering mode.\n");
548 	}
549 }
550 
551 static int mlx4_en_get_qp(struct mlx4_en_priv *priv)
552 {
553 	struct mlx4_en_dev *mdev = priv->mdev;
554 	struct mlx4_dev *dev = mdev->dev;
555 	int index = 0;
556 	int err = 0;
557 	int *qpn = &priv->base_qpn;
558 	u64 mac = mlx4_mac_to_u64(IF_LLADDR(priv->dev));
559 
560 	en_dbg(DRV, priv, "Registering MAC: %pM for adding\n",
561 	       IF_LLADDR(priv->dev));
562 	index = mlx4_register_mac(dev, priv->port, mac);
563 	if (index < 0) {
564 		err = index;
565 		en_err(priv, "Failed adding MAC: %pM\n",
566 		       IF_LLADDR(priv->dev));
567 		return err;
568 	}
569 
570 	if (dev->caps.steering_mode == MLX4_STEERING_MODE_A0) {
571 		int base_qpn = mlx4_get_base_qpn(dev, priv->port);
572 		*qpn = base_qpn + index;
573 		return 0;
574 	}
575 
576 	err = mlx4_qp_reserve_range(dev, 1, 1, qpn, MLX4_RESERVE_A0_QP);
577 	en_dbg(DRV, priv, "Reserved qp %d\n", *qpn);
578 	if (err) {
579 		en_err(priv, "Failed to reserve qp for mac registration\n");
580 		mlx4_unregister_mac(dev, priv->port, mac);
581 		return err;
582 	}
583 
584 	return 0;
585 }
586 
587 static void mlx4_en_put_qp(struct mlx4_en_priv *priv)
588 {
589 	struct mlx4_en_dev *mdev = priv->mdev;
590 	struct mlx4_dev *dev = mdev->dev;
591 	int qpn = priv->base_qpn;
592 
593 	if (dev->caps.steering_mode == MLX4_STEERING_MODE_A0) {
594 		u64 mac = mlx4_mac_to_u64(IF_LLADDR(priv->dev));
595 		en_dbg(DRV, priv, "Registering MAC: %pM for deleting\n",
596 		       IF_LLADDR(priv->dev));
597 		mlx4_unregister_mac(dev, priv->port, mac);
598 	} else {
599 		en_dbg(DRV, priv, "Releasing qp: port %d, qpn %d\n",
600 		       priv->port, qpn);
601 		mlx4_qp_release_range(dev, qpn, 1);
602 		priv->flags &= ~MLX4_EN_FLAG_FORCE_PROMISC;
603 	}
604 }
605 
606 static void mlx4_en_clear_uclist(struct net_device *dev)
607 {
608 	struct mlx4_en_priv *priv = netdev_priv(dev);
609 	struct mlx4_en_addr_list *tmp, *uc_to_del;
610 
611 	list_for_each_entry_safe(uc_to_del, tmp, &priv->uc_list, list) {
612 		list_del(&uc_to_del->list);
613 		kfree(uc_to_del);
614 	}
615 }
616 
617 static void mlx4_en_cache_uclist(struct net_device *dev)
618 {
619 	struct mlx4_en_priv *priv = netdev_priv(dev);
620 	struct mlx4_en_addr_list *tmp;
621 	struct ifaddr *ifa;
622 
623 	mlx4_en_clear_uclist(dev);
624 
625 	if_addr_rlock(dev);
626 	TAILQ_FOREACH(ifa, &dev->if_addrhead, ifa_link) {
627 		if (ifa->ifa_addr->sa_family != AF_LINK)
628 			continue;
629 		if (((struct sockaddr_dl *)ifa->ifa_addr)->sdl_alen !=
630 				ETHER_ADDR_LEN)
631 			continue;
632 		tmp = kzalloc(sizeof(struct mlx4_en_addr_list), GFP_ATOMIC);
633 		if (tmp == NULL) {
634 			en_err(priv, "Failed to allocate address list\n");
635 			break;
636 		}
637 		memcpy(tmp->addr,
638 			LLADDR((struct sockaddr_dl *)ifa->ifa_addr), ETH_ALEN);
639 		list_add_tail(&tmp->list, &priv->uc_list);
640 	}
641 	if_addr_runlock(dev);
642 }
643 
644 static void mlx4_en_clear_mclist(struct net_device *dev)
645 {
646 	struct mlx4_en_priv *priv = netdev_priv(dev);
647 	struct mlx4_en_addr_list *tmp, *mc_to_del;
648 
649 	list_for_each_entry_safe(mc_to_del, tmp, &priv->mc_list, list) {
650 		list_del(&mc_to_del->list);
651 		kfree(mc_to_del);
652 	}
653 }
654 
655 static void mlx4_en_cache_mclist(struct net_device *dev)
656 {
657 	struct mlx4_en_priv *priv = netdev_priv(dev);
658 	struct mlx4_en_addr_list *tmp;
659 	struct ifmultiaddr *ifma;
660 
661 	mlx4_en_clear_mclist(dev);
662 
663 	if_maddr_rlock(dev);
664 	TAILQ_FOREACH(ifma, &dev->if_multiaddrs, ifma_link) {
665 		if (ifma->ifma_addr->sa_family != AF_LINK)
666 			continue;
667 		if (((struct sockaddr_dl *)ifma->ifma_addr)->sdl_alen !=
668 				ETHER_ADDR_LEN)
669 			continue;
670 		tmp = kzalloc(sizeof(struct mlx4_en_addr_list), GFP_ATOMIC);
671 		if (tmp == NULL) {
672 			en_err(priv, "Failed to allocate address list\n");
673 			break;
674 		}
675 		memcpy(tmp->addr,
676 			LLADDR((struct sockaddr_dl *)ifma->ifma_addr), ETH_ALEN);
677 		list_add_tail(&tmp->list, &priv->mc_list);
678 	}
679 	if_maddr_runlock(dev);
680 }
681 
682 static void update_addr_list_flags(struct mlx4_en_priv *priv,
683 				struct list_head *dst,
684 				struct list_head *src)
685 {
686 	struct mlx4_en_addr_list *dst_tmp, *src_tmp, *new_mc;
687 	bool found;
688 
689 	/* Find all the entries that should be removed from dst,
690 	 * These are the entries that are not found in src
691 	 */
692 	list_for_each_entry(dst_tmp, dst, list) {
693 		found = false;
694 		list_for_each_entry(src_tmp, src, list) {
695 			if (!memcmp(dst_tmp->addr, src_tmp->addr, ETH_ALEN)) {
696 				found = true;
697 				break;
698 			}
699 		}
700 		if (!found)
701 			dst_tmp->action = MLX4_ADDR_LIST_REM;
702 	}
703 
704 	/* Add entries that exist in src but not in dst
705 	 * mark them as need to add
706 	 */
707 	list_for_each_entry(src_tmp, src, list) {
708 		found = false;
709 		list_for_each_entry(dst_tmp, dst, list) {
710 			if (!memcmp(dst_tmp->addr, src_tmp->addr, ETH_ALEN)) {
711 				dst_tmp->action = MLX4_ADDR_LIST_NONE;
712 				found = true;
713 				break;
714 			}
715 		}
716 		if (!found) {
717 			new_mc = kmalloc(sizeof(struct mlx4_en_addr_list),
718 					 GFP_KERNEL);
719 			if (!new_mc) {
720 				en_err(priv, "Failed to allocate current multicast list\n");
721 				return;
722 			}
723 			memcpy(new_mc, src_tmp,
724 			       sizeof(struct mlx4_en_addr_list));
725 			new_mc->action = MLX4_ADDR_LIST_ADD;
726 			list_add_tail(&new_mc->list, dst);
727 		}
728 	}
729 }
730 
731 static void mlx4_en_set_rx_mode(struct net_device *dev)
732 {
733 	struct mlx4_en_priv *priv = netdev_priv(dev);
734 
735 	if (!priv->port_up)
736 		return;
737 
738 	queue_work(priv->mdev->workqueue, &priv->rx_mode_task);
739 }
740 
741 static void mlx4_en_set_promisc_mode(struct mlx4_en_priv *priv,
742 				     struct mlx4_en_dev *mdev)
743 {
744 	int err = 0;
745 
746 	if (!(priv->flags & MLX4_EN_FLAG_PROMISC)) {
747 		priv->flags |= MLX4_EN_FLAG_PROMISC;
748 
749 		/* Enable promiscouos mode */
750 		switch (mdev->dev->caps.steering_mode) {
751 		case MLX4_STEERING_MODE_DEVICE_MANAGED:
752 			err = mlx4_flow_steer_promisc_add(mdev->dev,
753 							  priv->port,
754 							  priv->base_qpn,
755 							  MLX4_FS_ALL_DEFAULT);
756 			if (err)
757 				en_err(priv, "Failed enabling promiscuous mode\n");
758 			priv->flags |= MLX4_EN_FLAG_MC_PROMISC;
759 			break;
760 
761 		case MLX4_STEERING_MODE_B0:
762 			err = mlx4_unicast_promisc_add(mdev->dev,
763 						       priv->base_qpn,
764 						       priv->port);
765 			if (err)
766 				en_err(priv, "Failed enabling unicast promiscuous mode\n");
767 
768 			/* Add the default qp number as multicast
769 			 * promisc
770 			 */
771 			if (!(priv->flags & MLX4_EN_FLAG_MC_PROMISC)) {
772 				err = mlx4_multicast_promisc_add(mdev->dev,
773 								 priv->base_qpn,
774 								 priv->port);
775 				if (err)
776 					en_err(priv, "Failed enabling multicast promiscuous mode\n");
777 				priv->flags |= MLX4_EN_FLAG_MC_PROMISC;
778 			}
779 			break;
780 
781 		case MLX4_STEERING_MODE_A0:
782 			err = mlx4_SET_PORT_qpn_calc(mdev->dev,
783 						     priv->port,
784 						     priv->base_qpn,
785 						     1);
786 			if (err)
787 				en_err(priv, "Failed enabling promiscuous mode\n");
788 			break;
789 		}
790 
791 		/* Disable port multicast filter (unconditionally) */
792 		err = mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, 0,
793 					  0, MLX4_MCAST_DISABLE);
794 		if (err)
795 			en_err(priv, "Failed disabling multicast filter\n");
796 	}
797 }
798 
799 static void mlx4_en_clear_promisc_mode(struct mlx4_en_priv *priv,
800 				       struct mlx4_en_dev *mdev)
801 {
802 	int err = 0;
803 
804 	priv->flags &= ~MLX4_EN_FLAG_PROMISC;
805 
806 	/* Disable promiscouos mode */
807 	switch (mdev->dev->caps.steering_mode) {
808 	case MLX4_STEERING_MODE_DEVICE_MANAGED:
809 		err = mlx4_flow_steer_promisc_remove(mdev->dev,
810 						     priv->port,
811 						     MLX4_FS_ALL_DEFAULT);
812 		if (err)
813 			en_err(priv, "Failed disabling promiscuous mode\n");
814 		priv->flags &= ~MLX4_EN_FLAG_MC_PROMISC;
815 		break;
816 
817 	case MLX4_STEERING_MODE_B0:
818 		err = mlx4_unicast_promisc_remove(mdev->dev,
819 						  priv->base_qpn,
820 						  priv->port);
821 		if (err)
822 			en_err(priv, "Failed disabling unicast promiscuous mode\n");
823 		/* Disable Multicast promisc */
824 		if (priv->flags & MLX4_EN_FLAG_MC_PROMISC) {
825 			err = mlx4_multicast_promisc_remove(mdev->dev,
826 							    priv->base_qpn,
827 							    priv->port);
828 			if (err)
829 				en_err(priv, "Failed disabling multicast promiscuous mode\n");
830 			priv->flags &= ~MLX4_EN_FLAG_MC_PROMISC;
831 		}
832 		break;
833 
834 	case MLX4_STEERING_MODE_A0:
835 		err = mlx4_SET_PORT_qpn_calc(mdev->dev,
836 					     priv->port,
837 					     priv->base_qpn, 0);
838 		if (err)
839 			en_err(priv, "Failed disabling promiscuous mode\n");
840 		break;
841 	}
842 }
843 
844 static void mlx4_en_do_multicast(struct mlx4_en_priv *priv,
845 				 struct net_device *dev,
846 				 struct mlx4_en_dev *mdev)
847 {
848 	struct mlx4_en_addr_list *addr_list, *tmp;
849 	u8 mc_list[16] = {0};
850 	int err = 0;
851 	u64 mcast_addr = 0;
852 
853 
854 	/* Enable/disable the multicast filter according to IFF_ALLMULTI */
855 	if (dev->if_flags & IFF_ALLMULTI) {
856 		err = mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, 0,
857 					  0, MLX4_MCAST_DISABLE);
858 		if (err)
859 			en_err(priv, "Failed disabling multicast filter\n");
860 
861 		/* Add the default qp number as multicast promisc */
862 		if (!(priv->flags & MLX4_EN_FLAG_MC_PROMISC)) {
863 			switch (mdev->dev->caps.steering_mode) {
864 			case MLX4_STEERING_MODE_DEVICE_MANAGED:
865 				err = mlx4_flow_steer_promisc_add(mdev->dev,
866 								  priv->port,
867 								  priv->base_qpn,
868 								  MLX4_FS_MC_DEFAULT);
869 				break;
870 
871 			case MLX4_STEERING_MODE_B0:
872 				err = mlx4_multicast_promisc_add(mdev->dev,
873 								 priv->base_qpn,
874 								 priv->port);
875 				break;
876 
877 			case MLX4_STEERING_MODE_A0:
878 				break;
879 			}
880 			if (err)
881 				en_err(priv, "Failed entering multicast promisc mode\n");
882 			priv->flags |= MLX4_EN_FLAG_MC_PROMISC;
883 		}
884 	} else {
885 		/* Disable Multicast promisc */
886 		if (priv->flags & MLX4_EN_FLAG_MC_PROMISC) {
887 			switch (mdev->dev->caps.steering_mode) {
888 			case MLX4_STEERING_MODE_DEVICE_MANAGED:
889 				err = mlx4_flow_steer_promisc_remove(mdev->dev,
890 								     priv->port,
891 								     MLX4_FS_MC_DEFAULT);
892 				break;
893 
894 			case MLX4_STEERING_MODE_B0:
895 				err = mlx4_multicast_promisc_remove(mdev->dev,
896 								    priv->base_qpn,
897 								    priv->port);
898 				break;
899 
900 			case MLX4_STEERING_MODE_A0:
901 				break;
902 			}
903 			if (err)
904 				en_err(priv, "Failed disabling multicast promiscuous mode\n");
905 			priv->flags &= ~MLX4_EN_FLAG_MC_PROMISC;
906 		}
907 
908 		/* Update unicast list */
909 		mlx4_en_cache_uclist(dev);
910 
911 		update_addr_list_flags(priv, &priv->curr_uc_list, &priv->uc_list);
912 
913 		list_for_each_entry_safe(addr_list, tmp, &priv->curr_uc_list, list) {
914 			if (addr_list->action == MLX4_ADDR_LIST_REM) {
915 				mlx4_en_uc_steer_release(priv, addr_list->addr,
916 							       priv->rss_map.indir_qp.qpn,
917 							       addr_list->reg_id);
918 				/* remove from list */
919 				list_del(&addr_list->list);
920 				kfree(addr_list);
921 			} else if (addr_list->action == MLX4_ADDR_LIST_ADD) {
922 				err = mlx4_en_uc_steer_add(priv, addr_list->addr,
923 							   &priv->rss_map.indir_qp.qpn,
924 							   &addr_list->reg_id);
925 				if (err)
926 					en_err(priv, "Fail to add unicast address\n");
927 			}
928 		}
929 
930 		err = mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, 0,
931 					  0, MLX4_MCAST_DISABLE);
932 		if (err)
933 			en_err(priv, "Failed disabling multicast filter\n");
934 
935 		/* Flush mcast filter and init it with broadcast address */
936 		mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, ETH_BCAST,
937 				    1, MLX4_MCAST_CONFIG);
938 
939 		/* Update multicast list - we cache all addresses so they won't
940 		 * change while HW is updated holding the command semaphor */
941 		mlx4_en_cache_mclist(dev);
942 		list_for_each_entry(addr_list, &priv->mc_list, list) {
943 			mcast_addr = mlx4_mac_to_u64(addr_list->addr);
944 			mlx4_SET_MCAST_FLTR(mdev->dev, priv->port,
945 					mcast_addr, 0, MLX4_MCAST_CONFIG);
946 		}
947 		err = mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, 0,
948 					  0, MLX4_MCAST_ENABLE);
949 		if (err)
950 			en_err(priv, "Failed enabling multicast filter\n");
951 
952 		update_addr_list_flags(priv, &priv->curr_mc_list, &priv->mc_list);
953 
954 		list_for_each_entry_safe(addr_list, tmp, &priv->curr_mc_list, list) {
955 			if (addr_list->action == MLX4_ADDR_LIST_REM) {
956 				/* detach this address and delete from list */
957 				memcpy(&mc_list[10], addr_list->addr, ETH_ALEN);
958 				mc_list[5] = priv->port;
959 				err = mlx4_multicast_detach(mdev->dev,
960 							    &priv->rss_map.indir_qp,
961 							    mc_list,
962 							    MLX4_PROT_ETH,
963 							    addr_list->reg_id);
964 				if (err)
965 					en_err(priv, "Fail to detach multicast address\n");
966 
967 				if (addr_list->tunnel_reg_id) {
968 					err = mlx4_flow_detach(priv->mdev->dev, addr_list->tunnel_reg_id);
969 					if (err)
970 						en_err(priv, "Failed to detach multicast address\n");
971 				}
972 
973 				/* remove from list */
974 				list_del(&addr_list->list);
975 				kfree(addr_list);
976 			} else if (addr_list->action == MLX4_ADDR_LIST_ADD) {
977 				/* attach the address */
978 				memcpy(&mc_list[10], addr_list->addr, ETH_ALEN);
979 				/* needed for B0 steering support */
980 				mc_list[5] = priv->port;
981 				err = mlx4_multicast_attach(mdev->dev,
982 							    &priv->rss_map.indir_qp,
983 							    mc_list,
984 							    priv->port, 0,
985 							    MLX4_PROT_ETH,
986 							    &addr_list->reg_id);
987 				if (err)
988 					en_err(priv, "Fail to attach multicast address\n");
989 
990 				err = mlx4_en_tunnel_steer_add(priv, &mc_list[10], priv->base_qpn,
991 							       &addr_list->tunnel_reg_id);
992 				if (err)
993 					en_err(priv, "Failed to attach multicast address\n");
994 			}
995 		}
996 	}
997 }
998 
999 static void mlx4_en_do_set_rx_mode(struct work_struct *work)
1000 {
1001 	struct mlx4_en_priv *priv = container_of(work, struct mlx4_en_priv,
1002 						 rx_mode_task);
1003 	struct mlx4_en_dev *mdev = priv->mdev;
1004 	struct net_device *dev = priv->dev;
1005 
1006 	mutex_lock(&mdev->state_lock);
1007 	if (!mdev->device_up) {
1008 		en_dbg(HW, priv, "Card is not up, ignoring rx mode change.\n");
1009 		goto out;
1010 	}
1011 	if (!priv->port_up) {
1012 		en_dbg(HW, priv, "Port is down, ignoring rx mode change.\n");
1013 		goto out;
1014 	}
1015 	if (!mlx4_en_QUERY_PORT(mdev, priv->port)) {
1016 		if (priv->port_state.link_state) {
1017 			priv->last_link_state = MLX4_DEV_EVENT_PORT_UP;
1018 			/* update netif baudrate */
1019 			priv->dev->if_baudrate =
1020 			    IF_Mbps(priv->port_state.link_speed);
1021 			/* Important note: the following call for if_link_state_change
1022 			 * is needed for interface up scenario (start port, link state
1023 			 * change) */
1024 			if_link_state_change(priv->dev, LINK_STATE_UP);
1025 			en_dbg(HW, priv, "Link Up\n");
1026 		}
1027 	}
1028 
1029 	/* Promsicuous mode: disable all filters */
1030 	if ((dev->if_flags & IFF_PROMISC) ||
1031 	    (priv->flags & MLX4_EN_FLAG_FORCE_PROMISC)) {
1032 		mlx4_en_set_promisc_mode(priv, mdev);
1033 		goto out;
1034 	}
1035 
1036 	/* Not in promiscuous mode */
1037 	if (priv->flags & MLX4_EN_FLAG_PROMISC)
1038 		mlx4_en_clear_promisc_mode(priv, mdev);
1039 
1040 	mlx4_en_do_multicast(priv, dev, mdev);
1041 out:
1042 	mutex_unlock(&mdev->state_lock);
1043 }
1044 
1045 static void mlx4_en_watchdog_timeout(void *arg)
1046 {
1047         struct mlx4_en_priv *priv = arg;
1048         struct mlx4_en_dev *mdev = priv->mdev;
1049 
1050         en_dbg(DRV, priv, "Scheduling watchdog\n");
1051         queue_work(mdev->workqueue, &priv->watchdog_task);
1052         if (priv->port_up)
1053                 callout_reset(&priv->watchdog_timer, MLX4_EN_WATCHDOG_TIMEOUT,
1054                                 mlx4_en_watchdog_timeout, priv);
1055 }
1056 
1057 
1058 
1059 static void mlx4_en_set_default_moderation(struct mlx4_en_priv *priv)
1060 {
1061 	struct mlx4_en_cq *cq;
1062 	int i;
1063 
1064 	/* If we haven't received a specific coalescing setting
1065 	 * (module param), we set the moderation parameters as follows:
1066 	 * - moder_cnt is set to the number of mtu sized packets to
1067 	 *   satisfy our coalescing target.
1068 	 * - moder_time is set to a fixed value.
1069 	 */
1070 	priv->rx_frames = MLX4_EN_RX_COAL_TARGET;
1071 	priv->rx_usecs = MLX4_EN_RX_COAL_TIME;
1072 	priv->tx_frames = MLX4_EN_TX_COAL_PKTS;
1073 	priv->tx_usecs = MLX4_EN_TX_COAL_TIME;
1074 	en_dbg(INTR, priv, "Default coalesing params for mtu: %u - "
1075 	       "rx_frames:%d rx_usecs:%d\n",
1076 	       (unsigned)priv->dev->if_mtu, priv->rx_frames, priv->rx_usecs);
1077 
1078 	/* Setup cq moderation params */
1079 	for (i = 0; i < priv->rx_ring_num; i++) {
1080 		cq = priv->rx_cq[i];
1081 		cq->moder_cnt = priv->rx_frames;
1082 		cq->moder_time = priv->rx_usecs;
1083 		priv->last_moder_time[i] = MLX4_EN_AUTO_CONF;
1084 		priv->last_moder_packets[i] = 0;
1085 		priv->last_moder_bytes[i] = 0;
1086 	}
1087 
1088 	for (i = 0; i < priv->tx_ring_num; i++) {
1089 		cq = priv->tx_cq[i];
1090 		cq->moder_cnt = priv->tx_frames;
1091 		cq->moder_time = priv->tx_usecs;
1092 	}
1093 
1094 	/* Reset auto-moderation params */
1095 	priv->pkt_rate_low = MLX4_EN_RX_RATE_LOW;
1096 	priv->rx_usecs_low = MLX4_EN_RX_COAL_TIME_LOW;
1097 	priv->pkt_rate_high = MLX4_EN_RX_RATE_HIGH;
1098 	priv->rx_usecs_high = MLX4_EN_RX_COAL_TIME_HIGH;
1099 	priv->sample_interval = MLX4_EN_SAMPLE_INTERVAL;
1100 	priv->adaptive_rx_coal = 1;
1101 	priv->last_moder_jiffies = 0;
1102 	priv->last_moder_tx_packets = 0;
1103 }
1104 
1105 static void mlx4_en_auto_moderation(struct mlx4_en_priv *priv)
1106 {
1107 	unsigned long period = (unsigned long) (jiffies - priv->last_moder_jiffies);
1108 	struct mlx4_en_cq *cq;
1109 	unsigned long packets;
1110 	unsigned long rate;
1111 	unsigned long avg_pkt_size;
1112 	unsigned long rx_packets;
1113 	unsigned long rx_bytes;
1114 	unsigned long rx_pkt_diff;
1115 	int moder_time;
1116 	int ring, err;
1117 
1118 	if (!priv->adaptive_rx_coal || period < priv->sample_interval * HZ)
1119 		return;
1120 
1121 	for (ring = 0; ring < priv->rx_ring_num; ring++) {
1122                 spin_lock(&priv->stats_lock);
1123 		rx_packets = priv->rx_ring[ring]->packets;
1124 		rx_bytes = priv->rx_ring[ring]->bytes;
1125 		spin_unlock(&priv->stats_lock);
1126 
1127 		rx_pkt_diff = ((unsigned long) (rx_packets -
1128 				priv->last_moder_packets[ring]));
1129 		packets = rx_pkt_diff;
1130 		rate = packets * HZ / period;
1131 		avg_pkt_size = packets ? ((unsigned long) (rx_bytes -
1132 				priv->last_moder_bytes[ring])) / packets : 0;
1133 
1134 		/* Apply auto-moderation only when packet rate
1135 		 * exceeds a rate that it matters */
1136 		if (rate > (MLX4_EN_RX_RATE_THRESH / priv->rx_ring_num) &&
1137 		    avg_pkt_size > MLX4_EN_AVG_PKT_SMALL) {
1138 			if (rate < priv->pkt_rate_low)
1139 				moder_time = priv->rx_usecs_low;
1140 			else if (rate > priv->pkt_rate_high)
1141 				moder_time = priv->rx_usecs_high;
1142 			else
1143 				moder_time = (rate - priv->pkt_rate_low) *
1144 					(priv->rx_usecs_high - priv->rx_usecs_low) /
1145 					(priv->pkt_rate_high - priv->pkt_rate_low) +
1146 					priv->rx_usecs_low;
1147 		} else {
1148 			moder_time = priv->rx_usecs_low;
1149 		}
1150 
1151 		if (moder_time != priv->last_moder_time[ring]) {
1152 			priv->last_moder_time[ring] = moder_time;
1153 			cq = priv->rx_cq[ring];
1154 			cq->moder_time = moder_time;
1155 			cq->moder_cnt = priv->rx_frames;
1156 			err = mlx4_en_set_cq_moder(priv, cq);
1157 			if (err)
1158 				en_err(priv, "Failed modifying moderation for cq:%d\n",
1159 				       ring);
1160 		}
1161 		priv->last_moder_packets[ring] = rx_packets;
1162 		priv->last_moder_bytes[ring] = rx_bytes;
1163 	}
1164 
1165 	priv->last_moder_jiffies = jiffies;
1166 }
1167 
1168 static void mlx4_en_do_get_stats(struct work_struct *work)
1169 {
1170 	struct delayed_work *delay = to_delayed_work(work);
1171 	struct mlx4_en_priv *priv = container_of(delay, struct mlx4_en_priv,
1172 						 stats_task);
1173 	struct mlx4_en_dev *mdev = priv->mdev;
1174 	int err;
1175 
1176 	mutex_lock(&mdev->state_lock);
1177 	if (mdev->device_up) {
1178 		if (priv->port_up) {
1179 			if (mlx4_is_slave(mdev->dev))
1180 				err = mlx4_en_get_vport_stats(mdev, priv->port);
1181 			else
1182 				err = mlx4_en_DUMP_ETH_STATS(mdev, priv->port, 0);
1183 			if (err)
1184 				en_dbg(HW, priv, "Could not update stats\n");
1185 
1186 			mlx4_en_auto_moderation(priv);
1187 		}
1188 
1189 		queue_delayed_work(mdev->workqueue, &priv->stats_task, STATS_DELAY);
1190 	}
1191 	mutex_unlock(&mdev->state_lock);
1192 }
1193 
1194 /* mlx4_en_service_task - Run service task for tasks that needed to be done
1195  * periodically
1196  */
1197 static void mlx4_en_service_task(struct work_struct *work)
1198 {
1199 	struct delayed_work *delay = to_delayed_work(work);
1200 	struct mlx4_en_priv *priv = container_of(delay, struct mlx4_en_priv,
1201 						 service_task);
1202 	struct mlx4_en_dev *mdev = priv->mdev;
1203 
1204 	mutex_lock(&mdev->state_lock);
1205 	if (mdev->device_up) {
1206 		queue_delayed_work(mdev->workqueue, &priv->service_task,
1207 				   SERVICE_TASK_DELAY);
1208 	}
1209 	mutex_unlock(&mdev->state_lock);
1210 }
1211 
1212 static void mlx4_en_linkstate(struct work_struct *work)
1213 {
1214 	struct mlx4_en_priv *priv = container_of(work, struct mlx4_en_priv,
1215 						 linkstate_task);
1216 	struct mlx4_en_dev *mdev = priv->mdev;
1217 	int linkstate = priv->link_state;
1218 
1219 	mutex_lock(&mdev->state_lock);
1220 	/* If observable port state changed set carrier state and
1221 	 * report to system log */
1222 	if (priv->last_link_state != linkstate) {
1223 		if (linkstate == MLX4_DEV_EVENT_PORT_DOWN) {
1224 			en_info(priv, "Link Down\n");
1225 			if_link_state_change(priv->dev, LINK_STATE_DOWN);
1226 			/* update netif baudrate */
1227 			priv->dev->if_baudrate = 0;
1228 
1229 		/* make sure the port is up before notifying the OS.
1230 		 * This is tricky since we get here on INIT_PORT and
1231 		 * in such case we can't tell the OS the port is up.
1232 		 * To solve this there is a call to if_link_state_change
1233 		 * in set_rx_mode.
1234 		 * */
1235 		} else if (priv->port_up && (linkstate == MLX4_DEV_EVENT_PORT_UP)){
1236 			if (mlx4_en_QUERY_PORT(priv->mdev, priv->port))
1237 				en_info(priv, "Query port failed\n");
1238 			priv->dev->if_baudrate =
1239 			    IF_Mbps(priv->port_state.link_speed);
1240 			en_info(priv, "Link Up\n");
1241 			if_link_state_change(priv->dev, LINK_STATE_UP);
1242 		}
1243 	}
1244 	priv->last_link_state = linkstate;
1245 	mutex_unlock(&mdev->state_lock);
1246 }
1247 
1248 
1249 int mlx4_en_start_port(struct net_device *dev)
1250 {
1251 	struct mlx4_en_priv *priv = netdev_priv(dev);
1252 	struct mlx4_en_dev *mdev = priv->mdev;
1253 	struct mlx4_en_cq *cq;
1254 	struct mlx4_en_tx_ring *tx_ring;
1255 	int rx_index = 0;
1256 	int tx_index = 0;
1257 	int err = 0;
1258 	int i;
1259 	int j;
1260 	u8 mc_list[16] = {0};
1261 
1262 
1263 	if (priv->port_up) {
1264 		en_dbg(DRV, priv, "start port called while port already up\n");
1265 		return 0;
1266 	}
1267 
1268 	INIT_LIST_HEAD(&priv->mc_list);
1269 	INIT_LIST_HEAD(&priv->uc_list);
1270 	INIT_LIST_HEAD(&priv->curr_mc_list);
1271 	INIT_LIST_HEAD(&priv->curr_uc_list);
1272 	INIT_LIST_HEAD(&priv->ethtool_list);
1273 
1274 	/* Calculate Rx buf size */
1275 	dev->if_mtu = min(dev->if_mtu, priv->max_mtu);
1276         mlx4_en_calc_rx_buf(dev);
1277 	en_dbg(DRV, priv, "Rx buf size:%d\n", priv->rx_mb_size);
1278 
1279 	/* Configure rx cq's and rings */
1280 	err = mlx4_en_activate_rx_rings(priv);
1281 	if (err) {
1282 		en_err(priv, "Failed to activate RX rings\n");
1283 		return err;
1284 	}
1285 	for (i = 0; i < priv->rx_ring_num; i++) {
1286 		cq = priv->rx_cq[i];
1287 
1288 		mlx4_en_cq_init_lock(cq);
1289 		err = mlx4_en_activate_cq(priv, cq, i);
1290 		if (err) {
1291 			en_err(priv, "Failed activating Rx CQ\n");
1292 			goto cq_err;
1293 		}
1294 		for (j = 0; j < cq->size; j++)
1295 			cq->buf[j].owner_sr_opcode = MLX4_CQE_OWNER_MASK;
1296 		err = mlx4_en_set_cq_moder(priv, cq);
1297 		if (err) {
1298 			en_err(priv, "Failed setting cq moderation parameters");
1299 			mlx4_en_deactivate_cq(priv, cq);
1300 			goto cq_err;
1301 		}
1302 		mlx4_en_arm_cq(priv, cq);
1303 		priv->rx_ring[i]->cqn = cq->mcq.cqn;
1304 		++rx_index;
1305 	}
1306 
1307 	/* Set qp number */
1308 	en_dbg(DRV, priv, "Getting qp number for port %d\n", priv->port);
1309 	err = mlx4_en_get_qp(priv);
1310 	if (err) {
1311 		en_err(priv, "Failed getting eth qp\n");
1312 		goto cq_err;
1313 	}
1314 	mdev->mac_removed[priv->port] = 0;
1315 
1316 	priv->counter_index =
1317 			mlx4_get_default_counter_index(mdev->dev, priv->port);
1318 
1319 	err = mlx4_en_config_rss_steer(priv);
1320 	if (err) {
1321 		en_err(priv, "Failed configuring rss steering\n");
1322 		goto mac_err;
1323 	}
1324 
1325 	err = mlx4_en_create_drop_qp(priv);
1326 	if (err)
1327 		goto rss_err;
1328 
1329 	/* Configure tx cq's and rings */
1330 	for (i = 0; i < priv->tx_ring_num; i++) {
1331 		/* Configure cq */
1332 		cq = priv->tx_cq[i];
1333 		err = mlx4_en_activate_cq(priv, cq, i);
1334 		if (err) {
1335 			en_err(priv, "Failed activating Tx CQ\n");
1336 			goto tx_err;
1337 		}
1338 		err = mlx4_en_set_cq_moder(priv, cq);
1339 		if (err) {
1340 			en_err(priv, "Failed setting cq moderation parameters");
1341 			mlx4_en_deactivate_cq(priv, cq);
1342 			goto tx_err;
1343 		}
1344 		en_dbg(DRV, priv, "Resetting index of collapsed CQ:%d to -1\n", i);
1345 		cq->buf->wqe_index = cpu_to_be16(0xffff);
1346 
1347 		/* Configure ring */
1348 		tx_ring = priv->tx_ring[i];
1349 
1350 		err = mlx4_en_activate_tx_ring(priv, tx_ring, cq->mcq.cqn,
1351 					       i / priv->num_tx_rings_p_up);
1352 		if (err) {
1353 			en_err(priv, "Failed activating Tx ring %d\n", i);
1354 			mlx4_en_deactivate_cq(priv, cq);
1355 			goto tx_err;
1356 		}
1357 
1358 		/* Arm CQ for TX completions */
1359 		mlx4_en_arm_cq(priv, cq);
1360 
1361 		/* Set initial ownership of all Tx TXBBs to SW (1) */
1362 		for (j = 0; j < tx_ring->buf_size; j += STAMP_STRIDE)
1363 			*((u32 *) (tx_ring->buf + j)) = INIT_OWNER_BIT;
1364 		++tx_index;
1365 	}
1366 
1367 	/* Configure port */
1368 	err = mlx4_SET_PORT_general(mdev->dev, priv->port,
1369 				    priv->rx_mb_size,
1370 				    priv->prof->tx_pause,
1371 				    priv->prof->tx_ppp,
1372 				    priv->prof->rx_pause,
1373 				    priv->prof->rx_ppp);
1374 	if (err) {
1375 		en_err(priv, "Failed setting port general configurations for port %d, with error %d\n",
1376 		       priv->port, err);
1377 		goto tx_err;
1378 	}
1379 	/* Set default qp number */
1380 	err = mlx4_SET_PORT_qpn_calc(mdev->dev, priv->port, priv->base_qpn, 0);
1381 	if (err) {
1382 		en_err(priv, "Failed setting default qp numbers\n");
1383 		goto tx_err;
1384 	}
1385 
1386 	/* Init port */
1387 	en_dbg(HW, priv, "Initializing port\n");
1388 	err = mlx4_INIT_PORT(mdev->dev, priv->port);
1389 	if (err) {
1390 		en_err(priv, "Failed Initializing port\n");
1391 		goto tx_err;
1392 	}
1393 
1394 	/* Attach rx QP to bradcast address */
1395 	memset(&mc_list[10], 0xff, ETH_ALEN);
1396 	mc_list[5] = priv->port; /* needed for B0 steering support */
1397 	if (mlx4_multicast_attach(mdev->dev, &priv->rss_map.indir_qp, mc_list,
1398 				  priv->port, 0, MLX4_PROT_ETH,
1399 				  &priv->broadcast_id))
1400 		mlx4_warn(mdev, "Failed Attaching Broadcast\n");
1401 
1402 	/* Must redo promiscuous mode setup. */
1403 	priv->flags &= ~(MLX4_EN_FLAG_PROMISC | MLX4_EN_FLAG_MC_PROMISC);
1404 
1405 	/* Schedule multicast task to populate multicast list */
1406 	queue_work(mdev->workqueue, &priv->rx_mode_task);
1407 
1408 	priv->port_up = true;
1409 
1410         /* Enable the queues. */
1411         dev->if_drv_flags &= ~IFF_DRV_OACTIVE;
1412         dev->if_drv_flags |= IFF_DRV_RUNNING;
1413 #ifdef CONFIG_DEBUG_FS
1414 	mlx4_en_create_debug_files(priv);
1415 #endif
1416         callout_reset(&priv->watchdog_timer, MLX4_EN_WATCHDOG_TIMEOUT,
1417                     mlx4_en_watchdog_timeout, priv);
1418 
1419 
1420 	return 0;
1421 
1422 tx_err:
1423 	while (tx_index--) {
1424 		mlx4_en_deactivate_tx_ring(priv, priv->tx_ring[tx_index]);
1425 		mlx4_en_deactivate_cq(priv, priv->tx_cq[tx_index]);
1426 	}
1427 	mlx4_en_destroy_drop_qp(priv);
1428 rss_err:
1429 	mlx4_en_release_rss_steer(priv);
1430 mac_err:
1431 	mlx4_en_put_qp(priv);
1432 cq_err:
1433 	while (rx_index--)
1434 		mlx4_en_deactivate_cq(priv, priv->rx_cq[rx_index]);
1435 	for (i = 0; i < priv->rx_ring_num; i++)
1436 		mlx4_en_deactivate_rx_ring(priv, priv->rx_ring[i]);
1437 
1438 	return err; /* need to close devices */
1439 }
1440 
1441 
1442 void mlx4_en_stop_port(struct net_device *dev)
1443 {
1444 	struct mlx4_en_priv *priv = netdev_priv(dev);
1445 	struct mlx4_en_dev *mdev = priv->mdev;
1446 	struct mlx4_en_addr_list *addr_list, *tmp;
1447 	int i;
1448 	u8 mc_list[16] = {0};
1449 
1450 	if (!priv->port_up) {
1451 		en_dbg(DRV, priv, "stop port called while port already down\n");
1452 		return;
1453 	}
1454 
1455 #ifdef CONFIG_DEBUG_FS
1456 	mlx4_en_delete_debug_files(priv);
1457 #endif
1458 
1459 	/* close port*/
1460 	mlx4_CLOSE_PORT(mdev->dev, priv->port);
1461 
1462 	/* Set port as not active */
1463 	priv->port_up = false;
1464 	priv->counter_index = MLX4_SINK_COUNTER_INDEX(mdev->dev);
1465 
1466 	/* Promsicuous mode */
1467 	if (mdev->dev->caps.steering_mode ==
1468 	    MLX4_STEERING_MODE_DEVICE_MANAGED) {
1469 		priv->flags &= ~(MLX4_EN_FLAG_PROMISC |
1470 				 MLX4_EN_FLAG_MC_PROMISC);
1471 		mlx4_flow_steer_promisc_remove(mdev->dev,
1472 					       priv->port,
1473 					       MLX4_FS_ALL_DEFAULT);
1474 		mlx4_flow_steer_promisc_remove(mdev->dev,
1475 					       priv->port,
1476 					       MLX4_FS_MC_DEFAULT);
1477 	} else if (priv->flags & MLX4_EN_FLAG_PROMISC) {
1478 		priv->flags &= ~MLX4_EN_FLAG_PROMISC;
1479 
1480 		/* Disable promiscouos mode */
1481 		mlx4_unicast_promisc_remove(mdev->dev, priv->base_qpn,
1482 					    priv->port);
1483 
1484 		/* Disable Multicast promisc */
1485 		if (priv->flags & MLX4_EN_FLAG_MC_PROMISC) {
1486 			mlx4_multicast_promisc_remove(mdev->dev, priv->base_qpn,
1487 						      priv->port);
1488 			priv->flags &= ~MLX4_EN_FLAG_MC_PROMISC;
1489 		}
1490 	}
1491 
1492 	/* Detach All unicasts */
1493 	list_for_each_entry(addr_list, &priv->curr_uc_list, list) {
1494 		mlx4_en_uc_steer_release(priv, addr_list->addr,
1495 					 priv->rss_map.indir_qp.qpn,
1496 					 addr_list->reg_id);
1497 	}
1498 	mlx4_en_clear_uclist(dev);
1499 	list_for_each_entry_safe(addr_list, tmp, &priv->curr_uc_list, list) {
1500 		list_del(&addr_list->list);
1501 		kfree(addr_list);
1502 	}
1503 
1504 	/* Detach All multicasts */
1505 	memset(&mc_list[10], 0xff, ETH_ALEN);
1506 	mc_list[5] = priv->port; /* needed for B0 steering support */
1507 	mlx4_multicast_detach(mdev->dev, &priv->rss_map.indir_qp, mc_list,
1508 			      MLX4_PROT_ETH, priv->broadcast_id);
1509 	list_for_each_entry(addr_list, &priv->curr_mc_list, list) {
1510 		memcpy(&mc_list[10], addr_list->addr, ETH_ALEN);
1511 		mc_list[5] = priv->port;
1512 		mlx4_multicast_detach(mdev->dev, &priv->rss_map.indir_qp,
1513 				      mc_list, MLX4_PROT_ETH, addr_list->reg_id);
1514 	}
1515 	mlx4_en_clear_mclist(dev);
1516 	list_for_each_entry_safe(addr_list, tmp, &priv->curr_mc_list, list) {
1517 		list_del(&addr_list->list);
1518 		kfree(addr_list);
1519 	}
1520 
1521 	/* Flush multicast filter */
1522 	mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, 0, 1, MLX4_MCAST_CONFIG);
1523 	mlx4_en_destroy_drop_qp(priv);
1524 
1525 	/* Free TX Rings */
1526 	for (i = 0; i < priv->tx_ring_num; i++) {
1527 		mlx4_en_deactivate_tx_ring(priv, priv->tx_ring[i]);
1528 		mlx4_en_deactivate_cq(priv, priv->tx_cq[i]);
1529 	}
1530 	msleep(10);
1531 
1532 	for (i = 0; i < priv->tx_ring_num; i++)
1533 		mlx4_en_free_tx_buf(dev, priv->tx_ring[i]);
1534 
1535 	/* Free RSS qps */
1536 	mlx4_en_release_rss_steer(priv);
1537 
1538 	/* Unregister Mac address for the port */
1539 	mlx4_en_put_qp(priv);
1540 	mdev->mac_removed[priv->port] = 1;
1541 
1542 	/* Free RX Rings */
1543 	for (i = 0; i < priv->rx_ring_num; i++) {
1544 		struct mlx4_en_cq *cq = priv->rx_cq[i];
1545 		mlx4_en_deactivate_rx_ring(priv, priv->rx_ring[i]);
1546 		mlx4_en_deactivate_cq(priv, cq);
1547 	}
1548 
1549         callout_stop(&priv->watchdog_timer);
1550 
1551         dev->if_drv_flags &= ~(IFF_DRV_RUNNING | IFF_DRV_OACTIVE);
1552 }
1553 
1554 static void mlx4_en_restart(struct work_struct *work)
1555 {
1556 	struct mlx4_en_priv *priv = container_of(work, struct mlx4_en_priv,
1557 						 watchdog_task);
1558 	struct mlx4_en_dev *mdev = priv->mdev;
1559 	struct net_device *dev = priv->dev;
1560 	struct mlx4_en_tx_ring *ring;
1561 	int i;
1562 
1563 
1564 	if (priv->blocked == 0 || priv->port_up == 0)
1565 		return;
1566 	for (i = 0; i < priv->tx_ring_num; i++) {
1567 		ring = priv->tx_ring[i];
1568 		if (ring->blocked &&
1569 				ring->watchdog_time + MLX4_EN_WATCHDOG_TIMEOUT < ticks)
1570 			goto reset;
1571 	}
1572 	return;
1573 
1574 reset:
1575 	priv->port_stats.tx_timeout++;
1576 	en_dbg(DRV, priv, "Watchdog task called for port %d\n", priv->port);
1577 
1578 	mutex_lock(&mdev->state_lock);
1579 	if (priv->port_up) {
1580 		mlx4_en_stop_port(dev);
1581                 //for (i = 0; i < priv->tx_ring_num; i++)
1582                 //        netdev_tx_reset_queue(priv->tx_ring[i]->tx_queue);
1583 		if (mlx4_en_start_port(dev))
1584 			en_err(priv, "Failed restarting port %d\n", priv->port);
1585 	}
1586 	mutex_unlock(&mdev->state_lock);
1587 }
1588 
1589 static void mlx4_en_clear_stats(struct net_device *dev)
1590 {
1591 	struct mlx4_en_priv *priv = netdev_priv(dev);
1592 	struct mlx4_en_dev *mdev = priv->mdev;
1593 	int i;
1594 
1595 	if (!mlx4_is_slave(mdev->dev))
1596 		if (mlx4_en_DUMP_ETH_STATS(mdev, priv->port, 1))
1597 			en_dbg(HW, priv, "Failed dumping statistics\n");
1598 
1599 	memset(&priv->pstats, 0, sizeof(priv->pstats));
1600 	memset(&priv->pkstats, 0, sizeof(priv->pkstats));
1601 	memset(&priv->port_stats, 0, sizeof(priv->port_stats));
1602 	memset(&priv->vport_stats, 0, sizeof(priv->vport_stats));
1603 
1604 	for (i = 0; i < priv->tx_ring_num; i++) {
1605 		priv->tx_ring[i]->bytes = 0;
1606 		priv->tx_ring[i]->packets = 0;
1607 		priv->tx_ring[i]->tx_csum = 0;
1608 		priv->tx_ring[i]->oversized_packets = 0;
1609 	}
1610 	for (i = 0; i < priv->rx_ring_num; i++) {
1611 		priv->rx_ring[i]->bytes = 0;
1612 		priv->rx_ring[i]->packets = 0;
1613 		priv->rx_ring[i]->csum_ok = 0;
1614 		priv->rx_ring[i]->csum_none = 0;
1615 	}
1616 }
1617 
1618 static void mlx4_en_open(void* arg)
1619 {
1620 
1621         struct mlx4_en_priv *priv;
1622         struct mlx4_en_dev *mdev;
1623         struct net_device *dev;
1624         int err = 0;
1625 
1626         priv = arg;
1627         mdev = priv->mdev;
1628         dev = priv->dev;
1629 
1630 
1631 	mutex_lock(&mdev->state_lock);
1632 
1633 	if (!mdev->device_up) {
1634 		en_err(priv, "Cannot open - device down/disabled\n");
1635 		goto out;
1636 	}
1637 
1638 	/* Reset HW statistics and SW counters */
1639 	mlx4_en_clear_stats(dev);
1640 
1641 	err = mlx4_en_start_port(dev);
1642 	if (err)
1643 		en_err(priv, "Failed starting port:%d\n", priv->port);
1644 
1645 out:
1646 	mutex_unlock(&mdev->state_lock);
1647 	return;
1648 }
1649 
1650 void mlx4_en_free_resources(struct mlx4_en_priv *priv)
1651 {
1652 	int i;
1653 
1654 #ifdef CONFIG_RFS_ACCEL
1655 	if (priv->dev->rx_cpu_rmap) {
1656 		free_irq_cpu_rmap(priv->dev->rx_cpu_rmap);
1657 		priv->dev->rx_cpu_rmap = NULL;
1658 	}
1659 #endif
1660 
1661 	for (i = 0; i < priv->tx_ring_num; i++) {
1662 		if (priv->tx_ring && priv->tx_ring[i])
1663 			mlx4_en_destroy_tx_ring(priv, &priv->tx_ring[i]);
1664 		if (priv->tx_cq && priv->tx_cq[i])
1665 			mlx4_en_destroy_cq(priv, &priv->tx_cq[i]);
1666 	}
1667 
1668 	for (i = 0; i < priv->rx_ring_num; i++) {
1669 		if (priv->rx_ring[i])
1670 			mlx4_en_destroy_rx_ring(priv, &priv->rx_ring[i],
1671 				priv->prof->rx_ring_size, priv->stride);
1672 		if (priv->rx_cq[i])
1673 			mlx4_en_destroy_cq(priv, &priv->rx_cq[i]);
1674 	}
1675 
1676 	if (priv->stat_sysctl != NULL)
1677 		sysctl_ctx_free(&priv->stat_ctx);
1678 }
1679 
1680 int mlx4_en_alloc_resources(struct mlx4_en_priv *priv)
1681 {
1682 	struct mlx4_en_port_profile *prof = priv->prof;
1683 	int i;
1684 	int node = 0;
1685 
1686 	/* Create rx Rings */
1687 	for (i = 0; i < priv->rx_ring_num; i++) {
1688 		if (mlx4_en_create_cq(priv, &priv->rx_cq[i],
1689 				      prof->rx_ring_size, i, RX, node))
1690 			goto err;
1691 
1692 		if (mlx4_en_create_rx_ring(priv, &priv->rx_ring[i],
1693 					   prof->rx_ring_size, node))
1694 			goto err;
1695 	}
1696 
1697 	/* Create tx Rings */
1698 	for (i = 0; i < priv->tx_ring_num; i++) {
1699 		if (mlx4_en_create_cq(priv, &priv->tx_cq[i],
1700 				      prof->tx_ring_size, i, TX, node))
1701 			goto err;
1702 
1703 		if (mlx4_en_create_tx_ring(priv, &priv->tx_ring[i],
1704 					   prof->tx_ring_size, TXBB_SIZE, node, i))
1705 			goto err;
1706 	}
1707 
1708 #ifdef CONFIG_RFS_ACCEL
1709 	priv->dev->rx_cpu_rmap = alloc_irq_cpu_rmap(priv->rx_ring_num);
1710 	if (!priv->dev->rx_cpu_rmap)
1711 		goto err;
1712 #endif
1713         /* Re-create stat sysctls in case the number of rings changed. */
1714 	mlx4_en_sysctl_stat(priv);
1715 	return 0;
1716 
1717 err:
1718 	en_err(priv, "Failed to allocate NIC resources\n");
1719 	for (i = 0; i < priv->rx_ring_num; i++) {
1720 		if (priv->rx_ring[i])
1721 			mlx4_en_destroy_rx_ring(priv, &priv->rx_ring[i],
1722 						prof->rx_ring_size,
1723 						priv->stride);
1724 		if (priv->rx_cq[i])
1725 			mlx4_en_destroy_cq(priv, &priv->rx_cq[i]);
1726 	}
1727 	for (i = 0; i < priv->tx_ring_num; i++) {
1728 		if (priv->tx_ring[i])
1729 			mlx4_en_destroy_tx_ring(priv, &priv->tx_ring[i]);
1730 		if (priv->tx_cq[i])
1731 			mlx4_en_destroy_cq(priv, &priv->tx_cq[i]);
1732 	}
1733 	priv->port_up = false;
1734 	return -ENOMEM;
1735 }
1736 
1737 struct en_port_attribute {
1738 	struct attribute attr;
1739 	ssize_t (*show)(struct en_port *, struct en_port_attribute *, char *buf);
1740 	ssize_t (*store)(struct en_port *, struct en_port_attribute *, char *buf, size_t count);
1741 };
1742 
1743 #define PORT_ATTR_RO(_name) \
1744 struct en_port_attribute en_port_attr_##_name = __ATTR_RO(_name)
1745 
1746 #define EN_PORT_ATTR(_name, _mode, _show, _store) \
1747 struct en_port_attribute en_port_attr_##_name = __ATTR(_name, _mode, _show, _store)
1748 
1749 void mlx4_en_destroy_netdev(struct net_device *dev)
1750 {
1751 	struct mlx4_en_priv *priv = netdev_priv(dev);
1752 	struct mlx4_en_dev *mdev = priv->mdev;
1753 
1754 	en_dbg(DRV, priv, "Destroying netdev on port:%d\n", priv->port);
1755 
1756 	/* don't allow more IOCTLs */
1757 	priv->gone = 1;
1758 
1759 	/* XXX wait a bit to allow IOCTL handlers to complete */
1760 	pause("W", hz);
1761 
1762 	if (priv->vlan_attach != NULL)
1763 		EVENTHANDLER_DEREGISTER(vlan_config, priv->vlan_attach);
1764 	if (priv->vlan_detach != NULL)
1765 		EVENTHANDLER_DEREGISTER(vlan_unconfig, priv->vlan_detach);
1766 
1767 	/* Unregister device - this will close the port if it was up */
1768 	if (priv->registered) {
1769 		mutex_lock(&mdev->state_lock);
1770 		ether_ifdetach(dev);
1771 		mutex_unlock(&mdev->state_lock);
1772 	}
1773 
1774 	mutex_lock(&mdev->state_lock);
1775 	mlx4_en_stop_port(dev);
1776 	mutex_unlock(&mdev->state_lock);
1777 
1778 	if (priv->allocated)
1779 		mlx4_free_hwq_res(mdev->dev, &priv->res, MLX4_EN_PAGE_SIZE);
1780 
1781 	cancel_delayed_work(&priv->stats_task);
1782 	cancel_delayed_work(&priv->service_task);
1783 	/* flush any pending task for this netdev */
1784 	flush_workqueue(mdev->workqueue);
1785         callout_drain(&priv->watchdog_timer);
1786 
1787 	/* Detach the netdev so tasks would not attempt to access it */
1788 	mutex_lock(&mdev->state_lock);
1789 	mdev->pndev[priv->port] = NULL;
1790 	mutex_unlock(&mdev->state_lock);
1791 
1792 
1793 	mlx4_en_free_resources(priv);
1794 
1795 	/* freeing the sysctl conf cannot be called from within mlx4_en_free_resources */
1796 	if (priv->conf_sysctl != NULL)
1797 		sysctl_ctx_free(&priv->conf_ctx);
1798 
1799 	kfree(priv->tx_ring);
1800 	kfree(priv->tx_cq);
1801 
1802         kfree(priv);
1803         if_free(dev);
1804 
1805 }
1806 
1807 static int mlx4_en_change_mtu(struct net_device *dev, int new_mtu)
1808 {
1809 	struct mlx4_en_priv *priv = netdev_priv(dev);
1810 	struct mlx4_en_dev *mdev = priv->mdev;
1811 	int err = 0;
1812 
1813 	en_dbg(DRV, priv, "Change MTU called - current:%u new:%u\n",
1814 	       (unsigned)dev->if_mtu, (unsigned)new_mtu);
1815 
1816 	if ((new_mtu < MLX4_EN_MIN_MTU) || (new_mtu > priv->max_mtu)) {
1817 		en_err(priv, "Bad MTU size:%d, max %u.\n", new_mtu,
1818 		    priv->max_mtu);
1819 		return -EPERM;
1820 	}
1821 	mutex_lock(&mdev->state_lock);
1822 	dev->if_mtu = new_mtu;
1823 	if (dev->if_drv_flags & IFF_DRV_RUNNING) {
1824 		if (!mdev->device_up) {
1825 			/* NIC is probably restarting - let watchdog task reset
1826 			 *                          * the port */
1827 			en_dbg(DRV, priv, "Change MTU called with card down!?\n");
1828 		} else {
1829 			mlx4_en_stop_port(dev);
1830 			err = mlx4_en_start_port(dev);
1831 			if (err) {
1832 				en_err(priv, "Failed restarting port:%d\n",
1833 						priv->port);
1834 				queue_work(mdev->workqueue, &priv->watchdog_task);
1835 			}
1836 		}
1837 	}
1838 	mutex_unlock(&mdev->state_lock);
1839 	return 0;
1840 }
1841 
1842 static int mlx4_en_calc_media(struct mlx4_en_priv *priv)
1843 {
1844 	int trans_type;
1845 	int active;
1846 
1847 	active = IFM_ETHER;
1848 	if (priv->last_link_state == MLX4_DEV_EVENT_PORT_DOWN)
1849 		return (active);
1850 	active |= IFM_FDX;
1851 	trans_type = priv->port_state.transceiver;
1852 	/* XXX I don't know all of the transceiver values. */
1853 	switch (priv->port_state.link_speed) {
1854 	case 100:
1855 		active |= IFM_100_T;
1856 		break;
1857 	case 1000:
1858 		active |= IFM_1000_T;
1859 		break;
1860 	case 10000:
1861 		if (trans_type > 0 && trans_type <= 0xC)
1862 			active |= IFM_10G_SR;
1863 		else if (trans_type == 0x80 || trans_type == 0)
1864 			active |= IFM_10G_CX4;
1865 		break;
1866 	case 40000:
1867 		active |= IFM_40G_CR4;
1868 		break;
1869 	}
1870 	if (priv->prof->tx_pause)
1871 		active |= IFM_ETH_TXPAUSE;
1872 	if (priv->prof->rx_pause)
1873 		active |= IFM_ETH_RXPAUSE;
1874 
1875 	return (active);
1876 }
1877 
1878 static void mlx4_en_media_status(struct ifnet *dev, struct ifmediareq *ifmr)
1879 {
1880 	struct mlx4_en_priv *priv;
1881 
1882 	priv = dev->if_softc;
1883 	ifmr->ifm_status = IFM_AVALID;
1884 	if (priv->last_link_state != MLX4_DEV_EVENT_PORT_DOWN)
1885 		ifmr->ifm_status |= IFM_ACTIVE;
1886 	ifmr->ifm_active = mlx4_en_calc_media(priv);
1887 
1888 	return;
1889 }
1890 
1891 static int mlx4_en_media_change(struct ifnet *dev)
1892 {
1893 	struct mlx4_en_priv *priv;
1894         struct ifmedia *ifm;
1895 	int rxpause;
1896 	int txpause;
1897 	int error;
1898 
1899 	priv = dev->if_softc;
1900 	ifm = &priv->media;
1901 	rxpause = txpause = 0;
1902 	error = 0;
1903 
1904 	if (IFM_TYPE(ifm->ifm_media) != IFM_ETHER)
1905 		return (EINVAL);
1906         switch (IFM_SUBTYPE(ifm->ifm_media)) {
1907         case IFM_AUTO:
1908 		break;
1909 	case IFM_10G_SR:
1910 	case IFM_10G_CX4:
1911 	case IFM_1000_T:
1912 	case IFM_40G_CR4:
1913 		if ((IFM_SUBTYPE(ifm->ifm_media)
1914 			== IFM_SUBTYPE(mlx4_en_calc_media(priv)))
1915 			&& (ifm->ifm_media & IFM_FDX))
1916 			break;
1917 		/* Fallthrough */
1918 	default:
1919                 printf("%s: Only auto media type\n", if_name(dev));
1920                 return (EINVAL);
1921 	}
1922 	/* Allow user to set/clear pause */
1923 	if (IFM_OPTIONS(ifm->ifm_media) & IFM_ETH_RXPAUSE)
1924 		rxpause = 1;
1925 	if (IFM_OPTIONS(ifm->ifm_media) & IFM_ETH_TXPAUSE)
1926 		txpause = 1;
1927 	if (priv->prof->tx_pause != txpause || priv->prof->rx_pause != rxpause) {
1928 		priv->prof->tx_pause = txpause;
1929 		priv->prof->rx_pause = rxpause;
1930 		error = -mlx4_SET_PORT_general(priv->mdev->dev, priv->port,
1931 		     priv->rx_mb_size + ETHER_CRC_LEN, priv->prof->tx_pause,
1932 		     priv->prof->tx_ppp, priv->prof->rx_pause,
1933 		     priv->prof->rx_ppp);
1934 	}
1935 	return (error);
1936 }
1937 
1938 static int mlx4_en_ioctl(struct ifnet *dev, u_long command, caddr_t data)
1939 {
1940 	struct mlx4_en_priv *priv;
1941 	struct mlx4_en_dev *mdev;
1942 	struct ifreq *ifr;
1943 	int error;
1944 	int mask;
1945 	struct ifrsskey *ifrk;
1946 	const u32 *key;
1947 	struct ifrsshash *ifrh;
1948 	u8 rss_mask;
1949 
1950 	error = 0;
1951 	mask = 0;
1952 	priv = dev->if_softc;
1953 
1954 	/* check if detaching */
1955 	if (priv == NULL || priv->gone != 0)
1956 		return (ENXIO);
1957 
1958 	mdev = priv->mdev;
1959 	ifr = (struct ifreq *) data;
1960 
1961 	switch (command) {
1962 	case SIOCSIFMTU:
1963 		error = -mlx4_en_change_mtu(dev, ifr->ifr_mtu);
1964 		break;
1965 	case SIOCSIFFLAGS:
1966 		if (dev->if_flags & IFF_UP) {
1967 			if ((dev->if_drv_flags & IFF_DRV_RUNNING) == 0) {
1968 				mutex_lock(&mdev->state_lock);
1969 				mlx4_en_start_port(dev);
1970 				mutex_unlock(&mdev->state_lock);
1971 			} else {
1972 				mlx4_en_set_rx_mode(dev);
1973 			}
1974 		} else {
1975 			mutex_lock(&mdev->state_lock);
1976 			if (dev->if_drv_flags & IFF_DRV_RUNNING) {
1977 				mlx4_en_stop_port(dev);
1978 				if_link_state_change(dev, LINK_STATE_DOWN);
1979 			}
1980 			mutex_unlock(&mdev->state_lock);
1981 		}
1982 		break;
1983 	case SIOCADDMULTI:
1984 	case SIOCDELMULTI:
1985 		mlx4_en_set_rx_mode(dev);
1986 		break;
1987 	case SIOCSIFMEDIA:
1988 	case SIOCGIFMEDIA:
1989 		error = ifmedia_ioctl(dev, ifr, &priv->media, command);
1990 		break;
1991 	case SIOCSIFCAP:
1992 		mutex_lock(&mdev->state_lock);
1993 		mask = ifr->ifr_reqcap ^ dev->if_capenable;
1994 		if (mask & IFCAP_TXCSUM) {
1995 			dev->if_capenable ^= IFCAP_TXCSUM;
1996 			dev->if_hwassist ^= (CSUM_TCP | CSUM_UDP | CSUM_IP);
1997 
1998 			if (IFCAP_TSO4 & dev->if_capenable &&
1999 			    !(IFCAP_TXCSUM & dev->if_capenable)) {
2000 				dev->if_capenable &= ~IFCAP_TSO4;
2001 				dev->if_hwassist &= ~CSUM_IP_TSO;
2002 				if_printf(dev,
2003 				    "tso4 disabled due to -txcsum.\n");
2004 			}
2005 		}
2006 		if (mask & IFCAP_TXCSUM_IPV6) {
2007 			dev->if_capenable ^= IFCAP_TXCSUM_IPV6;
2008 			dev->if_hwassist ^= (CSUM_UDP_IPV6 | CSUM_TCP_IPV6);
2009 
2010 			if (IFCAP_TSO6 & dev->if_capenable &&
2011 			    !(IFCAP_TXCSUM_IPV6 & dev->if_capenable)) {
2012 				dev->if_capenable &= ~IFCAP_TSO6;
2013 				dev->if_hwassist &= ~CSUM_IP6_TSO;
2014 				if_printf(dev,
2015 				    "tso6 disabled due to -txcsum6.\n");
2016 			}
2017 		}
2018 		if (mask & IFCAP_RXCSUM)
2019 			dev->if_capenable ^= IFCAP_RXCSUM;
2020 		if (mask & IFCAP_RXCSUM_IPV6)
2021 			dev->if_capenable ^= IFCAP_RXCSUM_IPV6;
2022 
2023 		if (mask & IFCAP_TSO4) {
2024 			if (!(IFCAP_TSO4 & dev->if_capenable) &&
2025 			    !(IFCAP_TXCSUM & dev->if_capenable)) {
2026 				if_printf(dev, "enable txcsum first.\n");
2027 				error = EAGAIN;
2028 				goto out;
2029 			}
2030 			dev->if_capenable ^= IFCAP_TSO4;
2031 			dev->if_hwassist ^= CSUM_IP_TSO;
2032 		}
2033 		if (mask & IFCAP_TSO6) {
2034 			if (!(IFCAP_TSO6 & dev->if_capenable) &&
2035 			    !(IFCAP_TXCSUM_IPV6 & dev->if_capenable)) {
2036 				if_printf(dev, "enable txcsum6 first.\n");
2037 				error = EAGAIN;
2038 				goto out;
2039 			}
2040 			dev->if_capenable ^= IFCAP_TSO6;
2041 			dev->if_hwassist ^= CSUM_IP6_TSO;
2042 		}
2043 		if (mask & IFCAP_LRO)
2044 			dev->if_capenable ^= IFCAP_LRO;
2045 		if (mask & IFCAP_VLAN_HWTAGGING)
2046 			dev->if_capenable ^= IFCAP_VLAN_HWTAGGING;
2047 		if (mask & IFCAP_VLAN_HWFILTER)
2048 			dev->if_capenable ^= IFCAP_VLAN_HWFILTER;
2049 		if (mask & IFCAP_WOL_MAGIC)
2050 			dev->if_capenable ^= IFCAP_WOL_MAGIC;
2051 		if (dev->if_drv_flags & IFF_DRV_RUNNING)
2052 			mlx4_en_start_port(dev);
2053 out:
2054 		mutex_unlock(&mdev->state_lock);
2055 		VLAN_CAPABILITIES(dev);
2056 		break;
2057 #if __FreeBSD_version >= 1100036
2058 	case SIOCGI2C: {
2059 		struct ifi2creq i2c;
2060 
2061 		error = copyin(ifr_data_get_ptr(ifr), &i2c, sizeof(i2c));
2062 		if (error)
2063 			break;
2064 		if (i2c.len > sizeof(i2c.data)) {
2065 			error = EINVAL;
2066 			break;
2067 		}
2068 		/*
2069 		 * Note that we ignore i2c.addr here. The driver hardcodes
2070 		 * the address to 0x50, while standard expects it to be 0xA0.
2071 		 */
2072 		error = mlx4_get_module_info(mdev->dev, priv->port,
2073 		    i2c.offset, i2c.len, i2c.data);
2074 		if (error < 0) {
2075 			error = -error;
2076 			break;
2077 		}
2078 		error = copyout(&i2c, ifr_data_get_ptr(ifr), sizeof(i2c));
2079 		break;
2080 	}
2081 #endif
2082 	case SIOCGIFRSSKEY:
2083 		ifrk = (struct ifrsskey *)data;
2084 		ifrk->ifrk_func = RSS_FUNC_TOEPLITZ;
2085 		mutex_lock(&mdev->state_lock);
2086 		key = mlx4_en_get_rss_key(priv, &ifrk->ifrk_keylen);
2087 		if (ifrk->ifrk_keylen > RSS_KEYLEN)
2088 			error = EINVAL;
2089 		else
2090 			memcpy(ifrk->ifrk_key, key, ifrk->ifrk_keylen);
2091 		mutex_unlock(&mdev->state_lock);
2092 		break;
2093 
2094 	case SIOCGIFRSSHASH:
2095 		mutex_lock(&mdev->state_lock);
2096 		rss_mask = mlx4_en_get_rss_mask(priv);
2097 		mutex_unlock(&mdev->state_lock);
2098 		ifrh = (struct ifrsshash *)data;
2099 		ifrh->ifrh_func = RSS_FUNC_TOEPLITZ;
2100 		ifrh->ifrh_types = 0;
2101 		if (rss_mask & MLX4_RSS_IPV4)
2102 			ifrh->ifrh_types |= RSS_TYPE_IPV4;
2103 		if (rss_mask & MLX4_RSS_TCP_IPV4)
2104 			ifrh->ifrh_types |= RSS_TYPE_TCP_IPV4;
2105 		if (rss_mask & MLX4_RSS_IPV6)
2106 			ifrh->ifrh_types |= RSS_TYPE_IPV6;
2107 		if (rss_mask & MLX4_RSS_TCP_IPV6)
2108 			ifrh->ifrh_types |= RSS_TYPE_TCP_IPV6;
2109 		if (rss_mask & MLX4_RSS_UDP_IPV4)
2110 			ifrh->ifrh_types |= RSS_TYPE_UDP_IPV4;
2111 		if (rss_mask & MLX4_RSS_UDP_IPV6)
2112 			ifrh->ifrh_types |= RSS_TYPE_UDP_IPV6;
2113 		break;
2114 
2115 	default:
2116 		error = ether_ioctl(dev, command, data);
2117 		break;
2118 	}
2119 
2120 	return (error);
2121 }
2122 
2123 
2124 int mlx4_en_init_netdev(struct mlx4_en_dev *mdev, int port,
2125 			struct mlx4_en_port_profile *prof)
2126 {
2127 	struct net_device *dev;
2128 	struct mlx4_en_priv *priv;
2129 	uint8_t dev_addr[ETHER_ADDR_LEN];
2130 	int err;
2131 	int i;
2132 
2133 	priv = kzalloc(sizeof(*priv), GFP_KERNEL);
2134 	dev = priv->dev = if_alloc(IFT_ETHER);
2135 	if (dev == NULL) {
2136 		en_err(priv, "Net device allocation failed\n");
2137 		kfree(priv);
2138 		return -ENOMEM;
2139 	}
2140 	dev->if_softc = priv;
2141 	if_initname(dev, "mlxen", (device_get_unit(
2142 	    mdev->pdev->dev.bsddev) * MLX4_MAX_PORTS) + port - 1);
2143 	dev->if_mtu = ETHERMTU;
2144 	dev->if_init = mlx4_en_open;
2145 	dev->if_flags = IFF_BROADCAST | IFF_SIMPLEX | IFF_MULTICAST;
2146 	dev->if_ioctl = mlx4_en_ioctl;
2147 	dev->if_transmit = mlx4_en_transmit;
2148 	dev->if_qflush = mlx4_en_qflush;
2149 	dev->if_snd.ifq_maxlen = prof->tx_ring_size;
2150 
2151 	/*
2152 	 * Initialize driver private data
2153 	 */
2154 	priv->counter_index = 0xff;
2155 	spin_lock_init(&priv->stats_lock);
2156 	INIT_WORK(&priv->rx_mode_task, mlx4_en_do_set_rx_mode);
2157 	INIT_WORK(&priv->watchdog_task, mlx4_en_restart);
2158 	INIT_WORK(&priv->linkstate_task, mlx4_en_linkstate);
2159 	INIT_DELAYED_WORK(&priv->stats_task, mlx4_en_do_get_stats);
2160 	INIT_DELAYED_WORK(&priv->service_task, mlx4_en_service_task);
2161 	callout_init(&priv->watchdog_timer, 1);
2162 #ifdef CONFIG_RFS_ACCEL
2163 	INIT_LIST_HEAD(&priv->filters);
2164 	spin_lock_init(&priv->filters_lock);
2165 #endif
2166 
2167 	priv->msg_enable = MLX4_EN_MSG_LEVEL;
2168 	priv->dev = dev;
2169 	priv->mdev = mdev;
2170 	priv->ddev = &mdev->pdev->dev;
2171 	priv->prof = prof;
2172 	priv->port = port;
2173 	priv->port_up = false;
2174 	priv->flags = prof->flags;
2175 
2176 	priv->num_tx_rings_p_up = mdev->profile.num_tx_rings_p_up;
2177 	priv->tx_ring_num = prof->tx_ring_num;
2178 	priv->tx_ring = kcalloc(MAX_TX_RINGS,
2179 				sizeof(struct mlx4_en_tx_ring *), GFP_KERNEL);
2180 	if (!priv->tx_ring) {
2181 		err = -ENOMEM;
2182 		goto out;
2183 	}
2184 	priv->tx_cq = kcalloc(sizeof(struct mlx4_en_cq *), MAX_TX_RINGS,
2185 			GFP_KERNEL);
2186 	if (!priv->tx_cq) {
2187 		err = -ENOMEM;
2188 		goto out;
2189 	}
2190 
2191 	priv->rx_ring_num = prof->rx_ring_num;
2192 	priv->cqe_factor = (mdev->dev->caps.cqe_size == 64) ? 1 : 0;
2193 	priv->mac_index = -1;
2194 	priv->last_ifq_jiffies = 0;
2195 	priv->if_counters_rx_errors = 0;
2196 	priv->if_counters_rx_no_buffer = 0;
2197 #ifdef CONFIG_MLX4_EN_DCB
2198 	if (!mlx4_is_slave(priv->mdev->dev)) {
2199 		priv->dcbx_cap = DCB_CAP_DCBX_HOST;
2200 		priv->flags |= MLX4_EN_FLAG_DCB_ENABLED;
2201 		if (mdev->dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_ETS_CFG) {
2202 			dev->dcbnl_ops = &mlx4_en_dcbnl_ops;
2203 		} else {
2204 			en_info(priv, "QoS disabled - no HW support\n");
2205 			dev->dcbnl_ops = &mlx4_en_dcbnl_pfc_ops;
2206 		}
2207 	}
2208 #endif
2209 
2210 	/* Query for default mac and max mtu */
2211 	priv->max_mtu = mdev->dev->caps.eth_mtu_cap[priv->port];
2212         priv->mac = mdev->dev->caps.def_mac[priv->port];
2213         if (ILLEGAL_MAC(priv->mac)) {
2214 #if BITS_PER_LONG == 64
2215                 en_err(priv, "Port: %d, invalid mac burned: 0x%lx, quiting\n",
2216                                 priv->port, priv->mac);
2217 #elif BITS_PER_LONG == 32
2218                 en_err(priv, "Port: %d, invalid mac burned: 0x%llx, quiting\n",
2219                                 priv->port, priv->mac);
2220 #endif
2221                 err = -EINVAL;
2222                 goto out;
2223         }
2224 
2225 	priv->stride = roundup_pow_of_two(sizeof(struct mlx4_en_rx_desc) +
2226 					  DS_SIZE);
2227 
2228 	mlx4_en_sysctl_conf(priv);
2229 
2230 	err = mlx4_en_alloc_resources(priv);
2231 	if (err)
2232 		goto out;
2233 
2234 	/* Allocate page for receive rings */
2235 	err = mlx4_alloc_hwq_res(mdev->dev, &priv->res,
2236 				MLX4_EN_PAGE_SIZE, MLX4_EN_PAGE_SIZE);
2237 	if (err) {
2238 		en_err(priv, "Failed to allocate page for rx qps\n");
2239 		goto out;
2240 	}
2241 	priv->allocated = 1;
2242 
2243 	/*
2244 	 * Set driver features
2245 	 */
2246 	dev->if_capabilities |= IFCAP_HWCSUM | IFCAP_HWCSUM_IPV6;
2247 	dev->if_capabilities |= IFCAP_VLAN_MTU | IFCAP_VLAN_HWTAGGING;
2248 	dev->if_capabilities |= IFCAP_VLAN_HWCSUM | IFCAP_VLAN_HWFILTER;
2249 	dev->if_capabilities |= IFCAP_LINKSTATE | IFCAP_JUMBO_MTU;
2250 	dev->if_capabilities |= IFCAP_LRO;
2251 	dev->if_capabilities |= IFCAP_HWSTATS;
2252 
2253 	if (mdev->LSO_support)
2254 		dev->if_capabilities |= IFCAP_TSO4 | IFCAP_TSO6 | IFCAP_VLAN_HWTSO;
2255 
2256 #if __FreeBSD_version >= 1100000
2257 	/* set TSO limits so that we don't have to drop TX packets */
2258 	dev->if_hw_tsomax = MLX4_EN_TX_MAX_PAYLOAD_SIZE - (ETHER_HDR_LEN + ETHER_VLAN_ENCAP_LEN) /* hdr */;
2259 	dev->if_hw_tsomaxsegcount = MLX4_EN_TX_MAX_MBUF_FRAGS - 1 /* hdr */;
2260 	dev->if_hw_tsomaxsegsize = MLX4_EN_TX_MAX_MBUF_SIZE;
2261 #endif
2262 
2263 	dev->if_capenable = dev->if_capabilities;
2264 
2265 	dev->if_hwassist = 0;
2266 	if (dev->if_capenable & (IFCAP_TSO4 | IFCAP_TSO6))
2267 		dev->if_hwassist |= CSUM_TSO;
2268 	if (dev->if_capenable & IFCAP_TXCSUM)
2269 		dev->if_hwassist |= (CSUM_TCP | CSUM_UDP | CSUM_IP);
2270 	if (dev->if_capenable & IFCAP_TXCSUM_IPV6)
2271 		dev->if_hwassist |= (CSUM_UDP_IPV6 | CSUM_TCP_IPV6);
2272 
2273 
2274         /* Register for VLAN events */
2275 	priv->vlan_attach = EVENTHANDLER_REGISTER(vlan_config,
2276             mlx4_en_vlan_rx_add_vid, priv, EVENTHANDLER_PRI_FIRST);
2277 	priv->vlan_detach = EVENTHANDLER_REGISTER(vlan_unconfig,
2278             mlx4_en_vlan_rx_kill_vid, priv, EVENTHANDLER_PRI_FIRST);
2279 
2280 	mdev->pndev[priv->port] = dev;
2281 
2282 	priv->last_link_state = MLX4_DEV_EVENT_PORT_DOWN;
2283         mlx4_en_set_default_moderation(priv);
2284 
2285 	/* Set default MAC */
2286 	for (i = 0; i < ETHER_ADDR_LEN; i++)
2287 		dev_addr[ETHER_ADDR_LEN - 1 - i] = (u8) (priv->mac >> (8 * i));
2288 
2289 
2290 	ether_ifattach(dev, dev_addr);
2291 	if_link_state_change(dev, LINK_STATE_DOWN);
2292 	ifmedia_init(&priv->media, IFM_IMASK | IFM_ETH_FMASK,
2293 	    mlx4_en_media_change, mlx4_en_media_status);
2294 	ifmedia_add(&priv->media, IFM_ETHER | IFM_FDX | IFM_1000_T, 0, NULL);
2295 	ifmedia_add(&priv->media, IFM_ETHER | IFM_FDX | IFM_10G_SR, 0, NULL);
2296 	ifmedia_add(&priv->media, IFM_ETHER | IFM_FDX | IFM_10G_CX4, 0, NULL);
2297 	ifmedia_add(&priv->media, IFM_ETHER | IFM_FDX | IFM_40G_CR4, 0, NULL);
2298 	ifmedia_add(&priv->media, IFM_ETHER | IFM_AUTO, 0, NULL);
2299 	ifmedia_set(&priv->media, IFM_ETHER | IFM_AUTO);
2300 
2301 	en_warn(priv, "Using %d TX rings\n", prof->tx_ring_num);
2302 	en_warn(priv, "Using %d RX rings\n", prof->rx_ring_num);
2303 
2304 	priv->registered = 1;
2305 
2306         en_warn(priv, "Using %d TX rings\n", prof->tx_ring_num);
2307         en_warn(priv, "Using %d RX rings\n", prof->rx_ring_num);
2308 
2309 
2310 	priv->rx_mb_size = dev->if_mtu + ETH_HLEN + VLAN_HLEN + ETH_FCS_LEN;
2311 	err = mlx4_SET_PORT_general(mdev->dev, priv->port,
2312 				    priv->rx_mb_size,
2313 				    prof->tx_pause, prof->tx_ppp,
2314 				    prof->rx_pause, prof->rx_ppp);
2315 	if (err) {
2316 		en_err(priv, "Failed setting port general configurations "
2317 		       "for port %d, with error %d\n", priv->port, err);
2318 		goto out;
2319 	}
2320 
2321 	/* Init port */
2322 	en_warn(priv, "Initializing port\n");
2323 	err = mlx4_INIT_PORT(mdev->dev, priv->port);
2324 	if (err) {
2325 		en_err(priv, "Failed Initializing port\n");
2326 		goto out;
2327 	}
2328 
2329 	queue_delayed_work(mdev->workqueue, &priv->stats_task, STATS_DELAY);
2330 
2331         if (mdev->dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_TS)
2332                 queue_delayed_work(mdev->workqueue, &priv->service_task, SERVICE_TASK_DELAY);
2333 
2334 	return 0;
2335 
2336 out:
2337 	mlx4_en_destroy_netdev(dev);
2338 	return err;
2339 }
2340 
2341 static int mlx4_en_set_ring_size(struct net_device *dev,
2342     int rx_size, int tx_size)
2343 {
2344         struct mlx4_en_priv *priv = netdev_priv(dev);
2345         struct mlx4_en_dev *mdev = priv->mdev;
2346         int port_up = 0;
2347         int err = 0;
2348 
2349         rx_size = roundup_pow_of_two(rx_size);
2350         rx_size = max_t(u32, rx_size, MLX4_EN_MIN_RX_SIZE);
2351         rx_size = min_t(u32, rx_size, MLX4_EN_MAX_RX_SIZE);
2352         tx_size = roundup_pow_of_two(tx_size);
2353         tx_size = max_t(u32, tx_size, MLX4_EN_MIN_TX_SIZE);
2354         tx_size = min_t(u32, tx_size, MLX4_EN_MAX_TX_SIZE);
2355 
2356         if (rx_size == (priv->port_up ?
2357             priv->rx_ring[0]->actual_size : priv->rx_ring[0]->size) &&
2358             tx_size == priv->tx_ring[0]->size)
2359                 return 0;
2360         mutex_lock(&mdev->state_lock);
2361         if (priv->port_up) {
2362                 port_up = 1;
2363                 mlx4_en_stop_port(dev);
2364         }
2365         mlx4_en_free_resources(priv);
2366         priv->prof->tx_ring_size = tx_size;
2367         priv->prof->rx_ring_size = rx_size;
2368         err = mlx4_en_alloc_resources(priv);
2369         if (err) {
2370                 en_err(priv, "Failed reallocating port resources\n");
2371                 goto out;
2372         }
2373         if (port_up) {
2374                 err = mlx4_en_start_port(dev);
2375                 if (err)
2376                         en_err(priv, "Failed starting port\n");
2377         }
2378 out:
2379         mutex_unlock(&mdev->state_lock);
2380         return err;
2381 }
2382 static int mlx4_en_set_rx_ring_size(SYSCTL_HANDLER_ARGS)
2383 {
2384         struct mlx4_en_priv *priv;
2385         int size;
2386         int error;
2387 
2388         priv = arg1;
2389         size = priv->prof->rx_ring_size;
2390         error = sysctl_handle_int(oidp, &size, 0, req);
2391         if (error || !req->newptr)
2392                 return (error);
2393         error = -mlx4_en_set_ring_size(priv->dev, size,
2394             priv->prof->tx_ring_size);
2395         return (error);
2396 }
2397 
2398 static int mlx4_en_set_tx_ring_size(SYSCTL_HANDLER_ARGS)
2399 {
2400         struct mlx4_en_priv *priv;
2401         int size;
2402         int error;
2403 
2404         priv = arg1;
2405         size = priv->prof->tx_ring_size;
2406         error = sysctl_handle_int(oidp, &size, 0, req);
2407         if (error || !req->newptr)
2408                 return (error);
2409         error = -mlx4_en_set_ring_size(priv->dev, priv->prof->rx_ring_size,
2410             size);
2411 
2412         return (error);
2413 }
2414 
2415 static int mlx4_en_get_module_info(struct net_device *dev,
2416 				   struct ethtool_modinfo *modinfo)
2417 {
2418 	struct mlx4_en_priv *priv = netdev_priv(dev);
2419 	struct mlx4_en_dev *mdev = priv->mdev;
2420 	int ret;
2421 	u8 data[4];
2422 
2423 	/* Read first 2 bytes to get Module & REV ID */
2424 	ret = mlx4_get_module_info(mdev->dev, priv->port,
2425 				   0/*offset*/, 2/*size*/, data);
2426 
2427 	if (ret < 2) {
2428 		en_err(priv, "Failed to read eeprom module first two bytes, error: 0x%x\n", -ret);
2429 		return -EIO;
2430 	}
2431 
2432 	switch (data[0] /* identifier */) {
2433 	case MLX4_MODULE_ID_QSFP:
2434 		modinfo->type = ETH_MODULE_SFF_8436;
2435 		modinfo->eeprom_len = ETH_MODULE_SFF_8436_LEN;
2436 		break;
2437 	case MLX4_MODULE_ID_QSFP_PLUS:
2438 		if (data[1] >= 0x3) { /* revision id */
2439 			modinfo->type = ETH_MODULE_SFF_8636;
2440 			modinfo->eeprom_len = ETH_MODULE_SFF_8636_LEN;
2441 		} else {
2442 			modinfo->type = ETH_MODULE_SFF_8436;
2443 			modinfo->eeprom_len = ETH_MODULE_SFF_8436_LEN;
2444 		}
2445 		break;
2446 	case MLX4_MODULE_ID_QSFP28:
2447 		modinfo->type = ETH_MODULE_SFF_8636;
2448 		modinfo->eeprom_len = ETH_MODULE_SFF_8636_LEN;
2449 		break;
2450 	case MLX4_MODULE_ID_SFP:
2451 		modinfo->type = ETH_MODULE_SFF_8472;
2452 		modinfo->eeprom_len = ETH_MODULE_SFF_8472_LEN;
2453 		break;
2454 	default:
2455 		en_err(priv, "mlx4_en_get_module_info :  Not recognized cable type\n");
2456 		return -EINVAL;
2457 	}
2458 
2459 	return 0;
2460 }
2461 
2462 static int mlx4_en_get_module_eeprom(struct net_device *dev,
2463 				     struct ethtool_eeprom *ee,
2464 				     u8 *data)
2465 {
2466 	struct mlx4_en_priv *priv = netdev_priv(dev);
2467 	struct mlx4_en_dev *mdev = priv->mdev;
2468 	int offset = ee->offset;
2469 	int i = 0, ret;
2470 
2471 	if (ee->len == 0)
2472 		return -EINVAL;
2473 
2474 	memset(data, 0, ee->len);
2475 
2476 	while (i < ee->len) {
2477 		en_dbg(DRV, priv,
2478 		       "mlx4_get_module_info i(%d) offset(%d) len(%d)\n",
2479 		       i, offset, ee->len - i);
2480 
2481 		ret = mlx4_get_module_info(mdev->dev, priv->port,
2482 					   offset, ee->len - i, data + i);
2483 
2484 		if (!ret) /* Done reading */
2485 			return 0;
2486 
2487 		if (ret < 0) {
2488 			en_err(priv,
2489 			       "mlx4_get_module_info i(%d) offset(%d) bytes_to_read(%d) - FAILED (0x%x)\n",
2490 			       i, offset, ee->len - i, ret);
2491 			return -1;
2492 		}
2493 
2494 		i += ret;
2495 		offset += ret;
2496 	}
2497 	return 0;
2498 }
2499 
2500 static void mlx4_en_print_eeprom(u8 *data, __u32 len)
2501 {
2502 	int		i;
2503 	int		j = 0;
2504 	int		row = 0;
2505 	const int	NUM_OF_BYTES = 16;
2506 
2507 	printf("\nOffset\t\tValues\n");
2508 	printf("------\t\t------\n");
2509 	while(row < len){
2510 		printf("0x%04x\t\t",row);
2511 		for(i=0; i < NUM_OF_BYTES; i++){
2512 			printf("%02x ", data[j]);
2513 			row++;
2514 			j++;
2515 		}
2516 		printf("\n");
2517 	}
2518 }
2519 
2520 /* Read cable EEPROM module information by first inspecting the first
2521  * two bytes to get the length and then read the rest of the information.
2522  * The information is printed to dmesg. */
2523 static int mlx4_en_read_eeprom(SYSCTL_HANDLER_ARGS)
2524 {
2525 
2526 	u8*		data;
2527 	int		error;
2528 	int		result = 0;
2529 	struct		mlx4_en_priv *priv;
2530 	struct		net_device *dev;
2531 	struct		ethtool_modinfo modinfo;
2532 	struct		ethtool_eeprom ee;
2533 
2534 	error = sysctl_handle_int(oidp, &result, 0, req);
2535 	if (error || !req->newptr)
2536 		return (error);
2537 
2538 	if (result == 1) {
2539 		priv = arg1;
2540 		dev = priv->dev;
2541 		data = kmalloc(PAGE_SIZE, GFP_KERNEL);
2542 
2543 		error = mlx4_en_get_module_info(dev, &modinfo);
2544 		if (error) {
2545 			en_err(priv,
2546 			       "mlx4_en_get_module_info returned with error - FAILED (0x%x)\n",
2547 			       -error);
2548 			goto out;
2549 		}
2550 
2551 		ee.len = modinfo.eeprom_len;
2552 		ee.offset = 0;
2553 
2554 		error = mlx4_en_get_module_eeprom(dev, &ee, data);
2555 		if (error) {
2556 			en_err(priv,
2557 			       "mlx4_en_get_module_eeprom returned with error - FAILED (0x%x)\n",
2558 			       -error);
2559 			/* Continue printing partial information in case of an error */
2560 		}
2561 
2562 		/* EEPROM information will be printed in dmesg */
2563 		mlx4_en_print_eeprom(data, ee.len);
2564 out:
2565 		kfree(data);
2566 	}
2567 	/* Return zero to prevent sysctl failure. */
2568 	return (0);
2569 }
2570 
2571 static int mlx4_en_set_tx_ppp(SYSCTL_HANDLER_ARGS)
2572 {
2573         struct mlx4_en_priv *priv;
2574         int ppp;
2575         int error;
2576 
2577         priv = arg1;
2578         ppp = priv->prof->tx_ppp;
2579         error = sysctl_handle_int(oidp, &ppp, 0, req);
2580         if (error || !req->newptr)
2581                 return (error);
2582         if (ppp > 0xff || ppp < 0)
2583                 return (-EINVAL);
2584         priv->prof->tx_ppp = ppp;
2585         error = -mlx4_SET_PORT_general(priv->mdev->dev, priv->port,
2586                                        priv->rx_mb_size + ETHER_CRC_LEN,
2587                                        priv->prof->tx_pause,
2588                                        priv->prof->tx_ppp,
2589                                        priv->prof->rx_pause,
2590                                        priv->prof->rx_ppp);
2591 
2592         return (error);
2593 }
2594 
2595 static int mlx4_en_set_rx_ppp(SYSCTL_HANDLER_ARGS)
2596 {
2597         struct mlx4_en_priv *priv;
2598         struct mlx4_en_dev *mdev;
2599         int ppp;
2600         int error;
2601         int port_up;
2602 
2603         port_up = 0;
2604         priv = arg1;
2605         mdev = priv->mdev;
2606         ppp = priv->prof->rx_ppp;
2607         error = sysctl_handle_int(oidp, &ppp, 0, req);
2608         if (error || !req->newptr)
2609                 return (error);
2610         if (ppp > 0xff || ppp < 0)
2611                 return (-EINVAL);
2612         /* See if we have to change the number of tx queues. */
2613         if (!ppp != !priv->prof->rx_ppp) {
2614                 mutex_lock(&mdev->state_lock);
2615                 if (priv->port_up) {
2616                         port_up = 1;
2617                         mlx4_en_stop_port(priv->dev);
2618                 }
2619                 mlx4_en_free_resources(priv);
2620                 priv->prof->rx_ppp = ppp;
2621                 error = -mlx4_en_alloc_resources(priv);
2622                 if (error)
2623                         en_err(priv, "Failed reallocating port resources\n");
2624                 if (error == 0 && port_up) {
2625                         error = -mlx4_en_start_port(priv->dev);
2626                         if (error)
2627                                 en_err(priv, "Failed starting port\n");
2628                 }
2629                 mutex_unlock(&mdev->state_lock);
2630                 return (error);
2631 
2632         }
2633         priv->prof->rx_ppp = ppp;
2634         error = -mlx4_SET_PORT_general(priv->mdev->dev, priv->port,
2635                                        priv->rx_mb_size + ETHER_CRC_LEN,
2636                                        priv->prof->tx_pause,
2637                                        priv->prof->tx_ppp,
2638                                        priv->prof->rx_pause,
2639                                        priv->prof->rx_ppp);
2640 
2641         return (error);
2642 }
2643 
2644 static void mlx4_en_sysctl_conf(struct mlx4_en_priv *priv)
2645 {
2646         struct net_device *dev;
2647         struct sysctl_ctx_list *ctx;
2648         struct sysctl_oid *node;
2649         struct sysctl_oid_list *node_list;
2650         struct sysctl_oid *coal;
2651         struct sysctl_oid_list *coal_list;
2652 	const char *pnameunit;
2653         dev = priv->dev;
2654         ctx = &priv->conf_ctx;
2655 	pnameunit = device_get_nameunit(priv->mdev->pdev->dev.bsddev);
2656 
2657         sysctl_ctx_init(ctx);
2658         priv->conf_sysctl = SYSCTL_ADD_NODE(ctx, SYSCTL_STATIC_CHILDREN(_hw),
2659             OID_AUTO, dev->if_xname, CTLFLAG_RD, 0, "mlx4 10gig ethernet");
2660         node = SYSCTL_ADD_NODE(ctx, SYSCTL_CHILDREN(priv->conf_sysctl), OID_AUTO,
2661             "conf", CTLFLAG_RD, NULL, "Configuration");
2662         node_list = SYSCTL_CHILDREN(node);
2663 
2664         SYSCTL_ADD_UINT(ctx, node_list, OID_AUTO, "msg_enable",
2665             CTLFLAG_RW, &priv->msg_enable, 0,
2666             "Driver message enable bitfield");
2667         SYSCTL_ADD_UINT(ctx, node_list, OID_AUTO, "rx_rings",
2668             CTLFLAG_RD, &priv->rx_ring_num, 0,
2669             "Number of receive rings");
2670         SYSCTL_ADD_UINT(ctx, node_list, OID_AUTO, "tx_rings",
2671             CTLFLAG_RD, &priv->tx_ring_num, 0,
2672             "Number of transmit rings");
2673         SYSCTL_ADD_PROC(ctx, node_list, OID_AUTO, "rx_size",
2674             CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, priv, 0,
2675             mlx4_en_set_rx_ring_size, "I", "Receive ring size");
2676         SYSCTL_ADD_PROC(ctx, node_list, OID_AUTO, "tx_size",
2677             CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, priv, 0,
2678             mlx4_en_set_tx_ring_size, "I", "Transmit ring size");
2679         SYSCTL_ADD_PROC(ctx, node_list, OID_AUTO, "tx_ppp",
2680             CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, priv, 0,
2681             mlx4_en_set_tx_ppp, "I", "TX Per-priority pause");
2682         SYSCTL_ADD_PROC(ctx, node_list, OID_AUTO, "rx_ppp",
2683             CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, priv, 0,
2684             mlx4_en_set_rx_ppp, "I", "RX Per-priority pause");
2685         SYSCTL_ADD_UINT(ctx, node_list, OID_AUTO, "port_num",
2686             CTLFLAG_RD, &priv->port, 0,
2687             "Port Number");
2688         SYSCTL_ADD_STRING(ctx, node_list, OID_AUTO, "device_name",
2689 	    CTLFLAG_RD, __DECONST(void *, pnameunit), 0,
2690 	    "PCI device name");
2691         /* Add coalescer configuration. */
2692         coal = SYSCTL_ADD_NODE(ctx, node_list, OID_AUTO,
2693             "coalesce", CTLFLAG_RD, NULL, "Interrupt coalesce configuration");
2694         coal_list = SYSCTL_CHILDREN(coal);
2695         SYSCTL_ADD_UINT(ctx, coal_list, OID_AUTO, "pkt_rate_low",
2696             CTLFLAG_RW, &priv->pkt_rate_low, 0,
2697             "Packets per-second for minimum delay");
2698         SYSCTL_ADD_UINT(ctx, coal_list, OID_AUTO, "rx_usecs_low",
2699             CTLFLAG_RW, &priv->rx_usecs_low, 0,
2700             "Minimum RX delay in micro-seconds");
2701         SYSCTL_ADD_UINT(ctx, coal_list, OID_AUTO, "pkt_rate_high",
2702             CTLFLAG_RW, &priv->pkt_rate_high, 0,
2703             "Packets per-second for maximum delay");
2704         SYSCTL_ADD_UINT(ctx, coal_list, OID_AUTO, "rx_usecs_high",
2705             CTLFLAG_RW, &priv->rx_usecs_high, 0,
2706             "Maximum RX delay in micro-seconds");
2707         SYSCTL_ADD_UINT(ctx, coal_list, OID_AUTO, "sample_interval",
2708             CTLFLAG_RW, &priv->sample_interval, 0,
2709             "adaptive frequency in units of HZ ticks");
2710         SYSCTL_ADD_UINT(ctx, coal_list, OID_AUTO, "adaptive_rx_coal",
2711             CTLFLAG_RW, &priv->adaptive_rx_coal, 0,
2712             "Enable adaptive rx coalescing");
2713 	/* EEPROM support */
2714 	SYSCTL_ADD_PROC(ctx, node_list, OID_AUTO, "eeprom_info",
2715 	    CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, priv, 0,
2716 	    mlx4_en_read_eeprom, "I", "EEPROM information");
2717 }
2718 
2719 static void mlx4_en_sysctl_stat(struct mlx4_en_priv *priv)
2720 {
2721 	struct sysctl_ctx_list *ctx;
2722 	struct sysctl_oid_list *node_list;
2723 	struct sysctl_oid *ring_node;
2724 	struct sysctl_oid_list *ring_list;
2725 	struct mlx4_en_tx_ring *tx_ring;
2726 	struct mlx4_en_rx_ring *rx_ring;
2727 	char namebuf[128];
2728 	int i;
2729 
2730 	ctx = &priv->stat_ctx;
2731 	sysctl_ctx_init(ctx);
2732 	priv->stat_sysctl = SYSCTL_ADD_NODE(ctx, SYSCTL_CHILDREN(priv->conf_sysctl), OID_AUTO,
2733 	    "stat", CTLFLAG_RD, NULL, "Statistics");
2734 	node_list = SYSCTL_CHILDREN(priv->stat_sysctl);
2735 
2736 #ifdef MLX4_EN_PERF_STAT
2737 	SYSCTL_ADD_UINT(ctx, node_list, OID_AUTO, "tx_poll", CTLFLAG_RD,
2738 	    &priv->pstats.tx_poll, "TX Poll calls");
2739 	SYSCTL_ADD_QUAD(ctx, node_list, OID_AUTO, "tx_pktsz_avg", CTLFLAG_RD,
2740 	    &priv->pstats.tx_pktsz_avg, "TX average packet size");
2741 	SYSCTL_ADD_UINT(ctx, node_list, OID_AUTO, "inflight_avg", CTLFLAG_RD,
2742 	    &priv->pstats.inflight_avg, "TX average packets in-flight");
2743 	SYSCTL_ADD_UINT(ctx, node_list, OID_AUTO, "tx_coal_avg", CTLFLAG_RD,
2744 	    &priv->pstats.tx_coal_avg, "TX average coalesced completions");
2745 	SYSCTL_ADD_UINT(ctx, node_list, OID_AUTO, "rx_coal_avg", CTLFLAG_RD,
2746 	    &priv->pstats.rx_coal_avg, "RX average coalesced completions");
2747 #endif
2748 
2749 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "tso_packets", CTLFLAG_RD,
2750 	    &priv->port_stats.tso_packets, 0, "TSO packets sent");
2751 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "queue_stopped", CTLFLAG_RD,
2752 	    &priv->port_stats.queue_stopped, 0, "Queue full");
2753 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "wake_queue", CTLFLAG_RD,
2754 	    &priv->port_stats.wake_queue, 0, "Queue resumed after full");
2755 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "tx_timeout", CTLFLAG_RD,
2756 	    &priv->port_stats.tx_timeout, 0, "Transmit timeouts");
2757 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "tx_oversized_packets", CTLFLAG_RD,
2758 	    &priv->port_stats.oversized_packets, 0, "TX oversized packets, m_defrag failed");
2759 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_alloc_failed", CTLFLAG_RD,
2760 	    &priv->port_stats.rx_alloc_failed, 0, "RX failed to allocate mbuf");
2761 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_chksum_good", CTLFLAG_RD,
2762             &priv->port_stats.rx_chksum_good, 0, "RX checksum offload success");
2763 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_chksum_none", CTLFLAG_RD,
2764 	    &priv->port_stats.rx_chksum_none, 0, "RX without checksum offload");
2765 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "tx_chksum_offload",
2766 	    CTLFLAG_RD, &priv->port_stats.tx_chksum_offload, 0,
2767 	    "TX checksum offloads");
2768 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "defrag_attempts",
2769 	    CTLFLAG_RD, &priv->port_stats.defrag_attempts, 0,
2770 	    "Oversized chains defragged");
2771 
2772 	/* Could strdup the names and add in a loop.  This is simpler. */
2773 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_bytes", CTLFLAG_RD,
2774 	    &priv->pkstats.rx_bytes, 0, "RX Bytes");
2775 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_packets", CTLFLAG_RD,
2776 	    &priv->pkstats.rx_packets, 0, "RX packets");
2777 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_multicast_packets", CTLFLAG_RD,
2778 	    &priv->pkstats.rx_multicast_packets, 0, "RX Multicast Packets");
2779 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_broadcast_packets", CTLFLAG_RD,
2780 	    &priv->pkstats.rx_broadcast_packets, 0, "RX Broadcast Packets");
2781 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_errors", CTLFLAG_RD,
2782 	    &priv->pkstats.rx_errors, 0, "RX Errors");
2783 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_dropped", CTLFLAG_RD,
2784 	    &priv->pkstats.rx_dropped, 0, "RX Dropped");
2785 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_length_errors", CTLFLAG_RD,
2786 	    &priv->pkstats.rx_length_errors, 0, "RX Length Errors");
2787 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_over_errors", CTLFLAG_RD,
2788 	    &priv->pkstats.rx_over_errors, 0, "RX Over Errors");
2789 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_crc_errors", CTLFLAG_RD,
2790 	    &priv->pkstats.rx_crc_errors, 0, "RX CRC Errors");
2791 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_jabbers", CTLFLAG_RD,
2792 	    &priv->pkstats.rx_jabbers, 0, "RX Jabbers");
2793 
2794 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_in_range_length_error", CTLFLAG_RD,
2795 	    &priv->pkstats.rx_in_range_length_error, 0, "RX IN_Range Length Error");
2796 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_out_range_length_error",
2797 	    CTLFLAG_RD, &priv->pkstats.rx_out_range_length_error, 0,
2798 	    "RX Out Range Length Error");
2799 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_lt_64_bytes_packets", CTLFLAG_RD,
2800 	    &priv->pkstats.rx_lt_64_bytes_packets, 0, "RX Lt 64 Bytes Packets");
2801 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_127_bytes_packets", CTLFLAG_RD,
2802 	    &priv->pkstats.rx_127_bytes_packets, 0, "RX 127 bytes Packets");
2803 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_255_bytes_packets", CTLFLAG_RD,
2804 	    &priv->pkstats.rx_255_bytes_packets, 0, "RX 255 bytes Packets");
2805 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_511_bytes_packets", CTLFLAG_RD,
2806 	    &priv->pkstats.rx_511_bytes_packets, 0, "RX 511 bytes Packets");
2807 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_1023_bytes_packets", CTLFLAG_RD,
2808 	    &priv->pkstats.rx_1023_bytes_packets, 0, "RX 1023 bytes Packets");
2809 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_1518_bytes_packets", CTLFLAG_RD,
2810 	    &priv->pkstats.rx_1518_bytes_packets, 0, "RX 1518 bytes Packets");
2811 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_1522_bytes_packets", CTLFLAG_RD,
2812 	    &priv->pkstats.rx_1522_bytes_packets, 0, "RX 1522 bytes Packets");
2813 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_1548_bytes_packets", CTLFLAG_RD,
2814 	    &priv->pkstats.rx_1548_bytes_packets, 0, "RX 1548 bytes Packets");
2815 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "rx_gt_1548_bytes_packets", CTLFLAG_RD,
2816 	    &priv->pkstats.rx_gt_1548_bytes_packets, 0,
2817 	    "RX Greater Then 1548 bytes Packets");
2818 
2819 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "tx_packets", CTLFLAG_RD,
2820 	    &priv->pkstats.tx_packets, 0, "TX packets");
2821 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "tx_bytes", CTLFLAG_RD,
2822 	    &priv->pkstats.tx_bytes, 0, "TX Bytes");
2823 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "tx_multicast_packets", CTLFLAG_RD,
2824 	    &priv->pkstats.tx_multicast_packets, 0, "TX Multicast Packets");
2825 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "tx_broadcast_packets", CTLFLAG_RD,
2826 	    &priv->pkstats.tx_broadcast_packets, 0, "TX Broadcast Packets");
2827 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "tx_errors", CTLFLAG_RD,
2828 	    &priv->pkstats.tx_errors, 0, "TX Errors");
2829 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "tx_dropped", CTLFLAG_RD,
2830 	    &priv->pkstats.tx_dropped, 0, "TX Dropped");
2831 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "tx_lt_64_bytes_packets", CTLFLAG_RD,
2832 	    &priv->pkstats.tx_lt_64_bytes_packets, 0, "TX Less Then 64 Bytes Packets");
2833 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "tx_127_bytes_packets", CTLFLAG_RD,
2834 	    &priv->pkstats.tx_127_bytes_packets, 0, "TX 127 Bytes Packets");
2835 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "tx_255_bytes_packets", CTLFLAG_RD,
2836 	    &priv->pkstats.tx_255_bytes_packets, 0, "TX 255 Bytes Packets");
2837 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "tx_511_bytes_packets", CTLFLAG_RD,
2838 	    &priv->pkstats.tx_511_bytes_packets, 0, "TX 511 Bytes Packets");
2839 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "tx_1023_bytes_packets", CTLFLAG_RD,
2840 	    &priv->pkstats.tx_1023_bytes_packets, 0, "TX 1023 Bytes Packets");
2841 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "tx_1518_bytes_packets", CTLFLAG_RD,
2842 	    &priv->pkstats.tx_1518_bytes_packets, 0, "TX 1518 Bytes Packets");
2843 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "tx_1522_bytes_packets", CTLFLAG_RD,
2844 	    &priv->pkstats.tx_1522_bytes_packets, 0, "TX 1522 Bytes Packets");
2845 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "tx_1548_bytes_packets", CTLFLAG_RD,
2846 	    &priv->pkstats.tx_1548_bytes_packets, 0, "TX 1548 Bytes Packets");
2847 	SYSCTL_ADD_U64(ctx, node_list, OID_AUTO, "tx_gt_1548_bytes_packets", CTLFLAG_RD,
2848 	    &priv->pkstats.tx_gt_1548_bytes_packets, 0,
2849 	    "TX Greater Then 1548 Bytes Packets");
2850 
2851 	for (i = 0; i < priv->tx_ring_num; i++) {
2852 		tx_ring = priv->tx_ring[i];
2853 		snprintf(namebuf, sizeof(namebuf), "tx_ring%d", i);
2854 		ring_node = SYSCTL_ADD_NODE(ctx, node_list, OID_AUTO, namebuf,
2855 		    CTLFLAG_RD, NULL, "TX Ring");
2856 		ring_list = SYSCTL_CHILDREN(ring_node);
2857 		SYSCTL_ADD_U64(ctx, ring_list, OID_AUTO, "packets",
2858 		    CTLFLAG_RD, &tx_ring->packets, 0, "TX packets");
2859 		SYSCTL_ADD_U64(ctx, ring_list, OID_AUTO, "bytes",
2860 		    CTLFLAG_RD, &tx_ring->bytes, 0, "TX bytes");
2861 		SYSCTL_ADD_U64(ctx, ring_list, OID_AUTO, "tso_packets",
2862 		    CTLFLAG_RD, &tx_ring->tso_packets, 0, "TSO packets");
2863 		SYSCTL_ADD_U64(ctx, ring_list, OID_AUTO, "defrag_attempts",
2864 		    CTLFLAG_RD, &tx_ring->defrag_attempts, 0,
2865 		    "Oversized chains defragged");
2866 	}
2867 
2868 	for (i = 0; i < priv->rx_ring_num; i++) {
2869 		rx_ring = priv->rx_ring[i];
2870 		snprintf(namebuf, sizeof(namebuf), "rx_ring%d", i);
2871 		ring_node = SYSCTL_ADD_NODE(ctx, node_list, OID_AUTO, namebuf,
2872 		    CTLFLAG_RD, NULL, "RX Ring");
2873 		ring_list = SYSCTL_CHILDREN(ring_node);
2874 		SYSCTL_ADD_U64(ctx, ring_list, OID_AUTO, "packets",
2875 		    CTLFLAG_RD, &rx_ring->packets, 0, "RX packets");
2876 		SYSCTL_ADD_U64(ctx, ring_list, OID_AUTO, "bytes",
2877 		    CTLFLAG_RD, &rx_ring->bytes, 0, "RX bytes");
2878 		SYSCTL_ADD_U64(ctx, ring_list, OID_AUTO, "error",
2879 		    CTLFLAG_RD, &rx_ring->errors, 0, "RX soft errors");
2880 	}
2881 }
2882