1 /*
2  * Copyright (c) 2007 Mellanox Technologies. All rights reserved.
3  *
4  * This software is available to you under a choice of one of two
5  * licenses.  You may choose to be licensed under the terms of the GNU
6  * General Public License (GPL) Version 2, available from the file
7  * COPYING in the main directory of this source tree, or the
8  * OpenIB.org BSD license below:
9  *
10  *     Redistribution and use in source and binary forms, with or
11  *     without modification, are permitted provided that the following
12  *     conditions are met:
13  *
14  *      - Redistributions of source code must retain the above
15  *        copyright notice, this list of conditions and the following
16  *        disclaimer.
17  *
18  *      - Redistributions in binary form must reproduce the above
19  *        copyright notice, this list of conditions and the following
20  *        disclaimer in the documentation and/or other materials
21  *        provided with the distribution.
22  *
23  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
24  * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
25  * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
26  * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
27  * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
28  * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
29  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
30  * SOFTWARE.
31  *
32  */
33 
34 #include <linux/etherdevice.h>
35 #include <linux/tcp.h>
36 #include <linux/if_vlan.h>
37 #include <linux/delay.h>
38 #include <linux/slab.h>
39 #include <linux/hash.h>
40 #include <net/ip.h>
41 
42 #include <linux/mlx4/driver.h>
43 #include <linux/mlx4/device.h>
44 #include <linux/mlx4/cmd.h>
45 #include <linux/mlx4/cq.h>
46 
47 #include "mlx4_en.h"
48 #include "en_port.h"
49 
50 int mlx4_en_setup_tc(struct net_device *dev, u8 up)
51 {
52 	struct mlx4_en_priv *priv = netdev_priv(dev);
53 	int i;
54 	unsigned int offset = 0;
55 
56 	if (up && up != MLX4_EN_NUM_UP)
57 		return -EINVAL;
58 
59 	netdev_set_num_tc(dev, up);
60 
61 	/* Partition Tx queues evenly amongst UP's */
62 	for (i = 0; i < up; i++) {
63 		netdev_set_tc_queue(dev, i, priv->num_tx_rings_p_up, offset);
64 		offset += priv->num_tx_rings_p_up;
65 	}
66 
67 	return 0;
68 }
69 
70 #ifdef CONFIG_RFS_ACCEL
71 
72 struct mlx4_en_filter {
73 	struct list_head next;
74 	struct work_struct work;
75 
76 	__be32 src_ip;
77 	__be32 dst_ip;
78 	__be16 src_port;
79 	__be16 dst_port;
80 
81 	int rxq_index;
82 	struct mlx4_en_priv *priv;
83 	u32 flow_id;			/* RFS infrastructure id */
84 	int id;				/* mlx4_en driver id */
85 	u64 reg_id;			/* Flow steering API id */
86 	u8 activated;			/* Used to prevent expiry before filter
87 					 * is attached
88 					 */
89 	struct hlist_node filter_chain;
90 };
91 
92 static void mlx4_en_filter_rfs_expire(struct mlx4_en_priv *priv);
93 
94 static void mlx4_en_filter_work(struct work_struct *work)
95 {
96 	struct mlx4_en_filter *filter = container_of(work,
97 						     struct mlx4_en_filter,
98 						     work);
99 	struct mlx4_en_priv *priv = filter->priv;
100 	struct mlx4_spec_list spec_tcp = {
101 		.id = MLX4_NET_TRANS_RULE_ID_TCP,
102 		{
103 			.tcp_udp = {
104 				.dst_port = filter->dst_port,
105 				.dst_port_msk = (__force __be16)-1,
106 				.src_port = filter->src_port,
107 				.src_port_msk = (__force __be16)-1,
108 			},
109 		},
110 	};
111 	struct mlx4_spec_list spec_ip = {
112 		.id = MLX4_NET_TRANS_RULE_ID_IPV4,
113 		{
114 			.ipv4 = {
115 				.dst_ip = filter->dst_ip,
116 				.dst_ip_msk = (__force __be32)-1,
117 				.src_ip = filter->src_ip,
118 				.src_ip_msk = (__force __be32)-1,
119 			},
120 		},
121 	};
122 	struct mlx4_spec_list spec_eth = {
123 		.id = MLX4_NET_TRANS_RULE_ID_ETH,
124 	};
125 	struct mlx4_net_trans_rule rule = {
126 		.list = LIST_HEAD_INIT(rule.list),
127 		.queue_mode = MLX4_NET_TRANS_Q_LIFO,
128 		.exclusive = 1,
129 		.allow_loopback = 1,
130 		.promisc_mode = MLX4_FS_PROMISC_NONE,
131 		.port = priv->port,
132 		.priority = MLX4_DOMAIN_RFS,
133 	};
134 	int rc;
135 	__be64 mac;
136 	__be64 mac_mask = cpu_to_be64(MLX4_MAC_MASK << 16);
137 
138 	list_add_tail(&spec_eth.list, &rule.list);
139 	list_add_tail(&spec_ip.list, &rule.list);
140 	list_add_tail(&spec_tcp.list, &rule.list);
141 
142 	mac = cpu_to_be64((priv->mac & MLX4_MAC_MASK) << 16);
143 
144 	rule.qpn = priv->rss_map.qps[filter->rxq_index].qpn;
145 	memcpy(spec_eth.eth.dst_mac, &mac, ETH_ALEN);
146 	memcpy(spec_eth.eth.dst_mac_msk, &mac_mask, ETH_ALEN);
147 
148 	filter->activated = 0;
149 
150 	if (filter->reg_id) {
151 		rc = mlx4_flow_detach(priv->mdev->dev, filter->reg_id);
152 		if (rc && rc != -ENOENT)
153 			en_err(priv, "Error detaching flow. rc = %d\n", rc);
154 	}
155 
156 	rc = mlx4_flow_attach(priv->mdev->dev, &rule, &filter->reg_id);
157 	if (rc)
158 		en_err(priv, "Error attaching flow. err = %d\n", rc);
159 
160 	mlx4_en_filter_rfs_expire(priv);
161 
162 	filter->activated = 1;
163 }
164 
165 static inline struct hlist_head *
166 filter_hash_bucket(struct mlx4_en_priv *priv, __be32 src_ip, __be32 dst_ip,
167 		   __be16 src_port, __be16 dst_port)
168 {
169 	unsigned long l;
170 	int bucket_idx;
171 
172 	l = (__force unsigned long)src_port |
173 	    ((__force unsigned long)dst_port << 2);
174 	l ^= (__force unsigned long)(src_ip ^ dst_ip);
175 
176 	bucket_idx = hash_long(l, MLX4_EN_FILTER_HASH_SHIFT);
177 
178 	return &priv->filter_hash[bucket_idx];
179 }
180 
181 static struct mlx4_en_filter *
182 mlx4_en_filter_alloc(struct mlx4_en_priv *priv, int rxq_index, __be32 src_ip,
183 		     __be32 dst_ip, __be16 src_port, __be16 dst_port,
184 		     u32 flow_id)
185 {
186 	struct mlx4_en_filter *filter = NULL;
187 
188 	filter = kzalloc(sizeof(struct mlx4_en_filter), GFP_ATOMIC);
189 	if (!filter)
190 		return NULL;
191 
192 	filter->priv = priv;
193 	filter->rxq_index = rxq_index;
194 	INIT_WORK(&filter->work, mlx4_en_filter_work);
195 
196 	filter->src_ip = src_ip;
197 	filter->dst_ip = dst_ip;
198 	filter->src_port = src_port;
199 	filter->dst_port = dst_port;
200 
201 	filter->flow_id = flow_id;
202 
203 	filter->id = priv->last_filter_id++ % RPS_NO_FILTER;
204 
205 	list_add_tail(&filter->next, &priv->filters);
206 	hlist_add_head(&filter->filter_chain,
207 		       filter_hash_bucket(priv, src_ip, dst_ip, src_port,
208 					  dst_port));
209 
210 	return filter;
211 }
212 
213 static void mlx4_en_filter_free(struct mlx4_en_filter *filter)
214 {
215 	struct mlx4_en_priv *priv = filter->priv;
216 	int rc;
217 
218 	list_del(&filter->next);
219 
220 	rc = mlx4_flow_detach(priv->mdev->dev, filter->reg_id);
221 	if (rc && rc != -ENOENT)
222 		en_err(priv, "Error detaching flow. rc = %d\n", rc);
223 
224 	kfree(filter);
225 }
226 
227 static inline struct mlx4_en_filter *
228 mlx4_en_filter_find(struct mlx4_en_priv *priv, __be32 src_ip, __be32 dst_ip,
229 		    __be16 src_port, __be16 dst_port)
230 {
231 	struct hlist_node *elem;
232 	struct mlx4_en_filter *filter;
233 	struct mlx4_en_filter *ret = NULL;
234 
235 	hlist_for_each_entry(filter, elem,
236 			     filter_hash_bucket(priv, src_ip, dst_ip,
237 						src_port, dst_port),
238 			     filter_chain) {
239 		if (filter->src_ip == src_ip &&
240 		    filter->dst_ip == dst_ip &&
241 		    filter->src_port == src_port &&
242 		    filter->dst_port == dst_port) {
243 			ret = filter;
244 			break;
245 		}
246 	}
247 
248 	return ret;
249 }
250 
251 static int
252 mlx4_en_filter_rfs(struct net_device *net_dev, const struct sk_buff *skb,
253 		   u16 rxq_index, u32 flow_id)
254 {
255 	struct mlx4_en_priv *priv = netdev_priv(net_dev);
256 	struct mlx4_en_filter *filter;
257 	const struct iphdr *ip;
258 	const __be16 *ports;
259 	__be32 src_ip;
260 	__be32 dst_ip;
261 	__be16 src_port;
262 	__be16 dst_port;
263 	int nhoff = skb_network_offset(skb);
264 	int ret = 0;
265 
266 	if (skb->protocol != htons(ETH_P_IP))
267 		return -EPROTONOSUPPORT;
268 
269 	ip = (const struct iphdr *)(skb->data + nhoff);
270 	if (ip_is_fragment(ip))
271 		return -EPROTONOSUPPORT;
272 
273 	ports = (const __be16 *)(skb->data + nhoff + 4 * ip->ihl);
274 
275 	src_ip = ip->saddr;
276 	dst_ip = ip->daddr;
277 	src_port = ports[0];
278 	dst_port = ports[1];
279 
280 	if (ip->protocol != IPPROTO_TCP)
281 		return -EPROTONOSUPPORT;
282 
283 	spin_lock_bh(&priv->filters_lock);
284 	filter = mlx4_en_filter_find(priv, src_ip, dst_ip, src_port, dst_port);
285 	if (filter) {
286 		if (filter->rxq_index == rxq_index)
287 			goto out;
288 
289 		filter->rxq_index = rxq_index;
290 	} else {
291 		filter = mlx4_en_filter_alloc(priv, rxq_index,
292 					      src_ip, dst_ip,
293 					      src_port, dst_port, flow_id);
294 		if (!filter) {
295 			ret = -ENOMEM;
296 			goto err;
297 		}
298 	}
299 
300 	queue_work(priv->mdev->workqueue, &filter->work);
301 
302 out:
303 	ret = filter->id;
304 err:
305 	spin_unlock_bh(&priv->filters_lock);
306 
307 	return ret;
308 }
309 
310 void mlx4_en_cleanup_filters(struct mlx4_en_priv *priv,
311 			     struct mlx4_en_rx_ring *rx_ring)
312 {
313 	struct mlx4_en_filter *filter, *tmp;
314 	LIST_HEAD(del_list);
315 
316 	spin_lock_bh(&priv->filters_lock);
317 	list_for_each_entry_safe(filter, tmp, &priv->filters, next) {
318 		list_move(&filter->next, &del_list);
319 		hlist_del(&filter->filter_chain);
320 	}
321 	spin_unlock_bh(&priv->filters_lock);
322 
323 	list_for_each_entry_safe(filter, tmp, &del_list, next) {
324 		cancel_work_sync(&filter->work);
325 		mlx4_en_filter_free(filter);
326 	}
327 }
328 
329 static void mlx4_en_filter_rfs_expire(struct mlx4_en_priv *priv)
330 {
331 	struct mlx4_en_filter *filter = NULL, *tmp, *last_filter = NULL;
332 	LIST_HEAD(del_list);
333 	int i = 0;
334 
335 	spin_lock_bh(&priv->filters_lock);
336 	list_for_each_entry_safe(filter, tmp, &priv->filters, next) {
337 		if (i > MLX4_EN_FILTER_EXPIRY_QUOTA)
338 			break;
339 
340 		if (filter->activated &&
341 		    !work_pending(&filter->work) &&
342 		    rps_may_expire_flow(priv->dev,
343 					filter->rxq_index, filter->flow_id,
344 					filter->id)) {
345 			list_move(&filter->next, &del_list);
346 			hlist_del(&filter->filter_chain);
347 		} else
348 			last_filter = filter;
349 
350 		i++;
351 	}
352 
353 	if (last_filter && (&last_filter->next != priv->filters.next))
354 		list_move(&priv->filters, &last_filter->next);
355 
356 	spin_unlock_bh(&priv->filters_lock);
357 
358 	list_for_each_entry_safe(filter, tmp, &del_list, next)
359 		mlx4_en_filter_free(filter);
360 }
361 #endif
362 
363 static int mlx4_en_vlan_rx_add_vid(struct net_device *dev, unsigned short vid)
364 {
365 	struct mlx4_en_priv *priv = netdev_priv(dev);
366 	struct mlx4_en_dev *mdev = priv->mdev;
367 	int err;
368 	int idx;
369 
370 	en_dbg(HW, priv, "adding VLAN:%d\n", vid);
371 
372 	set_bit(vid, priv->active_vlans);
373 
374 	/* Add VID to port VLAN filter */
375 	mutex_lock(&mdev->state_lock);
376 	if (mdev->device_up && priv->port_up) {
377 		err = mlx4_SET_VLAN_FLTR(mdev->dev, priv);
378 		if (err)
379 			en_err(priv, "Failed configuring VLAN filter\n");
380 	}
381 	if (mlx4_register_vlan(mdev->dev, priv->port, vid, &idx))
382 		en_err(priv, "failed adding vlan %d\n", vid);
383 	mutex_unlock(&mdev->state_lock);
384 
385 	return 0;
386 }
387 
388 static int mlx4_en_vlan_rx_kill_vid(struct net_device *dev, unsigned short vid)
389 {
390 	struct mlx4_en_priv *priv = netdev_priv(dev);
391 	struct mlx4_en_dev *mdev = priv->mdev;
392 	int err;
393 	int idx;
394 
395 	en_dbg(HW, priv, "Killing VID:%d\n", vid);
396 
397 	clear_bit(vid, priv->active_vlans);
398 
399 	/* Remove VID from port VLAN filter */
400 	mutex_lock(&mdev->state_lock);
401 	if (!mlx4_find_cached_vlan(mdev->dev, priv->port, vid, &idx))
402 		mlx4_unregister_vlan(mdev->dev, priv->port, idx);
403 	else
404 		en_err(priv, "could not find vid %d in cache\n", vid);
405 
406 	if (mdev->device_up && priv->port_up) {
407 		err = mlx4_SET_VLAN_FLTR(mdev->dev, priv);
408 		if (err)
409 			en_err(priv, "Failed configuring VLAN filter\n");
410 	}
411 	mutex_unlock(&mdev->state_lock);
412 
413 	return 0;
414 }
415 
416 u64 mlx4_en_mac_to_u64(u8 *addr)
417 {
418 	u64 mac = 0;
419 	int i;
420 
421 	for (i = 0; i < ETH_ALEN; i++) {
422 		mac <<= 8;
423 		mac |= addr[i];
424 	}
425 	return mac;
426 }
427 
428 static int mlx4_en_set_mac(struct net_device *dev, void *addr)
429 {
430 	struct mlx4_en_priv *priv = netdev_priv(dev);
431 	struct mlx4_en_dev *mdev = priv->mdev;
432 	struct sockaddr *saddr = addr;
433 
434 	if (!is_valid_ether_addr(saddr->sa_data))
435 		return -EADDRNOTAVAIL;
436 
437 	memcpy(dev->dev_addr, saddr->sa_data, ETH_ALEN);
438 	priv->mac = mlx4_en_mac_to_u64(dev->dev_addr);
439 	queue_work(mdev->workqueue, &priv->mac_task);
440 	return 0;
441 }
442 
443 static void mlx4_en_do_set_mac(struct work_struct *work)
444 {
445 	struct mlx4_en_priv *priv = container_of(work, struct mlx4_en_priv,
446 						 mac_task);
447 	struct mlx4_en_dev *mdev = priv->mdev;
448 	int err = 0;
449 
450 	mutex_lock(&mdev->state_lock);
451 	if (priv->port_up) {
452 		/* Remove old MAC and insert the new one */
453 		err = mlx4_replace_mac(mdev->dev, priv->port,
454 				       priv->base_qpn, priv->mac);
455 		if (err)
456 			en_err(priv, "Failed changing HW MAC address\n");
457 	} else
458 		en_dbg(HW, priv, "Port is down while "
459 				 "registering mac, exiting...\n");
460 
461 	mutex_unlock(&mdev->state_lock);
462 }
463 
464 static void mlx4_en_clear_list(struct net_device *dev)
465 {
466 	struct mlx4_en_priv *priv = netdev_priv(dev);
467 	struct mlx4_en_mc_list *tmp, *mc_to_del;
468 
469 	list_for_each_entry_safe(mc_to_del, tmp, &priv->mc_list, list) {
470 		list_del(&mc_to_del->list);
471 		kfree(mc_to_del);
472 	}
473 }
474 
475 static void mlx4_en_cache_mclist(struct net_device *dev)
476 {
477 	struct mlx4_en_priv *priv = netdev_priv(dev);
478 	struct netdev_hw_addr *ha;
479 	struct mlx4_en_mc_list *tmp;
480 
481 	mlx4_en_clear_list(dev);
482 	netdev_for_each_mc_addr(ha, dev) {
483 		tmp = kzalloc(sizeof(struct mlx4_en_mc_list), GFP_ATOMIC);
484 		if (!tmp) {
485 			en_err(priv, "failed to allocate multicast list\n");
486 			mlx4_en_clear_list(dev);
487 			return;
488 		}
489 		memcpy(tmp->addr, ha->addr, ETH_ALEN);
490 		list_add_tail(&tmp->list, &priv->mc_list);
491 	}
492 }
493 
494 static void update_mclist_flags(struct mlx4_en_priv *priv,
495 				struct list_head *dst,
496 				struct list_head *src)
497 {
498 	struct mlx4_en_mc_list *dst_tmp, *src_tmp, *new_mc;
499 	bool found;
500 
501 	/* Find all the entries that should be removed from dst,
502 	 * These are the entries that are not found in src
503 	 */
504 	list_for_each_entry(dst_tmp, dst, list) {
505 		found = false;
506 		list_for_each_entry(src_tmp, src, list) {
507 			if (!memcmp(dst_tmp->addr, src_tmp->addr, ETH_ALEN)) {
508 				found = true;
509 				break;
510 			}
511 		}
512 		if (!found)
513 			dst_tmp->action = MCLIST_REM;
514 	}
515 
516 	/* Add entries that exist in src but not in dst
517 	 * mark them as need to add
518 	 */
519 	list_for_each_entry(src_tmp, src, list) {
520 		found = false;
521 		list_for_each_entry(dst_tmp, dst, list) {
522 			if (!memcmp(dst_tmp->addr, src_tmp->addr, ETH_ALEN)) {
523 				dst_tmp->action = MCLIST_NONE;
524 				found = true;
525 				break;
526 			}
527 		}
528 		if (!found) {
529 			new_mc = kmalloc(sizeof(struct mlx4_en_mc_list),
530 					 GFP_KERNEL);
531 			if (!new_mc) {
532 				en_err(priv, "Failed to allocate current multicast list\n");
533 				return;
534 			}
535 			memcpy(new_mc, src_tmp,
536 			       sizeof(struct mlx4_en_mc_list));
537 			new_mc->action = MCLIST_ADD;
538 			list_add_tail(&new_mc->list, dst);
539 		}
540 	}
541 }
542 
543 static void mlx4_en_set_multicast(struct net_device *dev)
544 {
545 	struct mlx4_en_priv *priv = netdev_priv(dev);
546 
547 	if (!priv->port_up)
548 		return;
549 
550 	queue_work(priv->mdev->workqueue, &priv->mcast_task);
551 }
552 
553 static void mlx4_en_do_set_multicast(struct work_struct *work)
554 {
555 	struct mlx4_en_priv *priv = container_of(work, struct mlx4_en_priv,
556 						 mcast_task);
557 	struct mlx4_en_dev *mdev = priv->mdev;
558 	struct net_device *dev = priv->dev;
559 	struct mlx4_en_mc_list *mclist, *tmp;
560 	u64 mcast_addr = 0;
561 	u8 mc_list[16] = {0};
562 	int err = 0;
563 
564 	mutex_lock(&mdev->state_lock);
565 	if (!mdev->device_up) {
566 		en_dbg(HW, priv, "Card is not up, "
567 				 "ignoring multicast change.\n");
568 		goto out;
569 	}
570 	if (!priv->port_up) {
571 		en_dbg(HW, priv, "Port is down, "
572 				 "ignoring  multicast change.\n");
573 		goto out;
574 	}
575 
576 	if (!netif_carrier_ok(dev)) {
577 		if (!mlx4_en_QUERY_PORT(mdev, priv->port)) {
578 			if (priv->port_state.link_state) {
579 				priv->last_link_state = MLX4_DEV_EVENT_PORT_UP;
580 				netif_carrier_on(dev);
581 				en_dbg(LINK, priv, "Link Up\n");
582 			}
583 		}
584 	}
585 
586 	/*
587 	 * Promsicuous mode: disable all filters
588 	 */
589 
590 	if (dev->flags & IFF_PROMISC) {
591 		if (!(priv->flags & MLX4_EN_FLAG_PROMISC)) {
592 			if (netif_msg_rx_status(priv))
593 				en_warn(priv, "Entering promiscuous mode\n");
594 			priv->flags |= MLX4_EN_FLAG_PROMISC;
595 
596 			/* Enable promiscouos mode */
597 			switch (mdev->dev->caps.steering_mode) {
598 			case MLX4_STEERING_MODE_DEVICE_MANAGED:
599 				err = mlx4_flow_steer_promisc_add(mdev->dev,
600 								  priv->port,
601 								  priv->base_qpn,
602 								  MLX4_FS_PROMISC_UPLINK);
603 				if (err)
604 					en_err(priv, "Failed enabling promiscuous mode\n");
605 				priv->flags |= MLX4_EN_FLAG_MC_PROMISC;
606 				break;
607 
608 			case MLX4_STEERING_MODE_B0:
609 				err = mlx4_unicast_promisc_add(mdev->dev,
610 							       priv->base_qpn,
611 							       priv->port);
612 				if (err)
613 					en_err(priv, "Failed enabling unicast promiscuous mode\n");
614 
615 				/* Add the default qp number as multicast
616 				 * promisc
617 				 */
618 				if (!(priv->flags & MLX4_EN_FLAG_MC_PROMISC)) {
619 					err = mlx4_multicast_promisc_add(mdev->dev,
620 									 priv->base_qpn,
621 									 priv->port);
622 					if (err)
623 						en_err(priv, "Failed enabling multicast promiscuous mode\n");
624 					priv->flags |= MLX4_EN_FLAG_MC_PROMISC;
625 				}
626 				break;
627 
628 			case MLX4_STEERING_MODE_A0:
629 				err = mlx4_SET_PORT_qpn_calc(mdev->dev,
630 							     priv->port,
631 							     priv->base_qpn,
632 							     1);
633 				if (err)
634 					en_err(priv, "Failed enabling promiscuous mode\n");
635 				break;
636 			}
637 
638 			/* Disable port multicast filter (unconditionally) */
639 			err = mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, 0,
640 						  0, MLX4_MCAST_DISABLE);
641 			if (err)
642 				en_err(priv, "Failed disabling "
643 					     "multicast filter\n");
644 
645 			/* Disable port VLAN filter */
646 			err = mlx4_SET_VLAN_FLTR(mdev->dev, priv);
647 			if (err)
648 				en_err(priv, "Failed disabling VLAN filter\n");
649 		}
650 		goto out;
651 	}
652 
653 	/*
654 	 * Not in promiscuous mode
655 	 */
656 
657 	if (priv->flags & MLX4_EN_FLAG_PROMISC) {
658 		if (netif_msg_rx_status(priv))
659 			en_warn(priv, "Leaving promiscuous mode\n");
660 		priv->flags &= ~MLX4_EN_FLAG_PROMISC;
661 
662 		/* Disable promiscouos mode */
663 		switch (mdev->dev->caps.steering_mode) {
664 		case MLX4_STEERING_MODE_DEVICE_MANAGED:
665 			err = mlx4_flow_steer_promisc_remove(mdev->dev,
666 							     priv->port,
667 							     MLX4_FS_PROMISC_UPLINK);
668 			if (err)
669 				en_err(priv, "Failed disabling promiscuous mode\n");
670 			priv->flags &= ~MLX4_EN_FLAG_MC_PROMISC;
671 			break;
672 
673 		case MLX4_STEERING_MODE_B0:
674 			err = mlx4_unicast_promisc_remove(mdev->dev,
675 							  priv->base_qpn,
676 							  priv->port);
677 			if (err)
678 				en_err(priv, "Failed disabling unicast promiscuous mode\n");
679 			/* Disable Multicast promisc */
680 			if (priv->flags & MLX4_EN_FLAG_MC_PROMISC) {
681 				err = mlx4_multicast_promisc_remove(mdev->dev,
682 								    priv->base_qpn,
683 								    priv->port);
684 				if (err)
685 					en_err(priv, "Failed disabling multicast promiscuous mode\n");
686 				priv->flags &= ~MLX4_EN_FLAG_MC_PROMISC;
687 			}
688 			break;
689 
690 		case MLX4_STEERING_MODE_A0:
691 			err = mlx4_SET_PORT_qpn_calc(mdev->dev,
692 						     priv->port,
693 						     priv->base_qpn, 0);
694 			if (err)
695 				en_err(priv, "Failed disabling promiscuous mode\n");
696 			break;
697 		}
698 
699 		/* Enable port VLAN filter */
700 		err = mlx4_SET_VLAN_FLTR(mdev->dev, priv);
701 		if (err)
702 			en_err(priv, "Failed enabling VLAN filter\n");
703 	}
704 
705 	/* Enable/disable the multicast filter according to IFF_ALLMULTI */
706 	if (dev->flags & IFF_ALLMULTI) {
707 		err = mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, 0,
708 					  0, MLX4_MCAST_DISABLE);
709 		if (err)
710 			en_err(priv, "Failed disabling multicast filter\n");
711 
712 		/* Add the default qp number as multicast promisc */
713 		if (!(priv->flags & MLX4_EN_FLAG_MC_PROMISC)) {
714 			switch (mdev->dev->caps.steering_mode) {
715 			case MLX4_STEERING_MODE_DEVICE_MANAGED:
716 				err = mlx4_flow_steer_promisc_add(mdev->dev,
717 								  priv->port,
718 								  priv->base_qpn,
719 								  MLX4_FS_PROMISC_ALL_MULTI);
720 				break;
721 
722 			case MLX4_STEERING_MODE_B0:
723 				err = mlx4_multicast_promisc_add(mdev->dev,
724 								 priv->base_qpn,
725 								 priv->port);
726 				break;
727 
728 			case MLX4_STEERING_MODE_A0:
729 				break;
730 			}
731 			if (err)
732 				en_err(priv, "Failed entering multicast promisc mode\n");
733 			priv->flags |= MLX4_EN_FLAG_MC_PROMISC;
734 		}
735 	} else {
736 		/* Disable Multicast promisc */
737 		if (priv->flags & MLX4_EN_FLAG_MC_PROMISC) {
738 			switch (mdev->dev->caps.steering_mode) {
739 			case MLX4_STEERING_MODE_DEVICE_MANAGED:
740 				err = mlx4_flow_steer_promisc_remove(mdev->dev,
741 								     priv->port,
742 								     MLX4_FS_PROMISC_ALL_MULTI);
743 				break;
744 
745 			case MLX4_STEERING_MODE_B0:
746 				err = mlx4_multicast_promisc_remove(mdev->dev,
747 								    priv->base_qpn,
748 								    priv->port);
749 				break;
750 
751 			case MLX4_STEERING_MODE_A0:
752 				break;
753 			}
754 			if (err)
755 				en_err(priv, "Failed disabling multicast promiscuous mode\n");
756 			priv->flags &= ~MLX4_EN_FLAG_MC_PROMISC;
757 		}
758 
759 		err = mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, 0,
760 					  0, MLX4_MCAST_DISABLE);
761 		if (err)
762 			en_err(priv, "Failed disabling multicast filter\n");
763 
764 		/* Flush mcast filter and init it with broadcast address */
765 		mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, ETH_BCAST,
766 				    1, MLX4_MCAST_CONFIG);
767 
768 		/* Update multicast list - we cache all addresses so they won't
769 		 * change while HW is updated holding the command semaphor */
770 		netif_tx_lock_bh(dev);
771 		mlx4_en_cache_mclist(dev);
772 		netif_tx_unlock_bh(dev);
773 		list_for_each_entry(mclist, &priv->mc_list, list) {
774 			mcast_addr = mlx4_en_mac_to_u64(mclist->addr);
775 			mlx4_SET_MCAST_FLTR(mdev->dev, priv->port,
776 					    mcast_addr, 0, MLX4_MCAST_CONFIG);
777 		}
778 		err = mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, 0,
779 					  0, MLX4_MCAST_ENABLE);
780 		if (err)
781 			en_err(priv, "Failed enabling multicast filter\n");
782 
783 		update_mclist_flags(priv, &priv->curr_list, &priv->mc_list);
784 		list_for_each_entry_safe(mclist, tmp, &priv->curr_list, list) {
785 			if (mclist->action == MCLIST_REM) {
786 				/* detach this address and delete from list */
787 				memcpy(&mc_list[10], mclist->addr, ETH_ALEN);
788 				mc_list[5] = priv->port;
789 				err = mlx4_multicast_detach(mdev->dev,
790 							    &priv->rss_map.indir_qp,
791 							    mc_list,
792 							    MLX4_PROT_ETH,
793 							    mclist->reg_id);
794 				if (err)
795 					en_err(priv, "Fail to detach multicast address\n");
796 
797 				/* remove from list */
798 				list_del(&mclist->list);
799 				kfree(mclist);
800 			} else if (mclist->action == MCLIST_ADD) {
801 				/* attach the address */
802 				memcpy(&mc_list[10], mclist->addr, ETH_ALEN);
803 				/* needed for B0 steering support */
804 				mc_list[5] = priv->port;
805 				err = mlx4_multicast_attach(mdev->dev,
806 							    &priv->rss_map.indir_qp,
807 							    mc_list,
808 							    priv->port, 0,
809 							    MLX4_PROT_ETH,
810 							    &mclist->reg_id);
811 				if (err)
812 					en_err(priv, "Fail to attach multicast address\n");
813 
814 			}
815 		}
816 	}
817 out:
818 	mutex_unlock(&mdev->state_lock);
819 }
820 
821 #ifdef CONFIG_NET_POLL_CONTROLLER
822 static void mlx4_en_netpoll(struct net_device *dev)
823 {
824 	struct mlx4_en_priv *priv = netdev_priv(dev);
825 	struct mlx4_en_cq *cq;
826 	unsigned long flags;
827 	int i;
828 
829 	for (i = 0; i < priv->rx_ring_num; i++) {
830 		cq = &priv->rx_cq[i];
831 		spin_lock_irqsave(&cq->lock, flags);
832 		napi_synchronize(&cq->napi);
833 		mlx4_en_process_rx_cq(dev, cq, 0);
834 		spin_unlock_irqrestore(&cq->lock, flags);
835 	}
836 }
837 #endif
838 
839 static void mlx4_en_tx_timeout(struct net_device *dev)
840 {
841 	struct mlx4_en_priv *priv = netdev_priv(dev);
842 	struct mlx4_en_dev *mdev = priv->mdev;
843 
844 	if (netif_msg_timer(priv))
845 		en_warn(priv, "Tx timeout called on port:%d\n", priv->port);
846 
847 	priv->port_stats.tx_timeout++;
848 	en_dbg(DRV, priv, "Scheduling watchdog\n");
849 	queue_work(mdev->workqueue, &priv->watchdog_task);
850 }
851 
852 
853 static struct net_device_stats *mlx4_en_get_stats(struct net_device *dev)
854 {
855 	struct mlx4_en_priv *priv = netdev_priv(dev);
856 
857 	spin_lock_bh(&priv->stats_lock);
858 	memcpy(&priv->ret_stats, &priv->stats, sizeof(priv->stats));
859 	spin_unlock_bh(&priv->stats_lock);
860 
861 	return &priv->ret_stats;
862 }
863 
864 static void mlx4_en_set_default_moderation(struct mlx4_en_priv *priv)
865 {
866 	struct mlx4_en_cq *cq;
867 	int i;
868 
869 	/* If we haven't received a specific coalescing setting
870 	 * (module param), we set the moderation parameters as follows:
871 	 * - moder_cnt is set to the number of mtu sized packets to
872 	 *   satisfy our coalescing target.
873 	 * - moder_time is set to a fixed value.
874 	 */
875 	priv->rx_frames = MLX4_EN_RX_COAL_TARGET;
876 	priv->rx_usecs = MLX4_EN_RX_COAL_TIME;
877 	priv->tx_frames = MLX4_EN_TX_COAL_PKTS;
878 	priv->tx_usecs = MLX4_EN_TX_COAL_TIME;
879 	en_dbg(INTR, priv, "Default coalesing params for mtu:%d - "
880 			   "rx_frames:%d rx_usecs:%d\n",
881 		 priv->dev->mtu, priv->rx_frames, priv->rx_usecs);
882 
883 	/* Setup cq moderation params */
884 	for (i = 0; i < priv->rx_ring_num; i++) {
885 		cq = &priv->rx_cq[i];
886 		cq->moder_cnt = priv->rx_frames;
887 		cq->moder_time = priv->rx_usecs;
888 		priv->last_moder_time[i] = MLX4_EN_AUTO_CONF;
889 		priv->last_moder_packets[i] = 0;
890 		priv->last_moder_bytes[i] = 0;
891 	}
892 
893 	for (i = 0; i < priv->tx_ring_num; i++) {
894 		cq = &priv->tx_cq[i];
895 		cq->moder_cnt = priv->tx_frames;
896 		cq->moder_time = priv->tx_usecs;
897 	}
898 
899 	/* Reset auto-moderation params */
900 	priv->pkt_rate_low = MLX4_EN_RX_RATE_LOW;
901 	priv->rx_usecs_low = MLX4_EN_RX_COAL_TIME_LOW;
902 	priv->pkt_rate_high = MLX4_EN_RX_RATE_HIGH;
903 	priv->rx_usecs_high = MLX4_EN_RX_COAL_TIME_HIGH;
904 	priv->sample_interval = MLX4_EN_SAMPLE_INTERVAL;
905 	priv->adaptive_rx_coal = 1;
906 	priv->last_moder_jiffies = 0;
907 	priv->last_moder_tx_packets = 0;
908 }
909 
910 static void mlx4_en_auto_moderation(struct mlx4_en_priv *priv)
911 {
912 	unsigned long period = (unsigned long) (jiffies - priv->last_moder_jiffies);
913 	struct mlx4_en_cq *cq;
914 	unsigned long packets;
915 	unsigned long rate;
916 	unsigned long avg_pkt_size;
917 	unsigned long rx_packets;
918 	unsigned long rx_bytes;
919 	unsigned long rx_pkt_diff;
920 	int moder_time;
921 	int ring, err;
922 
923 	if (!priv->adaptive_rx_coal || period < priv->sample_interval * HZ)
924 		return;
925 
926 	for (ring = 0; ring < priv->rx_ring_num; ring++) {
927 		spin_lock_bh(&priv->stats_lock);
928 		rx_packets = priv->rx_ring[ring].packets;
929 		rx_bytes = priv->rx_ring[ring].bytes;
930 		spin_unlock_bh(&priv->stats_lock);
931 
932 		rx_pkt_diff = ((unsigned long) (rx_packets -
933 				priv->last_moder_packets[ring]));
934 		packets = rx_pkt_diff;
935 		rate = packets * HZ / period;
936 		avg_pkt_size = packets ? ((unsigned long) (rx_bytes -
937 				priv->last_moder_bytes[ring])) / packets : 0;
938 
939 		/* Apply auto-moderation only when packet rate
940 		 * exceeds a rate that it matters */
941 		if (rate > (MLX4_EN_RX_RATE_THRESH / priv->rx_ring_num) &&
942 		    avg_pkt_size > MLX4_EN_AVG_PKT_SMALL) {
943 			if (rate < priv->pkt_rate_low)
944 				moder_time = priv->rx_usecs_low;
945 			else if (rate > priv->pkt_rate_high)
946 				moder_time = priv->rx_usecs_high;
947 			else
948 				moder_time = (rate - priv->pkt_rate_low) *
949 					(priv->rx_usecs_high - priv->rx_usecs_low) /
950 					(priv->pkt_rate_high - priv->pkt_rate_low) +
951 					priv->rx_usecs_low;
952 		} else {
953 			moder_time = priv->rx_usecs_low;
954 		}
955 
956 		if (moder_time != priv->last_moder_time[ring]) {
957 			priv->last_moder_time[ring] = moder_time;
958 			cq = &priv->rx_cq[ring];
959 			cq->moder_time = moder_time;
960 			err = mlx4_en_set_cq_moder(priv, cq);
961 			if (err)
962 				en_err(priv, "Failed modifying moderation "
963 					     "for cq:%d\n", ring);
964 		}
965 		priv->last_moder_packets[ring] = rx_packets;
966 		priv->last_moder_bytes[ring] = rx_bytes;
967 	}
968 
969 	priv->last_moder_jiffies = jiffies;
970 }
971 
972 static void mlx4_en_do_get_stats(struct work_struct *work)
973 {
974 	struct delayed_work *delay = to_delayed_work(work);
975 	struct mlx4_en_priv *priv = container_of(delay, struct mlx4_en_priv,
976 						 stats_task);
977 	struct mlx4_en_dev *mdev = priv->mdev;
978 	int err;
979 
980 	err = mlx4_en_DUMP_ETH_STATS(mdev, priv->port, 0);
981 	if (err)
982 		en_dbg(HW, priv, "Could not update stats\n");
983 
984 	mutex_lock(&mdev->state_lock);
985 	if (mdev->device_up) {
986 		if (priv->port_up)
987 			mlx4_en_auto_moderation(priv);
988 
989 		queue_delayed_work(mdev->workqueue, &priv->stats_task, STATS_DELAY);
990 	}
991 	if (mdev->mac_removed[MLX4_MAX_PORTS + 1 - priv->port]) {
992 		queue_work(mdev->workqueue, &priv->mac_task);
993 		mdev->mac_removed[MLX4_MAX_PORTS + 1 - priv->port] = 0;
994 	}
995 	mutex_unlock(&mdev->state_lock);
996 }
997 
998 static void mlx4_en_linkstate(struct work_struct *work)
999 {
1000 	struct mlx4_en_priv *priv = container_of(work, struct mlx4_en_priv,
1001 						 linkstate_task);
1002 	struct mlx4_en_dev *mdev = priv->mdev;
1003 	int linkstate = priv->link_state;
1004 
1005 	mutex_lock(&mdev->state_lock);
1006 	/* If observable port state changed set carrier state and
1007 	 * report to system log */
1008 	if (priv->last_link_state != linkstate) {
1009 		if (linkstate == MLX4_DEV_EVENT_PORT_DOWN) {
1010 			en_info(priv, "Link Down\n");
1011 			netif_carrier_off(priv->dev);
1012 		} else {
1013 			en_info(priv, "Link Up\n");
1014 			netif_carrier_on(priv->dev);
1015 		}
1016 	}
1017 	priv->last_link_state = linkstate;
1018 	mutex_unlock(&mdev->state_lock);
1019 }
1020 
1021 
1022 int mlx4_en_start_port(struct net_device *dev)
1023 {
1024 	struct mlx4_en_priv *priv = netdev_priv(dev);
1025 	struct mlx4_en_dev *mdev = priv->mdev;
1026 	struct mlx4_en_cq *cq;
1027 	struct mlx4_en_tx_ring *tx_ring;
1028 	int rx_index = 0;
1029 	int tx_index = 0;
1030 	int err = 0;
1031 	int i;
1032 	int j;
1033 	u8 mc_list[16] = {0};
1034 
1035 	if (priv->port_up) {
1036 		en_dbg(DRV, priv, "start port called while port already up\n");
1037 		return 0;
1038 	}
1039 
1040 	INIT_LIST_HEAD(&priv->mc_list);
1041 	INIT_LIST_HEAD(&priv->curr_list);
1042 
1043 	/* Calculate Rx buf size */
1044 	dev->mtu = min(dev->mtu, priv->max_mtu);
1045 	mlx4_en_calc_rx_buf(dev);
1046 	en_dbg(DRV, priv, "Rx buf size:%d\n", priv->rx_skb_size);
1047 
1048 	/* Configure rx cq's and rings */
1049 	err = mlx4_en_activate_rx_rings(priv);
1050 	if (err) {
1051 		en_err(priv, "Failed to activate RX rings\n");
1052 		return err;
1053 	}
1054 	for (i = 0; i < priv->rx_ring_num; i++) {
1055 		cq = &priv->rx_cq[i];
1056 
1057 		err = mlx4_en_activate_cq(priv, cq, i);
1058 		if (err) {
1059 			en_err(priv, "Failed activating Rx CQ\n");
1060 			goto cq_err;
1061 		}
1062 		for (j = 0; j < cq->size; j++)
1063 			cq->buf[j].owner_sr_opcode = MLX4_CQE_OWNER_MASK;
1064 		err = mlx4_en_set_cq_moder(priv, cq);
1065 		if (err) {
1066 			en_err(priv, "Failed setting cq moderation parameters");
1067 			mlx4_en_deactivate_cq(priv, cq);
1068 			goto cq_err;
1069 		}
1070 		mlx4_en_arm_cq(priv, cq);
1071 		priv->rx_ring[i].cqn = cq->mcq.cqn;
1072 		++rx_index;
1073 	}
1074 
1075 	/* Set qp number */
1076 	en_dbg(DRV, priv, "Getting qp number for port %d\n", priv->port);
1077 	err = mlx4_get_eth_qp(mdev->dev, priv->port,
1078 				priv->mac, &priv->base_qpn);
1079 	if (err) {
1080 		en_err(priv, "Failed getting eth qp\n");
1081 		goto cq_err;
1082 	}
1083 	mdev->mac_removed[priv->port] = 0;
1084 
1085 	err = mlx4_en_config_rss_steer(priv);
1086 	if (err) {
1087 		en_err(priv, "Failed configuring rss steering\n");
1088 		goto mac_err;
1089 	}
1090 
1091 	err = mlx4_en_create_drop_qp(priv);
1092 	if (err)
1093 		goto rss_err;
1094 
1095 	/* Configure tx cq's and rings */
1096 	for (i = 0; i < priv->tx_ring_num; i++) {
1097 		/* Configure cq */
1098 		cq = &priv->tx_cq[i];
1099 		err = mlx4_en_activate_cq(priv, cq, i);
1100 		if (err) {
1101 			en_err(priv, "Failed allocating Tx CQ\n");
1102 			goto tx_err;
1103 		}
1104 		err = mlx4_en_set_cq_moder(priv, cq);
1105 		if (err) {
1106 			en_err(priv, "Failed setting cq moderation parameters");
1107 			mlx4_en_deactivate_cq(priv, cq);
1108 			goto tx_err;
1109 		}
1110 		en_dbg(DRV, priv, "Resetting index of collapsed CQ:%d to -1\n", i);
1111 		cq->buf->wqe_index = cpu_to_be16(0xffff);
1112 
1113 		/* Configure ring */
1114 		tx_ring = &priv->tx_ring[i];
1115 		err = mlx4_en_activate_tx_ring(priv, tx_ring, cq->mcq.cqn,
1116 			i / priv->num_tx_rings_p_up);
1117 		if (err) {
1118 			en_err(priv, "Failed allocating Tx ring\n");
1119 			mlx4_en_deactivate_cq(priv, cq);
1120 			goto tx_err;
1121 		}
1122 		tx_ring->tx_queue = netdev_get_tx_queue(dev, i);
1123 
1124 		/* Arm CQ for TX completions */
1125 		mlx4_en_arm_cq(priv, cq);
1126 
1127 		/* Set initial ownership of all Tx TXBBs to SW (1) */
1128 		for (j = 0; j < tx_ring->buf_size; j += STAMP_STRIDE)
1129 			*((u32 *) (tx_ring->buf + j)) = 0xffffffff;
1130 		++tx_index;
1131 	}
1132 
1133 	/* Configure port */
1134 	err = mlx4_SET_PORT_general(mdev->dev, priv->port,
1135 				    priv->rx_skb_size + ETH_FCS_LEN,
1136 				    priv->prof->tx_pause,
1137 				    priv->prof->tx_ppp,
1138 				    priv->prof->rx_pause,
1139 				    priv->prof->rx_ppp);
1140 	if (err) {
1141 		en_err(priv, "Failed setting port general configurations "
1142 			     "for port %d, with error %d\n", priv->port, err);
1143 		goto tx_err;
1144 	}
1145 	/* Set default qp number */
1146 	err = mlx4_SET_PORT_qpn_calc(mdev->dev, priv->port, priv->base_qpn, 0);
1147 	if (err) {
1148 		en_err(priv, "Failed setting default qp numbers\n");
1149 		goto tx_err;
1150 	}
1151 
1152 	/* Init port */
1153 	en_dbg(HW, priv, "Initializing port\n");
1154 	err = mlx4_INIT_PORT(mdev->dev, priv->port);
1155 	if (err) {
1156 		en_err(priv, "Failed Initializing port\n");
1157 		goto tx_err;
1158 	}
1159 
1160 	/* Attach rx QP to bradcast address */
1161 	memset(&mc_list[10], 0xff, ETH_ALEN);
1162 	mc_list[5] = priv->port; /* needed for B0 steering support */
1163 	if (mlx4_multicast_attach(mdev->dev, &priv->rss_map.indir_qp, mc_list,
1164 				  priv->port, 0, MLX4_PROT_ETH,
1165 				  &priv->broadcast_id))
1166 		mlx4_warn(mdev, "Failed Attaching Broadcast\n");
1167 
1168 	/* Must redo promiscuous mode setup. */
1169 	priv->flags &= ~(MLX4_EN_FLAG_PROMISC | MLX4_EN_FLAG_MC_PROMISC);
1170 	if (mdev->dev->caps.steering_mode ==
1171 	    MLX4_STEERING_MODE_DEVICE_MANAGED) {
1172 		mlx4_flow_steer_promisc_remove(mdev->dev,
1173 					       priv->port,
1174 					       MLX4_FS_PROMISC_UPLINK);
1175 		mlx4_flow_steer_promisc_remove(mdev->dev,
1176 					       priv->port,
1177 					       MLX4_FS_PROMISC_ALL_MULTI);
1178 	}
1179 
1180 	/* Schedule multicast task to populate multicast list */
1181 	queue_work(mdev->workqueue, &priv->mcast_task);
1182 
1183 	mlx4_set_stats_bitmap(mdev->dev, &priv->stats_bitmap);
1184 
1185 	priv->port_up = true;
1186 	netif_tx_start_all_queues(dev);
1187 	return 0;
1188 
1189 tx_err:
1190 	while (tx_index--) {
1191 		mlx4_en_deactivate_tx_ring(priv, &priv->tx_ring[tx_index]);
1192 		mlx4_en_deactivate_cq(priv, &priv->tx_cq[tx_index]);
1193 	}
1194 	mlx4_en_destroy_drop_qp(priv);
1195 rss_err:
1196 	mlx4_en_release_rss_steer(priv);
1197 mac_err:
1198 	mlx4_put_eth_qp(mdev->dev, priv->port, priv->mac, priv->base_qpn);
1199 cq_err:
1200 	while (rx_index--)
1201 		mlx4_en_deactivate_cq(priv, &priv->rx_cq[rx_index]);
1202 	for (i = 0; i < priv->rx_ring_num; i++)
1203 		mlx4_en_deactivate_rx_ring(priv, &priv->rx_ring[i]);
1204 
1205 	return err; /* need to close devices */
1206 }
1207 
1208 
1209 void mlx4_en_stop_port(struct net_device *dev)
1210 {
1211 	struct mlx4_en_priv *priv = netdev_priv(dev);
1212 	struct mlx4_en_dev *mdev = priv->mdev;
1213 	struct mlx4_en_mc_list *mclist, *tmp;
1214 	int i;
1215 	u8 mc_list[16] = {0};
1216 
1217 	if (!priv->port_up) {
1218 		en_dbg(DRV, priv, "stop port called while port already down\n");
1219 		return;
1220 	}
1221 
1222 	/* Synchronize with tx routine */
1223 	netif_tx_lock_bh(dev);
1224 	netif_tx_stop_all_queues(dev);
1225 	netif_tx_unlock_bh(dev);
1226 
1227 	/* Set port as not active */
1228 	priv->port_up = false;
1229 
1230 	/* Detach All multicasts */
1231 	memset(&mc_list[10], 0xff, ETH_ALEN);
1232 	mc_list[5] = priv->port; /* needed for B0 steering support */
1233 	mlx4_multicast_detach(mdev->dev, &priv->rss_map.indir_qp, mc_list,
1234 			      MLX4_PROT_ETH, priv->broadcast_id);
1235 	list_for_each_entry(mclist, &priv->curr_list, list) {
1236 		memcpy(&mc_list[10], mclist->addr, ETH_ALEN);
1237 		mc_list[5] = priv->port;
1238 		mlx4_multicast_detach(mdev->dev, &priv->rss_map.indir_qp,
1239 				      mc_list, MLX4_PROT_ETH, mclist->reg_id);
1240 	}
1241 	mlx4_en_clear_list(dev);
1242 	list_for_each_entry_safe(mclist, tmp, &priv->curr_list, list) {
1243 		list_del(&mclist->list);
1244 		kfree(mclist);
1245 	}
1246 
1247 	/* Flush multicast filter */
1248 	mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, 0, 1, MLX4_MCAST_CONFIG);
1249 
1250 	mlx4_en_destroy_drop_qp(priv);
1251 
1252 	/* Free TX Rings */
1253 	for (i = 0; i < priv->tx_ring_num; i++) {
1254 		mlx4_en_deactivate_tx_ring(priv, &priv->tx_ring[i]);
1255 		mlx4_en_deactivate_cq(priv, &priv->tx_cq[i]);
1256 	}
1257 	msleep(10);
1258 
1259 	for (i = 0; i < priv->tx_ring_num; i++)
1260 		mlx4_en_free_tx_buf(dev, &priv->tx_ring[i]);
1261 
1262 	/* Free RSS qps */
1263 	mlx4_en_release_rss_steer(priv);
1264 
1265 	/* Unregister Mac address for the port */
1266 	mlx4_put_eth_qp(mdev->dev, priv->port, priv->mac, priv->base_qpn);
1267 	mdev->mac_removed[priv->port] = 1;
1268 
1269 	/* Free RX Rings */
1270 	for (i = 0; i < priv->rx_ring_num; i++) {
1271 		mlx4_en_deactivate_rx_ring(priv, &priv->rx_ring[i]);
1272 		while (test_bit(NAPI_STATE_SCHED, &priv->rx_cq[i].napi.state))
1273 			msleep(1);
1274 		mlx4_en_deactivate_cq(priv, &priv->rx_cq[i]);
1275 	}
1276 
1277 	/* close port*/
1278 	mlx4_CLOSE_PORT(mdev->dev, priv->port);
1279 }
1280 
1281 static void mlx4_en_restart(struct work_struct *work)
1282 {
1283 	struct mlx4_en_priv *priv = container_of(work, struct mlx4_en_priv,
1284 						 watchdog_task);
1285 	struct mlx4_en_dev *mdev = priv->mdev;
1286 	struct net_device *dev = priv->dev;
1287 	int i;
1288 
1289 	en_dbg(DRV, priv, "Watchdog task called for port %d\n", priv->port);
1290 
1291 	mutex_lock(&mdev->state_lock);
1292 	if (priv->port_up) {
1293 		mlx4_en_stop_port(dev);
1294 		for (i = 0; i < priv->tx_ring_num; i++)
1295 			netdev_tx_reset_queue(priv->tx_ring[i].tx_queue);
1296 		if (mlx4_en_start_port(dev))
1297 			en_err(priv, "Failed restarting port %d\n", priv->port);
1298 	}
1299 	mutex_unlock(&mdev->state_lock);
1300 }
1301 
1302 static void mlx4_en_clear_stats(struct net_device *dev)
1303 {
1304 	struct mlx4_en_priv *priv = netdev_priv(dev);
1305 	struct mlx4_en_dev *mdev = priv->mdev;
1306 	int i;
1307 
1308 	if (mlx4_en_DUMP_ETH_STATS(mdev, priv->port, 1))
1309 		en_dbg(HW, priv, "Failed dumping statistics\n");
1310 
1311 	memset(&priv->stats, 0, sizeof(priv->stats));
1312 	memset(&priv->pstats, 0, sizeof(priv->pstats));
1313 	memset(&priv->pkstats, 0, sizeof(priv->pkstats));
1314 	memset(&priv->port_stats, 0, sizeof(priv->port_stats));
1315 
1316 	for (i = 0; i < priv->tx_ring_num; i++) {
1317 		priv->tx_ring[i].bytes = 0;
1318 		priv->tx_ring[i].packets = 0;
1319 		priv->tx_ring[i].tx_csum = 0;
1320 	}
1321 	for (i = 0; i < priv->rx_ring_num; i++) {
1322 		priv->rx_ring[i].bytes = 0;
1323 		priv->rx_ring[i].packets = 0;
1324 		priv->rx_ring[i].csum_ok = 0;
1325 		priv->rx_ring[i].csum_none = 0;
1326 	}
1327 }
1328 
1329 static int mlx4_en_open(struct net_device *dev)
1330 {
1331 	struct mlx4_en_priv *priv = netdev_priv(dev);
1332 	struct mlx4_en_dev *mdev = priv->mdev;
1333 	int err = 0;
1334 
1335 	mutex_lock(&mdev->state_lock);
1336 
1337 	if (!mdev->device_up) {
1338 		en_err(priv, "Cannot open - device down/disabled\n");
1339 		err = -EBUSY;
1340 		goto out;
1341 	}
1342 
1343 	/* Reset HW statistics and SW counters */
1344 	mlx4_en_clear_stats(dev);
1345 
1346 	err = mlx4_en_start_port(dev);
1347 	if (err)
1348 		en_err(priv, "Failed starting port:%d\n", priv->port);
1349 
1350 out:
1351 	mutex_unlock(&mdev->state_lock);
1352 	return err;
1353 }
1354 
1355 
1356 static int mlx4_en_close(struct net_device *dev)
1357 {
1358 	struct mlx4_en_priv *priv = netdev_priv(dev);
1359 	struct mlx4_en_dev *mdev = priv->mdev;
1360 
1361 	en_dbg(IFDOWN, priv, "Close port called\n");
1362 
1363 	mutex_lock(&mdev->state_lock);
1364 
1365 	mlx4_en_stop_port(dev);
1366 	netif_carrier_off(dev);
1367 
1368 	mutex_unlock(&mdev->state_lock);
1369 	return 0;
1370 }
1371 
1372 void mlx4_en_free_resources(struct mlx4_en_priv *priv)
1373 {
1374 	int i;
1375 
1376 #ifdef CONFIG_RFS_ACCEL
1377 	free_irq_cpu_rmap(priv->dev->rx_cpu_rmap);
1378 	priv->dev->rx_cpu_rmap = NULL;
1379 #endif
1380 
1381 	for (i = 0; i < priv->tx_ring_num; i++) {
1382 		if (priv->tx_ring[i].tx_info)
1383 			mlx4_en_destroy_tx_ring(priv, &priv->tx_ring[i]);
1384 		if (priv->tx_cq[i].buf)
1385 			mlx4_en_destroy_cq(priv, &priv->tx_cq[i]);
1386 	}
1387 
1388 	for (i = 0; i < priv->rx_ring_num; i++) {
1389 		if (priv->rx_ring[i].rx_info)
1390 			mlx4_en_destroy_rx_ring(priv, &priv->rx_ring[i],
1391 				priv->prof->rx_ring_size, priv->stride);
1392 		if (priv->rx_cq[i].buf)
1393 			mlx4_en_destroy_cq(priv, &priv->rx_cq[i]);
1394 	}
1395 
1396 	if (priv->base_tx_qpn) {
1397 		mlx4_qp_release_range(priv->mdev->dev, priv->base_tx_qpn, priv->tx_ring_num);
1398 		priv->base_tx_qpn = 0;
1399 	}
1400 }
1401 
1402 int mlx4_en_alloc_resources(struct mlx4_en_priv *priv)
1403 {
1404 	struct mlx4_en_port_profile *prof = priv->prof;
1405 	int i;
1406 	int err;
1407 
1408 	err = mlx4_qp_reserve_range(priv->mdev->dev, priv->tx_ring_num, 256, &priv->base_tx_qpn);
1409 	if (err) {
1410 		en_err(priv, "failed reserving range for TX rings\n");
1411 		return err;
1412 	}
1413 
1414 	/* Create tx Rings */
1415 	for (i = 0; i < priv->tx_ring_num; i++) {
1416 		if (mlx4_en_create_cq(priv, &priv->tx_cq[i],
1417 				      prof->tx_ring_size, i, TX))
1418 			goto err;
1419 
1420 		if (mlx4_en_create_tx_ring(priv, &priv->tx_ring[i], priv->base_tx_qpn + i,
1421 					   prof->tx_ring_size, TXBB_SIZE))
1422 			goto err;
1423 	}
1424 
1425 	/* Create rx Rings */
1426 	for (i = 0; i < priv->rx_ring_num; i++) {
1427 		if (mlx4_en_create_cq(priv, &priv->rx_cq[i],
1428 				      prof->rx_ring_size, i, RX))
1429 			goto err;
1430 
1431 		if (mlx4_en_create_rx_ring(priv, &priv->rx_ring[i],
1432 					   prof->rx_ring_size, priv->stride))
1433 			goto err;
1434 	}
1435 
1436 #ifdef CONFIG_RFS_ACCEL
1437 	priv->dev->rx_cpu_rmap = alloc_irq_cpu_rmap(priv->rx_ring_num);
1438 	if (!priv->dev->rx_cpu_rmap)
1439 		goto err;
1440 
1441 	INIT_LIST_HEAD(&priv->filters);
1442 	spin_lock_init(&priv->filters_lock);
1443 #endif
1444 
1445 	return 0;
1446 
1447 err:
1448 	en_err(priv, "Failed to allocate NIC resources\n");
1449 	return -ENOMEM;
1450 }
1451 
1452 
1453 void mlx4_en_destroy_netdev(struct net_device *dev)
1454 {
1455 	struct mlx4_en_priv *priv = netdev_priv(dev);
1456 	struct mlx4_en_dev *mdev = priv->mdev;
1457 
1458 	en_dbg(DRV, priv, "Destroying netdev on port:%d\n", priv->port);
1459 
1460 	/* Unregister device - this will close the port if it was up */
1461 	if (priv->registered)
1462 		unregister_netdev(dev);
1463 
1464 	if (priv->allocated)
1465 		mlx4_free_hwq_res(mdev->dev, &priv->res, MLX4_EN_PAGE_SIZE);
1466 
1467 	cancel_delayed_work(&priv->stats_task);
1468 	/* flush any pending task for this netdev */
1469 	flush_workqueue(mdev->workqueue);
1470 
1471 	/* Detach the netdev so tasks would not attempt to access it */
1472 	mutex_lock(&mdev->state_lock);
1473 	mdev->pndev[priv->port] = NULL;
1474 	mutex_unlock(&mdev->state_lock);
1475 
1476 	mlx4_en_free_resources(priv);
1477 
1478 	kfree(priv->tx_ring);
1479 	kfree(priv->tx_cq);
1480 
1481 	free_netdev(dev);
1482 }
1483 
1484 static int mlx4_en_change_mtu(struct net_device *dev, int new_mtu)
1485 {
1486 	struct mlx4_en_priv *priv = netdev_priv(dev);
1487 	struct mlx4_en_dev *mdev = priv->mdev;
1488 	int err = 0;
1489 
1490 	en_dbg(DRV, priv, "Change MTU called - current:%d new:%d\n",
1491 		 dev->mtu, new_mtu);
1492 
1493 	if ((new_mtu < MLX4_EN_MIN_MTU) || (new_mtu > priv->max_mtu)) {
1494 		en_err(priv, "Bad MTU size:%d.\n", new_mtu);
1495 		return -EPERM;
1496 	}
1497 	dev->mtu = new_mtu;
1498 
1499 	if (netif_running(dev)) {
1500 		mutex_lock(&mdev->state_lock);
1501 		if (!mdev->device_up) {
1502 			/* NIC is probably restarting - let watchdog task reset
1503 			 * the port */
1504 			en_dbg(DRV, priv, "Change MTU called with card down!?\n");
1505 		} else {
1506 			mlx4_en_stop_port(dev);
1507 			err = mlx4_en_start_port(dev);
1508 			if (err) {
1509 				en_err(priv, "Failed restarting port:%d\n",
1510 					 priv->port);
1511 				queue_work(mdev->workqueue, &priv->watchdog_task);
1512 			}
1513 		}
1514 		mutex_unlock(&mdev->state_lock);
1515 	}
1516 	return 0;
1517 }
1518 
1519 static int mlx4_en_set_features(struct net_device *netdev,
1520 		netdev_features_t features)
1521 {
1522 	struct mlx4_en_priv *priv = netdev_priv(netdev);
1523 
1524 	if (features & NETIF_F_LOOPBACK)
1525 		priv->ctrl_flags |= cpu_to_be32(MLX4_WQE_CTRL_FORCE_LOOPBACK);
1526 	else
1527 		priv->ctrl_flags &=
1528 			cpu_to_be32(~MLX4_WQE_CTRL_FORCE_LOOPBACK);
1529 
1530 	return 0;
1531 
1532 }
1533 
1534 static const struct net_device_ops mlx4_netdev_ops = {
1535 	.ndo_open		= mlx4_en_open,
1536 	.ndo_stop		= mlx4_en_close,
1537 	.ndo_start_xmit		= mlx4_en_xmit,
1538 	.ndo_select_queue	= mlx4_en_select_queue,
1539 	.ndo_get_stats		= mlx4_en_get_stats,
1540 	.ndo_set_rx_mode	= mlx4_en_set_multicast,
1541 	.ndo_set_mac_address	= mlx4_en_set_mac,
1542 	.ndo_validate_addr	= eth_validate_addr,
1543 	.ndo_change_mtu		= mlx4_en_change_mtu,
1544 	.ndo_tx_timeout		= mlx4_en_tx_timeout,
1545 	.ndo_vlan_rx_add_vid	= mlx4_en_vlan_rx_add_vid,
1546 	.ndo_vlan_rx_kill_vid	= mlx4_en_vlan_rx_kill_vid,
1547 #ifdef CONFIG_NET_POLL_CONTROLLER
1548 	.ndo_poll_controller	= mlx4_en_netpoll,
1549 #endif
1550 	.ndo_set_features	= mlx4_en_set_features,
1551 	.ndo_setup_tc		= mlx4_en_setup_tc,
1552 #ifdef CONFIG_RFS_ACCEL
1553 	.ndo_rx_flow_steer	= mlx4_en_filter_rfs,
1554 #endif
1555 };
1556 
1557 int mlx4_en_init_netdev(struct mlx4_en_dev *mdev, int port,
1558 			struct mlx4_en_port_profile *prof)
1559 {
1560 	struct net_device *dev;
1561 	struct mlx4_en_priv *priv;
1562 	int i;
1563 	int err;
1564 
1565 	dev = alloc_etherdev_mqs(sizeof(struct mlx4_en_priv),
1566 				 MAX_TX_RINGS, MAX_RX_RINGS);
1567 	if (dev == NULL)
1568 		return -ENOMEM;
1569 
1570 	netif_set_real_num_tx_queues(dev, prof->tx_ring_num);
1571 	netif_set_real_num_rx_queues(dev, prof->rx_ring_num);
1572 
1573 	SET_NETDEV_DEV(dev, &mdev->dev->pdev->dev);
1574 	dev->dev_id =  port - 1;
1575 
1576 	/*
1577 	 * Initialize driver private data
1578 	 */
1579 
1580 	priv = netdev_priv(dev);
1581 	memset(priv, 0, sizeof(struct mlx4_en_priv));
1582 	priv->dev = dev;
1583 	priv->mdev = mdev;
1584 	priv->ddev = &mdev->pdev->dev;
1585 	priv->prof = prof;
1586 	priv->port = port;
1587 	priv->port_up = false;
1588 	priv->flags = prof->flags;
1589 	priv->ctrl_flags = cpu_to_be32(MLX4_WQE_CTRL_CQ_UPDATE |
1590 			MLX4_WQE_CTRL_SOLICITED);
1591 	priv->num_tx_rings_p_up = mdev->profile.num_tx_rings_p_up;
1592 	priv->tx_ring_num = prof->tx_ring_num;
1593 
1594 	priv->tx_ring = kzalloc(sizeof(struct mlx4_en_tx_ring) * MAX_TX_RINGS,
1595 				GFP_KERNEL);
1596 	if (!priv->tx_ring) {
1597 		err = -ENOMEM;
1598 		goto out;
1599 	}
1600 	priv->tx_cq = kzalloc(sizeof(struct mlx4_en_cq) * MAX_RX_RINGS,
1601 			      GFP_KERNEL);
1602 	if (!priv->tx_cq) {
1603 		err = -ENOMEM;
1604 		goto out;
1605 	}
1606 	priv->rx_ring_num = prof->rx_ring_num;
1607 	priv->cqe_factor = (mdev->dev->caps.cqe_size == 64) ? 1 : 0;
1608 	priv->mac_index = -1;
1609 	priv->msg_enable = MLX4_EN_MSG_LEVEL;
1610 	spin_lock_init(&priv->stats_lock);
1611 	INIT_WORK(&priv->mcast_task, mlx4_en_do_set_multicast);
1612 	INIT_WORK(&priv->mac_task, mlx4_en_do_set_mac);
1613 	INIT_WORK(&priv->watchdog_task, mlx4_en_restart);
1614 	INIT_WORK(&priv->linkstate_task, mlx4_en_linkstate);
1615 	INIT_DELAYED_WORK(&priv->stats_task, mlx4_en_do_get_stats);
1616 #ifdef CONFIG_MLX4_EN_DCB
1617 	if (!mlx4_is_slave(priv->mdev->dev))
1618 		dev->dcbnl_ops = &mlx4_en_dcbnl_ops;
1619 #endif
1620 
1621 	/* Query for default mac and max mtu */
1622 	priv->max_mtu = mdev->dev->caps.eth_mtu_cap[priv->port];
1623 	priv->mac = mdev->dev->caps.def_mac[priv->port];
1624 	if (ILLEGAL_MAC(priv->mac)) {
1625 		en_err(priv, "Port: %d, invalid mac burned: 0x%llx, quiting\n",
1626 			 priv->port, priv->mac);
1627 		err = -EINVAL;
1628 		goto out;
1629 	}
1630 
1631 	priv->stride = roundup_pow_of_two(sizeof(struct mlx4_en_rx_desc) +
1632 					  DS_SIZE * MLX4_EN_MAX_RX_FRAGS);
1633 	err = mlx4_en_alloc_resources(priv);
1634 	if (err)
1635 		goto out;
1636 
1637 	/* Allocate page for receive rings */
1638 	err = mlx4_alloc_hwq_res(mdev->dev, &priv->res,
1639 				MLX4_EN_PAGE_SIZE, MLX4_EN_PAGE_SIZE);
1640 	if (err) {
1641 		en_err(priv, "Failed to allocate page for rx qps\n");
1642 		goto out;
1643 	}
1644 	priv->allocated = 1;
1645 
1646 	/*
1647 	 * Initialize netdev entry points
1648 	 */
1649 	dev->netdev_ops = &mlx4_netdev_ops;
1650 	dev->watchdog_timeo = MLX4_EN_WATCHDOG_TIMEOUT;
1651 	netif_set_real_num_tx_queues(dev, priv->tx_ring_num);
1652 	netif_set_real_num_rx_queues(dev, priv->rx_ring_num);
1653 
1654 	SET_ETHTOOL_OPS(dev, &mlx4_en_ethtool_ops);
1655 
1656 	/* Set defualt MAC */
1657 	dev->addr_len = ETH_ALEN;
1658 	for (i = 0; i < ETH_ALEN; i++) {
1659 		dev->dev_addr[ETH_ALEN - 1 - i] = (u8) (priv->mac >> (8 * i));
1660 		dev->perm_addr[ETH_ALEN - 1 - i] = (u8) (priv->mac >> (8 * i));
1661 	}
1662 
1663 	/*
1664 	 * Set driver features
1665 	 */
1666 	dev->hw_features = NETIF_F_SG | NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM;
1667 	if (mdev->LSO_support)
1668 		dev->hw_features |= NETIF_F_TSO | NETIF_F_TSO6;
1669 
1670 	dev->vlan_features = dev->hw_features;
1671 
1672 	dev->hw_features |= NETIF_F_RXCSUM | NETIF_F_RXHASH;
1673 	dev->features = dev->hw_features | NETIF_F_HIGHDMA |
1674 			NETIF_F_HW_VLAN_TX | NETIF_F_HW_VLAN_RX |
1675 			NETIF_F_HW_VLAN_FILTER;
1676 	dev->hw_features |= NETIF_F_LOOPBACK;
1677 
1678 	if (mdev->dev->caps.steering_mode ==
1679 	    MLX4_STEERING_MODE_DEVICE_MANAGED)
1680 		dev->hw_features |= NETIF_F_NTUPLE;
1681 
1682 	mdev->pndev[port] = dev;
1683 
1684 	netif_carrier_off(dev);
1685 	err = register_netdev(dev);
1686 	if (err) {
1687 		en_err(priv, "Netdev registration failed for port %d\n", port);
1688 		goto out;
1689 	}
1690 	priv->registered = 1;
1691 
1692 	en_warn(priv, "Using %d TX rings\n", prof->tx_ring_num);
1693 	en_warn(priv, "Using %d RX rings\n", prof->rx_ring_num);
1694 
1695 	/* Configure port */
1696 	mlx4_en_calc_rx_buf(dev);
1697 	err = mlx4_SET_PORT_general(mdev->dev, priv->port,
1698 				    priv->rx_skb_size + ETH_FCS_LEN,
1699 				    prof->tx_pause, prof->tx_ppp,
1700 				    prof->rx_pause, prof->rx_ppp);
1701 	if (err) {
1702 		en_err(priv, "Failed setting port general configurations "
1703 		       "for port %d, with error %d\n", priv->port, err);
1704 		goto out;
1705 	}
1706 
1707 	/* Init port */
1708 	en_warn(priv, "Initializing port\n");
1709 	err = mlx4_INIT_PORT(mdev->dev, priv->port);
1710 	if (err) {
1711 		en_err(priv, "Failed Initializing port\n");
1712 		goto out;
1713 	}
1714 	mlx4_en_set_default_moderation(priv);
1715 	queue_delayed_work(mdev->workqueue, &priv->stats_task, STATS_DELAY);
1716 	return 0;
1717 
1718 out:
1719 	mlx4_en_destroy_netdev(dev);
1720 	return err;
1721 }
1722 
1723