2 * Copyright (c) 2007 Mellanox Technologies. All rights reserved.
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:
10 * Redistribution and use in source and binary forms, with or
11 * without modification, are permitted provided that the following
14 * - Redistributions of source code must retain the above
15 * copyright notice, this list of conditions and the following
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.
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
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>
41 #include <net/busy_poll.h>
42 #include <net/vxlan.h>
44 #include <linux/mlx4/driver.h>
45 #include <linux/mlx4/device.h>
46 #include <linux/mlx4/cmd.h>
47 #include <linux/mlx4/cq.h>
52 int mlx4_en_setup_tc(struct net_device *dev, u8 up)
54 struct mlx4_en_priv *priv = netdev_priv(dev);
56 unsigned int offset = 0;
58 if (up && up != MLX4_EN_NUM_UP)
61 netdev_set_num_tc(dev, up);
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;
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)
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];
83 return LL_FLUSH_FAILED;
85 if (!mlx4_en_cq_lock_poll(cq))
88 done = mlx4_en_process_rx_cq(dev, cq, 4);
90 rx_ring->cleaned += done;
94 mlx4_en_cq_unlock_poll(cq);
98 #endif /* CONFIG_NET_RX_BUSY_POLL */
100 #ifdef CONFIG_RFS_ACCEL
102 struct mlx4_en_filter {
103 struct list_head next;
104 struct work_struct work;
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
120 struct hlist_node filter_chain;
123 static void mlx4_en_filter_rfs_expire(struct mlx4_en_priv *priv);
125 static enum mlx4_net_trans_rule_id mlx4_ip_proto_to_trans_rule_id(u8 ip_proto)
129 return MLX4_NET_TRANS_RULE_ID_UDP;
131 return MLX4_NET_TRANS_RULE_ID_TCP;
133 return -EPROTONOSUPPORT;
137 static void mlx4_en_filter_work(struct work_struct *work)
139 struct mlx4_en_filter *filter = container_of(work,
140 struct mlx4_en_filter,
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),
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,
154 struct mlx4_spec_list spec_ip = {
155 .id = MLX4_NET_TRANS_RULE_ID_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,
165 struct mlx4_spec_list spec_eth = {
166 .id = MLX4_NET_TRANS_RULE_ID_ETH,
168 struct mlx4_net_trans_rule rule = {
169 .list = LIST_HEAD_INIT(rule.list),
170 .queue_mode = MLX4_NET_TRANS_Q_LIFO,
173 .promisc_mode = MLX4_FS_REGULAR,
175 .priority = MLX4_DOMAIN_RFS,
178 __be64 mac_mask = cpu_to_be64(MLX4_MAC_MASK << 16);
180 if (spec_tcp_udp.id < 0) {
181 en_warn(priv, "RFS: ignoring unsupported ip protocol (%d)\n",
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);
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);
193 filter->activated = 0;
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);
201 rc = mlx4_flow_attach(priv->mdev->dev, &rule, &filter->reg_id);
203 en_err(priv, "Error attaching flow. err = %d\n", rc);
206 mlx4_en_filter_rfs_expire(priv);
208 filter->activated = 1;
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)
218 l = (__force unsigned long)src_port |
219 ((__force unsigned long)dst_port << 2);
220 l ^= (__force unsigned long)(src_ip ^ dst_ip);
222 bucket_idx = hash_long(l, MLX4_EN_FILTER_HASH_SHIFT);
224 return &priv->filter_hash[bucket_idx];
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)
232 struct mlx4_en_filter *filter = NULL;
234 filter = kzalloc(sizeof(struct mlx4_en_filter), GFP_ATOMIC);
239 filter->rxq_index = rxq_index;
240 INIT_WORK(&filter->work, mlx4_en_filter_work);
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;
248 filter->flow_id = flow_id;
250 filter->id = priv->last_filter_id++ % RPS_NO_FILTER;
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,
260 static void mlx4_en_filter_free(struct mlx4_en_filter *filter)
262 struct mlx4_en_priv *priv = filter->priv;
265 list_del(&filter->next);
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);
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)
278 struct mlx4_en_filter *filter;
279 struct mlx4_en_filter *ret = NULL;
281 hlist_for_each_entry(filter,
282 filter_hash_bucket(priv, src_ip, dst_ip,
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) {
299 mlx4_en_filter_rfs(struct net_device *net_dev, const struct sk_buff *skb,
300 u16 rxq_index, u32 flow_id)
302 struct mlx4_en_priv *priv = netdev_priv(net_dev);
303 struct mlx4_en_filter *filter;
304 const struct iphdr *ip;
311 int nhoff = skb_network_offset(skb);
314 if (skb->protocol != htons(ETH_P_IP))
315 return -EPROTONOSUPPORT;
317 ip = (const struct iphdr *)(skb->data + nhoff);
318 if (ip_is_fragment(ip))
319 return -EPROTONOSUPPORT;
321 if ((ip->protocol != IPPROTO_TCP) && (ip->protocol != IPPROTO_UDP))
322 return -EPROTONOSUPPORT;
323 ports = (const __be16 *)(skb->data + nhoff + 4 * ip->ihl);
325 ip_proto = ip->protocol;
331 spin_lock_bh(&priv->filters_lock);
332 filter = mlx4_en_filter_find(priv, src_ip, dst_ip, ip_proto,
335 if (filter->rxq_index == rxq_index)
338 filter->rxq_index = rxq_index;
340 filter = mlx4_en_filter_alloc(priv, rxq_index,
341 src_ip, dst_ip, ip_proto,
342 src_port, dst_port, flow_id);
349 queue_work(priv->mdev->workqueue, &filter->work);
354 spin_unlock_bh(&priv->filters_lock);
359 void mlx4_en_cleanup_filters(struct mlx4_en_priv *priv)
361 struct mlx4_en_filter *filter, *tmp;
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);
369 spin_unlock_bh(&priv->filters_lock);
371 list_for_each_entry_safe(filter, tmp, &del_list, next) {
372 cancel_work_sync(&filter->work);
373 mlx4_en_filter_free(filter);
377 static void mlx4_en_filter_rfs_expire(struct mlx4_en_priv *priv)
379 struct mlx4_en_filter *filter = NULL, *tmp, *last_filter = NULL;
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)
388 if (filter->activated &&
389 !work_pending(&filter->work) &&
390 rps_may_expire_flow(priv->dev,
391 filter->rxq_index, filter->flow_id,
393 list_move(&filter->next, &del_list);
394 hlist_del(&filter->filter_chain);
396 last_filter = filter;
401 if (last_filter && (&last_filter->next != priv->filters.next))
402 list_move(&priv->filters, &last_filter->next);
404 spin_unlock_bh(&priv->filters_lock);
406 list_for_each_entry_safe(filter, tmp, &del_list, next)
407 mlx4_en_filter_free(filter);
411 static int mlx4_en_vlan_rx_add_vid(struct net_device *dev,
412 __be16 proto, u16 vid)
414 struct mlx4_en_priv *priv = netdev_priv(dev);
415 struct mlx4_en_dev *mdev = priv->mdev;
419 en_dbg(HW, priv, "adding VLAN:%d\n", vid);
421 set_bit(vid, priv->active_vlans);
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);
428 en_err(priv, "Failed configuring VLAN filter\n");
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);
437 static int mlx4_en_vlan_rx_kill_vid(struct net_device *dev,
438 __be16 proto, u16 vid)
440 struct mlx4_en_priv *priv = netdev_priv(dev);
441 struct mlx4_en_dev *mdev = priv->mdev;
444 en_dbg(HW, priv, "Killing VID:%d\n", vid);
446 clear_bit(vid, priv->active_vlans);
448 /* Remove VID from port VLAN filter */
449 mutex_lock(&mdev->state_lock);
450 mlx4_unregister_vlan(mdev->dev, priv->port, vid);
452 if (mdev->device_up && priv->port_up) {
453 err = mlx4_SET_VLAN_FLTR(mdev->dev, priv);
455 en_err(priv, "Failed configuring VLAN filter\n");
457 mutex_unlock(&mdev->state_lock);
462 static void mlx4_en_u64_to_mac(unsigned char dst_mac[ETH_ALEN + 2], u64 src_mac)
465 for (i = ETH_ALEN - 1; i >= 0; --i) {
466 dst_mac[i] = src_mac & 0xff;
469 memset(&dst_mac[ETH_ALEN], 0, 2);
473 static int mlx4_en_tunnel_steer_add(struct mlx4_en_priv *priv, unsigned char *addr,
474 int qpn, u64 *reg_id)
477 struct mlx4_spec_list spec_eth_outer = { {NULL} };
478 struct mlx4_spec_list spec_vxlan = { {NULL} };
479 struct mlx4_spec_list spec_eth_inner = { {NULL} };
481 struct mlx4_net_trans_rule rule = {
482 .queue_mode = MLX4_NET_TRANS_Q_FIFO,
485 .promisc_mode = MLX4_FS_REGULAR,
486 .priority = MLX4_DOMAIN_NIC,
489 __be64 mac_mask = cpu_to_be64(MLX4_MAC_MASK << 16);
491 if (priv->mdev->dev->caps.tunnel_offload_mode != MLX4_TUNNEL_OFFLOAD_MODE_VXLAN)
492 return 0; /* do nothing */
494 rule.port = priv->port;
496 INIT_LIST_HEAD(&rule.list);
498 spec_eth_outer.id = MLX4_NET_TRANS_RULE_ID_ETH;
499 memcpy(spec_eth_outer.eth.dst_mac, addr, ETH_ALEN);
500 memcpy(spec_eth_outer.eth.dst_mac_msk, &mac_mask, ETH_ALEN);
502 spec_vxlan.id = MLX4_NET_TRANS_RULE_ID_VXLAN; /* any vxlan header */
503 spec_eth_inner.id = MLX4_NET_TRANS_RULE_ID_ETH; /* any inner eth header */
505 list_add_tail(&spec_eth_outer.list, &rule.list);
506 list_add_tail(&spec_vxlan.list, &rule.list);
507 list_add_tail(&spec_eth_inner.list, &rule.list);
509 err = mlx4_flow_attach(priv->mdev->dev, &rule, reg_id);
511 en_err(priv, "failed to add vxlan steering rule, err %d\n", err);
514 en_dbg(DRV, priv, "added vxlan steering rule, mac %pM reg_id %llx\n", addr, *reg_id);
519 static int mlx4_en_uc_steer_add(struct mlx4_en_priv *priv,
520 unsigned char *mac, int *qpn, u64 *reg_id)
522 struct mlx4_en_dev *mdev = priv->mdev;
523 struct mlx4_dev *dev = mdev->dev;
526 switch (dev->caps.steering_mode) {
527 case MLX4_STEERING_MODE_B0: {
532 memcpy(&gid[10], mac, ETH_ALEN);
535 err = mlx4_unicast_attach(dev, &qp, gid, 0, MLX4_PROT_ETH);
538 case MLX4_STEERING_MODE_DEVICE_MANAGED: {
539 struct mlx4_spec_list spec_eth = { {NULL} };
540 __be64 mac_mask = cpu_to_be64(MLX4_MAC_MASK << 16);
542 struct mlx4_net_trans_rule rule = {
543 .queue_mode = MLX4_NET_TRANS_Q_FIFO,
546 .promisc_mode = MLX4_FS_REGULAR,
547 .priority = MLX4_DOMAIN_NIC,
550 rule.port = priv->port;
552 INIT_LIST_HEAD(&rule.list);
554 spec_eth.id = MLX4_NET_TRANS_RULE_ID_ETH;
555 memcpy(spec_eth.eth.dst_mac, mac, ETH_ALEN);
556 memcpy(spec_eth.eth.dst_mac_msk, &mac_mask, ETH_ALEN);
557 list_add_tail(&spec_eth.list, &rule.list);
559 err = mlx4_flow_attach(dev, &rule, reg_id);
566 en_warn(priv, "Failed Attaching Unicast\n");
571 static void mlx4_en_uc_steer_release(struct mlx4_en_priv *priv,
572 unsigned char *mac, int qpn, u64 reg_id)
574 struct mlx4_en_dev *mdev = priv->mdev;
575 struct mlx4_dev *dev = mdev->dev;
577 switch (dev->caps.steering_mode) {
578 case MLX4_STEERING_MODE_B0: {
583 memcpy(&gid[10], mac, ETH_ALEN);
586 mlx4_unicast_detach(dev, &qp, gid, MLX4_PROT_ETH);
589 case MLX4_STEERING_MODE_DEVICE_MANAGED: {
590 mlx4_flow_detach(dev, reg_id);
594 en_err(priv, "Invalid steering mode.\n");
598 static int mlx4_en_get_qp(struct mlx4_en_priv *priv)
600 struct mlx4_en_dev *mdev = priv->mdev;
601 struct mlx4_dev *dev = mdev->dev;
602 struct mlx4_mac_entry *entry;
606 int *qpn = &priv->base_qpn;
607 u64 mac = mlx4_mac_to_u64(priv->dev->dev_addr);
609 en_dbg(DRV, priv, "Registering MAC: %pM for adding\n",
610 priv->dev->dev_addr);
611 index = mlx4_register_mac(dev, priv->port, mac);
614 en_err(priv, "Failed adding MAC: %pM\n",
615 priv->dev->dev_addr);
619 if (dev->caps.steering_mode == MLX4_STEERING_MODE_A0) {
620 int base_qpn = mlx4_get_base_qpn(dev, priv->port);
621 *qpn = base_qpn + index;
625 err = mlx4_qp_reserve_range(dev, 1, 1, qpn);
626 en_dbg(DRV, priv, "Reserved qp %d\n", *qpn);
628 en_err(priv, "Failed to reserve qp for mac registration\n");
632 err = mlx4_en_uc_steer_add(priv, priv->dev->dev_addr, qpn, ®_id);
636 err = mlx4_en_tunnel_steer_add(priv, priv->dev->dev_addr, *qpn,
637 &priv->tunnel_reg_id);
641 entry = kmalloc(sizeof(*entry), GFP_KERNEL);
646 memcpy(entry->mac, priv->dev->dev_addr, sizeof(entry->mac));
647 entry->reg_id = reg_id;
649 hlist_add_head_rcu(&entry->hlist,
650 &priv->mac_hash[entry->mac[MLX4_EN_MAC_HASH_IDX]]);
655 if (priv->tunnel_reg_id)
656 mlx4_flow_detach(priv->mdev->dev, priv->tunnel_reg_id);
658 mlx4_en_uc_steer_release(priv, priv->dev->dev_addr, *qpn, reg_id);
661 mlx4_qp_release_range(dev, *qpn, 1);
664 mlx4_unregister_mac(dev, priv->port, mac);
668 static void mlx4_en_put_qp(struct mlx4_en_priv *priv)
670 struct mlx4_en_dev *mdev = priv->mdev;
671 struct mlx4_dev *dev = mdev->dev;
672 int qpn = priv->base_qpn;
675 if (dev->caps.steering_mode == MLX4_STEERING_MODE_A0) {
676 mac = mlx4_mac_to_u64(priv->dev->dev_addr);
677 en_dbg(DRV, priv, "Registering MAC: %pM for deleting\n",
678 priv->dev->dev_addr);
679 mlx4_unregister_mac(dev, priv->port, mac);
681 struct mlx4_mac_entry *entry;
682 struct hlist_node *tmp;
683 struct hlist_head *bucket;
686 for (i = 0; i < MLX4_EN_MAC_HASH_SIZE; ++i) {
687 bucket = &priv->mac_hash[i];
688 hlist_for_each_entry_safe(entry, tmp, bucket, hlist) {
689 mac = mlx4_mac_to_u64(entry->mac);
690 en_dbg(DRV, priv, "Registering MAC: %pM for deleting\n",
692 mlx4_en_uc_steer_release(priv, entry->mac,
695 mlx4_unregister_mac(dev, priv->port, mac);
696 hlist_del_rcu(&entry->hlist);
697 kfree_rcu(entry, rcu);
701 if (priv->tunnel_reg_id) {
702 mlx4_flow_detach(priv->mdev->dev, priv->tunnel_reg_id);
703 priv->tunnel_reg_id = 0;
706 en_dbg(DRV, priv, "Releasing qp: port %d, qpn %d\n",
708 mlx4_qp_release_range(dev, qpn, 1);
709 priv->flags &= ~MLX4_EN_FLAG_FORCE_PROMISC;
713 static int mlx4_en_replace_mac(struct mlx4_en_priv *priv, int qpn,
714 unsigned char *new_mac, unsigned char *prev_mac)
716 struct mlx4_en_dev *mdev = priv->mdev;
717 struct mlx4_dev *dev = mdev->dev;
719 u64 new_mac_u64 = mlx4_mac_to_u64(new_mac);
721 if (dev->caps.steering_mode != MLX4_STEERING_MODE_A0) {
722 struct hlist_head *bucket;
723 unsigned int mac_hash;
724 struct mlx4_mac_entry *entry;
725 struct hlist_node *tmp;
726 u64 prev_mac_u64 = mlx4_mac_to_u64(prev_mac);
728 bucket = &priv->mac_hash[prev_mac[MLX4_EN_MAC_HASH_IDX]];
729 hlist_for_each_entry_safe(entry, tmp, bucket, hlist) {
730 if (ether_addr_equal_64bits(entry->mac, prev_mac)) {
731 mlx4_en_uc_steer_release(priv, entry->mac,
733 mlx4_unregister_mac(dev, priv->port,
735 hlist_del_rcu(&entry->hlist);
737 memcpy(entry->mac, new_mac, ETH_ALEN);
739 mac_hash = new_mac[MLX4_EN_MAC_HASH_IDX];
740 hlist_add_head_rcu(&entry->hlist,
741 &priv->mac_hash[mac_hash]);
742 mlx4_register_mac(dev, priv->port, new_mac_u64);
743 err = mlx4_en_uc_steer_add(priv, new_mac,
748 if (priv->tunnel_reg_id) {
749 mlx4_flow_detach(priv->mdev->dev, priv->tunnel_reg_id);
750 priv->tunnel_reg_id = 0;
752 err = mlx4_en_tunnel_steer_add(priv, new_mac, qpn,
753 &priv->tunnel_reg_id);
760 return __mlx4_replace_mac(dev, priv->port, qpn, new_mac_u64);
763 static int mlx4_en_do_set_mac(struct mlx4_en_priv *priv)
768 /* Remove old MAC and insert the new one */
769 err = mlx4_en_replace_mac(priv, priv->base_qpn,
770 priv->dev->dev_addr, priv->prev_mac);
772 en_err(priv, "Failed changing HW MAC address\n");
773 memcpy(priv->prev_mac, priv->dev->dev_addr,
774 sizeof(priv->prev_mac));
776 en_dbg(HW, priv, "Port is down while registering mac, exiting...\n");
781 static int mlx4_en_set_mac(struct net_device *dev, void *addr)
783 struct mlx4_en_priv *priv = netdev_priv(dev);
784 struct mlx4_en_dev *mdev = priv->mdev;
785 struct sockaddr *saddr = addr;
788 if (!is_valid_ether_addr(saddr->sa_data))
789 return -EADDRNOTAVAIL;
791 memcpy(dev->dev_addr, saddr->sa_data, ETH_ALEN);
793 mutex_lock(&mdev->state_lock);
794 err = mlx4_en_do_set_mac(priv);
795 mutex_unlock(&mdev->state_lock);
800 static void mlx4_en_clear_list(struct net_device *dev)
802 struct mlx4_en_priv *priv = netdev_priv(dev);
803 struct mlx4_en_mc_list *tmp, *mc_to_del;
805 list_for_each_entry_safe(mc_to_del, tmp, &priv->mc_list, list) {
806 list_del(&mc_to_del->list);
811 static void mlx4_en_cache_mclist(struct net_device *dev)
813 struct mlx4_en_priv *priv = netdev_priv(dev);
814 struct netdev_hw_addr *ha;
815 struct mlx4_en_mc_list *tmp;
817 mlx4_en_clear_list(dev);
818 netdev_for_each_mc_addr(ha, dev) {
819 tmp = kzalloc(sizeof(struct mlx4_en_mc_list), GFP_ATOMIC);
821 mlx4_en_clear_list(dev);
824 memcpy(tmp->addr, ha->addr, ETH_ALEN);
825 list_add_tail(&tmp->list, &priv->mc_list);
829 static void update_mclist_flags(struct mlx4_en_priv *priv,
830 struct list_head *dst,
831 struct list_head *src)
833 struct mlx4_en_mc_list *dst_tmp, *src_tmp, *new_mc;
836 /* Find all the entries that should be removed from dst,
837 * These are the entries that are not found in src
839 list_for_each_entry(dst_tmp, dst, list) {
841 list_for_each_entry(src_tmp, src, list) {
842 if (ether_addr_equal(dst_tmp->addr, src_tmp->addr)) {
848 dst_tmp->action = MCLIST_REM;
851 /* Add entries that exist in src but not in dst
852 * mark them as need to add
854 list_for_each_entry(src_tmp, src, list) {
856 list_for_each_entry(dst_tmp, dst, list) {
857 if (ether_addr_equal(dst_tmp->addr, src_tmp->addr)) {
858 dst_tmp->action = MCLIST_NONE;
864 new_mc = kmemdup(src_tmp,
865 sizeof(struct mlx4_en_mc_list),
870 new_mc->action = MCLIST_ADD;
871 list_add_tail(&new_mc->list, dst);
876 static void mlx4_en_set_rx_mode(struct net_device *dev)
878 struct mlx4_en_priv *priv = netdev_priv(dev);
883 queue_work(priv->mdev->workqueue, &priv->rx_mode_task);
886 static void mlx4_en_set_promisc_mode(struct mlx4_en_priv *priv,
887 struct mlx4_en_dev *mdev)
891 if (!(priv->flags & MLX4_EN_FLAG_PROMISC)) {
892 if (netif_msg_rx_status(priv))
893 en_warn(priv, "Entering promiscuous mode\n");
894 priv->flags |= MLX4_EN_FLAG_PROMISC;
896 /* Enable promiscouos mode */
897 switch (mdev->dev->caps.steering_mode) {
898 case MLX4_STEERING_MODE_DEVICE_MANAGED:
899 err = mlx4_flow_steer_promisc_add(mdev->dev,
902 MLX4_FS_ALL_DEFAULT);
904 en_err(priv, "Failed enabling promiscuous mode\n");
905 priv->flags |= MLX4_EN_FLAG_MC_PROMISC;
908 case MLX4_STEERING_MODE_B0:
909 err = mlx4_unicast_promisc_add(mdev->dev,
913 en_err(priv, "Failed enabling unicast promiscuous mode\n");
915 /* Add the default qp number as multicast
918 if (!(priv->flags & MLX4_EN_FLAG_MC_PROMISC)) {
919 err = mlx4_multicast_promisc_add(mdev->dev,
923 en_err(priv, "Failed enabling multicast promiscuous mode\n");
924 priv->flags |= MLX4_EN_FLAG_MC_PROMISC;
928 case MLX4_STEERING_MODE_A0:
929 err = mlx4_SET_PORT_qpn_calc(mdev->dev,
934 en_err(priv, "Failed enabling promiscuous mode\n");
938 /* Disable port multicast filter (unconditionally) */
939 err = mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, 0,
940 0, MLX4_MCAST_DISABLE);
942 en_err(priv, "Failed disabling multicast filter\n");
944 /* Disable port VLAN filter */
945 err = mlx4_SET_VLAN_FLTR(mdev->dev, priv);
947 en_err(priv, "Failed disabling VLAN filter\n");
951 static void mlx4_en_clear_promisc_mode(struct mlx4_en_priv *priv,
952 struct mlx4_en_dev *mdev)
956 if (netif_msg_rx_status(priv))
957 en_warn(priv, "Leaving promiscuous mode\n");
958 priv->flags &= ~MLX4_EN_FLAG_PROMISC;
960 /* Disable promiscouos mode */
961 switch (mdev->dev->caps.steering_mode) {
962 case MLX4_STEERING_MODE_DEVICE_MANAGED:
963 err = mlx4_flow_steer_promisc_remove(mdev->dev,
965 MLX4_FS_ALL_DEFAULT);
967 en_err(priv, "Failed disabling promiscuous mode\n");
968 priv->flags &= ~MLX4_EN_FLAG_MC_PROMISC;
971 case MLX4_STEERING_MODE_B0:
972 err = mlx4_unicast_promisc_remove(mdev->dev,
976 en_err(priv, "Failed disabling unicast promiscuous mode\n");
977 /* Disable Multicast promisc */
978 if (priv->flags & MLX4_EN_FLAG_MC_PROMISC) {
979 err = mlx4_multicast_promisc_remove(mdev->dev,
983 en_err(priv, "Failed disabling multicast promiscuous mode\n");
984 priv->flags &= ~MLX4_EN_FLAG_MC_PROMISC;
988 case MLX4_STEERING_MODE_A0:
989 err = mlx4_SET_PORT_qpn_calc(mdev->dev,
993 en_err(priv, "Failed disabling promiscuous mode\n");
997 /* Enable port VLAN filter */
998 err = mlx4_SET_VLAN_FLTR(mdev->dev, priv);
1000 en_err(priv, "Failed enabling VLAN filter\n");
1003 static void mlx4_en_do_multicast(struct mlx4_en_priv *priv,
1004 struct net_device *dev,
1005 struct mlx4_en_dev *mdev)
1007 struct mlx4_en_mc_list *mclist, *tmp;
1009 u8 mc_list[16] = {0};
1012 /* Enable/disable the multicast filter according to IFF_ALLMULTI */
1013 if (dev->flags & IFF_ALLMULTI) {
1014 err = mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, 0,
1015 0, MLX4_MCAST_DISABLE);
1017 en_err(priv, "Failed disabling multicast filter\n");
1019 /* Add the default qp number as multicast promisc */
1020 if (!(priv->flags & MLX4_EN_FLAG_MC_PROMISC)) {
1021 switch (mdev->dev->caps.steering_mode) {
1022 case MLX4_STEERING_MODE_DEVICE_MANAGED:
1023 err = mlx4_flow_steer_promisc_add(mdev->dev,
1026 MLX4_FS_MC_DEFAULT);
1029 case MLX4_STEERING_MODE_B0:
1030 err = mlx4_multicast_promisc_add(mdev->dev,
1035 case MLX4_STEERING_MODE_A0:
1039 en_err(priv, "Failed entering multicast promisc mode\n");
1040 priv->flags |= MLX4_EN_FLAG_MC_PROMISC;
1043 /* Disable Multicast promisc */
1044 if (priv->flags & MLX4_EN_FLAG_MC_PROMISC) {
1045 switch (mdev->dev->caps.steering_mode) {
1046 case MLX4_STEERING_MODE_DEVICE_MANAGED:
1047 err = mlx4_flow_steer_promisc_remove(mdev->dev,
1049 MLX4_FS_MC_DEFAULT);
1052 case MLX4_STEERING_MODE_B0:
1053 err = mlx4_multicast_promisc_remove(mdev->dev,
1058 case MLX4_STEERING_MODE_A0:
1062 en_err(priv, "Failed disabling multicast promiscuous mode\n");
1063 priv->flags &= ~MLX4_EN_FLAG_MC_PROMISC;
1066 err = mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, 0,
1067 0, MLX4_MCAST_DISABLE);
1069 en_err(priv, "Failed disabling multicast filter\n");
1071 /* Flush mcast filter and init it with broadcast address */
1072 mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, ETH_BCAST,
1073 1, MLX4_MCAST_CONFIG);
1075 /* Update multicast list - we cache all addresses so they won't
1076 * change while HW is updated holding the command semaphor */
1077 netif_addr_lock_bh(dev);
1078 mlx4_en_cache_mclist(dev);
1079 netif_addr_unlock_bh(dev);
1080 list_for_each_entry(mclist, &priv->mc_list, list) {
1081 mcast_addr = mlx4_mac_to_u64(mclist->addr);
1082 mlx4_SET_MCAST_FLTR(mdev->dev, priv->port,
1083 mcast_addr, 0, MLX4_MCAST_CONFIG);
1085 err = mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, 0,
1086 0, MLX4_MCAST_ENABLE);
1088 en_err(priv, "Failed enabling multicast filter\n");
1090 update_mclist_flags(priv, &priv->curr_list, &priv->mc_list);
1091 list_for_each_entry_safe(mclist, tmp, &priv->curr_list, list) {
1092 if (mclist->action == MCLIST_REM) {
1093 /* detach this address and delete from list */
1094 memcpy(&mc_list[10], mclist->addr, ETH_ALEN);
1095 mc_list[5] = priv->port;
1096 err = mlx4_multicast_detach(mdev->dev,
1097 &priv->rss_map.indir_qp,
1102 en_err(priv, "Fail to detach multicast address\n");
1104 if (mclist->tunnel_reg_id) {
1105 err = mlx4_flow_detach(priv->mdev->dev, mclist->tunnel_reg_id);
1107 en_err(priv, "Failed to detach multicast address\n");
1110 /* remove from list */
1111 list_del(&mclist->list);
1113 } else if (mclist->action == MCLIST_ADD) {
1114 /* attach the address */
1115 memcpy(&mc_list[10], mclist->addr, ETH_ALEN);
1116 /* needed for B0 steering support */
1117 mc_list[5] = priv->port;
1118 err = mlx4_multicast_attach(mdev->dev,
1119 &priv->rss_map.indir_qp,
1125 en_err(priv, "Fail to attach multicast address\n");
1127 err = mlx4_en_tunnel_steer_add(priv, &mc_list[10], priv->base_qpn,
1128 &mclist->tunnel_reg_id);
1130 en_err(priv, "Failed to attach multicast address\n");
1136 static void mlx4_en_do_uc_filter(struct mlx4_en_priv *priv,
1137 struct net_device *dev,
1138 struct mlx4_en_dev *mdev)
1140 struct netdev_hw_addr *ha;
1141 struct mlx4_mac_entry *entry;
1142 struct hlist_node *tmp;
1146 struct hlist_head *bucket;
1151 /* Note that we do not need to protect our mac_hash traversal with rcu,
1152 * since all modification code is protected by mdev->state_lock
1155 /* find what to remove */
1156 for (i = 0; i < MLX4_EN_MAC_HASH_SIZE; ++i) {
1157 bucket = &priv->mac_hash[i];
1158 hlist_for_each_entry_safe(entry, tmp, bucket, hlist) {
1160 netdev_for_each_uc_addr(ha, dev) {
1161 if (ether_addr_equal_64bits(entry->mac,
1168 /* MAC address of the port is not in uc list */
1169 if (ether_addr_equal_64bits(entry->mac, dev->dev_addr))
1173 mac = mlx4_mac_to_u64(entry->mac);
1174 mlx4_en_uc_steer_release(priv, entry->mac,
1177 mlx4_unregister_mac(mdev->dev, priv->port, mac);
1179 hlist_del_rcu(&entry->hlist);
1180 kfree_rcu(entry, rcu);
1181 en_dbg(DRV, priv, "Removed MAC %pM on port:%d\n",
1182 entry->mac, priv->port);
1188 /* if we didn't remove anything, there is no use in trying to add
1189 * again once we are in a forced promisc mode state
1191 if ((priv->flags & MLX4_EN_FLAG_FORCE_PROMISC) && 0 == removed)
1194 prev_flags = priv->flags;
1195 priv->flags &= ~MLX4_EN_FLAG_FORCE_PROMISC;
1197 /* find what to add */
1198 netdev_for_each_uc_addr(ha, dev) {
1200 bucket = &priv->mac_hash[ha->addr[MLX4_EN_MAC_HASH_IDX]];
1201 hlist_for_each_entry(entry, bucket, hlist) {
1202 if (ether_addr_equal_64bits(entry->mac, ha->addr)) {
1209 entry = kmalloc(sizeof(*entry), GFP_KERNEL);
1211 en_err(priv, "Failed adding MAC %pM on port:%d (out of memory)\n",
1212 ha->addr, priv->port);
1213 priv->flags |= MLX4_EN_FLAG_FORCE_PROMISC;
1216 mac = mlx4_mac_to_u64(ha->addr);
1217 memcpy(entry->mac, ha->addr, ETH_ALEN);
1218 err = mlx4_register_mac(mdev->dev, priv->port, mac);
1220 en_err(priv, "Failed registering MAC %pM on port %d: %d\n",
1221 ha->addr, priv->port, err);
1223 priv->flags |= MLX4_EN_FLAG_FORCE_PROMISC;
1226 err = mlx4_en_uc_steer_add(priv, ha->addr,
1230 en_err(priv, "Failed adding MAC %pM on port %d: %d\n",
1231 ha->addr, priv->port, err);
1232 mlx4_unregister_mac(mdev->dev, priv->port, mac);
1234 priv->flags |= MLX4_EN_FLAG_FORCE_PROMISC;
1237 unsigned int mac_hash;
1238 en_dbg(DRV, priv, "Added MAC %pM on port:%d\n",
1239 ha->addr, priv->port);
1240 mac_hash = ha->addr[MLX4_EN_MAC_HASH_IDX];
1241 bucket = &priv->mac_hash[mac_hash];
1242 hlist_add_head_rcu(&entry->hlist, bucket);
1247 if (priv->flags & MLX4_EN_FLAG_FORCE_PROMISC) {
1248 en_warn(priv, "Forcing promiscuous mode on port:%d\n",
1250 } else if (prev_flags & MLX4_EN_FLAG_FORCE_PROMISC) {
1251 en_warn(priv, "Stop forcing promiscuous mode on port:%d\n",
1256 static void mlx4_en_do_set_rx_mode(struct work_struct *work)
1258 struct mlx4_en_priv *priv = container_of(work, struct mlx4_en_priv,
1260 struct mlx4_en_dev *mdev = priv->mdev;
1261 struct net_device *dev = priv->dev;
1263 mutex_lock(&mdev->state_lock);
1264 if (!mdev->device_up) {
1265 en_dbg(HW, priv, "Card is not up, ignoring rx mode change.\n");
1268 if (!priv->port_up) {
1269 en_dbg(HW, priv, "Port is down, ignoring rx mode change.\n");
1273 if (!netif_carrier_ok(dev)) {
1274 if (!mlx4_en_QUERY_PORT(mdev, priv->port)) {
1275 if (priv->port_state.link_state) {
1276 priv->last_link_state = MLX4_DEV_EVENT_PORT_UP;
1277 netif_carrier_on(dev);
1278 en_dbg(LINK, priv, "Link Up\n");
1283 if (dev->priv_flags & IFF_UNICAST_FLT)
1284 mlx4_en_do_uc_filter(priv, dev, mdev);
1286 /* Promsicuous mode: disable all filters */
1287 if ((dev->flags & IFF_PROMISC) ||
1288 (priv->flags & MLX4_EN_FLAG_FORCE_PROMISC)) {
1289 mlx4_en_set_promisc_mode(priv, mdev);
1293 /* Not in promiscuous mode */
1294 if (priv->flags & MLX4_EN_FLAG_PROMISC)
1295 mlx4_en_clear_promisc_mode(priv, mdev);
1297 mlx4_en_do_multicast(priv, dev, mdev);
1299 mutex_unlock(&mdev->state_lock);
1302 #ifdef CONFIG_NET_POLL_CONTROLLER
1303 static void mlx4_en_netpoll(struct net_device *dev)
1305 struct mlx4_en_priv *priv = netdev_priv(dev);
1306 struct mlx4_en_cq *cq;
1307 unsigned long flags;
1310 for (i = 0; i < priv->rx_ring_num; i++) {
1311 cq = priv->rx_cq[i];
1312 spin_lock_irqsave(&cq->lock, flags);
1313 napi_synchronize(&cq->napi);
1314 mlx4_en_process_rx_cq(dev, cq, 0);
1315 spin_unlock_irqrestore(&cq->lock, flags);
1320 static void mlx4_en_tx_timeout(struct net_device *dev)
1322 struct mlx4_en_priv *priv = netdev_priv(dev);
1323 struct mlx4_en_dev *mdev = priv->mdev;
1326 if (netif_msg_timer(priv))
1327 en_warn(priv, "Tx timeout called on port:%d\n", priv->port);
1329 for (i = 0; i < priv->tx_ring_num; i++) {
1330 if (!netif_tx_queue_stopped(netdev_get_tx_queue(dev, i)))
1332 en_warn(priv, "TX timeout on queue: %d, QP: 0x%x, CQ: 0x%x, Cons: 0x%x, Prod: 0x%x\n",
1333 i, priv->tx_ring[i]->qpn, priv->tx_ring[i]->cqn,
1334 priv->tx_ring[i]->cons, priv->tx_ring[i]->prod);
1337 priv->port_stats.tx_timeout++;
1338 en_dbg(DRV, priv, "Scheduling watchdog\n");
1339 queue_work(mdev->workqueue, &priv->watchdog_task);
1343 static struct net_device_stats *mlx4_en_get_stats(struct net_device *dev)
1345 struct mlx4_en_priv *priv = netdev_priv(dev);
1347 spin_lock_bh(&priv->stats_lock);
1348 memcpy(&priv->ret_stats, &priv->stats, sizeof(priv->stats));
1349 spin_unlock_bh(&priv->stats_lock);
1351 return &priv->ret_stats;
1354 static void mlx4_en_set_default_moderation(struct mlx4_en_priv *priv)
1356 struct mlx4_en_cq *cq;
1359 /* If we haven't received a specific coalescing setting
1360 * (module param), we set the moderation parameters as follows:
1361 * - moder_cnt is set to the number of mtu sized packets to
1362 * satisfy our coalescing target.
1363 * - moder_time is set to a fixed value.
1365 priv->rx_frames = MLX4_EN_RX_COAL_TARGET;
1366 priv->rx_usecs = MLX4_EN_RX_COAL_TIME;
1367 priv->tx_frames = MLX4_EN_TX_COAL_PKTS;
1368 priv->tx_usecs = MLX4_EN_TX_COAL_TIME;
1369 en_dbg(INTR, priv, "Default coalesing params for mtu:%d - rx_frames:%d rx_usecs:%d\n",
1370 priv->dev->mtu, priv->rx_frames, priv->rx_usecs);
1372 /* Setup cq moderation params */
1373 for (i = 0; i < priv->rx_ring_num; i++) {
1374 cq = priv->rx_cq[i];
1375 cq->moder_cnt = priv->rx_frames;
1376 cq->moder_time = priv->rx_usecs;
1377 priv->last_moder_time[i] = MLX4_EN_AUTO_CONF;
1378 priv->last_moder_packets[i] = 0;
1379 priv->last_moder_bytes[i] = 0;
1382 for (i = 0; i < priv->tx_ring_num; i++) {
1383 cq = priv->tx_cq[i];
1384 cq->moder_cnt = priv->tx_frames;
1385 cq->moder_time = priv->tx_usecs;
1388 /* Reset auto-moderation params */
1389 priv->pkt_rate_low = MLX4_EN_RX_RATE_LOW;
1390 priv->rx_usecs_low = MLX4_EN_RX_COAL_TIME_LOW;
1391 priv->pkt_rate_high = MLX4_EN_RX_RATE_HIGH;
1392 priv->rx_usecs_high = MLX4_EN_RX_COAL_TIME_HIGH;
1393 priv->sample_interval = MLX4_EN_SAMPLE_INTERVAL;
1394 priv->adaptive_rx_coal = 1;
1395 priv->last_moder_jiffies = 0;
1396 priv->last_moder_tx_packets = 0;
1399 static void mlx4_en_auto_moderation(struct mlx4_en_priv *priv)
1401 unsigned long period = (unsigned long) (jiffies - priv->last_moder_jiffies);
1402 struct mlx4_en_cq *cq;
1403 unsigned long packets;
1405 unsigned long avg_pkt_size;
1406 unsigned long rx_packets;
1407 unsigned long rx_bytes;
1408 unsigned long rx_pkt_diff;
1412 if (!priv->adaptive_rx_coal || period < priv->sample_interval * HZ)
1415 for (ring = 0; ring < priv->rx_ring_num; ring++) {
1416 spin_lock_bh(&priv->stats_lock);
1417 rx_packets = priv->rx_ring[ring]->packets;
1418 rx_bytes = priv->rx_ring[ring]->bytes;
1419 spin_unlock_bh(&priv->stats_lock);
1421 rx_pkt_diff = ((unsigned long) (rx_packets -
1422 priv->last_moder_packets[ring]));
1423 packets = rx_pkt_diff;
1424 rate = packets * HZ / period;
1425 avg_pkt_size = packets ? ((unsigned long) (rx_bytes -
1426 priv->last_moder_bytes[ring])) / packets : 0;
1428 /* Apply auto-moderation only when packet rate
1429 * exceeds a rate that it matters */
1430 if (rate > (MLX4_EN_RX_RATE_THRESH / priv->rx_ring_num) &&
1431 avg_pkt_size > MLX4_EN_AVG_PKT_SMALL) {
1432 if (rate < priv->pkt_rate_low)
1433 moder_time = priv->rx_usecs_low;
1434 else if (rate > priv->pkt_rate_high)
1435 moder_time = priv->rx_usecs_high;
1437 moder_time = (rate - priv->pkt_rate_low) *
1438 (priv->rx_usecs_high - priv->rx_usecs_low) /
1439 (priv->pkt_rate_high - priv->pkt_rate_low) +
1442 moder_time = priv->rx_usecs_low;
1445 if (moder_time != priv->last_moder_time[ring]) {
1446 priv->last_moder_time[ring] = moder_time;
1447 cq = priv->rx_cq[ring];
1448 cq->moder_time = moder_time;
1449 cq->moder_cnt = priv->rx_frames;
1450 err = mlx4_en_set_cq_moder(priv, cq);
1452 en_err(priv, "Failed modifying moderation for cq:%d\n",
1455 priv->last_moder_packets[ring] = rx_packets;
1456 priv->last_moder_bytes[ring] = rx_bytes;
1459 priv->last_moder_jiffies = jiffies;
1462 static void mlx4_en_do_get_stats(struct work_struct *work)
1464 struct delayed_work *delay = to_delayed_work(work);
1465 struct mlx4_en_priv *priv = container_of(delay, struct mlx4_en_priv,
1467 struct mlx4_en_dev *mdev = priv->mdev;
1470 mutex_lock(&mdev->state_lock);
1471 if (mdev->device_up) {
1472 if (priv->port_up) {
1473 err = mlx4_en_DUMP_ETH_STATS(mdev, priv->port, 0);
1475 en_dbg(HW, priv, "Could not update stats\n");
1477 mlx4_en_auto_moderation(priv);
1480 queue_delayed_work(mdev->workqueue, &priv->stats_task, STATS_DELAY);
1482 if (mdev->mac_removed[MLX4_MAX_PORTS + 1 - priv->port]) {
1483 mlx4_en_do_set_mac(priv);
1484 mdev->mac_removed[MLX4_MAX_PORTS + 1 - priv->port] = 0;
1486 mutex_unlock(&mdev->state_lock);
1489 /* mlx4_en_service_task - Run service task for tasks that needed to be done
1492 static void mlx4_en_service_task(struct work_struct *work)
1494 struct delayed_work *delay = to_delayed_work(work);
1495 struct mlx4_en_priv *priv = container_of(delay, struct mlx4_en_priv,
1497 struct mlx4_en_dev *mdev = priv->mdev;
1499 mutex_lock(&mdev->state_lock);
1500 if (mdev->device_up) {
1501 if (mdev->dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_TS)
1502 mlx4_en_ptp_overflow_check(mdev);
1504 queue_delayed_work(mdev->workqueue, &priv->service_task,
1505 SERVICE_TASK_DELAY);
1507 mutex_unlock(&mdev->state_lock);
1510 static void mlx4_en_linkstate(struct work_struct *work)
1512 struct mlx4_en_priv *priv = container_of(work, struct mlx4_en_priv,
1514 struct mlx4_en_dev *mdev = priv->mdev;
1515 int linkstate = priv->link_state;
1517 mutex_lock(&mdev->state_lock);
1518 /* If observable port state changed set carrier state and
1519 * report to system log */
1520 if (priv->last_link_state != linkstate) {
1521 if (linkstate == MLX4_DEV_EVENT_PORT_DOWN) {
1522 en_info(priv, "Link Down\n");
1523 netif_carrier_off(priv->dev);
1525 en_info(priv, "Link Up\n");
1526 netif_carrier_on(priv->dev);
1529 priv->last_link_state = linkstate;
1530 mutex_unlock(&mdev->state_lock);
1534 int mlx4_en_start_port(struct net_device *dev)
1536 struct mlx4_en_priv *priv = netdev_priv(dev);
1537 struct mlx4_en_dev *mdev = priv->mdev;
1538 struct mlx4_en_cq *cq;
1539 struct mlx4_en_tx_ring *tx_ring;
1545 u8 mc_list[16] = {0};
1547 if (priv->port_up) {
1548 en_dbg(DRV, priv, "start port called while port already up\n");
1552 INIT_LIST_HEAD(&priv->mc_list);
1553 INIT_LIST_HEAD(&priv->curr_list);
1554 INIT_LIST_HEAD(&priv->ethtool_list);
1555 memset(&priv->ethtool_rules[0], 0,
1556 sizeof(struct ethtool_flow_id) * MAX_NUM_OF_FS_RULES);
1558 /* Calculate Rx buf size */
1559 dev->mtu = min(dev->mtu, priv->max_mtu);
1560 mlx4_en_calc_rx_buf(dev);
1561 en_dbg(DRV, priv, "Rx buf size:%d\n", priv->rx_skb_size);
1563 /* Configure rx cq's and rings */
1564 err = mlx4_en_activate_rx_rings(priv);
1566 en_err(priv, "Failed to activate RX rings\n");
1569 for (i = 0; i < priv->rx_ring_num; i++) {
1570 cq = priv->rx_cq[i];
1572 mlx4_en_cq_init_lock(cq);
1574 err = mlx4_en_activate_cq(priv, cq, i);
1576 en_err(priv, "Failed activating Rx CQ\n");
1579 for (j = 0; j < cq->size; j++)
1580 cq->buf[j].owner_sr_opcode = MLX4_CQE_OWNER_MASK;
1581 err = mlx4_en_set_cq_moder(priv, cq);
1583 en_err(priv, "Failed setting cq moderation parameters");
1584 mlx4_en_deactivate_cq(priv, cq);
1587 mlx4_en_arm_cq(priv, cq);
1588 priv->rx_ring[i]->cqn = cq->mcq.cqn;
1593 en_dbg(DRV, priv, "Getting qp number for port %d\n", priv->port);
1594 err = mlx4_en_get_qp(priv);
1596 en_err(priv, "Failed getting eth qp\n");
1599 mdev->mac_removed[priv->port] = 0;
1601 err = mlx4_en_config_rss_steer(priv);
1603 en_err(priv, "Failed configuring rss steering\n");
1607 err = mlx4_en_create_drop_qp(priv);
1611 /* Configure tx cq's and rings */
1612 for (i = 0; i < priv->tx_ring_num; i++) {
1614 cq = priv->tx_cq[i];
1615 err = mlx4_en_activate_cq(priv, cq, i);
1617 en_err(priv, "Failed allocating Tx CQ\n");
1620 err = mlx4_en_set_cq_moder(priv, cq);
1622 en_err(priv, "Failed setting cq moderation parameters");
1623 mlx4_en_deactivate_cq(priv, cq);
1626 en_dbg(DRV, priv, "Resetting index of collapsed CQ:%d to -1\n", i);
1627 cq->buf->wqe_index = cpu_to_be16(0xffff);
1629 /* Configure ring */
1630 tx_ring = priv->tx_ring[i];
1631 err = mlx4_en_activate_tx_ring(priv, tx_ring, cq->mcq.cqn,
1632 i / priv->num_tx_rings_p_up);
1634 en_err(priv, "Failed allocating Tx ring\n");
1635 mlx4_en_deactivate_cq(priv, cq);
1638 tx_ring->tx_queue = netdev_get_tx_queue(dev, i);
1640 /* Arm CQ for TX completions */
1641 mlx4_en_arm_cq(priv, cq);
1643 /* Set initial ownership of all Tx TXBBs to SW (1) */
1644 for (j = 0; j < tx_ring->buf_size; j += STAMP_STRIDE)
1645 *((u32 *) (tx_ring->buf + j)) = 0xffffffff;
1649 /* Configure port */
1650 err = mlx4_SET_PORT_general(mdev->dev, priv->port,
1651 priv->rx_skb_size + ETH_FCS_LEN,
1652 priv->prof->tx_pause,
1654 priv->prof->rx_pause,
1655 priv->prof->rx_ppp);
1657 en_err(priv, "Failed setting port general configurations for port %d, with error %d\n",
1661 /* Set default qp number */
1662 err = mlx4_SET_PORT_qpn_calc(mdev->dev, priv->port, priv->base_qpn, 0);
1664 en_err(priv, "Failed setting default qp numbers\n");
1668 if (mdev->dev->caps.tunnel_offload_mode == MLX4_TUNNEL_OFFLOAD_MODE_VXLAN) {
1669 err = mlx4_SET_PORT_VXLAN(mdev->dev, priv->port, VXLAN_STEER_BY_OUTER_MAC, 1);
1671 en_err(priv, "Failed setting port L2 tunnel configuration, err %d\n",
1678 en_dbg(HW, priv, "Initializing port\n");
1679 err = mlx4_INIT_PORT(mdev->dev, priv->port);
1681 en_err(priv, "Failed Initializing port\n");
1685 /* Attach rx QP to bradcast address */
1686 memset(&mc_list[10], 0xff, ETH_ALEN);
1687 mc_list[5] = priv->port; /* needed for B0 steering support */
1688 if (mlx4_multicast_attach(mdev->dev, &priv->rss_map.indir_qp, mc_list,
1689 priv->port, 0, MLX4_PROT_ETH,
1690 &priv->broadcast_id))
1691 mlx4_warn(mdev, "Failed Attaching Broadcast\n");
1693 /* Must redo promiscuous mode setup. */
1694 priv->flags &= ~(MLX4_EN_FLAG_PROMISC | MLX4_EN_FLAG_MC_PROMISC);
1696 /* Schedule multicast task to populate multicast list */
1697 queue_work(mdev->workqueue, &priv->rx_mode_task);
1699 mlx4_set_stats_bitmap(mdev->dev, &priv->stats_bitmap);
1701 #ifdef CONFIG_MLX4_EN_VXLAN
1702 if (priv->mdev->dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_VXLAN_OFFLOADS)
1703 vxlan_get_rx_port(dev);
1705 priv->port_up = true;
1706 netif_tx_start_all_queues(dev);
1707 netif_device_attach(dev);
1712 while (tx_index--) {
1713 mlx4_en_deactivate_tx_ring(priv, priv->tx_ring[tx_index]);
1714 mlx4_en_deactivate_cq(priv, priv->tx_cq[tx_index]);
1716 mlx4_en_destroy_drop_qp(priv);
1718 mlx4_en_release_rss_steer(priv);
1720 mlx4_en_put_qp(priv);
1723 mlx4_en_deactivate_cq(priv, priv->rx_cq[rx_index]);
1724 for (i = 0; i < priv->rx_ring_num; i++)
1725 mlx4_en_deactivate_rx_ring(priv, priv->rx_ring[i]);
1727 return err; /* need to close devices */
1731 void mlx4_en_stop_port(struct net_device *dev, int detach)
1733 struct mlx4_en_priv *priv = netdev_priv(dev);
1734 struct mlx4_en_dev *mdev = priv->mdev;
1735 struct mlx4_en_mc_list *mclist, *tmp;
1736 struct ethtool_flow_id *flow, *tmp_flow;
1738 u8 mc_list[16] = {0};
1740 if (!priv->port_up) {
1741 en_dbg(DRV, priv, "stop port called while port already down\n");
1746 mlx4_CLOSE_PORT(mdev->dev, priv->port);
1748 /* Synchronize with tx routine */
1749 netif_tx_lock_bh(dev);
1751 netif_device_detach(dev);
1752 netif_tx_stop_all_queues(dev);
1753 netif_tx_unlock_bh(dev);
1755 netif_tx_disable(dev);
1757 /* Set port as not active */
1758 priv->port_up = false;
1760 /* Promsicuous mode */
1761 if (mdev->dev->caps.steering_mode ==
1762 MLX4_STEERING_MODE_DEVICE_MANAGED) {
1763 priv->flags &= ~(MLX4_EN_FLAG_PROMISC |
1764 MLX4_EN_FLAG_MC_PROMISC);
1765 mlx4_flow_steer_promisc_remove(mdev->dev,
1767 MLX4_FS_ALL_DEFAULT);
1768 mlx4_flow_steer_promisc_remove(mdev->dev,
1770 MLX4_FS_MC_DEFAULT);
1771 } else if (priv->flags & MLX4_EN_FLAG_PROMISC) {
1772 priv->flags &= ~MLX4_EN_FLAG_PROMISC;
1774 /* Disable promiscouos mode */
1775 mlx4_unicast_promisc_remove(mdev->dev, priv->base_qpn,
1778 /* Disable Multicast promisc */
1779 if (priv->flags & MLX4_EN_FLAG_MC_PROMISC) {
1780 mlx4_multicast_promisc_remove(mdev->dev, priv->base_qpn,
1782 priv->flags &= ~MLX4_EN_FLAG_MC_PROMISC;
1786 /* Detach All multicasts */
1787 memset(&mc_list[10], 0xff, ETH_ALEN);
1788 mc_list[5] = priv->port; /* needed for B0 steering support */
1789 mlx4_multicast_detach(mdev->dev, &priv->rss_map.indir_qp, mc_list,
1790 MLX4_PROT_ETH, priv->broadcast_id);
1791 list_for_each_entry(mclist, &priv->curr_list, list) {
1792 memcpy(&mc_list[10], mclist->addr, ETH_ALEN);
1793 mc_list[5] = priv->port;
1794 mlx4_multicast_detach(mdev->dev, &priv->rss_map.indir_qp,
1795 mc_list, MLX4_PROT_ETH, mclist->reg_id);
1796 if (mclist->tunnel_reg_id)
1797 mlx4_flow_detach(mdev->dev, mclist->tunnel_reg_id);
1799 mlx4_en_clear_list(dev);
1800 list_for_each_entry_safe(mclist, tmp, &priv->curr_list, list) {
1801 list_del(&mclist->list);
1805 /* Flush multicast filter */
1806 mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, 0, 1, MLX4_MCAST_CONFIG);
1808 /* Remove flow steering rules for the port*/
1809 if (mdev->dev->caps.steering_mode ==
1810 MLX4_STEERING_MODE_DEVICE_MANAGED) {
1812 list_for_each_entry_safe(flow, tmp_flow,
1813 &priv->ethtool_list, list) {
1814 mlx4_flow_detach(mdev->dev, flow->id);
1815 list_del(&flow->list);
1819 mlx4_en_destroy_drop_qp(priv);
1822 for (i = 0; i < priv->tx_ring_num; i++) {
1823 mlx4_en_deactivate_tx_ring(priv, priv->tx_ring[i]);
1824 mlx4_en_deactivate_cq(priv, priv->tx_cq[i]);
1828 for (i = 0; i < priv->tx_ring_num; i++)
1829 mlx4_en_free_tx_buf(dev, priv->tx_ring[i]);
1832 mlx4_en_release_rss_steer(priv);
1834 /* Unregister Mac address for the port */
1835 mlx4_en_put_qp(priv);
1836 if (!(mdev->dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_REASSIGN_MAC_EN))
1837 mdev->mac_removed[priv->port] = 1;
1840 for (i = 0; i < priv->rx_ring_num; i++) {
1841 struct mlx4_en_cq *cq = priv->rx_cq[i];
1844 while (!mlx4_en_cq_lock_napi(cq)) {
1845 pr_info("CQ %d locked\n", i);
1850 while (test_bit(NAPI_STATE_SCHED, &cq->napi.state))
1852 mlx4_en_deactivate_rx_ring(priv, priv->rx_ring[i]);
1853 mlx4_en_deactivate_cq(priv, cq);
1857 static void mlx4_en_restart(struct work_struct *work)
1859 struct mlx4_en_priv *priv = container_of(work, struct mlx4_en_priv,
1861 struct mlx4_en_dev *mdev = priv->mdev;
1862 struct net_device *dev = priv->dev;
1864 en_dbg(DRV, priv, "Watchdog task called for port %d\n", priv->port);
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);
1872 mutex_unlock(&mdev->state_lock);
1875 static void mlx4_en_clear_stats(struct net_device *dev)
1877 struct mlx4_en_priv *priv = netdev_priv(dev);
1878 struct mlx4_en_dev *mdev = priv->mdev;
1881 if (mlx4_en_DUMP_ETH_STATS(mdev, priv->port, 1))
1882 en_dbg(HW, priv, "Failed dumping statistics\n");
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));
1889 for (i = 0; i < priv->tx_ring_num; i++) {
1890 priv->tx_ring[i]->bytes = 0;
1891 priv->tx_ring[i]->packets = 0;
1892 priv->tx_ring[i]->tx_csum = 0;
1894 for (i = 0; i < priv->rx_ring_num; i++) {
1895 priv->rx_ring[i]->bytes = 0;
1896 priv->rx_ring[i]->packets = 0;
1897 priv->rx_ring[i]->csum_ok = 0;
1898 priv->rx_ring[i]->csum_none = 0;
1902 static int mlx4_en_open(struct net_device *dev)
1904 struct mlx4_en_priv *priv = netdev_priv(dev);
1905 struct mlx4_en_dev *mdev = priv->mdev;
1908 mutex_lock(&mdev->state_lock);
1910 if (!mdev->device_up) {
1911 en_err(priv, "Cannot open - device down/disabled\n");
1916 /* Reset HW statistics and SW counters */
1917 mlx4_en_clear_stats(dev);
1919 err = mlx4_en_start_port(dev);
1921 en_err(priv, "Failed starting port:%d\n", priv->port);
1924 mutex_unlock(&mdev->state_lock);
1929 static int mlx4_en_close(struct net_device *dev)
1931 struct mlx4_en_priv *priv = netdev_priv(dev);
1932 struct mlx4_en_dev *mdev = priv->mdev;
1934 en_dbg(IFDOWN, priv, "Close port called\n");
1936 mutex_lock(&mdev->state_lock);
1938 mlx4_en_stop_port(dev, 0);
1939 netif_carrier_off(dev);
1941 mutex_unlock(&mdev->state_lock);
1945 void mlx4_en_free_resources(struct mlx4_en_priv *priv)
1949 #ifdef CONFIG_RFS_ACCEL
1950 free_irq_cpu_rmap(priv->dev->rx_cpu_rmap);
1951 priv->dev->rx_cpu_rmap = NULL;
1954 for (i = 0; i < priv->tx_ring_num; i++) {
1955 if (priv->tx_ring && priv->tx_ring[i])
1956 mlx4_en_destroy_tx_ring(priv, &priv->tx_ring[i]);
1957 if (priv->tx_cq && priv->tx_cq[i])
1958 mlx4_en_destroy_cq(priv, &priv->tx_cq[i]);
1961 for (i = 0; i < priv->rx_ring_num; i++) {
1962 if (priv->rx_ring[i])
1963 mlx4_en_destroy_rx_ring(priv, &priv->rx_ring[i],
1964 priv->prof->rx_ring_size, priv->stride);
1966 mlx4_en_destroy_cq(priv, &priv->rx_cq[i]);
1969 if (priv->base_tx_qpn) {
1970 mlx4_qp_release_range(priv->mdev->dev, priv->base_tx_qpn, priv->tx_ring_num);
1971 priv->base_tx_qpn = 0;
1975 int mlx4_en_alloc_resources(struct mlx4_en_priv *priv)
1977 struct mlx4_en_port_profile *prof = priv->prof;
1982 err = mlx4_qp_reserve_range(priv->mdev->dev, priv->tx_ring_num, 256, &priv->base_tx_qpn);
1984 en_err(priv, "failed reserving range for TX rings\n");
1988 /* Create tx Rings */
1989 for (i = 0; i < priv->tx_ring_num; i++) {
1990 node = cpu_to_node(i % num_online_cpus());
1991 if (mlx4_en_create_cq(priv, &priv->tx_cq[i],
1992 prof->tx_ring_size, i, TX, node))
1995 if (mlx4_en_create_tx_ring(priv, &priv->tx_ring[i],
1996 priv->base_tx_qpn + i,
1997 prof->tx_ring_size, TXBB_SIZE,
2002 /* Create rx Rings */
2003 for (i = 0; i < priv->rx_ring_num; i++) {
2004 node = cpu_to_node(i % num_online_cpus());
2005 if (mlx4_en_create_cq(priv, &priv->rx_cq[i],
2006 prof->rx_ring_size, i, RX, node))
2009 if (mlx4_en_create_rx_ring(priv, &priv->rx_ring[i],
2010 prof->rx_ring_size, priv->stride,
2015 #ifdef CONFIG_RFS_ACCEL
2016 if (priv->mdev->dev->caps.comp_pool) {
2017 priv->dev->rx_cpu_rmap = alloc_irq_cpu_rmap(priv->mdev->dev->caps.comp_pool);
2018 if (!priv->dev->rx_cpu_rmap)
2026 en_err(priv, "Failed to allocate NIC resources\n");
2027 for (i = 0; i < priv->rx_ring_num; i++) {
2028 if (priv->rx_ring[i])
2029 mlx4_en_destroy_rx_ring(priv, &priv->rx_ring[i],
2033 mlx4_en_destroy_cq(priv, &priv->rx_cq[i]);
2035 for (i = 0; i < priv->tx_ring_num; i++) {
2036 if (priv->tx_ring[i])
2037 mlx4_en_destroy_tx_ring(priv, &priv->tx_ring[i]);
2039 mlx4_en_destroy_cq(priv, &priv->tx_cq[i]);
2045 void mlx4_en_destroy_netdev(struct net_device *dev)
2047 struct mlx4_en_priv *priv = netdev_priv(dev);
2048 struct mlx4_en_dev *mdev = priv->mdev;
2050 en_dbg(DRV, priv, "Destroying netdev on port:%d\n", priv->port);
2052 /* Unregister device - this will close the port if it was up */
2053 if (priv->registered)
2054 unregister_netdev(dev);
2056 if (priv->allocated)
2057 mlx4_free_hwq_res(mdev->dev, &priv->res, MLX4_EN_PAGE_SIZE);
2059 cancel_delayed_work(&priv->stats_task);
2060 cancel_delayed_work(&priv->service_task);
2061 /* flush any pending task for this netdev */
2062 flush_workqueue(mdev->workqueue);
2064 /* Detach the netdev so tasks would not attempt to access it */
2065 mutex_lock(&mdev->state_lock);
2066 mdev->pndev[priv->port] = NULL;
2067 mutex_unlock(&mdev->state_lock);
2069 mlx4_en_free_resources(priv);
2071 kfree(priv->tx_ring);
2077 static int mlx4_en_change_mtu(struct net_device *dev, int new_mtu)
2079 struct mlx4_en_priv *priv = netdev_priv(dev);
2080 struct mlx4_en_dev *mdev = priv->mdev;
2083 en_dbg(DRV, priv, "Change MTU called - current:%d new:%d\n",
2086 if ((new_mtu < MLX4_EN_MIN_MTU) || (new_mtu > priv->max_mtu)) {
2087 en_err(priv, "Bad MTU size:%d.\n", new_mtu);
2092 if (netif_running(dev)) {
2093 mutex_lock(&mdev->state_lock);
2094 if (!mdev->device_up) {
2095 /* NIC is probably restarting - let watchdog task reset
2097 en_dbg(DRV, priv, "Change MTU called with card down!?\n");
2099 mlx4_en_stop_port(dev, 1);
2100 err = mlx4_en_start_port(dev);
2102 en_err(priv, "Failed restarting port:%d\n",
2104 queue_work(mdev->workqueue, &priv->watchdog_task);
2107 mutex_unlock(&mdev->state_lock);
2112 static int mlx4_en_hwtstamp_set(struct net_device *dev, struct ifreq *ifr)
2114 struct mlx4_en_priv *priv = netdev_priv(dev);
2115 struct mlx4_en_dev *mdev = priv->mdev;
2116 struct hwtstamp_config config;
2118 if (copy_from_user(&config, ifr->ifr_data, sizeof(config)))
2121 /* reserved for future extensions */
2125 /* device doesn't support time stamping */
2126 if (!(mdev->dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_TS))
2129 /* TX HW timestamp */
2130 switch (config.tx_type) {
2131 case HWTSTAMP_TX_OFF:
2132 case HWTSTAMP_TX_ON:
2138 /* RX HW timestamp */
2139 switch (config.rx_filter) {
2140 case HWTSTAMP_FILTER_NONE:
2142 case HWTSTAMP_FILTER_ALL:
2143 case HWTSTAMP_FILTER_SOME:
2144 case HWTSTAMP_FILTER_PTP_V1_L4_EVENT:
2145 case HWTSTAMP_FILTER_PTP_V1_L4_SYNC:
2146 case HWTSTAMP_FILTER_PTP_V1_L4_DELAY_REQ:
2147 case HWTSTAMP_FILTER_PTP_V2_L4_EVENT:
2148 case HWTSTAMP_FILTER_PTP_V2_L4_SYNC:
2149 case HWTSTAMP_FILTER_PTP_V2_L4_DELAY_REQ:
2150 case HWTSTAMP_FILTER_PTP_V2_L2_EVENT:
2151 case HWTSTAMP_FILTER_PTP_V2_L2_SYNC:
2152 case HWTSTAMP_FILTER_PTP_V2_L2_DELAY_REQ:
2153 case HWTSTAMP_FILTER_PTP_V2_EVENT:
2154 case HWTSTAMP_FILTER_PTP_V2_SYNC:
2155 case HWTSTAMP_FILTER_PTP_V2_DELAY_REQ:
2156 config.rx_filter = HWTSTAMP_FILTER_ALL;
2162 if (mlx4_en_timestamp_config(dev, config.tx_type, config.rx_filter)) {
2163 config.tx_type = HWTSTAMP_TX_OFF;
2164 config.rx_filter = HWTSTAMP_FILTER_NONE;
2167 return copy_to_user(ifr->ifr_data, &config,
2168 sizeof(config)) ? -EFAULT : 0;
2171 static int mlx4_en_hwtstamp_get(struct net_device *dev, struct ifreq *ifr)
2173 struct mlx4_en_priv *priv = netdev_priv(dev);
2175 return copy_to_user(ifr->ifr_data, &priv->hwtstamp_config,
2176 sizeof(priv->hwtstamp_config)) ? -EFAULT : 0;
2179 static int mlx4_en_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
2183 return mlx4_en_hwtstamp_set(dev, ifr);
2185 return mlx4_en_hwtstamp_get(dev, ifr);
2191 static int mlx4_en_set_features(struct net_device *netdev,
2192 netdev_features_t features)
2194 struct mlx4_en_priv *priv = netdev_priv(netdev);
2196 if (features & NETIF_F_LOOPBACK)
2197 priv->ctrl_flags |= cpu_to_be32(MLX4_WQE_CTRL_FORCE_LOOPBACK);
2200 cpu_to_be32(~MLX4_WQE_CTRL_FORCE_LOOPBACK);
2202 mlx4_en_update_loopback_state(netdev, features);
2208 static int mlx4_en_set_vf_mac(struct net_device *dev, int queue, u8 *mac)
2210 struct mlx4_en_priv *en_priv = netdev_priv(dev);
2211 struct mlx4_en_dev *mdev = en_priv->mdev;
2212 u64 mac_u64 = mlx4_mac_to_u64(mac);
2214 if (!is_valid_ether_addr(mac))
2217 return mlx4_set_vf_mac(mdev->dev, en_priv->port, queue, mac_u64);
2220 static int mlx4_en_set_vf_vlan(struct net_device *dev, int vf, u16 vlan, u8 qos)
2222 struct mlx4_en_priv *en_priv = netdev_priv(dev);
2223 struct mlx4_en_dev *mdev = en_priv->mdev;
2225 return mlx4_set_vf_vlan(mdev->dev, en_priv->port, vf, vlan, qos);
2228 static int mlx4_en_set_vf_spoofchk(struct net_device *dev, int vf, bool setting)
2230 struct mlx4_en_priv *en_priv = netdev_priv(dev);
2231 struct mlx4_en_dev *mdev = en_priv->mdev;
2233 return mlx4_set_vf_spoofchk(mdev->dev, en_priv->port, vf, setting);
2236 static int mlx4_en_get_vf_config(struct net_device *dev, int vf, struct ifla_vf_info *ivf)
2238 struct mlx4_en_priv *en_priv = netdev_priv(dev);
2239 struct mlx4_en_dev *mdev = en_priv->mdev;
2241 return mlx4_get_vf_config(mdev->dev, en_priv->port, vf, ivf);
2244 static int mlx4_en_set_vf_link_state(struct net_device *dev, int vf, int link_state)
2246 struct mlx4_en_priv *en_priv = netdev_priv(dev);
2247 struct mlx4_en_dev *mdev = en_priv->mdev;
2249 return mlx4_set_vf_link_state(mdev->dev, en_priv->port, vf, link_state);
2252 #define PORT_ID_BYTE_LEN 8
2253 static int mlx4_en_get_phys_port_id(struct net_device *dev,
2254 struct netdev_phys_port_id *ppid)
2256 struct mlx4_en_priv *priv = netdev_priv(dev);
2257 struct mlx4_dev *mdev = priv->mdev->dev;
2259 u64 phys_port_id = mdev->caps.phys_port_id[priv->port];
2264 ppid->id_len = sizeof(phys_port_id);
2265 for (i = PORT_ID_BYTE_LEN - 1; i >= 0; --i) {
2266 ppid->id[i] = phys_port_id & 0xff;
2272 #ifdef CONFIG_MLX4_EN_VXLAN
2273 static void mlx4_en_add_vxlan_offloads(struct work_struct *work)
2276 struct mlx4_en_priv *priv = container_of(work, struct mlx4_en_priv,
2279 ret = mlx4_config_vxlan_port(priv->mdev->dev, priv->vxlan_port);
2283 ret = mlx4_SET_PORT_VXLAN(priv->mdev->dev, priv->port,
2284 VXLAN_STEER_BY_OUTER_MAC, 1);
2287 en_err(priv, "failed setting L2 tunnel configuration ret %d\n", ret);
2290 static void mlx4_en_del_vxlan_offloads(struct work_struct *work)
2293 struct mlx4_en_priv *priv = container_of(work, struct mlx4_en_priv,
2296 ret = mlx4_SET_PORT_VXLAN(priv->mdev->dev, priv->port,
2297 VXLAN_STEER_BY_OUTER_MAC, 0);
2299 en_err(priv, "failed setting L2 tunnel configuration ret %d\n", ret);
2301 priv->vxlan_port = 0;
2304 static void mlx4_en_add_vxlan_port(struct net_device *dev,
2305 sa_family_t sa_family, __be16 port)
2307 struct mlx4_en_priv *priv = netdev_priv(dev);
2308 __be16 current_port;
2310 if (!(priv->mdev->dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_VXLAN_OFFLOADS))
2313 if (sa_family == AF_INET6)
2316 current_port = priv->vxlan_port;
2317 if (current_port && current_port != port) {
2318 en_warn(priv, "vxlan port %d configured, can't add port %d\n",
2319 ntohs(current_port), ntohs(port));
2323 priv->vxlan_port = port;
2324 queue_work(priv->mdev->workqueue, &priv->vxlan_add_task);
2327 static void mlx4_en_del_vxlan_port(struct net_device *dev,
2328 sa_family_t sa_family, __be16 port)
2330 struct mlx4_en_priv *priv = netdev_priv(dev);
2331 __be16 current_port;
2333 if (priv->mdev->dev->caps.tunnel_offload_mode != MLX4_TUNNEL_OFFLOAD_MODE_VXLAN)
2336 if (sa_family == AF_INET6)
2339 current_port = priv->vxlan_port;
2340 if (current_port != port) {
2341 en_dbg(DRV, priv, "vxlan port %d isn't configured, ignoring\n", ntohs(port));
2345 queue_work(priv->mdev->workqueue, &priv->vxlan_del_task);
2349 static const struct net_device_ops mlx4_netdev_ops = {
2350 .ndo_open = mlx4_en_open,
2351 .ndo_stop = mlx4_en_close,
2352 .ndo_start_xmit = mlx4_en_xmit,
2353 .ndo_select_queue = mlx4_en_select_queue,
2354 .ndo_get_stats = mlx4_en_get_stats,
2355 .ndo_set_rx_mode = mlx4_en_set_rx_mode,
2356 .ndo_set_mac_address = mlx4_en_set_mac,
2357 .ndo_validate_addr = eth_validate_addr,
2358 .ndo_change_mtu = mlx4_en_change_mtu,
2359 .ndo_do_ioctl = mlx4_en_ioctl,
2360 .ndo_tx_timeout = mlx4_en_tx_timeout,
2361 .ndo_vlan_rx_add_vid = mlx4_en_vlan_rx_add_vid,
2362 .ndo_vlan_rx_kill_vid = mlx4_en_vlan_rx_kill_vid,
2363 #ifdef CONFIG_NET_POLL_CONTROLLER
2364 .ndo_poll_controller = mlx4_en_netpoll,
2366 .ndo_set_features = mlx4_en_set_features,
2367 .ndo_setup_tc = mlx4_en_setup_tc,
2368 #ifdef CONFIG_RFS_ACCEL
2369 .ndo_rx_flow_steer = mlx4_en_filter_rfs,
2371 #ifdef CONFIG_NET_RX_BUSY_POLL
2372 .ndo_busy_poll = mlx4_en_low_latency_recv,
2374 .ndo_get_phys_port_id = mlx4_en_get_phys_port_id,
2375 #ifdef CONFIG_MLX4_EN_VXLAN
2376 .ndo_add_vxlan_port = mlx4_en_add_vxlan_port,
2377 .ndo_del_vxlan_port = mlx4_en_del_vxlan_port,
2381 static const struct net_device_ops mlx4_netdev_ops_master = {
2382 .ndo_open = mlx4_en_open,
2383 .ndo_stop = mlx4_en_close,
2384 .ndo_start_xmit = mlx4_en_xmit,
2385 .ndo_select_queue = mlx4_en_select_queue,
2386 .ndo_get_stats = mlx4_en_get_stats,
2387 .ndo_set_rx_mode = mlx4_en_set_rx_mode,
2388 .ndo_set_mac_address = mlx4_en_set_mac,
2389 .ndo_validate_addr = eth_validate_addr,
2390 .ndo_change_mtu = mlx4_en_change_mtu,
2391 .ndo_tx_timeout = mlx4_en_tx_timeout,
2392 .ndo_vlan_rx_add_vid = mlx4_en_vlan_rx_add_vid,
2393 .ndo_vlan_rx_kill_vid = mlx4_en_vlan_rx_kill_vid,
2394 .ndo_set_vf_mac = mlx4_en_set_vf_mac,
2395 .ndo_set_vf_vlan = mlx4_en_set_vf_vlan,
2396 .ndo_set_vf_spoofchk = mlx4_en_set_vf_spoofchk,
2397 .ndo_set_vf_link_state = mlx4_en_set_vf_link_state,
2398 .ndo_get_vf_config = mlx4_en_get_vf_config,
2399 #ifdef CONFIG_NET_POLL_CONTROLLER
2400 .ndo_poll_controller = mlx4_en_netpoll,
2402 .ndo_set_features = mlx4_en_set_features,
2403 .ndo_setup_tc = mlx4_en_setup_tc,
2404 #ifdef CONFIG_RFS_ACCEL
2405 .ndo_rx_flow_steer = mlx4_en_filter_rfs,
2407 .ndo_get_phys_port_id = mlx4_en_get_phys_port_id,
2410 int mlx4_en_init_netdev(struct mlx4_en_dev *mdev, int port,
2411 struct mlx4_en_port_profile *prof)
2413 struct net_device *dev;
2414 struct mlx4_en_priv *priv;
2419 dev = alloc_etherdev_mqs(sizeof(struct mlx4_en_priv),
2420 MAX_TX_RINGS, MAX_RX_RINGS);
2424 netif_set_real_num_tx_queues(dev, prof->tx_ring_num);
2425 netif_set_real_num_rx_queues(dev, prof->rx_ring_num);
2427 SET_NETDEV_DEV(dev, &mdev->dev->pdev->dev);
2428 dev->dev_port = port - 1;
2431 * Initialize driver private data
2434 priv = netdev_priv(dev);
2435 memset(priv, 0, sizeof(struct mlx4_en_priv));
2438 priv->ddev = &mdev->pdev->dev;
2441 priv->port_up = false;
2442 priv->flags = prof->flags;
2443 priv->ctrl_flags = cpu_to_be32(MLX4_WQE_CTRL_CQ_UPDATE |
2444 MLX4_WQE_CTRL_SOLICITED);
2445 priv->num_tx_rings_p_up = mdev->profile.num_tx_rings_p_up;
2446 priv->tx_ring_num = prof->tx_ring_num;
2448 priv->tx_ring = kzalloc(sizeof(struct mlx4_en_tx_ring *) * MAX_TX_RINGS,
2450 if (!priv->tx_ring) {
2454 priv->tx_cq = kzalloc(sizeof(struct mlx4_en_cq *) * MAX_TX_RINGS,
2460 priv->rx_ring_num = prof->rx_ring_num;
2461 priv->cqe_factor = (mdev->dev->caps.cqe_size == 64) ? 1 : 0;
2462 priv->mac_index = -1;
2463 priv->msg_enable = MLX4_EN_MSG_LEVEL;
2464 spin_lock_init(&priv->stats_lock);
2465 INIT_WORK(&priv->rx_mode_task, mlx4_en_do_set_rx_mode);
2466 INIT_WORK(&priv->watchdog_task, mlx4_en_restart);
2467 INIT_WORK(&priv->linkstate_task, mlx4_en_linkstate);
2468 INIT_DELAYED_WORK(&priv->stats_task, mlx4_en_do_get_stats);
2469 INIT_DELAYED_WORK(&priv->service_task, mlx4_en_service_task);
2470 #ifdef CONFIG_MLX4_EN_VXLAN
2471 INIT_WORK(&priv->vxlan_add_task, mlx4_en_add_vxlan_offloads);
2472 INIT_WORK(&priv->vxlan_del_task, mlx4_en_del_vxlan_offloads);
2474 #ifdef CONFIG_MLX4_EN_DCB
2475 if (!mlx4_is_slave(priv->mdev->dev)) {
2476 if (mdev->dev->caps.flags & MLX4_DEV_CAP_FLAG_SET_ETH_SCHED) {
2477 dev->dcbnl_ops = &mlx4_en_dcbnl_ops;
2479 en_info(priv, "enabling only PFC DCB ops\n");
2480 dev->dcbnl_ops = &mlx4_en_dcbnl_pfc_ops;
2485 for (i = 0; i < MLX4_EN_MAC_HASH_SIZE; ++i)
2486 INIT_HLIST_HEAD(&priv->mac_hash[i]);
2488 /* Query for default mac and max mtu */
2489 priv->max_mtu = mdev->dev->caps.eth_mtu_cap[priv->port];
2491 /* Set default MAC */
2492 dev->addr_len = ETH_ALEN;
2493 mlx4_en_u64_to_mac(dev->dev_addr, mdev->dev->caps.def_mac[priv->port]);
2494 if (!is_valid_ether_addr(dev->dev_addr)) {
2495 if (mlx4_is_slave(priv->mdev->dev)) {
2496 eth_hw_addr_random(dev);
2497 en_warn(priv, "Assigned random MAC address %pM\n", dev->dev_addr);
2498 mac_u64 = mlx4_mac_to_u64(dev->dev_addr);
2499 mdev->dev->caps.def_mac[priv->port] = mac_u64;
2501 en_err(priv, "Port: %d, invalid mac burned: %pM, quiting\n",
2502 priv->port, dev->dev_addr);
2508 memcpy(priv->prev_mac, dev->dev_addr, sizeof(priv->prev_mac));
2510 priv->stride = roundup_pow_of_two(sizeof(struct mlx4_en_rx_desc) +
2511 DS_SIZE * MLX4_EN_MAX_RX_FRAGS);
2512 err = mlx4_en_alloc_resources(priv);
2516 #ifdef CONFIG_RFS_ACCEL
2517 INIT_LIST_HEAD(&priv->filters);
2518 spin_lock_init(&priv->filters_lock);
2521 /* Initialize time stamping config */
2522 priv->hwtstamp_config.flags = 0;
2523 priv->hwtstamp_config.tx_type = HWTSTAMP_TX_OFF;
2524 priv->hwtstamp_config.rx_filter = HWTSTAMP_FILTER_NONE;
2526 /* Allocate page for receive rings */
2527 err = mlx4_alloc_hwq_res(mdev->dev, &priv->res,
2528 MLX4_EN_PAGE_SIZE, MLX4_EN_PAGE_SIZE);
2530 en_err(priv, "Failed to allocate page for rx qps\n");
2533 priv->allocated = 1;
2536 * Initialize netdev entry points
2538 if (mlx4_is_master(priv->mdev->dev))
2539 dev->netdev_ops = &mlx4_netdev_ops_master;
2541 dev->netdev_ops = &mlx4_netdev_ops;
2542 dev->watchdog_timeo = MLX4_EN_WATCHDOG_TIMEOUT;
2543 netif_set_real_num_tx_queues(dev, priv->tx_ring_num);
2544 netif_set_real_num_rx_queues(dev, priv->rx_ring_num);
2546 SET_ETHTOOL_OPS(dev, &mlx4_en_ethtool_ops);
2549 * Set driver features
2551 dev->hw_features = NETIF_F_SG | NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM;
2552 if (mdev->LSO_support)
2553 dev->hw_features |= NETIF_F_TSO | NETIF_F_TSO6;
2555 dev->vlan_features = dev->hw_features;
2557 dev->hw_features |= NETIF_F_RXCSUM | NETIF_F_RXHASH;
2558 dev->features = dev->hw_features | NETIF_F_HIGHDMA |
2559 NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_CTAG_RX |
2560 NETIF_F_HW_VLAN_CTAG_FILTER;
2561 dev->hw_features |= NETIF_F_LOOPBACK;
2563 if (mdev->dev->caps.steering_mode ==
2564 MLX4_STEERING_MODE_DEVICE_MANAGED)
2565 dev->hw_features |= NETIF_F_NTUPLE;
2567 if (mdev->dev->caps.steering_mode != MLX4_STEERING_MODE_A0)
2568 dev->priv_flags |= IFF_UNICAST_FLT;
2570 if (mdev->dev->caps.tunnel_offload_mode == MLX4_TUNNEL_OFFLOAD_MODE_VXLAN) {
2571 dev->hw_enc_features |= NETIF_F_IP_CSUM | NETIF_F_RXCSUM |
2572 NETIF_F_TSO | NETIF_F_GSO_UDP_TUNNEL;
2573 dev->hw_features |= NETIF_F_GSO_UDP_TUNNEL;
2574 dev->features |= NETIF_F_GSO_UDP_TUNNEL;
2577 mdev->pndev[port] = dev;
2579 netif_carrier_off(dev);
2580 mlx4_en_set_default_moderation(priv);
2582 err = register_netdev(dev);
2584 en_err(priv, "Netdev registration failed for port %d\n", port);
2587 priv->registered = 1;
2589 en_warn(priv, "Using %d TX rings\n", prof->tx_ring_num);
2590 en_warn(priv, "Using %d RX rings\n", prof->rx_ring_num);
2592 mlx4_en_update_loopback_state(priv->dev, priv->dev->features);
2594 /* Configure port */
2595 mlx4_en_calc_rx_buf(dev);
2596 err = mlx4_SET_PORT_general(mdev->dev, priv->port,
2597 priv->rx_skb_size + ETH_FCS_LEN,
2598 prof->tx_pause, prof->tx_ppp,
2599 prof->rx_pause, prof->rx_ppp);
2601 en_err(priv, "Failed setting port general configurations "
2602 "for port %d, with error %d\n", priv->port, err);
2606 if (mdev->dev->caps.tunnel_offload_mode == MLX4_TUNNEL_OFFLOAD_MODE_VXLAN) {
2607 err = mlx4_SET_PORT_VXLAN(mdev->dev, priv->port, VXLAN_STEER_BY_OUTER_MAC, 1);
2609 en_err(priv, "Failed setting port L2 tunnel configuration, err %d\n",
2616 en_warn(priv, "Initializing port\n");
2617 err = mlx4_INIT_PORT(mdev->dev, priv->port);
2619 en_err(priv, "Failed Initializing port\n");
2622 queue_delayed_work(mdev->workqueue, &priv->stats_task, STATS_DELAY);
2624 if (mdev->dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_TS)
2625 queue_delayed_work(mdev->workqueue, &priv->service_task,
2626 SERVICE_TASK_DELAY);
2631 mlx4_en_destroy_netdev(dev);