Add the rt linux 4.1.3-rt3 as base
[kvmfornfv.git] / kernel / drivers / net / ethernet / mellanox / mlx4 / en_netdev.c
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 #include <net/busy_poll.h>
42 #include <net/vxlan.h>
43
44 #include <linux/mlx4/driver.h>
45 #include <linux/mlx4/device.h>
46 #include <linux/mlx4/cmd.h>
47 #include <linux/mlx4/cq.h>
48
49 #include "mlx4_en.h"
50 #include "en_port.h"
51
52 int mlx4_en_setup_tc(struct net_device *dev, u8 up)
53 {
54         struct mlx4_en_priv *priv = netdev_priv(dev);
55         int i;
56         unsigned int offset = 0;
57
58         if (up && up != MLX4_EN_NUM_UP)
59                 return -EINVAL;
60
61         netdev_set_num_tc(dev, up);
62
63         /* Partition Tx queues evenly amongst UP's */
64         for (i = 0; i < up; i++) {
65                 netdev_set_tc_queue(dev, i, priv->num_tx_rings_p_up, offset);
66                 offset += priv->num_tx_rings_p_up;
67         }
68
69         return 0;
70 }
71
72 #ifdef CONFIG_NET_RX_BUSY_POLL
73 /* must be called with local_bh_disable()d */
74 static int mlx4_en_low_latency_recv(struct napi_struct *napi)
75 {
76         struct mlx4_en_cq *cq = container_of(napi, struct mlx4_en_cq, napi);
77         struct net_device *dev = cq->dev;
78         struct mlx4_en_priv *priv = netdev_priv(dev);
79         struct mlx4_en_rx_ring *rx_ring = priv->rx_ring[cq->ring];
80         int done;
81
82         if (!priv->port_up)
83                 return LL_FLUSH_FAILED;
84
85         if (!mlx4_en_cq_lock_poll(cq))
86                 return LL_FLUSH_BUSY;
87
88         done = mlx4_en_process_rx_cq(dev, cq, 4);
89         if (likely(done))
90                 rx_ring->cleaned += done;
91         else
92                 rx_ring->misses++;
93
94         mlx4_en_cq_unlock_poll(cq);
95
96         return done;
97 }
98 #endif  /* CONFIG_NET_RX_BUSY_POLL */
99
100 #ifdef CONFIG_RFS_ACCEL
101
102 struct mlx4_en_filter {
103         struct list_head next;
104         struct work_struct work;
105
106         u8     ip_proto;
107         __be32 src_ip;
108         __be32 dst_ip;
109         __be16 src_port;
110         __be16 dst_port;
111
112         int rxq_index;
113         struct mlx4_en_priv *priv;
114         u32 flow_id;                    /* RFS infrastructure id */
115         int id;                         /* mlx4_en driver id */
116         u64 reg_id;                     /* Flow steering API id */
117         u8 activated;                   /* Used to prevent expiry before filter
118                                          * is attached
119                                          */
120         struct hlist_node filter_chain;
121 };
122
123 static void mlx4_en_filter_rfs_expire(struct mlx4_en_priv *priv);
124
125 static enum mlx4_net_trans_rule_id mlx4_ip_proto_to_trans_rule_id(u8 ip_proto)
126 {
127         switch (ip_proto) {
128         case IPPROTO_UDP:
129                 return MLX4_NET_TRANS_RULE_ID_UDP;
130         case IPPROTO_TCP:
131                 return MLX4_NET_TRANS_RULE_ID_TCP;
132         default:
133                 return MLX4_NET_TRANS_RULE_NUM;
134         }
135 };
136
137 static void mlx4_en_filter_work(struct work_struct *work)
138 {
139         struct mlx4_en_filter *filter = container_of(work,
140                                                      struct mlx4_en_filter,
141                                                      work);
142         struct mlx4_en_priv *priv = filter->priv;
143         struct mlx4_spec_list spec_tcp_udp = {
144                 .id = mlx4_ip_proto_to_trans_rule_id(filter->ip_proto),
145                 {
146                         .tcp_udp = {
147                                 .dst_port = filter->dst_port,
148                                 .dst_port_msk = (__force __be16)-1,
149                                 .src_port = filter->src_port,
150                                 .src_port_msk = (__force __be16)-1,
151                         },
152                 },
153         };
154         struct mlx4_spec_list spec_ip = {
155                 .id = MLX4_NET_TRANS_RULE_ID_IPV4,
156                 {
157                         .ipv4 = {
158                                 .dst_ip = filter->dst_ip,
159                                 .dst_ip_msk = (__force __be32)-1,
160                                 .src_ip = filter->src_ip,
161                                 .src_ip_msk = (__force __be32)-1,
162                         },
163                 },
164         };
165         struct mlx4_spec_list spec_eth = {
166                 .id = MLX4_NET_TRANS_RULE_ID_ETH,
167         };
168         struct mlx4_net_trans_rule rule = {
169                 .list = LIST_HEAD_INIT(rule.list),
170                 .queue_mode = MLX4_NET_TRANS_Q_LIFO,
171                 .exclusive = 1,
172                 .allow_loopback = 1,
173                 .promisc_mode = MLX4_FS_REGULAR,
174                 .port = priv->port,
175                 .priority = MLX4_DOMAIN_RFS,
176         };
177         int rc;
178         __be64 mac_mask = cpu_to_be64(MLX4_MAC_MASK << 16);
179
180         if (spec_tcp_udp.id >= MLX4_NET_TRANS_RULE_NUM) {
181                 en_warn(priv, "RFS: ignoring unsupported ip protocol (%d)\n",
182                         filter->ip_proto);
183                 goto ignore;
184         }
185         list_add_tail(&spec_eth.list, &rule.list);
186         list_add_tail(&spec_ip.list, &rule.list);
187         list_add_tail(&spec_tcp_udp.list, &rule.list);
188
189         rule.qpn = priv->rss_map.qps[filter->rxq_index].qpn;
190         memcpy(spec_eth.eth.dst_mac, priv->dev->dev_addr, ETH_ALEN);
191         memcpy(spec_eth.eth.dst_mac_msk, &mac_mask, ETH_ALEN);
192
193         filter->activated = 0;
194
195         if (filter->reg_id) {
196                 rc = mlx4_flow_detach(priv->mdev->dev, filter->reg_id);
197                 if (rc && rc != -ENOENT)
198                         en_err(priv, "Error detaching flow. rc = %d\n", rc);
199         }
200
201         rc = mlx4_flow_attach(priv->mdev->dev, &rule, &filter->reg_id);
202         if (rc)
203                 en_err(priv, "Error attaching flow. err = %d\n", rc);
204
205 ignore:
206         mlx4_en_filter_rfs_expire(priv);
207
208         filter->activated = 1;
209 }
210
211 static inline struct hlist_head *
212 filter_hash_bucket(struct mlx4_en_priv *priv, __be32 src_ip, __be32 dst_ip,
213                    __be16 src_port, __be16 dst_port)
214 {
215         unsigned long l;
216         int bucket_idx;
217
218         l = (__force unsigned long)src_port |
219             ((__force unsigned long)dst_port << 2);
220         l ^= (__force unsigned long)(src_ip ^ dst_ip);
221
222         bucket_idx = hash_long(l, MLX4_EN_FILTER_HASH_SHIFT);
223
224         return &priv->filter_hash[bucket_idx];
225 }
226
227 static struct mlx4_en_filter *
228 mlx4_en_filter_alloc(struct mlx4_en_priv *priv, int rxq_index, __be32 src_ip,
229                      __be32 dst_ip, u8 ip_proto, __be16 src_port,
230                      __be16 dst_port, u32 flow_id)
231 {
232         struct mlx4_en_filter *filter = NULL;
233
234         filter = kzalloc(sizeof(struct mlx4_en_filter), GFP_ATOMIC);
235         if (!filter)
236                 return NULL;
237
238         filter->priv = priv;
239         filter->rxq_index = rxq_index;
240         INIT_WORK(&filter->work, mlx4_en_filter_work);
241
242         filter->src_ip = src_ip;
243         filter->dst_ip = dst_ip;
244         filter->ip_proto = ip_proto;
245         filter->src_port = src_port;
246         filter->dst_port = dst_port;
247
248         filter->flow_id = flow_id;
249
250         filter->id = priv->last_filter_id++ % RPS_NO_FILTER;
251
252         list_add_tail(&filter->next, &priv->filters);
253         hlist_add_head(&filter->filter_chain,
254                        filter_hash_bucket(priv, src_ip, dst_ip, src_port,
255                                           dst_port));
256
257         return filter;
258 }
259
260 static void mlx4_en_filter_free(struct mlx4_en_filter *filter)
261 {
262         struct mlx4_en_priv *priv = filter->priv;
263         int rc;
264
265         list_del(&filter->next);
266
267         rc = mlx4_flow_detach(priv->mdev->dev, filter->reg_id);
268         if (rc && rc != -ENOENT)
269                 en_err(priv, "Error detaching flow. rc = %d\n", rc);
270
271         kfree(filter);
272 }
273
274 static inline struct mlx4_en_filter *
275 mlx4_en_filter_find(struct mlx4_en_priv *priv, __be32 src_ip, __be32 dst_ip,
276                     u8 ip_proto, __be16 src_port, __be16 dst_port)
277 {
278         struct mlx4_en_filter *filter;
279         struct mlx4_en_filter *ret = NULL;
280
281         hlist_for_each_entry(filter,
282                              filter_hash_bucket(priv, src_ip, dst_ip,
283                                                 src_port, dst_port),
284                              filter_chain) {
285                 if (filter->src_ip == src_ip &&
286                     filter->dst_ip == dst_ip &&
287                     filter->ip_proto == ip_proto &&
288                     filter->src_port == src_port &&
289                     filter->dst_port == dst_port) {
290                         ret = filter;
291                         break;
292                 }
293         }
294
295         return ret;
296 }
297
298 static int
299 mlx4_en_filter_rfs(struct net_device *net_dev, const struct sk_buff *skb,
300                    u16 rxq_index, u32 flow_id)
301 {
302         struct mlx4_en_priv *priv = netdev_priv(net_dev);
303         struct mlx4_en_filter *filter;
304         const struct iphdr *ip;
305         const __be16 *ports;
306         u8 ip_proto;
307         __be32 src_ip;
308         __be32 dst_ip;
309         __be16 src_port;
310         __be16 dst_port;
311         int nhoff = skb_network_offset(skb);
312         int ret = 0;
313
314         if (skb->protocol != htons(ETH_P_IP))
315                 return -EPROTONOSUPPORT;
316
317         ip = (const struct iphdr *)(skb->data + nhoff);
318         if (ip_is_fragment(ip))
319                 return -EPROTONOSUPPORT;
320
321         if ((ip->protocol != IPPROTO_TCP) && (ip->protocol != IPPROTO_UDP))
322                 return -EPROTONOSUPPORT;
323         ports = (const __be16 *)(skb->data + nhoff + 4 * ip->ihl);
324
325         ip_proto = ip->protocol;
326         src_ip = ip->saddr;
327         dst_ip = ip->daddr;
328         src_port = ports[0];
329         dst_port = ports[1];
330
331         spin_lock_bh(&priv->filters_lock);
332         filter = mlx4_en_filter_find(priv, src_ip, dst_ip, ip_proto,
333                                      src_port, dst_port);
334         if (filter) {
335                 if (filter->rxq_index == rxq_index)
336                         goto out;
337
338                 filter->rxq_index = rxq_index;
339         } else {
340                 filter = mlx4_en_filter_alloc(priv, rxq_index,
341                                               src_ip, dst_ip, ip_proto,
342                                               src_port, dst_port, flow_id);
343                 if (!filter) {
344                         ret = -ENOMEM;
345                         goto err;
346                 }
347         }
348
349         queue_work(priv->mdev->workqueue, &filter->work);
350
351 out:
352         ret = filter->id;
353 err:
354         spin_unlock_bh(&priv->filters_lock);
355
356         return ret;
357 }
358
359 void mlx4_en_cleanup_filters(struct mlx4_en_priv *priv)
360 {
361         struct mlx4_en_filter *filter, *tmp;
362         LIST_HEAD(del_list);
363
364         spin_lock_bh(&priv->filters_lock);
365         list_for_each_entry_safe(filter, tmp, &priv->filters, next) {
366                 list_move(&filter->next, &del_list);
367                 hlist_del(&filter->filter_chain);
368         }
369         spin_unlock_bh(&priv->filters_lock);
370
371         list_for_each_entry_safe(filter, tmp, &del_list, next) {
372                 cancel_work_sync(&filter->work);
373                 mlx4_en_filter_free(filter);
374         }
375 }
376
377 static void mlx4_en_filter_rfs_expire(struct mlx4_en_priv *priv)
378 {
379         struct mlx4_en_filter *filter = NULL, *tmp, *last_filter = NULL;
380         LIST_HEAD(del_list);
381         int i = 0;
382
383         spin_lock_bh(&priv->filters_lock);
384         list_for_each_entry_safe(filter, tmp, &priv->filters, next) {
385                 if (i > MLX4_EN_FILTER_EXPIRY_QUOTA)
386                         break;
387
388                 if (filter->activated &&
389                     !work_pending(&filter->work) &&
390                     rps_may_expire_flow(priv->dev,
391                                         filter->rxq_index, filter->flow_id,
392                                         filter->id)) {
393                         list_move(&filter->next, &del_list);
394                         hlist_del(&filter->filter_chain);
395                 } else
396                         last_filter = filter;
397
398                 i++;
399         }
400
401         if (last_filter && (&last_filter->next != priv->filters.next))
402                 list_move(&priv->filters, &last_filter->next);
403
404         spin_unlock_bh(&priv->filters_lock);
405
406         list_for_each_entry_safe(filter, tmp, &del_list, next)
407                 mlx4_en_filter_free(filter);
408 }
409 #endif
410
411 static int mlx4_en_vlan_rx_add_vid(struct net_device *dev,
412                                    __be16 proto, u16 vid)
413 {
414         struct mlx4_en_priv *priv = netdev_priv(dev);
415         struct mlx4_en_dev *mdev = priv->mdev;
416         int err;
417         int idx;
418
419         en_dbg(HW, priv, "adding VLAN:%d\n", vid);
420
421         set_bit(vid, priv->active_vlans);
422
423         /* Add VID to port VLAN filter */
424         mutex_lock(&mdev->state_lock);
425         if (mdev->device_up && priv->port_up) {
426                 err = mlx4_SET_VLAN_FLTR(mdev->dev, priv);
427                 if (err)
428                         en_err(priv, "Failed configuring VLAN filter\n");
429         }
430         if (mlx4_register_vlan(mdev->dev, priv->port, vid, &idx))
431                 en_dbg(HW, priv, "failed adding vlan %d\n", vid);
432         mutex_unlock(&mdev->state_lock);
433
434         return 0;
435 }
436
437 static int mlx4_en_vlan_rx_kill_vid(struct net_device *dev,
438                                     __be16 proto, u16 vid)
439 {
440         struct mlx4_en_priv *priv = netdev_priv(dev);
441         struct mlx4_en_dev *mdev = priv->mdev;
442         int err;
443
444         en_dbg(HW, priv, "Killing VID:%d\n", vid);
445
446         clear_bit(vid, priv->active_vlans);
447
448         /* Remove VID from port VLAN filter */
449         mutex_lock(&mdev->state_lock);
450         mlx4_unregister_vlan(mdev->dev, priv->port, vid);
451
452         if (mdev->device_up && priv->port_up) {
453                 err = mlx4_SET_VLAN_FLTR(mdev->dev, priv);
454                 if (err)
455                         en_err(priv, "Failed configuring VLAN filter\n");
456         }
457         mutex_unlock(&mdev->state_lock);
458
459         return 0;
460 }
461
462 static void mlx4_en_u64_to_mac(unsigned char dst_mac[ETH_ALEN + 2], u64 src_mac)
463 {
464         int i;
465         for (i = ETH_ALEN - 1; i >= 0; --i) {
466                 dst_mac[i] = src_mac & 0xff;
467                 src_mac >>= 8;
468         }
469         memset(&dst_mac[ETH_ALEN], 0, 2);
470 }
471
472
473 static int mlx4_en_tunnel_steer_add(struct mlx4_en_priv *priv, unsigned char *addr,
474                                     int qpn, u64 *reg_id)
475 {
476         int err;
477
478         if (priv->mdev->dev->caps.tunnel_offload_mode != MLX4_TUNNEL_OFFLOAD_MODE_VXLAN ||
479             priv->mdev->dev->caps.dmfs_high_steer_mode == MLX4_STEERING_DMFS_A0_STATIC)
480                 return 0; /* do nothing */
481
482         err = mlx4_tunnel_steer_add(priv->mdev->dev, addr, priv->port, qpn,
483                                     MLX4_DOMAIN_NIC, reg_id);
484         if (err) {
485                 en_err(priv, "failed to add vxlan steering rule, err %d\n", err);
486                 return err;
487         }
488         en_dbg(DRV, priv, "added vxlan steering rule, mac %pM reg_id %llx\n", addr, *reg_id);
489         return 0;
490 }
491
492
493 static int mlx4_en_uc_steer_add(struct mlx4_en_priv *priv,
494                                 unsigned char *mac, int *qpn, u64 *reg_id)
495 {
496         struct mlx4_en_dev *mdev = priv->mdev;
497         struct mlx4_dev *dev = mdev->dev;
498         int err;
499
500         switch (dev->caps.steering_mode) {
501         case MLX4_STEERING_MODE_B0: {
502                 struct mlx4_qp qp;
503                 u8 gid[16] = {0};
504
505                 qp.qpn = *qpn;
506                 memcpy(&gid[10], mac, ETH_ALEN);
507                 gid[5] = priv->port;
508
509                 err = mlx4_unicast_attach(dev, &qp, gid, 0, MLX4_PROT_ETH);
510                 break;
511         }
512         case MLX4_STEERING_MODE_DEVICE_MANAGED: {
513                 struct mlx4_spec_list spec_eth = { {NULL} };
514                 __be64 mac_mask = cpu_to_be64(MLX4_MAC_MASK << 16);
515
516                 struct mlx4_net_trans_rule rule = {
517                         .queue_mode = MLX4_NET_TRANS_Q_FIFO,
518                         .exclusive = 0,
519                         .allow_loopback = 1,
520                         .promisc_mode = MLX4_FS_REGULAR,
521                         .priority = MLX4_DOMAIN_NIC,
522                 };
523
524                 rule.port = priv->port;
525                 rule.qpn = *qpn;
526                 INIT_LIST_HEAD(&rule.list);
527
528                 spec_eth.id = MLX4_NET_TRANS_RULE_ID_ETH;
529                 memcpy(spec_eth.eth.dst_mac, mac, ETH_ALEN);
530                 memcpy(spec_eth.eth.dst_mac_msk, &mac_mask, ETH_ALEN);
531                 list_add_tail(&spec_eth.list, &rule.list);
532
533                 err = mlx4_flow_attach(dev, &rule, reg_id);
534                 break;
535         }
536         default:
537                 return -EINVAL;
538         }
539         if (err)
540                 en_warn(priv, "Failed Attaching Unicast\n");
541
542         return err;
543 }
544
545 static void mlx4_en_uc_steer_release(struct mlx4_en_priv *priv,
546                                      unsigned char *mac, int qpn, u64 reg_id)
547 {
548         struct mlx4_en_dev *mdev = priv->mdev;
549         struct mlx4_dev *dev = mdev->dev;
550
551         switch (dev->caps.steering_mode) {
552         case MLX4_STEERING_MODE_B0: {
553                 struct mlx4_qp qp;
554                 u8 gid[16] = {0};
555
556                 qp.qpn = qpn;
557                 memcpy(&gid[10], mac, ETH_ALEN);
558                 gid[5] = priv->port;
559
560                 mlx4_unicast_detach(dev, &qp, gid, MLX4_PROT_ETH);
561                 break;
562         }
563         case MLX4_STEERING_MODE_DEVICE_MANAGED: {
564                 mlx4_flow_detach(dev, reg_id);
565                 break;
566         }
567         default:
568                 en_err(priv, "Invalid steering mode.\n");
569         }
570 }
571
572 static int mlx4_en_get_qp(struct mlx4_en_priv *priv)
573 {
574         struct mlx4_en_dev *mdev = priv->mdev;
575         struct mlx4_dev *dev = mdev->dev;
576         struct mlx4_mac_entry *entry;
577         int index = 0;
578         int err = 0;
579         u64 reg_id = 0;
580         int *qpn = &priv->base_qpn;
581         u64 mac = mlx4_mac_to_u64(priv->dev->dev_addr);
582
583         en_dbg(DRV, priv, "Registering MAC: %pM for adding\n",
584                priv->dev->dev_addr);
585         index = mlx4_register_mac(dev, priv->port, mac);
586         if (index < 0) {
587                 err = index;
588                 en_err(priv, "Failed adding MAC: %pM\n",
589                        priv->dev->dev_addr);
590                 return err;
591         }
592
593         if (dev->caps.steering_mode == MLX4_STEERING_MODE_A0) {
594                 int base_qpn = mlx4_get_base_qpn(dev, priv->port);
595                 *qpn = base_qpn + index;
596                 return 0;
597         }
598
599         err = mlx4_qp_reserve_range(dev, 1, 1, qpn, MLX4_RESERVE_A0_QP);
600         en_dbg(DRV, priv, "Reserved qp %d\n", *qpn);
601         if (err) {
602                 en_err(priv, "Failed to reserve qp for mac registration\n");
603                 goto qp_err;
604         }
605
606         err = mlx4_en_uc_steer_add(priv, priv->dev->dev_addr, qpn, &reg_id);
607         if (err)
608                 goto steer_err;
609
610         err = mlx4_en_tunnel_steer_add(priv, priv->dev->dev_addr, *qpn,
611                                        &priv->tunnel_reg_id);
612         if (err)
613                 goto tunnel_err;
614
615         entry = kmalloc(sizeof(*entry), GFP_KERNEL);
616         if (!entry) {
617                 err = -ENOMEM;
618                 goto alloc_err;
619         }
620         memcpy(entry->mac, priv->dev->dev_addr, sizeof(entry->mac));
621         memcpy(priv->current_mac, entry->mac, sizeof(priv->current_mac));
622         entry->reg_id = reg_id;
623
624         hlist_add_head_rcu(&entry->hlist,
625                            &priv->mac_hash[entry->mac[MLX4_EN_MAC_HASH_IDX]]);
626
627         return 0;
628
629 alloc_err:
630         if (priv->tunnel_reg_id)
631                 mlx4_flow_detach(priv->mdev->dev, priv->tunnel_reg_id);
632 tunnel_err:
633         mlx4_en_uc_steer_release(priv, priv->dev->dev_addr, *qpn, reg_id);
634
635 steer_err:
636         mlx4_qp_release_range(dev, *qpn, 1);
637
638 qp_err:
639         mlx4_unregister_mac(dev, priv->port, mac);
640         return err;
641 }
642
643 static void mlx4_en_put_qp(struct mlx4_en_priv *priv)
644 {
645         struct mlx4_en_dev *mdev = priv->mdev;
646         struct mlx4_dev *dev = mdev->dev;
647         int qpn = priv->base_qpn;
648         u64 mac;
649
650         if (dev->caps.steering_mode == MLX4_STEERING_MODE_A0) {
651                 mac = mlx4_mac_to_u64(priv->dev->dev_addr);
652                 en_dbg(DRV, priv, "Registering MAC: %pM for deleting\n",
653                        priv->dev->dev_addr);
654                 mlx4_unregister_mac(dev, priv->port, mac);
655         } else {
656                 struct mlx4_mac_entry *entry;
657                 struct hlist_node *tmp;
658                 struct hlist_head *bucket;
659                 unsigned int i;
660
661                 for (i = 0; i < MLX4_EN_MAC_HASH_SIZE; ++i) {
662                         bucket = &priv->mac_hash[i];
663                         hlist_for_each_entry_safe(entry, tmp, bucket, hlist) {
664                                 mac = mlx4_mac_to_u64(entry->mac);
665                                 en_dbg(DRV, priv, "Registering MAC: %pM for deleting\n",
666                                        entry->mac);
667                                 mlx4_en_uc_steer_release(priv, entry->mac,
668                                                          qpn, entry->reg_id);
669
670                                 mlx4_unregister_mac(dev, priv->port, mac);
671                                 hlist_del_rcu(&entry->hlist);
672                                 kfree_rcu(entry, rcu);
673                         }
674                 }
675
676                 if (priv->tunnel_reg_id) {
677                         mlx4_flow_detach(priv->mdev->dev, priv->tunnel_reg_id);
678                         priv->tunnel_reg_id = 0;
679                 }
680
681                 en_dbg(DRV, priv, "Releasing qp: port %d, qpn %d\n",
682                        priv->port, qpn);
683                 mlx4_qp_release_range(dev, qpn, 1);
684                 priv->flags &= ~MLX4_EN_FLAG_FORCE_PROMISC;
685         }
686 }
687
688 static int mlx4_en_replace_mac(struct mlx4_en_priv *priv, int qpn,
689                                unsigned char *new_mac, unsigned char *prev_mac)
690 {
691         struct mlx4_en_dev *mdev = priv->mdev;
692         struct mlx4_dev *dev = mdev->dev;
693         int err = 0;
694         u64 new_mac_u64 = mlx4_mac_to_u64(new_mac);
695
696         if (dev->caps.steering_mode != MLX4_STEERING_MODE_A0) {
697                 struct hlist_head *bucket;
698                 unsigned int mac_hash;
699                 struct mlx4_mac_entry *entry;
700                 struct hlist_node *tmp;
701                 u64 prev_mac_u64 = mlx4_mac_to_u64(prev_mac);
702
703                 bucket = &priv->mac_hash[prev_mac[MLX4_EN_MAC_HASH_IDX]];
704                 hlist_for_each_entry_safe(entry, tmp, bucket, hlist) {
705                         if (ether_addr_equal_64bits(entry->mac, prev_mac)) {
706                                 mlx4_en_uc_steer_release(priv, entry->mac,
707                                                          qpn, entry->reg_id);
708                                 mlx4_unregister_mac(dev, priv->port,
709                                                     prev_mac_u64);
710                                 hlist_del_rcu(&entry->hlist);
711                                 synchronize_rcu();
712                                 memcpy(entry->mac, new_mac, ETH_ALEN);
713                                 entry->reg_id = 0;
714                                 mac_hash = new_mac[MLX4_EN_MAC_HASH_IDX];
715                                 hlist_add_head_rcu(&entry->hlist,
716                                                    &priv->mac_hash[mac_hash]);
717                                 mlx4_register_mac(dev, priv->port, new_mac_u64);
718                                 err = mlx4_en_uc_steer_add(priv, new_mac,
719                                                            &qpn,
720                                                            &entry->reg_id);
721                                 if (err)
722                                         return err;
723                                 if (priv->tunnel_reg_id) {
724                                         mlx4_flow_detach(priv->mdev->dev, priv->tunnel_reg_id);
725                                         priv->tunnel_reg_id = 0;
726                                 }
727                                 err = mlx4_en_tunnel_steer_add(priv, new_mac, qpn,
728                                                                &priv->tunnel_reg_id);
729                                 return err;
730                         }
731                 }
732                 return -EINVAL;
733         }
734
735         return __mlx4_replace_mac(dev, priv->port, qpn, new_mac_u64);
736 }
737
738 static int mlx4_en_do_set_mac(struct mlx4_en_priv *priv,
739                               unsigned char new_mac[ETH_ALEN + 2])
740 {
741         int err = 0;
742
743         if (priv->port_up) {
744                 /* Remove old MAC and insert the new one */
745                 err = mlx4_en_replace_mac(priv, priv->base_qpn,
746                                           new_mac, priv->current_mac);
747                 if (err)
748                         en_err(priv, "Failed changing HW MAC address\n");
749         } else
750                 en_dbg(HW, priv, "Port is down while registering mac, exiting...\n");
751
752         if (!err)
753                 memcpy(priv->current_mac, new_mac, sizeof(priv->current_mac));
754
755         return err;
756 }
757
758 static int mlx4_en_set_mac(struct net_device *dev, void *addr)
759 {
760         struct mlx4_en_priv *priv = netdev_priv(dev);
761         struct mlx4_en_dev *mdev = priv->mdev;
762         struct sockaddr *saddr = addr;
763         unsigned char new_mac[ETH_ALEN + 2];
764         int err;
765
766         if (!is_valid_ether_addr(saddr->sa_data))
767                 return -EADDRNOTAVAIL;
768
769         mutex_lock(&mdev->state_lock);
770         memcpy(new_mac, saddr->sa_data, ETH_ALEN);
771         err = mlx4_en_do_set_mac(priv, new_mac);
772         if (!err)
773                 memcpy(dev->dev_addr, saddr->sa_data, ETH_ALEN);
774         mutex_unlock(&mdev->state_lock);
775
776         return err;
777 }
778
779 static void mlx4_en_clear_list(struct net_device *dev)
780 {
781         struct mlx4_en_priv *priv = netdev_priv(dev);
782         struct mlx4_en_mc_list *tmp, *mc_to_del;
783
784         list_for_each_entry_safe(mc_to_del, tmp, &priv->mc_list, list) {
785                 list_del(&mc_to_del->list);
786                 kfree(mc_to_del);
787         }
788 }
789
790 static void mlx4_en_cache_mclist(struct net_device *dev)
791 {
792         struct mlx4_en_priv *priv = netdev_priv(dev);
793         struct netdev_hw_addr *ha;
794         struct mlx4_en_mc_list *tmp;
795
796         mlx4_en_clear_list(dev);
797         netdev_for_each_mc_addr(ha, dev) {
798                 tmp = kzalloc(sizeof(struct mlx4_en_mc_list), GFP_ATOMIC);
799                 if (!tmp) {
800                         mlx4_en_clear_list(dev);
801                         return;
802                 }
803                 memcpy(tmp->addr, ha->addr, ETH_ALEN);
804                 list_add_tail(&tmp->list, &priv->mc_list);
805         }
806 }
807
808 static void update_mclist_flags(struct mlx4_en_priv *priv,
809                                 struct list_head *dst,
810                                 struct list_head *src)
811 {
812         struct mlx4_en_mc_list *dst_tmp, *src_tmp, *new_mc;
813         bool found;
814
815         /* Find all the entries that should be removed from dst,
816          * These are the entries that are not found in src
817          */
818         list_for_each_entry(dst_tmp, dst, list) {
819                 found = false;
820                 list_for_each_entry(src_tmp, src, list) {
821                         if (ether_addr_equal(dst_tmp->addr, src_tmp->addr)) {
822                                 found = true;
823                                 break;
824                         }
825                 }
826                 if (!found)
827                         dst_tmp->action = MCLIST_REM;
828         }
829
830         /* Add entries that exist in src but not in dst
831          * mark them as need to add
832          */
833         list_for_each_entry(src_tmp, src, list) {
834                 found = false;
835                 list_for_each_entry(dst_tmp, dst, list) {
836                         if (ether_addr_equal(dst_tmp->addr, src_tmp->addr)) {
837                                 dst_tmp->action = MCLIST_NONE;
838                                 found = true;
839                                 break;
840                         }
841                 }
842                 if (!found) {
843                         new_mc = kmemdup(src_tmp,
844                                          sizeof(struct mlx4_en_mc_list),
845                                          GFP_KERNEL);
846                         if (!new_mc)
847                                 return;
848
849                         new_mc->action = MCLIST_ADD;
850                         list_add_tail(&new_mc->list, dst);
851                 }
852         }
853 }
854
855 static void mlx4_en_set_rx_mode(struct net_device *dev)
856 {
857         struct mlx4_en_priv *priv = netdev_priv(dev);
858
859         if (!priv->port_up)
860                 return;
861
862         queue_work(priv->mdev->workqueue, &priv->rx_mode_task);
863 }
864
865 static void mlx4_en_set_promisc_mode(struct mlx4_en_priv *priv,
866                                      struct mlx4_en_dev *mdev)
867 {
868         int err = 0;
869
870         if (!(priv->flags & MLX4_EN_FLAG_PROMISC)) {
871                 if (netif_msg_rx_status(priv))
872                         en_warn(priv, "Entering promiscuous mode\n");
873                 priv->flags |= MLX4_EN_FLAG_PROMISC;
874
875                 /* Enable promiscouos mode */
876                 switch (mdev->dev->caps.steering_mode) {
877                 case MLX4_STEERING_MODE_DEVICE_MANAGED:
878                         err = mlx4_flow_steer_promisc_add(mdev->dev,
879                                                           priv->port,
880                                                           priv->base_qpn,
881                                                           MLX4_FS_ALL_DEFAULT);
882                         if (err)
883                                 en_err(priv, "Failed enabling promiscuous mode\n");
884                         priv->flags |= MLX4_EN_FLAG_MC_PROMISC;
885                         break;
886
887                 case MLX4_STEERING_MODE_B0:
888                         err = mlx4_unicast_promisc_add(mdev->dev,
889                                                        priv->base_qpn,
890                                                        priv->port);
891                         if (err)
892                                 en_err(priv, "Failed enabling unicast promiscuous mode\n");
893
894                         /* Add the default qp number as multicast
895                          * promisc
896                          */
897                         if (!(priv->flags & MLX4_EN_FLAG_MC_PROMISC)) {
898                                 err = mlx4_multicast_promisc_add(mdev->dev,
899                                                                  priv->base_qpn,
900                                                                  priv->port);
901                                 if (err)
902                                         en_err(priv, "Failed enabling multicast promiscuous mode\n");
903                                 priv->flags |= MLX4_EN_FLAG_MC_PROMISC;
904                         }
905                         break;
906
907                 case MLX4_STEERING_MODE_A0:
908                         err = mlx4_SET_PORT_qpn_calc(mdev->dev,
909                                                      priv->port,
910                                                      priv->base_qpn,
911                                                      1);
912                         if (err)
913                                 en_err(priv, "Failed enabling promiscuous mode\n");
914                         break;
915                 }
916
917                 /* Disable port multicast filter (unconditionally) */
918                 err = mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, 0,
919                                           0, MLX4_MCAST_DISABLE);
920                 if (err)
921                         en_err(priv, "Failed disabling multicast filter\n");
922         }
923 }
924
925 static void mlx4_en_clear_promisc_mode(struct mlx4_en_priv *priv,
926                                        struct mlx4_en_dev *mdev)
927 {
928         int err = 0;
929
930         if (netif_msg_rx_status(priv))
931                 en_warn(priv, "Leaving promiscuous mode\n");
932         priv->flags &= ~MLX4_EN_FLAG_PROMISC;
933
934         /* Disable promiscouos mode */
935         switch (mdev->dev->caps.steering_mode) {
936         case MLX4_STEERING_MODE_DEVICE_MANAGED:
937                 err = mlx4_flow_steer_promisc_remove(mdev->dev,
938                                                      priv->port,
939                                                      MLX4_FS_ALL_DEFAULT);
940                 if (err)
941                         en_err(priv, "Failed disabling promiscuous mode\n");
942                 priv->flags &= ~MLX4_EN_FLAG_MC_PROMISC;
943                 break;
944
945         case MLX4_STEERING_MODE_B0:
946                 err = mlx4_unicast_promisc_remove(mdev->dev,
947                                                   priv->base_qpn,
948                                                   priv->port);
949                 if (err)
950                         en_err(priv, "Failed disabling unicast promiscuous mode\n");
951                 /* Disable Multicast promisc */
952                 if (priv->flags & MLX4_EN_FLAG_MC_PROMISC) {
953                         err = mlx4_multicast_promisc_remove(mdev->dev,
954                                                             priv->base_qpn,
955                                                             priv->port);
956                         if (err)
957                                 en_err(priv, "Failed disabling multicast promiscuous mode\n");
958                         priv->flags &= ~MLX4_EN_FLAG_MC_PROMISC;
959                 }
960                 break;
961
962         case MLX4_STEERING_MODE_A0:
963                 err = mlx4_SET_PORT_qpn_calc(mdev->dev,
964                                              priv->port,
965                                              priv->base_qpn, 0);
966                 if (err)
967                         en_err(priv, "Failed disabling promiscuous mode\n");
968                 break;
969         }
970 }
971
972 static void mlx4_en_do_multicast(struct mlx4_en_priv *priv,
973                                  struct net_device *dev,
974                                  struct mlx4_en_dev *mdev)
975 {
976         struct mlx4_en_mc_list *mclist, *tmp;
977         u64 mcast_addr = 0;
978         u8 mc_list[16] = {0};
979         int err = 0;
980
981         /* Enable/disable the multicast filter according to IFF_ALLMULTI */
982         if (dev->flags & IFF_ALLMULTI) {
983                 err = mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, 0,
984                                           0, MLX4_MCAST_DISABLE);
985                 if (err)
986                         en_err(priv, "Failed disabling multicast filter\n");
987
988                 /* Add the default qp number as multicast promisc */
989                 if (!(priv->flags & MLX4_EN_FLAG_MC_PROMISC)) {
990                         switch (mdev->dev->caps.steering_mode) {
991                         case MLX4_STEERING_MODE_DEVICE_MANAGED:
992                                 err = mlx4_flow_steer_promisc_add(mdev->dev,
993                                                                   priv->port,
994                                                                   priv->base_qpn,
995                                                                   MLX4_FS_MC_DEFAULT);
996                                 break;
997
998                         case MLX4_STEERING_MODE_B0:
999                                 err = mlx4_multicast_promisc_add(mdev->dev,
1000                                                                  priv->base_qpn,
1001                                                                  priv->port);
1002                                 break;
1003
1004                         case MLX4_STEERING_MODE_A0:
1005                                 break;
1006                         }
1007                         if (err)
1008                                 en_err(priv, "Failed entering multicast promisc mode\n");
1009                         priv->flags |= MLX4_EN_FLAG_MC_PROMISC;
1010                 }
1011         } else {
1012                 /* Disable Multicast promisc */
1013                 if (priv->flags & MLX4_EN_FLAG_MC_PROMISC) {
1014                         switch (mdev->dev->caps.steering_mode) {
1015                         case MLX4_STEERING_MODE_DEVICE_MANAGED:
1016                                 err = mlx4_flow_steer_promisc_remove(mdev->dev,
1017                                                                      priv->port,
1018                                                                      MLX4_FS_MC_DEFAULT);
1019                                 break;
1020
1021                         case MLX4_STEERING_MODE_B0:
1022                                 err = mlx4_multicast_promisc_remove(mdev->dev,
1023                                                                     priv->base_qpn,
1024                                                                     priv->port);
1025                                 break;
1026
1027                         case MLX4_STEERING_MODE_A0:
1028                                 break;
1029                         }
1030                         if (err)
1031                                 en_err(priv, "Failed disabling multicast promiscuous mode\n");
1032                         priv->flags &= ~MLX4_EN_FLAG_MC_PROMISC;
1033                 }
1034
1035                 err = mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, 0,
1036                                           0, MLX4_MCAST_DISABLE);
1037                 if (err)
1038                         en_err(priv, "Failed disabling multicast filter\n");
1039
1040                 /* Flush mcast filter and init it with broadcast address */
1041                 mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, ETH_BCAST,
1042                                     1, MLX4_MCAST_CONFIG);
1043
1044                 /* Update multicast list - we cache all addresses so they won't
1045                  * change while HW is updated holding the command semaphor */
1046                 netif_addr_lock_bh(dev);
1047                 mlx4_en_cache_mclist(dev);
1048                 netif_addr_unlock_bh(dev);
1049                 list_for_each_entry(mclist, &priv->mc_list, list) {
1050                         mcast_addr = mlx4_mac_to_u64(mclist->addr);
1051                         mlx4_SET_MCAST_FLTR(mdev->dev, priv->port,
1052                                             mcast_addr, 0, MLX4_MCAST_CONFIG);
1053                 }
1054                 err = mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, 0,
1055                                           0, MLX4_MCAST_ENABLE);
1056                 if (err)
1057                         en_err(priv, "Failed enabling multicast filter\n");
1058
1059                 update_mclist_flags(priv, &priv->curr_list, &priv->mc_list);
1060                 list_for_each_entry_safe(mclist, tmp, &priv->curr_list, list) {
1061                         if (mclist->action == MCLIST_REM) {
1062                                 /* detach this address and delete from list */
1063                                 memcpy(&mc_list[10], mclist->addr, ETH_ALEN);
1064                                 mc_list[5] = priv->port;
1065                                 err = mlx4_multicast_detach(mdev->dev,
1066                                                             &priv->rss_map.indir_qp,
1067                                                             mc_list,
1068                                                             MLX4_PROT_ETH,
1069                                                             mclist->reg_id);
1070                                 if (err)
1071                                         en_err(priv, "Fail to detach multicast address\n");
1072
1073                                 if (mclist->tunnel_reg_id) {
1074                                         err = mlx4_flow_detach(priv->mdev->dev, mclist->tunnel_reg_id);
1075                                         if (err)
1076                                                 en_err(priv, "Failed to detach multicast address\n");
1077                                 }
1078
1079                                 /* remove from list */
1080                                 list_del(&mclist->list);
1081                                 kfree(mclist);
1082                         } else if (mclist->action == MCLIST_ADD) {
1083                                 /* attach the address */
1084                                 memcpy(&mc_list[10], mclist->addr, ETH_ALEN);
1085                                 /* needed for B0 steering support */
1086                                 mc_list[5] = priv->port;
1087                                 err = mlx4_multicast_attach(mdev->dev,
1088                                                             &priv->rss_map.indir_qp,
1089                                                             mc_list,
1090                                                             priv->port, 0,
1091                                                             MLX4_PROT_ETH,
1092                                                             &mclist->reg_id);
1093                                 if (err)
1094                                         en_err(priv, "Fail to attach multicast address\n");
1095
1096                                 err = mlx4_en_tunnel_steer_add(priv, &mc_list[10], priv->base_qpn,
1097                                                                &mclist->tunnel_reg_id);
1098                                 if (err)
1099                                         en_err(priv, "Failed to attach multicast address\n");
1100                         }
1101                 }
1102         }
1103 }
1104
1105 static void mlx4_en_do_uc_filter(struct mlx4_en_priv *priv,
1106                                  struct net_device *dev,
1107                                  struct mlx4_en_dev *mdev)
1108 {
1109         struct netdev_hw_addr *ha;
1110         struct mlx4_mac_entry *entry;
1111         struct hlist_node *tmp;
1112         bool found;
1113         u64 mac;
1114         int err = 0;
1115         struct hlist_head *bucket;
1116         unsigned int i;
1117         int removed = 0;
1118         u32 prev_flags;
1119
1120         /* Note that we do not need to protect our mac_hash traversal with rcu,
1121          * since all modification code is protected by mdev->state_lock
1122          */
1123
1124         /* find what to remove */
1125         for (i = 0; i < MLX4_EN_MAC_HASH_SIZE; ++i) {
1126                 bucket = &priv->mac_hash[i];
1127                 hlist_for_each_entry_safe(entry, tmp, bucket, hlist) {
1128                         found = false;
1129                         netdev_for_each_uc_addr(ha, dev) {
1130                                 if (ether_addr_equal_64bits(entry->mac,
1131                                                             ha->addr)) {
1132                                         found = true;
1133                                         break;
1134                                 }
1135                         }
1136
1137                         /* MAC address of the port is not in uc list */
1138                         if (ether_addr_equal_64bits(entry->mac,
1139                                                     priv->current_mac))
1140                                 found = true;
1141
1142                         if (!found) {
1143                                 mac = mlx4_mac_to_u64(entry->mac);
1144                                 mlx4_en_uc_steer_release(priv, entry->mac,
1145                                                          priv->base_qpn,
1146                                                          entry->reg_id);
1147                                 mlx4_unregister_mac(mdev->dev, priv->port, mac);
1148
1149                                 hlist_del_rcu(&entry->hlist);
1150                                 kfree_rcu(entry, rcu);
1151                                 en_dbg(DRV, priv, "Removed MAC %pM on port:%d\n",
1152                                        entry->mac, priv->port);
1153                                 ++removed;
1154                         }
1155                 }
1156         }
1157
1158         /* if we didn't remove anything, there is no use in trying to add
1159          * again once we are in a forced promisc mode state
1160          */
1161         if ((priv->flags & MLX4_EN_FLAG_FORCE_PROMISC) && 0 == removed)
1162                 return;
1163
1164         prev_flags = priv->flags;
1165         priv->flags &= ~MLX4_EN_FLAG_FORCE_PROMISC;
1166
1167         /* find what to add */
1168         netdev_for_each_uc_addr(ha, dev) {
1169                 found = false;
1170                 bucket = &priv->mac_hash[ha->addr[MLX4_EN_MAC_HASH_IDX]];
1171                 hlist_for_each_entry(entry, bucket, hlist) {
1172                         if (ether_addr_equal_64bits(entry->mac, ha->addr)) {
1173                                 found = true;
1174                                 break;
1175                         }
1176                 }
1177
1178                 if (!found) {
1179                         entry = kmalloc(sizeof(*entry), GFP_KERNEL);
1180                         if (!entry) {
1181                                 en_err(priv, "Failed adding MAC %pM on port:%d (out of memory)\n",
1182                                        ha->addr, priv->port);
1183                                 priv->flags |= MLX4_EN_FLAG_FORCE_PROMISC;
1184                                 break;
1185                         }
1186                         mac = mlx4_mac_to_u64(ha->addr);
1187                         memcpy(entry->mac, ha->addr, ETH_ALEN);
1188                         err = mlx4_register_mac(mdev->dev, priv->port, mac);
1189                         if (err < 0) {
1190                                 en_err(priv, "Failed registering MAC %pM on port %d: %d\n",
1191                                        ha->addr, priv->port, err);
1192                                 kfree(entry);
1193                                 priv->flags |= MLX4_EN_FLAG_FORCE_PROMISC;
1194                                 break;
1195                         }
1196                         err = mlx4_en_uc_steer_add(priv, ha->addr,
1197                                                    &priv->base_qpn,
1198                                                    &entry->reg_id);
1199                         if (err) {
1200                                 en_err(priv, "Failed adding MAC %pM on port %d: %d\n",
1201                                        ha->addr, priv->port, err);
1202                                 mlx4_unregister_mac(mdev->dev, priv->port, mac);
1203                                 kfree(entry);
1204                                 priv->flags |= MLX4_EN_FLAG_FORCE_PROMISC;
1205                                 break;
1206                         } else {
1207                                 unsigned int mac_hash;
1208                                 en_dbg(DRV, priv, "Added MAC %pM on port:%d\n",
1209                                        ha->addr, priv->port);
1210                                 mac_hash = ha->addr[MLX4_EN_MAC_HASH_IDX];
1211                                 bucket = &priv->mac_hash[mac_hash];
1212                                 hlist_add_head_rcu(&entry->hlist, bucket);
1213                         }
1214                 }
1215         }
1216
1217         if (priv->flags & MLX4_EN_FLAG_FORCE_PROMISC) {
1218                 en_warn(priv, "Forcing promiscuous mode on port:%d\n",
1219                         priv->port);
1220         } else if (prev_flags & MLX4_EN_FLAG_FORCE_PROMISC) {
1221                 en_warn(priv, "Stop forcing promiscuous mode on port:%d\n",
1222                         priv->port);
1223         }
1224 }
1225
1226 static void mlx4_en_do_set_rx_mode(struct work_struct *work)
1227 {
1228         struct mlx4_en_priv *priv = container_of(work, struct mlx4_en_priv,
1229                                                  rx_mode_task);
1230         struct mlx4_en_dev *mdev = priv->mdev;
1231         struct net_device *dev = priv->dev;
1232
1233         mutex_lock(&mdev->state_lock);
1234         if (!mdev->device_up) {
1235                 en_dbg(HW, priv, "Card is not up, ignoring rx mode change.\n");
1236                 goto out;
1237         }
1238         if (!priv->port_up) {
1239                 en_dbg(HW, priv, "Port is down, ignoring rx mode change.\n");
1240                 goto out;
1241         }
1242
1243         if (!netif_carrier_ok(dev)) {
1244                 if (!mlx4_en_QUERY_PORT(mdev, priv->port)) {
1245                         if (priv->port_state.link_state) {
1246                                 priv->last_link_state = MLX4_DEV_EVENT_PORT_UP;
1247                                 netif_carrier_on(dev);
1248                                 en_dbg(LINK, priv, "Link Up\n");
1249                         }
1250                 }
1251         }
1252
1253         if (dev->priv_flags & IFF_UNICAST_FLT)
1254                 mlx4_en_do_uc_filter(priv, dev, mdev);
1255
1256         /* Promsicuous mode: disable all filters */
1257         if ((dev->flags & IFF_PROMISC) ||
1258             (priv->flags & MLX4_EN_FLAG_FORCE_PROMISC)) {
1259                 mlx4_en_set_promisc_mode(priv, mdev);
1260                 goto out;
1261         }
1262
1263         /* Not in promiscuous mode */
1264         if (priv->flags & MLX4_EN_FLAG_PROMISC)
1265                 mlx4_en_clear_promisc_mode(priv, mdev);
1266
1267         mlx4_en_do_multicast(priv, dev, mdev);
1268 out:
1269         mutex_unlock(&mdev->state_lock);
1270 }
1271
1272 #ifdef CONFIG_NET_POLL_CONTROLLER
1273 static void mlx4_en_netpoll(struct net_device *dev)
1274 {
1275         struct mlx4_en_priv *priv = netdev_priv(dev);
1276         struct mlx4_en_cq *cq;
1277         int i;
1278
1279         for (i = 0; i < priv->rx_ring_num; i++) {
1280                 cq = priv->rx_cq[i];
1281                 napi_schedule(&cq->napi);
1282         }
1283 }
1284 #endif
1285
1286 static void mlx4_en_tx_timeout(struct net_device *dev)
1287 {
1288         struct mlx4_en_priv *priv = netdev_priv(dev);
1289         struct mlx4_en_dev *mdev = priv->mdev;
1290         int i;
1291
1292         if (netif_msg_timer(priv))
1293                 en_warn(priv, "Tx timeout called on port:%d\n", priv->port);
1294
1295         for (i = 0; i < priv->tx_ring_num; i++) {
1296                 if (!netif_tx_queue_stopped(netdev_get_tx_queue(dev, i)))
1297                         continue;
1298                 en_warn(priv, "TX timeout on queue: %d, QP: 0x%x, CQ: 0x%x, Cons: 0x%x, Prod: 0x%x\n",
1299                         i, priv->tx_ring[i]->qpn, priv->tx_ring[i]->cqn,
1300                         priv->tx_ring[i]->cons, priv->tx_ring[i]->prod);
1301         }
1302
1303         priv->port_stats.tx_timeout++;
1304         en_dbg(DRV, priv, "Scheduling watchdog\n");
1305         queue_work(mdev->workqueue, &priv->watchdog_task);
1306 }
1307
1308
1309 static struct net_device_stats *mlx4_en_get_stats(struct net_device *dev)
1310 {
1311         struct mlx4_en_priv *priv = netdev_priv(dev);
1312
1313         spin_lock_bh(&priv->stats_lock);
1314         memcpy(&priv->ret_stats, &priv->stats, sizeof(priv->stats));
1315         spin_unlock_bh(&priv->stats_lock);
1316
1317         return &priv->ret_stats;
1318 }
1319
1320 static void mlx4_en_set_default_moderation(struct mlx4_en_priv *priv)
1321 {
1322         struct mlx4_en_cq *cq;
1323         int i;
1324
1325         /* If we haven't received a specific coalescing setting
1326          * (module param), we set the moderation parameters as follows:
1327          * - moder_cnt is set to the number of mtu sized packets to
1328          *   satisfy our coalescing target.
1329          * - moder_time is set to a fixed value.
1330          */
1331         priv->rx_frames = MLX4_EN_RX_COAL_TARGET;
1332         priv->rx_usecs = MLX4_EN_RX_COAL_TIME;
1333         priv->tx_frames = MLX4_EN_TX_COAL_PKTS;
1334         priv->tx_usecs = MLX4_EN_TX_COAL_TIME;
1335         en_dbg(INTR, priv, "Default coalesing params for mtu:%d - rx_frames:%d rx_usecs:%d\n",
1336                priv->dev->mtu, priv->rx_frames, priv->rx_usecs);
1337
1338         /* Setup cq moderation params */
1339         for (i = 0; i < priv->rx_ring_num; i++) {
1340                 cq = priv->rx_cq[i];
1341                 cq->moder_cnt = priv->rx_frames;
1342                 cq->moder_time = priv->rx_usecs;
1343                 priv->last_moder_time[i] = MLX4_EN_AUTO_CONF;
1344                 priv->last_moder_packets[i] = 0;
1345                 priv->last_moder_bytes[i] = 0;
1346         }
1347
1348         for (i = 0; i < priv->tx_ring_num; i++) {
1349                 cq = priv->tx_cq[i];
1350                 cq->moder_cnt = priv->tx_frames;
1351                 cq->moder_time = priv->tx_usecs;
1352         }
1353
1354         /* Reset auto-moderation params */
1355         priv->pkt_rate_low = MLX4_EN_RX_RATE_LOW;
1356         priv->rx_usecs_low = MLX4_EN_RX_COAL_TIME_LOW;
1357         priv->pkt_rate_high = MLX4_EN_RX_RATE_HIGH;
1358         priv->rx_usecs_high = MLX4_EN_RX_COAL_TIME_HIGH;
1359         priv->sample_interval = MLX4_EN_SAMPLE_INTERVAL;
1360         priv->adaptive_rx_coal = 1;
1361         priv->last_moder_jiffies = 0;
1362         priv->last_moder_tx_packets = 0;
1363 }
1364
1365 static void mlx4_en_auto_moderation(struct mlx4_en_priv *priv)
1366 {
1367         unsigned long period = (unsigned long) (jiffies - priv->last_moder_jiffies);
1368         struct mlx4_en_cq *cq;
1369         unsigned long packets;
1370         unsigned long rate;
1371         unsigned long avg_pkt_size;
1372         unsigned long rx_packets;
1373         unsigned long rx_bytes;
1374         unsigned long rx_pkt_diff;
1375         int moder_time;
1376         int ring, err;
1377
1378         if (!priv->adaptive_rx_coal || period < priv->sample_interval * HZ)
1379                 return;
1380
1381         for (ring = 0; ring < priv->rx_ring_num; ring++) {
1382                 spin_lock_bh(&priv->stats_lock);
1383                 rx_packets = priv->rx_ring[ring]->packets;
1384                 rx_bytes = priv->rx_ring[ring]->bytes;
1385                 spin_unlock_bh(&priv->stats_lock);
1386
1387                 rx_pkt_diff = ((unsigned long) (rx_packets -
1388                                 priv->last_moder_packets[ring]));
1389                 packets = rx_pkt_diff;
1390                 rate = packets * HZ / period;
1391                 avg_pkt_size = packets ? ((unsigned long) (rx_bytes -
1392                                 priv->last_moder_bytes[ring])) / packets : 0;
1393
1394                 /* Apply auto-moderation only when packet rate
1395                  * exceeds a rate that it matters */
1396                 if (rate > (MLX4_EN_RX_RATE_THRESH / priv->rx_ring_num) &&
1397                     avg_pkt_size > MLX4_EN_AVG_PKT_SMALL) {
1398                         if (rate < priv->pkt_rate_low)
1399                                 moder_time = priv->rx_usecs_low;
1400                         else if (rate > priv->pkt_rate_high)
1401                                 moder_time = priv->rx_usecs_high;
1402                         else
1403                                 moder_time = (rate - priv->pkt_rate_low) *
1404                                         (priv->rx_usecs_high - priv->rx_usecs_low) /
1405                                         (priv->pkt_rate_high - priv->pkt_rate_low) +
1406                                         priv->rx_usecs_low;
1407                 } else {
1408                         moder_time = priv->rx_usecs_low;
1409                 }
1410
1411                 if (moder_time != priv->last_moder_time[ring]) {
1412                         priv->last_moder_time[ring] = moder_time;
1413                         cq = priv->rx_cq[ring];
1414                         cq->moder_time = moder_time;
1415                         cq->moder_cnt = priv->rx_frames;
1416                         err = mlx4_en_set_cq_moder(priv, cq);
1417                         if (err)
1418                                 en_err(priv, "Failed modifying moderation for cq:%d\n",
1419                                        ring);
1420                 }
1421                 priv->last_moder_packets[ring] = rx_packets;
1422                 priv->last_moder_bytes[ring] = rx_bytes;
1423         }
1424
1425         priv->last_moder_jiffies = jiffies;
1426 }
1427
1428 static void mlx4_en_do_get_stats(struct work_struct *work)
1429 {
1430         struct delayed_work *delay = to_delayed_work(work);
1431         struct mlx4_en_priv *priv = container_of(delay, struct mlx4_en_priv,
1432                                                  stats_task);
1433         struct mlx4_en_dev *mdev = priv->mdev;
1434         int err;
1435
1436         mutex_lock(&mdev->state_lock);
1437         if (mdev->device_up) {
1438                 if (priv->port_up) {
1439                         err = mlx4_en_DUMP_ETH_STATS(mdev, priv->port, 0);
1440                         if (err)
1441                                 en_dbg(HW, priv, "Could not update stats\n");
1442
1443                         mlx4_en_auto_moderation(priv);
1444                 }
1445
1446                 queue_delayed_work(mdev->workqueue, &priv->stats_task, STATS_DELAY);
1447         }
1448         if (mdev->mac_removed[MLX4_MAX_PORTS + 1 - priv->port]) {
1449                 mlx4_en_do_set_mac(priv, priv->current_mac);
1450                 mdev->mac_removed[MLX4_MAX_PORTS + 1 - priv->port] = 0;
1451         }
1452         mutex_unlock(&mdev->state_lock);
1453 }
1454
1455 /* mlx4_en_service_task - Run service task for tasks that needed to be done
1456  * periodically
1457  */
1458 static void mlx4_en_service_task(struct work_struct *work)
1459 {
1460         struct delayed_work *delay = to_delayed_work(work);
1461         struct mlx4_en_priv *priv = container_of(delay, struct mlx4_en_priv,
1462                                                  service_task);
1463         struct mlx4_en_dev *mdev = priv->mdev;
1464
1465         mutex_lock(&mdev->state_lock);
1466         if (mdev->device_up) {
1467                 if (mdev->dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_TS)
1468                         mlx4_en_ptp_overflow_check(mdev);
1469
1470                 mlx4_en_recover_from_oom(priv);
1471                 queue_delayed_work(mdev->workqueue, &priv->service_task,
1472                                    SERVICE_TASK_DELAY);
1473         }
1474         mutex_unlock(&mdev->state_lock);
1475 }
1476
1477 static void mlx4_en_linkstate(struct work_struct *work)
1478 {
1479         struct mlx4_en_priv *priv = container_of(work, struct mlx4_en_priv,
1480                                                  linkstate_task);
1481         struct mlx4_en_dev *mdev = priv->mdev;
1482         int linkstate = priv->link_state;
1483
1484         mutex_lock(&mdev->state_lock);
1485         /* If observable port state changed set carrier state and
1486          * report to system log */
1487         if (priv->last_link_state != linkstate) {
1488                 if (linkstate == MLX4_DEV_EVENT_PORT_DOWN) {
1489                         en_info(priv, "Link Down\n");
1490                         netif_carrier_off(priv->dev);
1491                 } else {
1492                         en_info(priv, "Link Up\n");
1493                         netif_carrier_on(priv->dev);
1494                 }
1495         }
1496         priv->last_link_state = linkstate;
1497         mutex_unlock(&mdev->state_lock);
1498 }
1499
1500 static int mlx4_en_init_affinity_hint(struct mlx4_en_priv *priv, int ring_idx)
1501 {
1502         struct mlx4_en_rx_ring *ring = priv->rx_ring[ring_idx];
1503         int numa_node = priv->mdev->dev->numa_node;
1504
1505         if (!zalloc_cpumask_var(&ring->affinity_mask, GFP_KERNEL))
1506                 return -ENOMEM;
1507
1508         cpumask_set_cpu(cpumask_local_spread(ring_idx, numa_node),
1509                         ring->affinity_mask);
1510         return 0;
1511 }
1512
1513 static void mlx4_en_free_affinity_hint(struct mlx4_en_priv *priv, int ring_idx)
1514 {
1515         free_cpumask_var(priv->rx_ring[ring_idx]->affinity_mask);
1516 }
1517
1518 int mlx4_en_start_port(struct net_device *dev)
1519 {
1520         struct mlx4_en_priv *priv = netdev_priv(dev);
1521         struct mlx4_en_dev *mdev = priv->mdev;
1522         struct mlx4_en_cq *cq;
1523         struct mlx4_en_tx_ring *tx_ring;
1524         int rx_index = 0;
1525         int tx_index = 0;
1526         int err = 0;
1527         int i;
1528         int j;
1529         u8 mc_list[16] = {0};
1530
1531         if (priv->port_up) {
1532                 en_dbg(DRV, priv, "start port called while port already up\n");
1533                 return 0;
1534         }
1535
1536         INIT_LIST_HEAD(&priv->mc_list);
1537         INIT_LIST_HEAD(&priv->curr_list);
1538         INIT_LIST_HEAD(&priv->ethtool_list);
1539         memset(&priv->ethtool_rules[0], 0,
1540                sizeof(struct ethtool_flow_id) * MAX_NUM_OF_FS_RULES);
1541
1542         /* Calculate Rx buf size */
1543         dev->mtu = min(dev->mtu, priv->max_mtu);
1544         mlx4_en_calc_rx_buf(dev);
1545         en_dbg(DRV, priv, "Rx buf size:%d\n", priv->rx_skb_size);
1546
1547         /* Configure rx cq's and rings */
1548         err = mlx4_en_activate_rx_rings(priv);
1549         if (err) {
1550                 en_err(priv, "Failed to activate RX rings\n");
1551                 return err;
1552         }
1553         for (i = 0; i < priv->rx_ring_num; i++) {
1554                 cq = priv->rx_cq[i];
1555
1556                 mlx4_en_cq_init_lock(cq);
1557
1558                 err = mlx4_en_init_affinity_hint(priv, i);
1559                 if (err) {
1560                         en_err(priv, "Failed preparing IRQ affinity hint\n");
1561                         goto cq_err;
1562                 }
1563
1564                 err = mlx4_en_activate_cq(priv, cq, i);
1565                 if (err) {
1566                         en_err(priv, "Failed activating Rx CQ\n");
1567                         mlx4_en_free_affinity_hint(priv, i);
1568                         goto cq_err;
1569                 }
1570
1571                 for (j = 0; j < cq->size; j++) {
1572                         struct mlx4_cqe *cqe = NULL;
1573
1574                         cqe = mlx4_en_get_cqe(cq->buf, j, priv->cqe_size) +
1575                               priv->cqe_factor;
1576                         cqe->owner_sr_opcode = MLX4_CQE_OWNER_MASK;
1577                 }
1578
1579                 err = mlx4_en_set_cq_moder(priv, cq);
1580                 if (err) {
1581                         en_err(priv, "Failed setting cq moderation parameters\n");
1582                         mlx4_en_deactivate_cq(priv, cq);
1583                         mlx4_en_free_affinity_hint(priv, i);
1584                         goto cq_err;
1585                 }
1586                 mlx4_en_arm_cq(priv, cq);
1587                 priv->rx_ring[i]->cqn = cq->mcq.cqn;
1588                 ++rx_index;
1589         }
1590
1591         /* Set qp number */
1592         en_dbg(DRV, priv, "Getting qp number for port %d\n", priv->port);
1593         err = mlx4_en_get_qp(priv);
1594         if (err) {
1595                 en_err(priv, "Failed getting eth qp\n");
1596                 goto cq_err;
1597         }
1598         mdev->mac_removed[priv->port] = 0;
1599
1600         err = mlx4_en_config_rss_steer(priv);
1601         if (err) {
1602                 en_err(priv, "Failed configuring rss steering\n");
1603                 goto mac_err;
1604         }
1605
1606         err = mlx4_en_create_drop_qp(priv);
1607         if (err)
1608                 goto rss_err;
1609
1610         /* Configure tx cq's and rings */
1611         for (i = 0; i < priv->tx_ring_num; i++) {
1612                 /* Configure cq */
1613                 cq = priv->tx_cq[i];
1614                 err = mlx4_en_activate_cq(priv, cq, i);
1615                 if (err) {
1616                         en_err(priv, "Failed allocating Tx CQ\n");
1617                         goto tx_err;
1618                 }
1619                 err = mlx4_en_set_cq_moder(priv, cq);
1620                 if (err) {
1621                         en_err(priv, "Failed setting cq moderation parameters\n");
1622                         mlx4_en_deactivate_cq(priv, cq);
1623                         goto tx_err;
1624                 }
1625                 en_dbg(DRV, priv, "Resetting index of collapsed CQ:%d to -1\n", i);
1626                 cq->buf->wqe_index = cpu_to_be16(0xffff);
1627
1628                 /* Configure ring */
1629                 tx_ring = priv->tx_ring[i];
1630                 err = mlx4_en_activate_tx_ring(priv, tx_ring, cq->mcq.cqn,
1631                         i / priv->num_tx_rings_p_up);
1632                 if (err) {
1633                         en_err(priv, "Failed allocating Tx ring\n");
1634                         mlx4_en_deactivate_cq(priv, cq);
1635                         goto tx_err;
1636                 }
1637                 tx_ring->tx_queue = netdev_get_tx_queue(dev, i);
1638
1639                 /* Arm CQ for TX completions */
1640                 mlx4_en_arm_cq(priv, cq);
1641
1642                 /* Set initial ownership of all Tx TXBBs to SW (1) */
1643                 for (j = 0; j < tx_ring->buf_size; j += STAMP_STRIDE)
1644                         *((u32 *) (tx_ring->buf + j)) = 0xffffffff;
1645                 ++tx_index;
1646         }
1647
1648         /* Configure port */
1649         err = mlx4_SET_PORT_general(mdev->dev, priv->port,
1650                                     priv->rx_skb_size + ETH_FCS_LEN,
1651                                     priv->prof->tx_pause,
1652                                     priv->prof->tx_ppp,
1653                                     priv->prof->rx_pause,
1654                                     priv->prof->rx_ppp);
1655         if (err) {
1656                 en_err(priv, "Failed setting port general configurations for port %d, with error %d\n",
1657                        priv->port, err);
1658                 goto tx_err;
1659         }
1660         /* Set default qp number */
1661         err = mlx4_SET_PORT_qpn_calc(mdev->dev, priv->port, priv->base_qpn, 0);
1662         if (err) {
1663                 en_err(priv, "Failed setting default qp numbers\n");
1664                 goto tx_err;
1665         }
1666
1667         if (mdev->dev->caps.tunnel_offload_mode == MLX4_TUNNEL_OFFLOAD_MODE_VXLAN) {
1668                 err = mlx4_SET_PORT_VXLAN(mdev->dev, priv->port, VXLAN_STEER_BY_OUTER_MAC, 1);
1669                 if (err) {
1670                         en_err(priv, "Failed setting port L2 tunnel configuration, err %d\n",
1671                                err);
1672                         goto tx_err;
1673                 }
1674         }
1675
1676         /* Init port */
1677         en_dbg(HW, priv, "Initializing port\n");
1678         err = mlx4_INIT_PORT(mdev->dev, priv->port);
1679         if (err) {
1680                 en_err(priv, "Failed Initializing port\n");
1681                 goto tx_err;
1682         }
1683
1684         /* Attach rx QP to bradcast address */
1685         eth_broadcast_addr(&mc_list[10]);
1686         mc_list[5] = priv->port; /* needed for B0 steering support */
1687         if (mlx4_multicast_attach(mdev->dev, &priv->rss_map.indir_qp, mc_list,
1688                                   priv->port, 0, MLX4_PROT_ETH,
1689                                   &priv->broadcast_id))
1690                 mlx4_warn(mdev, "Failed Attaching Broadcast\n");
1691
1692         /* Must redo promiscuous mode setup. */
1693         priv->flags &= ~(MLX4_EN_FLAG_PROMISC | MLX4_EN_FLAG_MC_PROMISC);
1694
1695         /* Schedule multicast task to populate multicast list */
1696         queue_work(mdev->workqueue, &priv->rx_mode_task);
1697
1698 #ifdef CONFIG_MLX4_EN_VXLAN
1699         if (priv->mdev->dev->caps.tunnel_offload_mode == MLX4_TUNNEL_OFFLOAD_MODE_VXLAN)
1700                 vxlan_get_rx_port(dev);
1701 #endif
1702         priv->port_up = true;
1703         netif_tx_start_all_queues(dev);
1704         netif_device_attach(dev);
1705
1706         return 0;
1707
1708 tx_err:
1709         while (tx_index--) {
1710                 mlx4_en_deactivate_tx_ring(priv, priv->tx_ring[tx_index]);
1711                 mlx4_en_deactivate_cq(priv, priv->tx_cq[tx_index]);
1712         }
1713         mlx4_en_destroy_drop_qp(priv);
1714 rss_err:
1715         mlx4_en_release_rss_steer(priv);
1716 mac_err:
1717         mlx4_en_put_qp(priv);
1718 cq_err:
1719         while (rx_index--) {
1720                 mlx4_en_deactivate_cq(priv, priv->rx_cq[rx_index]);
1721                 mlx4_en_free_affinity_hint(priv, rx_index);
1722         }
1723         for (i = 0; i < priv->rx_ring_num; i++)
1724                 mlx4_en_deactivate_rx_ring(priv, priv->rx_ring[i]);
1725
1726         return err; /* need to close devices */
1727 }
1728
1729
1730 void mlx4_en_stop_port(struct net_device *dev, int detach)
1731 {
1732         struct mlx4_en_priv *priv = netdev_priv(dev);
1733         struct mlx4_en_dev *mdev = priv->mdev;
1734         struct mlx4_en_mc_list *mclist, *tmp;
1735         struct ethtool_flow_id *flow, *tmp_flow;
1736         int i;
1737         u8 mc_list[16] = {0};
1738
1739         if (!priv->port_up) {
1740                 en_dbg(DRV, priv, "stop port called while port already down\n");
1741                 return;
1742         }
1743
1744         /* close port*/
1745         mlx4_CLOSE_PORT(mdev->dev, priv->port);
1746
1747         /* Synchronize with tx routine */
1748         netif_tx_lock_bh(dev);
1749         if (detach)
1750                 netif_device_detach(dev);
1751         netif_tx_stop_all_queues(dev);
1752         netif_tx_unlock_bh(dev);
1753
1754         netif_tx_disable(dev);
1755
1756         /* Set port as not active */
1757         priv->port_up = false;
1758
1759         /* Promsicuous mode */
1760         if (mdev->dev->caps.steering_mode ==
1761             MLX4_STEERING_MODE_DEVICE_MANAGED) {
1762                 priv->flags &= ~(MLX4_EN_FLAG_PROMISC |
1763                                  MLX4_EN_FLAG_MC_PROMISC);
1764                 mlx4_flow_steer_promisc_remove(mdev->dev,
1765                                                priv->port,
1766                                                MLX4_FS_ALL_DEFAULT);
1767                 mlx4_flow_steer_promisc_remove(mdev->dev,
1768                                                priv->port,
1769                                                MLX4_FS_MC_DEFAULT);
1770         } else if (priv->flags & MLX4_EN_FLAG_PROMISC) {
1771                 priv->flags &= ~MLX4_EN_FLAG_PROMISC;
1772
1773                 /* Disable promiscouos mode */
1774                 mlx4_unicast_promisc_remove(mdev->dev, priv->base_qpn,
1775                                             priv->port);
1776
1777                 /* Disable Multicast promisc */
1778                 if (priv->flags & MLX4_EN_FLAG_MC_PROMISC) {
1779                         mlx4_multicast_promisc_remove(mdev->dev, priv->base_qpn,
1780                                                       priv->port);
1781                         priv->flags &= ~MLX4_EN_FLAG_MC_PROMISC;
1782                 }
1783         }
1784
1785         /* Detach All multicasts */
1786         eth_broadcast_addr(&mc_list[10]);
1787         mc_list[5] = priv->port; /* needed for B0 steering support */
1788         mlx4_multicast_detach(mdev->dev, &priv->rss_map.indir_qp, mc_list,
1789                               MLX4_PROT_ETH, priv->broadcast_id);
1790         list_for_each_entry(mclist, &priv->curr_list, list) {
1791                 memcpy(&mc_list[10], mclist->addr, ETH_ALEN);
1792                 mc_list[5] = priv->port;
1793                 mlx4_multicast_detach(mdev->dev, &priv->rss_map.indir_qp,
1794                                       mc_list, MLX4_PROT_ETH, mclist->reg_id);
1795                 if (mclist->tunnel_reg_id)
1796                         mlx4_flow_detach(mdev->dev, mclist->tunnel_reg_id);
1797         }
1798         mlx4_en_clear_list(dev);
1799         list_for_each_entry_safe(mclist, tmp, &priv->curr_list, list) {
1800                 list_del(&mclist->list);
1801                 kfree(mclist);
1802         }
1803
1804         /* Flush multicast filter */
1805         mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, 0, 1, MLX4_MCAST_CONFIG);
1806
1807         /* Remove flow steering rules for the port*/
1808         if (mdev->dev->caps.steering_mode ==
1809             MLX4_STEERING_MODE_DEVICE_MANAGED) {
1810                 ASSERT_RTNL();
1811                 list_for_each_entry_safe(flow, tmp_flow,
1812                                          &priv->ethtool_list, list) {
1813                         mlx4_flow_detach(mdev->dev, flow->id);
1814                         list_del(&flow->list);
1815                 }
1816         }
1817
1818         mlx4_en_destroy_drop_qp(priv);
1819
1820         /* Free TX Rings */
1821         for (i = 0; i < priv->tx_ring_num; i++) {
1822                 mlx4_en_deactivate_tx_ring(priv, priv->tx_ring[i]);
1823                 mlx4_en_deactivate_cq(priv, priv->tx_cq[i]);
1824         }
1825         msleep(10);
1826
1827         for (i = 0; i < priv->tx_ring_num; i++)
1828                 mlx4_en_free_tx_buf(dev, priv->tx_ring[i]);
1829
1830         /* Free RSS qps */
1831         mlx4_en_release_rss_steer(priv);
1832
1833         /* Unregister Mac address for the port */
1834         mlx4_en_put_qp(priv);
1835         if (!(mdev->dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_REASSIGN_MAC_EN))
1836                 mdev->mac_removed[priv->port] = 1;
1837
1838         /* Free RX Rings */
1839         for (i = 0; i < priv->rx_ring_num; i++) {
1840                 struct mlx4_en_cq *cq = priv->rx_cq[i];
1841
1842                 local_bh_disable();
1843                 while (!mlx4_en_cq_lock_napi(cq)) {
1844                         pr_info("CQ %d locked\n", i);
1845                         mdelay(1);
1846                 }
1847                 local_bh_enable();
1848
1849                 napi_synchronize(&cq->napi);
1850                 mlx4_en_deactivate_rx_ring(priv, priv->rx_ring[i]);
1851                 mlx4_en_deactivate_cq(priv, cq);
1852
1853                 mlx4_en_free_affinity_hint(priv, i);
1854         }
1855 }
1856
1857 static void mlx4_en_restart(struct work_struct *work)
1858 {
1859         struct mlx4_en_priv *priv = container_of(work, struct mlx4_en_priv,
1860                                                  watchdog_task);
1861         struct mlx4_en_dev *mdev = priv->mdev;
1862         struct net_device *dev = priv->dev;
1863
1864         en_dbg(DRV, priv, "Watchdog task called for port %d\n", priv->port);
1865
1866         mutex_lock(&mdev->state_lock);
1867         if (priv->port_up) {
1868                 mlx4_en_stop_port(dev, 1);
1869                 if (mlx4_en_start_port(dev))
1870                         en_err(priv, "Failed restarting port %d\n", priv->port);
1871         }
1872         mutex_unlock(&mdev->state_lock);
1873 }
1874
1875 static void mlx4_en_clear_stats(struct net_device *dev)
1876 {
1877         struct mlx4_en_priv *priv = netdev_priv(dev);
1878         struct mlx4_en_dev *mdev = priv->mdev;
1879         int i;
1880
1881         if (mlx4_en_DUMP_ETH_STATS(mdev, priv->port, 1))
1882                 en_dbg(HW, priv, "Failed dumping statistics\n");
1883
1884         memset(&priv->stats, 0, sizeof(priv->stats));
1885         memset(&priv->pstats, 0, sizeof(priv->pstats));
1886         memset(&priv->pkstats, 0, sizeof(priv->pkstats));
1887         memset(&priv->port_stats, 0, sizeof(priv->port_stats));
1888         memset(&priv->rx_flowstats, 0, sizeof(priv->rx_flowstats));
1889         memset(&priv->tx_flowstats, 0, sizeof(priv->tx_flowstats));
1890         memset(&priv->rx_priority_flowstats, 0,
1891                sizeof(priv->rx_priority_flowstats));
1892         memset(&priv->tx_priority_flowstats, 0,
1893                sizeof(priv->tx_priority_flowstats));
1894
1895         for (i = 0; i < priv->tx_ring_num; i++) {
1896                 priv->tx_ring[i]->bytes = 0;
1897                 priv->tx_ring[i]->packets = 0;
1898                 priv->tx_ring[i]->tx_csum = 0;
1899         }
1900         for (i = 0; i < priv->rx_ring_num; i++) {
1901                 priv->rx_ring[i]->bytes = 0;
1902                 priv->rx_ring[i]->packets = 0;
1903                 priv->rx_ring[i]->csum_ok = 0;
1904                 priv->rx_ring[i]->csum_none = 0;
1905                 priv->rx_ring[i]->csum_complete = 0;
1906         }
1907 }
1908
1909 static int mlx4_en_open(struct net_device *dev)
1910 {
1911         struct mlx4_en_priv *priv = netdev_priv(dev);
1912         struct mlx4_en_dev *mdev = priv->mdev;
1913         int err = 0;
1914
1915         mutex_lock(&mdev->state_lock);
1916
1917         if (!mdev->device_up) {
1918                 en_err(priv, "Cannot open - device down/disabled\n");
1919                 err = -EBUSY;
1920                 goto out;
1921         }
1922
1923         /* Reset HW statistics and SW counters */
1924         mlx4_en_clear_stats(dev);
1925
1926         err = mlx4_en_start_port(dev);
1927         if (err)
1928                 en_err(priv, "Failed starting port:%d\n", priv->port);
1929
1930 out:
1931         mutex_unlock(&mdev->state_lock);
1932         return err;
1933 }
1934
1935
1936 static int mlx4_en_close(struct net_device *dev)
1937 {
1938         struct mlx4_en_priv *priv = netdev_priv(dev);
1939         struct mlx4_en_dev *mdev = priv->mdev;
1940
1941         en_dbg(IFDOWN, priv, "Close port called\n");
1942
1943         mutex_lock(&mdev->state_lock);
1944
1945         mlx4_en_stop_port(dev, 0);
1946         netif_carrier_off(dev);
1947
1948         mutex_unlock(&mdev->state_lock);
1949         return 0;
1950 }
1951
1952 void mlx4_en_free_resources(struct mlx4_en_priv *priv)
1953 {
1954         int i;
1955
1956 #ifdef CONFIG_RFS_ACCEL
1957         free_irq_cpu_rmap(priv->dev->rx_cpu_rmap);
1958         priv->dev->rx_cpu_rmap = NULL;
1959 #endif
1960
1961         for (i = 0; i < priv->tx_ring_num; i++) {
1962                 if (priv->tx_ring && priv->tx_ring[i])
1963                         mlx4_en_destroy_tx_ring(priv, &priv->tx_ring[i]);
1964                 if (priv->tx_cq && priv->tx_cq[i])
1965                         mlx4_en_destroy_cq(priv, &priv->tx_cq[i]);
1966         }
1967
1968         for (i = 0; i < priv->rx_ring_num; i++) {
1969                 if (priv->rx_ring[i])
1970                         mlx4_en_destroy_rx_ring(priv, &priv->rx_ring[i],
1971                                 priv->prof->rx_ring_size, priv->stride);
1972                 if (priv->rx_cq[i])
1973                         mlx4_en_destroy_cq(priv, &priv->rx_cq[i]);
1974         }
1975
1976 }
1977
1978 int mlx4_en_alloc_resources(struct mlx4_en_priv *priv)
1979 {
1980         struct mlx4_en_port_profile *prof = priv->prof;
1981         int i;
1982         int node;
1983
1984         /* Create tx Rings */
1985         for (i = 0; i < priv->tx_ring_num; i++) {
1986                 node = cpu_to_node(i % num_online_cpus());
1987                 if (mlx4_en_create_cq(priv, &priv->tx_cq[i],
1988                                       prof->tx_ring_size, i, TX, node))
1989                         goto err;
1990
1991                 if (mlx4_en_create_tx_ring(priv, &priv->tx_ring[i],
1992                                            prof->tx_ring_size, TXBB_SIZE,
1993                                            node, i))
1994                         goto err;
1995         }
1996
1997         /* Create rx Rings */
1998         for (i = 0; i < priv->rx_ring_num; i++) {
1999                 node = cpu_to_node(i % num_online_cpus());
2000                 if (mlx4_en_create_cq(priv, &priv->rx_cq[i],
2001                                       prof->rx_ring_size, i, RX, node))
2002                         goto err;
2003
2004                 if (mlx4_en_create_rx_ring(priv, &priv->rx_ring[i],
2005                                            prof->rx_ring_size, priv->stride,
2006                                            node))
2007                         goto err;
2008         }
2009
2010 #ifdef CONFIG_RFS_ACCEL
2011         if (priv->mdev->dev->caps.comp_pool) {
2012                 priv->dev->rx_cpu_rmap = alloc_irq_cpu_rmap(priv->mdev->dev->caps.comp_pool);
2013                 if (!priv->dev->rx_cpu_rmap)
2014                         goto err;
2015         }
2016 #endif
2017
2018         return 0;
2019
2020 err:
2021         en_err(priv, "Failed to allocate NIC resources\n");
2022         for (i = 0; i < priv->rx_ring_num; i++) {
2023                 if (priv->rx_ring[i])
2024                         mlx4_en_destroy_rx_ring(priv, &priv->rx_ring[i],
2025                                                 prof->rx_ring_size,
2026                                                 priv->stride);
2027                 if (priv->rx_cq[i])
2028                         mlx4_en_destroy_cq(priv, &priv->rx_cq[i]);
2029         }
2030         for (i = 0; i < priv->tx_ring_num; i++) {
2031                 if (priv->tx_ring[i])
2032                         mlx4_en_destroy_tx_ring(priv, &priv->tx_ring[i]);
2033                 if (priv->tx_cq[i])
2034                         mlx4_en_destroy_cq(priv, &priv->tx_cq[i]);
2035         }
2036         return -ENOMEM;
2037 }
2038
2039
2040 void mlx4_en_destroy_netdev(struct net_device *dev)
2041 {
2042         struct mlx4_en_priv *priv = netdev_priv(dev);
2043         struct mlx4_en_dev *mdev = priv->mdev;
2044
2045         en_dbg(DRV, priv, "Destroying netdev on port:%d\n", priv->port);
2046
2047         /* Unregister device - this will close the port if it was up */
2048         if (priv->registered)
2049                 unregister_netdev(dev);
2050
2051         if (priv->allocated)
2052                 mlx4_free_hwq_res(mdev->dev, &priv->res, MLX4_EN_PAGE_SIZE);
2053
2054         cancel_delayed_work(&priv->stats_task);
2055         cancel_delayed_work(&priv->service_task);
2056         /* flush any pending task for this netdev */
2057         flush_workqueue(mdev->workqueue);
2058
2059         /* Detach the netdev so tasks would not attempt to access it */
2060         mutex_lock(&mdev->state_lock);
2061         mdev->pndev[priv->port] = NULL;
2062         mdev->upper[priv->port] = NULL;
2063         mutex_unlock(&mdev->state_lock);
2064
2065         mlx4_en_free_resources(priv);
2066
2067         kfree(priv->tx_ring);
2068         kfree(priv->tx_cq);
2069
2070         free_netdev(dev);
2071 }
2072
2073 static int mlx4_en_change_mtu(struct net_device *dev, int new_mtu)
2074 {
2075         struct mlx4_en_priv *priv = netdev_priv(dev);
2076         struct mlx4_en_dev *mdev = priv->mdev;
2077         int err = 0;
2078
2079         en_dbg(DRV, priv, "Change MTU called - current:%d new:%d\n",
2080                  dev->mtu, new_mtu);
2081
2082         if ((new_mtu < MLX4_EN_MIN_MTU) || (new_mtu > priv->max_mtu)) {
2083                 en_err(priv, "Bad MTU size:%d.\n", new_mtu);
2084                 return -EPERM;
2085         }
2086         dev->mtu = new_mtu;
2087
2088         if (netif_running(dev)) {
2089                 mutex_lock(&mdev->state_lock);
2090                 if (!mdev->device_up) {
2091                         /* NIC is probably restarting - let watchdog task reset
2092                          * the port */
2093                         en_dbg(DRV, priv, "Change MTU called with card down!?\n");
2094                 } else {
2095                         mlx4_en_stop_port(dev, 1);
2096                         err = mlx4_en_start_port(dev);
2097                         if (err) {
2098                                 en_err(priv, "Failed restarting port:%d\n",
2099                                          priv->port);
2100                                 queue_work(mdev->workqueue, &priv->watchdog_task);
2101                         }
2102                 }
2103                 mutex_unlock(&mdev->state_lock);
2104         }
2105         return 0;
2106 }
2107
2108 static int mlx4_en_hwtstamp_set(struct net_device *dev, struct ifreq *ifr)
2109 {
2110         struct mlx4_en_priv *priv = netdev_priv(dev);
2111         struct mlx4_en_dev *mdev = priv->mdev;
2112         struct hwtstamp_config config;
2113
2114         if (copy_from_user(&config, ifr->ifr_data, sizeof(config)))
2115                 return -EFAULT;
2116
2117         /* reserved for future extensions */
2118         if (config.flags)
2119                 return -EINVAL;
2120
2121         /* device doesn't support time stamping */
2122         if (!(mdev->dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_TS))
2123                 return -EINVAL;
2124
2125         /* TX HW timestamp */
2126         switch (config.tx_type) {
2127         case HWTSTAMP_TX_OFF:
2128         case HWTSTAMP_TX_ON:
2129                 break;
2130         default:
2131                 return -ERANGE;
2132         }
2133
2134         /* RX HW timestamp */
2135         switch (config.rx_filter) {
2136         case HWTSTAMP_FILTER_NONE:
2137                 break;
2138         case HWTSTAMP_FILTER_ALL:
2139         case HWTSTAMP_FILTER_SOME:
2140         case HWTSTAMP_FILTER_PTP_V1_L4_EVENT:
2141         case HWTSTAMP_FILTER_PTP_V1_L4_SYNC:
2142         case HWTSTAMP_FILTER_PTP_V1_L4_DELAY_REQ:
2143         case HWTSTAMP_FILTER_PTP_V2_L4_EVENT:
2144         case HWTSTAMP_FILTER_PTP_V2_L4_SYNC:
2145         case HWTSTAMP_FILTER_PTP_V2_L4_DELAY_REQ:
2146         case HWTSTAMP_FILTER_PTP_V2_L2_EVENT:
2147         case HWTSTAMP_FILTER_PTP_V2_L2_SYNC:
2148         case HWTSTAMP_FILTER_PTP_V2_L2_DELAY_REQ:
2149         case HWTSTAMP_FILTER_PTP_V2_EVENT:
2150         case HWTSTAMP_FILTER_PTP_V2_SYNC:
2151         case HWTSTAMP_FILTER_PTP_V2_DELAY_REQ:
2152                 config.rx_filter = HWTSTAMP_FILTER_ALL;
2153                 break;
2154         default:
2155                 return -ERANGE;
2156         }
2157
2158         if (mlx4_en_reset_config(dev, config, dev->features)) {
2159                 config.tx_type = HWTSTAMP_TX_OFF;
2160                 config.rx_filter = HWTSTAMP_FILTER_NONE;
2161         }
2162
2163         return copy_to_user(ifr->ifr_data, &config,
2164                             sizeof(config)) ? -EFAULT : 0;
2165 }
2166
2167 static int mlx4_en_hwtstamp_get(struct net_device *dev, struct ifreq *ifr)
2168 {
2169         struct mlx4_en_priv *priv = netdev_priv(dev);
2170
2171         return copy_to_user(ifr->ifr_data, &priv->hwtstamp_config,
2172                             sizeof(priv->hwtstamp_config)) ? -EFAULT : 0;
2173 }
2174
2175 static int mlx4_en_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
2176 {
2177         switch (cmd) {
2178         case SIOCSHWTSTAMP:
2179                 return mlx4_en_hwtstamp_set(dev, ifr);
2180         case SIOCGHWTSTAMP:
2181                 return mlx4_en_hwtstamp_get(dev, ifr);
2182         default:
2183                 return -EOPNOTSUPP;
2184         }
2185 }
2186
2187 static int mlx4_en_set_features(struct net_device *netdev,
2188                 netdev_features_t features)
2189 {
2190         struct mlx4_en_priv *priv = netdev_priv(netdev);
2191         bool reset = false;
2192         int ret = 0;
2193
2194         if (DEV_FEATURE_CHANGED(netdev, features, NETIF_F_RXFCS)) {
2195                 en_info(priv, "Turn %s RX-FCS\n",
2196                         (features & NETIF_F_RXFCS) ? "ON" : "OFF");
2197                 reset = true;
2198         }
2199
2200         if (DEV_FEATURE_CHANGED(netdev, features, NETIF_F_RXALL)) {
2201                 u8 ignore_fcs_value = (features & NETIF_F_RXALL) ? 1 : 0;
2202
2203                 en_info(priv, "Turn %s RX-ALL\n",
2204                         ignore_fcs_value ? "ON" : "OFF");
2205                 ret = mlx4_SET_PORT_fcs_check(priv->mdev->dev,
2206                                               priv->port, ignore_fcs_value);
2207                 if (ret)
2208                         return ret;
2209         }
2210
2211         if (DEV_FEATURE_CHANGED(netdev, features, NETIF_F_HW_VLAN_CTAG_RX)) {
2212                 en_info(priv, "Turn %s RX vlan strip offload\n",
2213                         (features & NETIF_F_HW_VLAN_CTAG_RX) ? "ON" : "OFF");
2214                 reset = true;
2215         }
2216
2217         if (DEV_FEATURE_CHANGED(netdev, features, NETIF_F_HW_VLAN_CTAG_TX))
2218                 en_info(priv, "Turn %s TX vlan strip offload\n",
2219                         (features & NETIF_F_HW_VLAN_CTAG_TX) ? "ON" : "OFF");
2220
2221         if (DEV_FEATURE_CHANGED(netdev, features, NETIF_F_LOOPBACK)) {
2222                 en_info(priv, "Turn %s loopback\n",
2223                         (features & NETIF_F_LOOPBACK) ? "ON" : "OFF");
2224                 mlx4_en_update_loopback_state(netdev, features);
2225         }
2226
2227         if (reset) {
2228                 ret = mlx4_en_reset_config(netdev, priv->hwtstamp_config,
2229                                            features);
2230                 if (ret)
2231                         return ret;
2232         }
2233
2234         return 0;
2235 }
2236
2237 static int mlx4_en_set_vf_mac(struct net_device *dev, int queue, u8 *mac)
2238 {
2239         struct mlx4_en_priv *en_priv = netdev_priv(dev);
2240         struct mlx4_en_dev *mdev = en_priv->mdev;
2241         u64 mac_u64 = mlx4_mac_to_u64(mac);
2242
2243         if (!is_valid_ether_addr(mac))
2244                 return -EINVAL;
2245
2246         return mlx4_set_vf_mac(mdev->dev, en_priv->port, queue, mac_u64);
2247 }
2248
2249 static int mlx4_en_set_vf_vlan(struct net_device *dev, int vf, u16 vlan, u8 qos)
2250 {
2251         struct mlx4_en_priv *en_priv = netdev_priv(dev);
2252         struct mlx4_en_dev *mdev = en_priv->mdev;
2253
2254         return mlx4_set_vf_vlan(mdev->dev, en_priv->port, vf, vlan, qos);
2255 }
2256
2257 static int mlx4_en_set_vf_rate(struct net_device *dev, int vf, int min_tx_rate,
2258                                int max_tx_rate)
2259 {
2260         struct mlx4_en_priv *en_priv = netdev_priv(dev);
2261         struct mlx4_en_dev *mdev = en_priv->mdev;
2262
2263         return mlx4_set_vf_rate(mdev->dev, en_priv->port, vf, min_tx_rate,
2264                                 max_tx_rate);
2265 }
2266
2267 static int mlx4_en_set_vf_spoofchk(struct net_device *dev, int vf, bool setting)
2268 {
2269         struct mlx4_en_priv *en_priv = netdev_priv(dev);
2270         struct mlx4_en_dev *mdev = en_priv->mdev;
2271
2272         return mlx4_set_vf_spoofchk(mdev->dev, en_priv->port, vf, setting);
2273 }
2274
2275 static int mlx4_en_get_vf_config(struct net_device *dev, int vf, struct ifla_vf_info *ivf)
2276 {
2277         struct mlx4_en_priv *en_priv = netdev_priv(dev);
2278         struct mlx4_en_dev *mdev = en_priv->mdev;
2279
2280         return mlx4_get_vf_config(mdev->dev, en_priv->port, vf, ivf);
2281 }
2282
2283 static int mlx4_en_set_vf_link_state(struct net_device *dev, int vf, int link_state)
2284 {
2285         struct mlx4_en_priv *en_priv = netdev_priv(dev);
2286         struct mlx4_en_dev *mdev = en_priv->mdev;
2287
2288         return mlx4_set_vf_link_state(mdev->dev, en_priv->port, vf, link_state);
2289 }
2290
2291 #define PORT_ID_BYTE_LEN 8
2292 static int mlx4_en_get_phys_port_id(struct net_device *dev,
2293                                     struct netdev_phys_item_id *ppid)
2294 {
2295         struct mlx4_en_priv *priv = netdev_priv(dev);
2296         struct mlx4_dev *mdev = priv->mdev->dev;
2297         int i;
2298         u64 phys_port_id = mdev->caps.phys_port_id[priv->port];
2299
2300         if (!phys_port_id)
2301                 return -EOPNOTSUPP;
2302
2303         ppid->id_len = sizeof(phys_port_id);
2304         for (i = PORT_ID_BYTE_LEN - 1; i >= 0; --i) {
2305                 ppid->id[i] =  phys_port_id & 0xff;
2306                 phys_port_id >>= 8;
2307         }
2308         return 0;
2309 }
2310
2311 #ifdef CONFIG_MLX4_EN_VXLAN
2312 static void mlx4_en_add_vxlan_offloads(struct work_struct *work)
2313 {
2314         int ret;
2315         struct mlx4_en_priv *priv = container_of(work, struct mlx4_en_priv,
2316                                                  vxlan_add_task);
2317
2318         ret = mlx4_config_vxlan_port(priv->mdev->dev, priv->vxlan_port);
2319         if (ret)
2320                 goto out;
2321
2322         ret = mlx4_SET_PORT_VXLAN(priv->mdev->dev, priv->port,
2323                                   VXLAN_STEER_BY_OUTER_MAC, 1);
2324 out:
2325         if (ret) {
2326                 en_err(priv, "failed setting L2 tunnel configuration ret %d\n", ret);
2327                 return;
2328         }
2329
2330         /* set offloads */
2331         priv->dev->hw_enc_features |= NETIF_F_IP_CSUM | NETIF_F_RXCSUM |
2332                                       NETIF_F_TSO | NETIF_F_GSO_UDP_TUNNEL;
2333         priv->dev->hw_features |= NETIF_F_GSO_UDP_TUNNEL;
2334         priv->dev->features    |= NETIF_F_GSO_UDP_TUNNEL;
2335 }
2336
2337 static void mlx4_en_del_vxlan_offloads(struct work_struct *work)
2338 {
2339         int ret;
2340         struct mlx4_en_priv *priv = container_of(work, struct mlx4_en_priv,
2341                                                  vxlan_del_task);
2342         /* unset offloads */
2343         priv->dev->hw_enc_features &= ~(NETIF_F_IP_CSUM | NETIF_F_RXCSUM |
2344                                       NETIF_F_TSO | NETIF_F_GSO_UDP_TUNNEL);
2345         priv->dev->hw_features &= ~NETIF_F_GSO_UDP_TUNNEL;
2346         priv->dev->features    &= ~NETIF_F_GSO_UDP_TUNNEL;
2347
2348         ret = mlx4_SET_PORT_VXLAN(priv->mdev->dev, priv->port,
2349                                   VXLAN_STEER_BY_OUTER_MAC, 0);
2350         if (ret)
2351                 en_err(priv, "failed setting L2 tunnel configuration ret %d\n", ret);
2352
2353         priv->vxlan_port = 0;
2354 }
2355
2356 static void mlx4_en_add_vxlan_port(struct  net_device *dev,
2357                                    sa_family_t sa_family, __be16 port)
2358 {
2359         struct mlx4_en_priv *priv = netdev_priv(dev);
2360         __be16 current_port;
2361
2362         if (priv->mdev->dev->caps.tunnel_offload_mode != MLX4_TUNNEL_OFFLOAD_MODE_VXLAN)
2363                 return;
2364
2365         if (sa_family == AF_INET6)
2366                 return;
2367
2368         current_port = priv->vxlan_port;
2369         if (current_port && current_port != port) {
2370                 en_warn(priv, "vxlan port %d configured, can't add port %d\n",
2371                         ntohs(current_port), ntohs(port));
2372                 return;
2373         }
2374
2375         priv->vxlan_port = port;
2376         queue_work(priv->mdev->workqueue, &priv->vxlan_add_task);
2377 }
2378
2379 static void mlx4_en_del_vxlan_port(struct  net_device *dev,
2380                                    sa_family_t sa_family, __be16 port)
2381 {
2382         struct mlx4_en_priv *priv = netdev_priv(dev);
2383         __be16 current_port;
2384
2385         if (priv->mdev->dev->caps.tunnel_offload_mode != MLX4_TUNNEL_OFFLOAD_MODE_VXLAN)
2386                 return;
2387
2388         if (sa_family == AF_INET6)
2389                 return;
2390
2391         current_port = priv->vxlan_port;
2392         if (current_port != port) {
2393                 en_dbg(DRV, priv, "vxlan port %d isn't configured, ignoring\n", ntohs(port));
2394                 return;
2395         }
2396
2397         queue_work(priv->mdev->workqueue, &priv->vxlan_del_task);
2398 }
2399
2400 static netdev_features_t mlx4_en_features_check(struct sk_buff *skb,
2401                                                 struct net_device *dev,
2402                                                 netdev_features_t features)
2403 {
2404         features = vlan_features_check(skb, features);
2405         return vxlan_features_check(skb, features);
2406 }
2407 #endif
2408
2409 static int mlx4_en_set_tx_maxrate(struct net_device *dev, int queue_index, u32 maxrate)
2410 {
2411         struct mlx4_en_priv *priv = netdev_priv(dev);
2412         struct mlx4_en_tx_ring *tx_ring = priv->tx_ring[queue_index];
2413         struct mlx4_update_qp_params params;
2414         int err;
2415
2416         if (!(priv->mdev->dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_QP_RATE_LIMIT))
2417                 return -EOPNOTSUPP;
2418
2419         /* rate provided to us in Mbs, check if it fits into 12 bits, if not use Gbs */
2420         if (maxrate >> 12) {
2421                 params.rate_unit = MLX4_QP_RATE_LIMIT_GBS;
2422                 params.rate_val  = maxrate / 1000;
2423         } else if (maxrate) {
2424                 params.rate_unit = MLX4_QP_RATE_LIMIT_MBS;
2425                 params.rate_val  = maxrate;
2426         } else { /* zero serves to revoke the QP rate-limitation */
2427                 params.rate_unit = 0;
2428                 params.rate_val  = 0;
2429         }
2430
2431         err = mlx4_update_qp(priv->mdev->dev, tx_ring->qpn, MLX4_UPDATE_QP_RATE_LIMIT,
2432                              &params);
2433         return err;
2434 }
2435
2436 static const struct net_device_ops mlx4_netdev_ops = {
2437         .ndo_open               = mlx4_en_open,
2438         .ndo_stop               = mlx4_en_close,
2439         .ndo_start_xmit         = mlx4_en_xmit,
2440         .ndo_select_queue       = mlx4_en_select_queue,
2441         .ndo_get_stats          = mlx4_en_get_stats,
2442         .ndo_set_rx_mode        = mlx4_en_set_rx_mode,
2443         .ndo_set_mac_address    = mlx4_en_set_mac,
2444         .ndo_validate_addr      = eth_validate_addr,
2445         .ndo_change_mtu         = mlx4_en_change_mtu,
2446         .ndo_do_ioctl           = mlx4_en_ioctl,
2447         .ndo_tx_timeout         = mlx4_en_tx_timeout,
2448         .ndo_vlan_rx_add_vid    = mlx4_en_vlan_rx_add_vid,
2449         .ndo_vlan_rx_kill_vid   = mlx4_en_vlan_rx_kill_vid,
2450 #ifdef CONFIG_NET_POLL_CONTROLLER
2451         .ndo_poll_controller    = mlx4_en_netpoll,
2452 #endif
2453         .ndo_set_features       = mlx4_en_set_features,
2454         .ndo_setup_tc           = mlx4_en_setup_tc,
2455 #ifdef CONFIG_RFS_ACCEL
2456         .ndo_rx_flow_steer      = mlx4_en_filter_rfs,
2457 #endif
2458 #ifdef CONFIG_NET_RX_BUSY_POLL
2459         .ndo_busy_poll          = mlx4_en_low_latency_recv,
2460 #endif
2461         .ndo_get_phys_port_id   = mlx4_en_get_phys_port_id,
2462 #ifdef CONFIG_MLX4_EN_VXLAN
2463         .ndo_add_vxlan_port     = mlx4_en_add_vxlan_port,
2464         .ndo_del_vxlan_port     = mlx4_en_del_vxlan_port,
2465         .ndo_features_check     = mlx4_en_features_check,
2466 #endif
2467         .ndo_set_tx_maxrate     = mlx4_en_set_tx_maxrate,
2468 };
2469
2470 static const struct net_device_ops mlx4_netdev_ops_master = {
2471         .ndo_open               = mlx4_en_open,
2472         .ndo_stop               = mlx4_en_close,
2473         .ndo_start_xmit         = mlx4_en_xmit,
2474         .ndo_select_queue       = mlx4_en_select_queue,
2475         .ndo_get_stats          = mlx4_en_get_stats,
2476         .ndo_set_rx_mode        = mlx4_en_set_rx_mode,
2477         .ndo_set_mac_address    = mlx4_en_set_mac,
2478         .ndo_validate_addr      = eth_validate_addr,
2479         .ndo_change_mtu         = mlx4_en_change_mtu,
2480         .ndo_tx_timeout         = mlx4_en_tx_timeout,
2481         .ndo_vlan_rx_add_vid    = mlx4_en_vlan_rx_add_vid,
2482         .ndo_vlan_rx_kill_vid   = mlx4_en_vlan_rx_kill_vid,
2483         .ndo_set_vf_mac         = mlx4_en_set_vf_mac,
2484         .ndo_set_vf_vlan        = mlx4_en_set_vf_vlan,
2485         .ndo_set_vf_rate        = mlx4_en_set_vf_rate,
2486         .ndo_set_vf_spoofchk    = mlx4_en_set_vf_spoofchk,
2487         .ndo_set_vf_link_state  = mlx4_en_set_vf_link_state,
2488         .ndo_get_vf_config      = mlx4_en_get_vf_config,
2489 #ifdef CONFIG_NET_POLL_CONTROLLER
2490         .ndo_poll_controller    = mlx4_en_netpoll,
2491 #endif
2492         .ndo_set_features       = mlx4_en_set_features,
2493         .ndo_setup_tc           = mlx4_en_setup_tc,
2494 #ifdef CONFIG_RFS_ACCEL
2495         .ndo_rx_flow_steer      = mlx4_en_filter_rfs,
2496 #endif
2497         .ndo_get_phys_port_id   = mlx4_en_get_phys_port_id,
2498 #ifdef CONFIG_MLX4_EN_VXLAN
2499         .ndo_add_vxlan_port     = mlx4_en_add_vxlan_port,
2500         .ndo_del_vxlan_port     = mlx4_en_del_vxlan_port,
2501         .ndo_features_check     = mlx4_en_features_check,
2502 #endif
2503         .ndo_set_tx_maxrate     = mlx4_en_set_tx_maxrate,
2504 };
2505
2506 struct mlx4_en_bond {
2507         struct work_struct work;
2508         struct mlx4_en_priv *priv;
2509         int is_bonded;
2510         struct mlx4_port_map port_map;
2511 };
2512
2513 static void mlx4_en_bond_work(struct work_struct *work)
2514 {
2515         struct mlx4_en_bond *bond = container_of(work,
2516                                                      struct mlx4_en_bond,
2517                                                      work);
2518         int err = 0;
2519         struct mlx4_dev *dev = bond->priv->mdev->dev;
2520
2521         if (bond->is_bonded) {
2522                 if (!mlx4_is_bonded(dev)) {
2523                         err = mlx4_bond(dev);
2524                         if (err)
2525                                 en_err(bond->priv, "Fail to bond device\n");
2526                 }
2527                 if (!err) {
2528                         err = mlx4_port_map_set(dev, &bond->port_map);
2529                         if (err)
2530                                 en_err(bond->priv, "Fail to set port map [%d][%d]: %d\n",
2531                                        bond->port_map.port1,
2532                                        bond->port_map.port2,
2533                                        err);
2534                 }
2535         } else if (mlx4_is_bonded(dev)) {
2536                 err = mlx4_unbond(dev);
2537                 if (err)
2538                         en_err(bond->priv, "Fail to unbond device\n");
2539         }
2540         dev_put(bond->priv->dev);
2541         kfree(bond);
2542 }
2543
2544 static int mlx4_en_queue_bond_work(struct mlx4_en_priv *priv, int is_bonded,
2545                                    u8 v2p_p1, u8 v2p_p2)
2546 {
2547         struct mlx4_en_bond *bond = NULL;
2548
2549         bond = kzalloc(sizeof(*bond), GFP_ATOMIC);
2550         if (!bond)
2551                 return -ENOMEM;
2552
2553         INIT_WORK(&bond->work, mlx4_en_bond_work);
2554         bond->priv = priv;
2555         bond->is_bonded = is_bonded;
2556         bond->port_map.port1 = v2p_p1;
2557         bond->port_map.port2 = v2p_p2;
2558         dev_hold(priv->dev);
2559         queue_work(priv->mdev->workqueue, &bond->work);
2560         return 0;
2561 }
2562
2563 int mlx4_en_netdev_event(struct notifier_block *this,
2564                          unsigned long event, void *ptr)
2565 {
2566         struct net_device *ndev = netdev_notifier_info_to_dev(ptr);
2567         u8 port = 0;
2568         struct mlx4_en_dev *mdev;
2569         struct mlx4_dev *dev;
2570         int i, num_eth_ports = 0;
2571         bool do_bond = true;
2572         struct mlx4_en_priv *priv;
2573         u8 v2p_port1 = 0;
2574         u8 v2p_port2 = 0;
2575
2576         if (!net_eq(dev_net(ndev), &init_net))
2577                 return NOTIFY_DONE;
2578
2579         mdev = container_of(this, struct mlx4_en_dev, nb);
2580         dev = mdev->dev;
2581
2582         /* Go into this mode only when two network devices set on two ports
2583          * of the same mlx4 device are slaves of the same bonding master
2584          */
2585         mlx4_foreach_port(i, dev, MLX4_PORT_TYPE_ETH) {
2586                 ++num_eth_ports;
2587                 if (!port && (mdev->pndev[i] == ndev))
2588                         port = i;
2589                 mdev->upper[i] = mdev->pndev[i] ?
2590                         netdev_master_upper_dev_get(mdev->pndev[i]) : NULL;
2591                 /* condition not met: network device is a slave */
2592                 if (!mdev->upper[i])
2593                         do_bond = false;
2594                 if (num_eth_ports < 2)
2595                         continue;
2596                 /* condition not met: same master */
2597                 if (mdev->upper[i] != mdev->upper[i-1])
2598                         do_bond = false;
2599         }
2600         /* condition not met: 2 salves */
2601         do_bond = (num_eth_ports ==  2) ? do_bond : false;
2602
2603         /* handle only events that come with enough info */
2604         if ((do_bond && (event != NETDEV_BONDING_INFO)) || !port)
2605                 return NOTIFY_DONE;
2606
2607         priv = netdev_priv(ndev);
2608         if (do_bond) {
2609                 struct netdev_notifier_bonding_info *notifier_info = ptr;
2610                 struct netdev_bonding_info *bonding_info =
2611                         &notifier_info->bonding_info;
2612
2613                 /* required mode 1, 2 or 4 */
2614                 if ((bonding_info->master.bond_mode != BOND_MODE_ACTIVEBACKUP) &&
2615                     (bonding_info->master.bond_mode != BOND_MODE_XOR) &&
2616                     (bonding_info->master.bond_mode != BOND_MODE_8023AD))
2617                         do_bond = false;
2618
2619                 /* require exactly 2 slaves */
2620                 if (bonding_info->master.num_slaves != 2)
2621                         do_bond = false;
2622
2623                 /* calc v2p */
2624                 if (do_bond) {
2625                         if (bonding_info->master.bond_mode ==
2626                             BOND_MODE_ACTIVEBACKUP) {
2627                                 /* in active-backup mode virtual ports are
2628                                  * mapped to the physical port of the active
2629                                  * slave */
2630                                 if (bonding_info->slave.state ==
2631                                     BOND_STATE_BACKUP) {
2632                                         if (port == 1) {
2633                                                 v2p_port1 = 2;
2634                                                 v2p_port2 = 2;
2635                                         } else {
2636                                                 v2p_port1 = 1;
2637                                                 v2p_port2 = 1;
2638                                         }
2639                                 } else { /* BOND_STATE_ACTIVE */
2640                                         if (port == 1) {
2641                                                 v2p_port1 = 1;
2642                                                 v2p_port2 = 1;
2643                                         } else {
2644                                                 v2p_port1 = 2;
2645                                                 v2p_port2 = 2;
2646                                         }
2647                                 }
2648                         } else { /* Active-Active */
2649                                 /* in active-active mode a virtual port is
2650                                  * mapped to the native physical port if and only
2651                                  * if the physical port is up */
2652                                 __s8 link = bonding_info->slave.link;
2653
2654                                 if (port == 1)
2655                                         v2p_port2 = 2;
2656                                 else
2657                                         v2p_port1 = 1;
2658                                 if ((link == BOND_LINK_UP) ||
2659                                     (link == BOND_LINK_FAIL)) {
2660                                         if (port == 1)
2661                                                 v2p_port1 = 1;
2662                                         else
2663                                                 v2p_port2 = 2;
2664                                 } else { /* BOND_LINK_DOWN || BOND_LINK_BACK */
2665                                         if (port == 1)
2666                                                 v2p_port1 = 2;
2667                                         else
2668                                                 v2p_port2 = 1;
2669                                 }
2670                         }
2671                 }
2672         }
2673
2674         mlx4_en_queue_bond_work(priv, do_bond,
2675                                 v2p_port1, v2p_port2);
2676
2677         return NOTIFY_DONE;
2678 }
2679
2680 void mlx4_en_update_pfc_stats_bitmap(struct mlx4_dev *dev,
2681                                      struct mlx4_en_stats_bitmap *stats_bitmap,
2682                                      u8 rx_ppp, u8 rx_pause,
2683                                      u8 tx_ppp, u8 tx_pause)
2684 {
2685         int last_i = NUM_MAIN_STATS + NUM_PORT_STATS;
2686
2687         if (!mlx4_is_slave(dev) &&
2688             (dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_FLOWSTATS_EN)) {
2689                 mutex_lock(&stats_bitmap->mutex);
2690                 bitmap_clear(stats_bitmap->bitmap, last_i, NUM_FLOW_STATS);
2691
2692                 if (rx_ppp)
2693                         bitmap_set(stats_bitmap->bitmap, last_i,
2694                                    NUM_FLOW_PRIORITY_STATS_RX);
2695                 last_i += NUM_FLOW_PRIORITY_STATS_RX;
2696
2697                 if (rx_pause && !(rx_ppp))
2698                         bitmap_set(stats_bitmap->bitmap, last_i,
2699                                    NUM_FLOW_STATS_RX);
2700                 last_i += NUM_FLOW_STATS_RX;
2701
2702                 if (tx_ppp)
2703                         bitmap_set(stats_bitmap->bitmap, last_i,
2704                                    NUM_FLOW_PRIORITY_STATS_TX);
2705                 last_i += NUM_FLOW_PRIORITY_STATS_TX;
2706
2707                 if (tx_pause && !(tx_ppp))
2708                         bitmap_set(stats_bitmap->bitmap, last_i,
2709                                    NUM_FLOW_STATS_TX);
2710                 last_i += NUM_FLOW_STATS_TX;
2711
2712                 mutex_unlock(&stats_bitmap->mutex);
2713         }
2714 }
2715
2716 void mlx4_en_set_stats_bitmap(struct mlx4_dev *dev,
2717                               struct mlx4_en_stats_bitmap *stats_bitmap,
2718                               u8 rx_ppp, u8 rx_pause,
2719                               u8 tx_ppp, u8 tx_pause)
2720 {
2721         int last_i = 0;
2722
2723         mutex_init(&stats_bitmap->mutex);
2724         bitmap_zero(stats_bitmap->bitmap, NUM_ALL_STATS);
2725
2726         if (mlx4_is_slave(dev)) {
2727                 bitmap_set(stats_bitmap->bitmap, last_i +
2728                                          MLX4_FIND_NETDEV_STAT(rx_packets), 1);
2729                 bitmap_set(stats_bitmap->bitmap, last_i +
2730                                          MLX4_FIND_NETDEV_STAT(tx_packets), 1);
2731                 bitmap_set(stats_bitmap->bitmap, last_i +
2732                                          MLX4_FIND_NETDEV_STAT(rx_bytes), 1);
2733                 bitmap_set(stats_bitmap->bitmap, last_i +
2734                                          MLX4_FIND_NETDEV_STAT(tx_bytes), 1);
2735                 bitmap_set(stats_bitmap->bitmap, last_i +
2736                                          MLX4_FIND_NETDEV_STAT(rx_dropped), 1);
2737                 bitmap_set(stats_bitmap->bitmap, last_i +
2738                                          MLX4_FIND_NETDEV_STAT(tx_dropped), 1);
2739         } else {
2740                 bitmap_set(stats_bitmap->bitmap, last_i, NUM_MAIN_STATS);
2741         }
2742         last_i += NUM_MAIN_STATS;
2743
2744         bitmap_set(stats_bitmap->bitmap, last_i, NUM_PORT_STATS);
2745         last_i += NUM_PORT_STATS;
2746
2747         mlx4_en_update_pfc_stats_bitmap(dev, stats_bitmap,
2748                                         rx_ppp, rx_pause,
2749                                         tx_ppp, tx_pause);
2750         last_i += NUM_FLOW_STATS;
2751
2752         if (!mlx4_is_slave(dev))
2753                 bitmap_set(stats_bitmap->bitmap, last_i, NUM_PKT_STATS);
2754 }
2755
2756 int mlx4_en_init_netdev(struct mlx4_en_dev *mdev, int port,
2757                         struct mlx4_en_port_profile *prof)
2758 {
2759         struct net_device *dev;
2760         struct mlx4_en_priv *priv;
2761         int i;
2762         int err;
2763         u64 mac_u64;
2764
2765         dev = alloc_etherdev_mqs(sizeof(struct mlx4_en_priv),
2766                                  MAX_TX_RINGS, MAX_RX_RINGS);
2767         if (dev == NULL)
2768                 return -ENOMEM;
2769
2770         netif_set_real_num_tx_queues(dev, prof->tx_ring_num);
2771         netif_set_real_num_rx_queues(dev, prof->rx_ring_num);
2772
2773         SET_NETDEV_DEV(dev, &mdev->dev->persist->pdev->dev);
2774         dev->dev_port = port - 1;
2775
2776         /*
2777          * Initialize driver private data
2778          */
2779
2780         priv = netdev_priv(dev);
2781         memset(priv, 0, sizeof(struct mlx4_en_priv));
2782         spin_lock_init(&priv->stats_lock);
2783         INIT_WORK(&priv->rx_mode_task, mlx4_en_do_set_rx_mode);
2784         INIT_WORK(&priv->watchdog_task, mlx4_en_restart);
2785         INIT_WORK(&priv->linkstate_task, mlx4_en_linkstate);
2786         INIT_DELAYED_WORK(&priv->stats_task, mlx4_en_do_get_stats);
2787         INIT_DELAYED_WORK(&priv->service_task, mlx4_en_service_task);
2788 #ifdef CONFIG_MLX4_EN_VXLAN
2789         INIT_WORK(&priv->vxlan_add_task, mlx4_en_add_vxlan_offloads);
2790         INIT_WORK(&priv->vxlan_del_task, mlx4_en_del_vxlan_offloads);
2791 #endif
2792 #ifdef CONFIG_RFS_ACCEL
2793         INIT_LIST_HEAD(&priv->filters);
2794         spin_lock_init(&priv->filters_lock);
2795 #endif
2796
2797         priv->dev = dev;
2798         priv->mdev = mdev;
2799         priv->ddev = &mdev->pdev->dev;
2800         priv->prof = prof;
2801         priv->port = port;
2802         priv->port_up = false;
2803         priv->flags = prof->flags;
2804         priv->pflags = MLX4_EN_PRIV_FLAGS_BLUEFLAME;
2805         priv->ctrl_flags = cpu_to_be32(MLX4_WQE_CTRL_CQ_UPDATE |
2806                         MLX4_WQE_CTRL_SOLICITED);
2807         priv->num_tx_rings_p_up = mdev->profile.num_tx_rings_p_up;
2808         priv->tx_ring_num = prof->tx_ring_num;
2809         priv->tx_work_limit = MLX4_EN_DEFAULT_TX_WORK;
2810         netdev_rss_key_fill(priv->rss_key, sizeof(priv->rss_key));
2811
2812         priv->tx_ring = kzalloc(sizeof(struct mlx4_en_tx_ring *) * MAX_TX_RINGS,
2813                                 GFP_KERNEL);
2814         if (!priv->tx_ring) {
2815                 err = -ENOMEM;
2816                 goto out;
2817         }
2818         priv->tx_cq = kzalloc(sizeof(struct mlx4_en_cq *) * MAX_TX_RINGS,
2819                               GFP_KERNEL);
2820         if (!priv->tx_cq) {
2821                 err = -ENOMEM;
2822                 goto out;
2823         }
2824         priv->rx_ring_num = prof->rx_ring_num;
2825         priv->cqe_factor = (mdev->dev->caps.cqe_size == 64) ? 1 : 0;
2826         priv->cqe_size = mdev->dev->caps.cqe_size;
2827         priv->mac_index = -1;
2828         priv->msg_enable = MLX4_EN_MSG_LEVEL;
2829 #ifdef CONFIG_MLX4_EN_DCB
2830         if (!mlx4_is_slave(priv->mdev->dev)) {
2831                 if (mdev->dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_ETS_CFG) {
2832                         dev->dcbnl_ops = &mlx4_en_dcbnl_ops;
2833                 } else {
2834                         en_info(priv, "enabling only PFC DCB ops\n");
2835                         dev->dcbnl_ops = &mlx4_en_dcbnl_pfc_ops;
2836                 }
2837         }
2838 #endif
2839
2840         for (i = 0; i < MLX4_EN_MAC_HASH_SIZE; ++i)
2841                 INIT_HLIST_HEAD(&priv->mac_hash[i]);
2842
2843         /* Query for default mac and max mtu */
2844         priv->max_mtu = mdev->dev->caps.eth_mtu_cap[priv->port];
2845
2846         if (mdev->dev->caps.rx_checksum_flags_port[priv->port] &
2847             MLX4_RX_CSUM_MODE_VAL_NON_TCP_UDP)
2848                 priv->flags |= MLX4_EN_FLAG_RX_CSUM_NON_TCP_UDP;
2849
2850         /* Set default MAC */
2851         dev->addr_len = ETH_ALEN;
2852         mlx4_en_u64_to_mac(dev->dev_addr, mdev->dev->caps.def_mac[priv->port]);
2853         if (!is_valid_ether_addr(dev->dev_addr)) {
2854                 if (mlx4_is_slave(priv->mdev->dev)) {
2855                         eth_hw_addr_random(dev);
2856                         en_warn(priv, "Assigned random MAC address %pM\n", dev->dev_addr);
2857                         mac_u64 = mlx4_mac_to_u64(dev->dev_addr);
2858                         mdev->dev->caps.def_mac[priv->port] = mac_u64;
2859                 } else {
2860                         en_err(priv, "Port: %d, invalid mac burned: %pM, quiting\n",
2861                                priv->port, dev->dev_addr);
2862                         err = -EINVAL;
2863                         goto out;
2864                 }
2865         }
2866
2867         memcpy(priv->current_mac, dev->dev_addr, sizeof(priv->current_mac));
2868
2869         priv->stride = roundup_pow_of_two(sizeof(struct mlx4_en_rx_desc) +
2870                                           DS_SIZE * MLX4_EN_MAX_RX_FRAGS);
2871         err = mlx4_en_alloc_resources(priv);
2872         if (err)
2873                 goto out;
2874
2875         /* Initialize time stamping config */
2876         priv->hwtstamp_config.flags = 0;
2877         priv->hwtstamp_config.tx_type = HWTSTAMP_TX_OFF;
2878         priv->hwtstamp_config.rx_filter = HWTSTAMP_FILTER_NONE;
2879
2880         /* Allocate page for receive rings */
2881         err = mlx4_alloc_hwq_res(mdev->dev, &priv->res,
2882                                 MLX4_EN_PAGE_SIZE, MLX4_EN_PAGE_SIZE);
2883         if (err) {
2884                 en_err(priv, "Failed to allocate page for rx qps\n");
2885                 goto out;
2886         }
2887         priv->allocated = 1;
2888
2889         /*
2890          * Initialize netdev entry points
2891          */
2892         if (mlx4_is_master(priv->mdev->dev))
2893                 dev->netdev_ops = &mlx4_netdev_ops_master;
2894         else
2895                 dev->netdev_ops = &mlx4_netdev_ops;
2896         dev->watchdog_timeo = MLX4_EN_WATCHDOG_TIMEOUT;
2897         netif_set_real_num_tx_queues(dev, priv->tx_ring_num);
2898         netif_set_real_num_rx_queues(dev, priv->rx_ring_num);
2899
2900         dev->ethtool_ops = &mlx4_en_ethtool_ops;
2901
2902         /*
2903          * Set driver features
2904          */
2905         dev->hw_features = NETIF_F_SG | NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM;
2906         if (mdev->LSO_support)
2907                 dev->hw_features |= NETIF_F_TSO | NETIF_F_TSO6;
2908
2909         dev->vlan_features = dev->hw_features;
2910
2911         dev->hw_features |= NETIF_F_RXCSUM | NETIF_F_RXHASH;
2912         dev->features = dev->hw_features | NETIF_F_HIGHDMA |
2913                         NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_CTAG_RX |
2914                         NETIF_F_HW_VLAN_CTAG_FILTER;
2915         dev->hw_features |= NETIF_F_LOOPBACK |
2916                         NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_CTAG_RX;
2917
2918         if (mdev->dev->caps.flags & MLX4_DEV_CAP_FLAG_FCS_KEEP)
2919                 dev->hw_features |= NETIF_F_RXFCS;
2920
2921         if (mdev->dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_IGNORE_FCS)
2922                 dev->hw_features |= NETIF_F_RXALL;
2923
2924         if (mdev->dev->caps.steering_mode ==
2925             MLX4_STEERING_MODE_DEVICE_MANAGED &&
2926             mdev->dev->caps.dmfs_high_steer_mode != MLX4_STEERING_DMFS_A0_STATIC)
2927                 dev->hw_features |= NETIF_F_NTUPLE;
2928
2929         if (mdev->dev->caps.steering_mode != MLX4_STEERING_MODE_A0)
2930                 dev->priv_flags |= IFF_UNICAST_FLT;
2931
2932         /* Setting a default hash function value */
2933         if (mdev->dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_RSS_TOP) {
2934                 priv->rss_hash_fn = ETH_RSS_HASH_TOP;
2935         } else if (mdev->dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_RSS_XOR) {
2936                 priv->rss_hash_fn = ETH_RSS_HASH_XOR;
2937         } else {
2938                 en_warn(priv,
2939                         "No RSS hash capabilities exposed, using Toeplitz\n");
2940                 priv->rss_hash_fn = ETH_RSS_HASH_TOP;
2941         }
2942
2943         mdev->pndev[port] = dev;
2944         mdev->upper[port] = NULL;
2945
2946         netif_carrier_off(dev);
2947         mlx4_en_set_default_moderation(priv);
2948
2949         en_warn(priv, "Using %d TX rings\n", prof->tx_ring_num);
2950         en_warn(priv, "Using %d RX rings\n", prof->rx_ring_num);
2951
2952         mlx4_en_update_loopback_state(priv->dev, priv->dev->features);
2953
2954         /* Configure port */
2955         mlx4_en_calc_rx_buf(dev);
2956         err = mlx4_SET_PORT_general(mdev->dev, priv->port,
2957                                     priv->rx_skb_size + ETH_FCS_LEN,
2958                                     prof->tx_pause, prof->tx_ppp,
2959                                     prof->rx_pause, prof->rx_ppp);
2960         if (err) {
2961                 en_err(priv, "Failed setting port general configurations for port %d, with error %d\n",
2962                        priv->port, err);
2963                 goto out;
2964         }
2965
2966         if (mdev->dev->caps.tunnel_offload_mode == MLX4_TUNNEL_OFFLOAD_MODE_VXLAN) {
2967                 err = mlx4_SET_PORT_VXLAN(mdev->dev, priv->port, VXLAN_STEER_BY_OUTER_MAC, 1);
2968                 if (err) {
2969                         en_err(priv, "Failed setting port L2 tunnel configuration, err %d\n",
2970                                err);
2971                         goto out;
2972                 }
2973         }
2974
2975         /* Init port */
2976         en_warn(priv, "Initializing port\n");
2977         err = mlx4_INIT_PORT(mdev->dev, priv->port);
2978         if (err) {
2979                 en_err(priv, "Failed Initializing port\n");
2980                 goto out;
2981         }
2982         queue_delayed_work(mdev->workqueue, &priv->stats_task, STATS_DELAY);
2983
2984         if (mdev->dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_TS)
2985                 queue_delayed_work(mdev->workqueue, &priv->service_task,
2986                                    SERVICE_TASK_DELAY);
2987
2988         mlx4_en_set_stats_bitmap(mdev->dev, &priv->stats_bitmap,
2989                                  mdev->profile.prof[priv->port].rx_ppp,
2990                                  mdev->profile.prof[priv->port].rx_pause,
2991                                  mdev->profile.prof[priv->port].tx_ppp,
2992                                  mdev->profile.prof[priv->port].tx_pause);
2993
2994         err = register_netdev(dev);
2995         if (err) {
2996                 en_err(priv, "Netdev registration failed for port %d\n", port);
2997                 goto out;
2998         }
2999
3000         priv->registered = 1;
3001
3002         return 0;
3003
3004 out:
3005         mlx4_en_destroy_netdev(dev);
3006         return err;
3007 }
3008
3009 int mlx4_en_reset_config(struct net_device *dev,
3010                          struct hwtstamp_config ts_config,
3011                          netdev_features_t features)
3012 {
3013         struct mlx4_en_priv *priv = netdev_priv(dev);
3014         struct mlx4_en_dev *mdev = priv->mdev;
3015         int port_up = 0;
3016         int err = 0;
3017
3018         if (priv->hwtstamp_config.tx_type == ts_config.tx_type &&
3019             priv->hwtstamp_config.rx_filter == ts_config.rx_filter &&
3020             !DEV_FEATURE_CHANGED(dev, features, NETIF_F_HW_VLAN_CTAG_RX) &&
3021             !DEV_FEATURE_CHANGED(dev, features, NETIF_F_RXFCS))
3022                 return 0; /* Nothing to change */
3023
3024         if (DEV_FEATURE_CHANGED(dev, features, NETIF_F_HW_VLAN_CTAG_RX) &&
3025             (features & NETIF_F_HW_VLAN_CTAG_RX) &&
3026             (priv->hwtstamp_config.rx_filter != HWTSTAMP_FILTER_NONE)) {
3027                 en_warn(priv, "Can't turn ON rx vlan offload while time-stamping rx filter is ON\n");
3028                 return -EINVAL;
3029         }
3030
3031         mutex_lock(&mdev->state_lock);
3032         if (priv->port_up) {
3033                 port_up = 1;
3034                 mlx4_en_stop_port(dev, 1);
3035         }
3036
3037         mlx4_en_free_resources(priv);
3038
3039         en_warn(priv, "Changing device configuration rx filter(%x) rx vlan(%x)\n",
3040                 ts_config.rx_filter, !!(features & NETIF_F_HW_VLAN_CTAG_RX));
3041
3042         priv->hwtstamp_config.tx_type = ts_config.tx_type;
3043         priv->hwtstamp_config.rx_filter = ts_config.rx_filter;
3044
3045         if (DEV_FEATURE_CHANGED(dev, features, NETIF_F_HW_VLAN_CTAG_RX)) {
3046                 if (features & NETIF_F_HW_VLAN_CTAG_RX)
3047                         dev->features |= NETIF_F_HW_VLAN_CTAG_RX;
3048                 else
3049                         dev->features &= ~NETIF_F_HW_VLAN_CTAG_RX;
3050         } else if (ts_config.rx_filter == HWTSTAMP_FILTER_NONE) {
3051                 /* RX time-stamping is OFF, update the RX vlan offload
3052                  * to the latest wanted state
3053                  */
3054                 if (dev->wanted_features & NETIF_F_HW_VLAN_CTAG_RX)
3055                         dev->features |= NETIF_F_HW_VLAN_CTAG_RX;
3056                 else
3057                         dev->features &= ~NETIF_F_HW_VLAN_CTAG_RX;
3058         }
3059
3060         if (DEV_FEATURE_CHANGED(dev, features, NETIF_F_RXFCS)) {
3061                 if (features & NETIF_F_RXFCS)
3062                         dev->features |= NETIF_F_RXFCS;
3063                 else
3064                         dev->features &= ~NETIF_F_RXFCS;
3065         }
3066
3067         /* RX vlan offload and RX time-stamping can't co-exist !
3068          * Regardless of the caller's choice,
3069          * Turn Off RX vlan offload in case of time-stamping is ON
3070          */
3071         if (ts_config.rx_filter != HWTSTAMP_FILTER_NONE) {
3072                 if (dev->features & NETIF_F_HW_VLAN_CTAG_RX)
3073                         en_warn(priv, "Turning off RX vlan offload since RX time-stamping is ON\n");
3074                 dev->features &= ~NETIF_F_HW_VLAN_CTAG_RX;
3075         }
3076
3077         err = mlx4_en_alloc_resources(priv);
3078         if (err) {
3079                 en_err(priv, "Failed reallocating port resources\n");
3080                 goto out;
3081         }
3082         if (port_up) {
3083                 err = mlx4_en_start_port(dev);
3084                 if (err)
3085                         en_err(priv, "Failed starting port\n");
3086         }
3087
3088 out:
3089         mutex_unlock(&mdev->state_lock);
3090         netdev_features_change(dev);
3091         return err;
3092 }