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 MLX4_NET_TRANS_RULE_NUM
;
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
>= MLX4_NET_TRANS_RULE_NUM
) {
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
)
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 */
482 err
= mlx4_tunnel_steer_add(priv
->mdev
->dev
, addr
, priv
->port
, qpn
,
483 MLX4_DOMAIN_NIC
, reg_id
);
485 en_err(priv
, "failed to add vxlan steering rule, err %d\n", err
);
488 en_dbg(DRV
, priv
, "added vxlan steering rule, mac %pM reg_id %llx\n", addr
, *reg_id
);
493 static int mlx4_en_uc_steer_add(struct mlx4_en_priv
*priv
,
494 unsigned char *mac
, int *qpn
, u64
*reg_id
)
496 struct mlx4_en_dev
*mdev
= priv
->mdev
;
497 struct mlx4_dev
*dev
= mdev
->dev
;
500 switch (dev
->caps
.steering_mode
) {
501 case MLX4_STEERING_MODE_B0
: {
506 memcpy(&gid
[10], mac
, ETH_ALEN
);
509 err
= mlx4_unicast_attach(dev
, &qp
, gid
, 0, MLX4_PROT_ETH
);
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);
516 struct mlx4_net_trans_rule rule
= {
517 .queue_mode
= MLX4_NET_TRANS_Q_FIFO
,
520 .promisc_mode
= MLX4_FS_REGULAR
,
521 .priority
= MLX4_DOMAIN_NIC
,
524 rule
.port
= priv
->port
;
526 INIT_LIST_HEAD(&rule
.list
);
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
);
533 err
= mlx4_flow_attach(dev
, &rule
, reg_id
);
540 en_warn(priv
, "Failed Attaching Unicast\n");
545 static void mlx4_en_uc_steer_release(struct mlx4_en_priv
*priv
,
546 unsigned char *mac
, int qpn
, u64 reg_id
)
548 struct mlx4_en_dev
*mdev
= priv
->mdev
;
549 struct mlx4_dev
*dev
= mdev
->dev
;
551 switch (dev
->caps
.steering_mode
) {
552 case MLX4_STEERING_MODE_B0
: {
557 memcpy(&gid
[10], mac
, ETH_ALEN
);
560 mlx4_unicast_detach(dev
, &qp
, gid
, MLX4_PROT_ETH
);
563 case MLX4_STEERING_MODE_DEVICE_MANAGED
: {
564 mlx4_flow_detach(dev
, reg_id
);
568 en_err(priv
, "Invalid steering mode.\n");
572 static int mlx4_en_get_qp(struct mlx4_en_priv
*priv
)
574 struct mlx4_en_dev
*mdev
= priv
->mdev
;
575 struct mlx4_dev
*dev
= mdev
->dev
;
576 struct mlx4_mac_entry
*entry
;
580 int *qpn
= &priv
->base_qpn
;
581 u64 mac
= mlx4_mac_to_u64(priv
->dev
->dev_addr
);
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
);
588 en_err(priv
, "Failed adding MAC: %pM\n",
589 priv
->dev
->dev_addr
);
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
;
599 err
= mlx4_qp_reserve_range(dev
, 1, 1, qpn
, MLX4_RESERVE_A0_QP
);
600 en_dbg(DRV
, priv
, "Reserved qp %d\n", *qpn
);
602 en_err(priv
, "Failed to reserve qp for mac registration\n");
606 err
= mlx4_en_uc_steer_add(priv
, priv
->dev
->dev_addr
, qpn
, ®_id
);
610 err
= mlx4_en_tunnel_steer_add(priv
, priv
->dev
->dev_addr
, *qpn
,
611 &priv
->tunnel_reg_id
);
615 entry
= kmalloc(sizeof(*entry
), GFP_KERNEL
);
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
;
624 hlist_add_head_rcu(&entry
->hlist
,
625 &priv
->mac_hash
[entry
->mac
[MLX4_EN_MAC_HASH_IDX
]]);
630 if (priv
->tunnel_reg_id
)
631 mlx4_flow_detach(priv
->mdev
->dev
, priv
->tunnel_reg_id
);
633 mlx4_en_uc_steer_release(priv
, priv
->dev
->dev_addr
, *qpn
, reg_id
);
636 mlx4_qp_release_range(dev
, *qpn
, 1);
639 mlx4_unregister_mac(dev
, priv
->port
, mac
);
643 static void mlx4_en_put_qp(struct mlx4_en_priv
*priv
)
645 struct mlx4_en_dev
*mdev
= priv
->mdev
;
646 struct mlx4_dev
*dev
= mdev
->dev
;
647 int qpn
= priv
->base_qpn
;
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
);
656 struct mlx4_mac_entry
*entry
;
657 struct hlist_node
*tmp
;
658 struct hlist_head
*bucket
;
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",
667 mlx4_en_uc_steer_release(priv
, entry
->mac
,
670 mlx4_unregister_mac(dev
, priv
->port
, mac
);
671 hlist_del_rcu(&entry
->hlist
);
672 kfree_rcu(entry
, rcu
);
676 if (priv
->tunnel_reg_id
) {
677 mlx4_flow_detach(priv
->mdev
->dev
, priv
->tunnel_reg_id
);
678 priv
->tunnel_reg_id
= 0;
681 en_dbg(DRV
, priv
, "Releasing qp: port %d, qpn %d\n",
683 mlx4_qp_release_range(dev
, qpn
, 1);
684 priv
->flags
&= ~MLX4_EN_FLAG_FORCE_PROMISC
;
688 static int mlx4_en_replace_mac(struct mlx4_en_priv
*priv
, int qpn
,
689 unsigned char *new_mac
, unsigned char *prev_mac
)
691 struct mlx4_en_dev
*mdev
= priv
->mdev
;
692 struct mlx4_dev
*dev
= mdev
->dev
;
694 u64 new_mac_u64
= mlx4_mac_to_u64(new_mac
);
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
);
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
,
708 mlx4_unregister_mac(dev
, priv
->port
,
710 hlist_del_rcu(&entry
->hlist
);
712 memcpy(entry
->mac
, new_mac
, ETH_ALEN
);
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
,
723 if (priv
->tunnel_reg_id
) {
724 mlx4_flow_detach(priv
->mdev
->dev
, priv
->tunnel_reg_id
);
725 priv
->tunnel_reg_id
= 0;
727 err
= mlx4_en_tunnel_steer_add(priv
, new_mac
, qpn
,
728 &priv
->tunnel_reg_id
);
735 return __mlx4_replace_mac(dev
, priv
->port
, qpn
, new_mac_u64
);
738 static int mlx4_en_do_set_mac(struct mlx4_en_priv
*priv
,
739 unsigned char new_mac
[ETH_ALEN
+ 2])
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
);
748 en_err(priv
, "Failed changing HW MAC address\n");
750 en_dbg(HW
, priv
, "Port is down while registering mac, exiting...\n");
753 memcpy(priv
->current_mac
, new_mac
, sizeof(priv
->current_mac
));
758 static int mlx4_en_set_mac(struct net_device
*dev
, void *addr
)
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];
766 if (!is_valid_ether_addr(saddr
->sa_data
))
767 return -EADDRNOTAVAIL
;
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
);
773 memcpy(dev
->dev_addr
, saddr
->sa_data
, ETH_ALEN
);
774 mutex_unlock(&mdev
->state_lock
);
779 static void mlx4_en_clear_list(struct net_device
*dev
)
781 struct mlx4_en_priv
*priv
= netdev_priv(dev
);
782 struct mlx4_en_mc_list
*tmp
, *mc_to_del
;
784 list_for_each_entry_safe(mc_to_del
, tmp
, &priv
->mc_list
, list
) {
785 list_del(&mc_to_del
->list
);
790 static void mlx4_en_cache_mclist(struct net_device
*dev
)
792 struct mlx4_en_priv
*priv
= netdev_priv(dev
);
793 struct netdev_hw_addr
*ha
;
794 struct mlx4_en_mc_list
*tmp
;
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
);
800 mlx4_en_clear_list(dev
);
803 memcpy(tmp
->addr
, ha
->addr
, ETH_ALEN
);
804 list_add_tail(&tmp
->list
, &priv
->mc_list
);
808 static void update_mclist_flags(struct mlx4_en_priv
*priv
,
809 struct list_head
*dst
,
810 struct list_head
*src
)
812 struct mlx4_en_mc_list
*dst_tmp
, *src_tmp
, *new_mc
;
815 /* Find all the entries that should be removed from dst,
816 * These are the entries that are not found in src
818 list_for_each_entry(dst_tmp
, dst
, list
) {
820 list_for_each_entry(src_tmp
, src
, list
) {
821 if (ether_addr_equal(dst_tmp
->addr
, src_tmp
->addr
)) {
827 dst_tmp
->action
= MCLIST_REM
;
830 /* Add entries that exist in src but not in dst
831 * mark them as need to add
833 list_for_each_entry(src_tmp
, src
, list
) {
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
;
843 new_mc
= kmemdup(src_tmp
,
844 sizeof(struct mlx4_en_mc_list
),
849 new_mc
->action
= MCLIST_ADD
;
850 list_add_tail(&new_mc
->list
, dst
);
855 static void mlx4_en_set_rx_mode(struct net_device
*dev
)
857 struct mlx4_en_priv
*priv
= netdev_priv(dev
);
862 queue_work(priv
->mdev
->workqueue
, &priv
->rx_mode_task
);
865 static void mlx4_en_set_promisc_mode(struct mlx4_en_priv
*priv
,
866 struct mlx4_en_dev
*mdev
)
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
;
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
,
881 MLX4_FS_ALL_DEFAULT
);
883 en_err(priv
, "Failed enabling promiscuous mode\n");
884 priv
->flags
|= MLX4_EN_FLAG_MC_PROMISC
;
887 case MLX4_STEERING_MODE_B0
:
888 err
= mlx4_unicast_promisc_add(mdev
->dev
,
892 en_err(priv
, "Failed enabling unicast promiscuous mode\n");
894 /* Add the default qp number as multicast
897 if (!(priv
->flags
& MLX4_EN_FLAG_MC_PROMISC
)) {
898 err
= mlx4_multicast_promisc_add(mdev
->dev
,
902 en_err(priv
, "Failed enabling multicast promiscuous mode\n");
903 priv
->flags
|= MLX4_EN_FLAG_MC_PROMISC
;
907 case MLX4_STEERING_MODE_A0
:
908 err
= mlx4_SET_PORT_qpn_calc(mdev
->dev
,
913 en_err(priv
, "Failed enabling promiscuous mode\n");
917 /* Disable port multicast filter (unconditionally) */
918 err
= mlx4_SET_MCAST_FLTR(mdev
->dev
, priv
->port
, 0,
919 0, MLX4_MCAST_DISABLE
);
921 en_err(priv
, "Failed disabling multicast filter\n");
925 static void mlx4_en_clear_promisc_mode(struct mlx4_en_priv
*priv
,
926 struct mlx4_en_dev
*mdev
)
930 if (netif_msg_rx_status(priv
))
931 en_warn(priv
, "Leaving promiscuous mode\n");
932 priv
->flags
&= ~MLX4_EN_FLAG_PROMISC
;
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
,
939 MLX4_FS_ALL_DEFAULT
);
941 en_err(priv
, "Failed disabling promiscuous mode\n");
942 priv
->flags
&= ~MLX4_EN_FLAG_MC_PROMISC
;
945 case MLX4_STEERING_MODE_B0
:
946 err
= mlx4_unicast_promisc_remove(mdev
->dev
,
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
,
957 en_err(priv
, "Failed disabling multicast promiscuous mode\n");
958 priv
->flags
&= ~MLX4_EN_FLAG_MC_PROMISC
;
962 case MLX4_STEERING_MODE_A0
:
963 err
= mlx4_SET_PORT_qpn_calc(mdev
->dev
,
967 en_err(priv
, "Failed disabling promiscuous mode\n");
972 static void mlx4_en_do_multicast(struct mlx4_en_priv
*priv
,
973 struct net_device
*dev
,
974 struct mlx4_en_dev
*mdev
)
976 struct mlx4_en_mc_list
*mclist
, *tmp
;
978 u8 mc_list
[16] = {0};
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
);
986 en_err(priv
, "Failed disabling multicast filter\n");
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
,
998 case MLX4_STEERING_MODE_B0
:
999 err
= mlx4_multicast_promisc_add(mdev
->dev
,
1004 case MLX4_STEERING_MODE_A0
:
1008 en_err(priv
, "Failed entering multicast promisc mode\n");
1009 priv
->flags
|= MLX4_EN_FLAG_MC_PROMISC
;
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
,
1018 MLX4_FS_MC_DEFAULT
);
1021 case MLX4_STEERING_MODE_B0
:
1022 err
= mlx4_multicast_promisc_remove(mdev
->dev
,
1027 case MLX4_STEERING_MODE_A0
:
1031 en_err(priv
, "Failed disabling multicast promiscuous mode\n");
1032 priv
->flags
&= ~MLX4_EN_FLAG_MC_PROMISC
;
1035 err
= mlx4_SET_MCAST_FLTR(mdev
->dev
, priv
->port
, 0,
1036 0, MLX4_MCAST_DISABLE
);
1038 en_err(priv
, "Failed disabling multicast filter\n");
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
);
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
);
1054 err
= mlx4_SET_MCAST_FLTR(mdev
->dev
, priv
->port
, 0,
1055 0, MLX4_MCAST_ENABLE
);
1057 en_err(priv
, "Failed enabling multicast filter\n");
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
,
1071 en_err(priv
, "Fail to detach multicast address\n");
1073 if (mclist
->tunnel_reg_id
) {
1074 err
= mlx4_flow_detach(priv
->mdev
->dev
, mclist
->tunnel_reg_id
);
1076 en_err(priv
, "Failed to detach multicast address\n");
1079 /* remove from list */
1080 list_del(&mclist
->list
);
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
,
1094 en_err(priv
, "Fail to attach multicast address\n");
1096 err
= mlx4_en_tunnel_steer_add(priv
, &mc_list
[10], priv
->base_qpn
,
1097 &mclist
->tunnel_reg_id
);
1099 en_err(priv
, "Failed to attach multicast address\n");
1105 static void mlx4_en_do_uc_filter(struct mlx4_en_priv
*priv
,
1106 struct net_device
*dev
,
1107 struct mlx4_en_dev
*mdev
)
1109 struct netdev_hw_addr
*ha
;
1110 struct mlx4_mac_entry
*entry
;
1111 struct hlist_node
*tmp
;
1115 struct hlist_head
*bucket
;
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
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
) {
1129 netdev_for_each_uc_addr(ha
, dev
) {
1130 if (ether_addr_equal_64bits(entry
->mac
,
1137 /* MAC address of the port is not in uc list */
1138 if (ether_addr_equal_64bits(entry
->mac
,
1143 mac
= mlx4_mac_to_u64(entry
->mac
);
1144 mlx4_en_uc_steer_release(priv
, entry
->mac
,
1147 mlx4_unregister_mac(mdev
->dev
, priv
->port
, mac
);
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
);
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
1161 if ((priv
->flags
& MLX4_EN_FLAG_FORCE_PROMISC
) && 0 == removed
)
1164 prev_flags
= priv
->flags
;
1165 priv
->flags
&= ~MLX4_EN_FLAG_FORCE_PROMISC
;
1167 /* find what to add */
1168 netdev_for_each_uc_addr(ha
, dev
) {
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
)) {
1179 entry
= kmalloc(sizeof(*entry
), GFP_KERNEL
);
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
;
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
);
1190 en_err(priv
, "Failed registering MAC %pM on port %d: %d\n",
1191 ha
->addr
, priv
->port
, err
);
1193 priv
->flags
|= MLX4_EN_FLAG_FORCE_PROMISC
;
1196 err
= mlx4_en_uc_steer_add(priv
, ha
->addr
,
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
);
1204 priv
->flags
|= MLX4_EN_FLAG_FORCE_PROMISC
;
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
);
1217 if (priv
->flags
& MLX4_EN_FLAG_FORCE_PROMISC
) {
1218 en_warn(priv
, "Forcing promiscuous mode on port:%d\n",
1220 } else if (prev_flags
& MLX4_EN_FLAG_FORCE_PROMISC
) {
1221 en_warn(priv
, "Stop forcing promiscuous mode on port:%d\n",
1226 static void mlx4_en_do_set_rx_mode(struct work_struct
*work
)
1228 struct mlx4_en_priv
*priv
= container_of(work
, struct mlx4_en_priv
,
1230 struct mlx4_en_dev
*mdev
= priv
->mdev
;
1231 struct net_device
*dev
= priv
->dev
;
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");
1238 if (!priv
->port_up
) {
1239 en_dbg(HW
, priv
, "Port is down, ignoring rx mode change.\n");
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");
1253 if (dev
->priv_flags
& IFF_UNICAST_FLT
)
1254 mlx4_en_do_uc_filter(priv
, dev
, mdev
);
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
);
1263 /* Not in promiscuous mode */
1264 if (priv
->flags
& MLX4_EN_FLAG_PROMISC
)
1265 mlx4_en_clear_promisc_mode(priv
, mdev
);
1267 mlx4_en_do_multicast(priv
, dev
, mdev
);
1269 mutex_unlock(&mdev
->state_lock
);
1272 #ifdef CONFIG_NET_POLL_CONTROLLER
1273 static void mlx4_en_netpoll(struct net_device
*dev
)
1275 struct mlx4_en_priv
*priv
= netdev_priv(dev
);
1276 struct mlx4_en_cq
*cq
;
1279 for (i
= 0; i
< priv
->rx_ring_num
; i
++) {
1280 cq
= priv
->rx_cq
[i
];
1281 napi_schedule(&cq
->napi
);
1286 static void mlx4_en_tx_timeout(struct net_device
*dev
)
1288 struct mlx4_en_priv
*priv
= netdev_priv(dev
);
1289 struct mlx4_en_dev
*mdev
= priv
->mdev
;
1292 if (netif_msg_timer(priv
))
1293 en_warn(priv
, "Tx timeout called on port:%d\n", priv
->port
);
1295 for (i
= 0; i
< priv
->tx_ring_num
; i
++) {
1296 if (!netif_tx_queue_stopped(netdev_get_tx_queue(dev
, i
)))
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
);
1303 priv
->port_stats
.tx_timeout
++;
1304 en_dbg(DRV
, priv
, "Scheduling watchdog\n");
1305 queue_work(mdev
->workqueue
, &priv
->watchdog_task
);
1309 static struct net_device_stats
*mlx4_en_get_stats(struct net_device
*dev
)
1311 struct mlx4_en_priv
*priv
= netdev_priv(dev
);
1313 spin_lock_bh(&priv
->stats_lock
);
1314 memcpy(&priv
->ret_stats
, &priv
->stats
, sizeof(priv
->stats
));
1315 spin_unlock_bh(&priv
->stats_lock
);
1317 return &priv
->ret_stats
;
1320 static void mlx4_en_set_default_moderation(struct mlx4_en_priv
*priv
)
1322 struct mlx4_en_cq
*cq
;
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.
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
);
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;
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
;
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;
1365 static void mlx4_en_auto_moderation(struct mlx4_en_priv
*priv
)
1367 unsigned long period
= (unsigned long) (jiffies
- priv
->last_moder_jiffies
);
1368 struct mlx4_en_cq
*cq
;
1369 unsigned long packets
;
1371 unsigned long avg_pkt_size
;
1372 unsigned long rx_packets
;
1373 unsigned long rx_bytes
;
1374 unsigned long rx_pkt_diff
;
1378 if (!priv
->adaptive_rx_coal
|| period
< priv
->sample_interval
* HZ
)
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
);
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;
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
;
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
) +
1408 moder_time
= priv
->rx_usecs_low
;
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
);
1418 en_err(priv
, "Failed modifying moderation for cq:%d\n",
1421 priv
->last_moder_packets
[ring
] = rx_packets
;
1422 priv
->last_moder_bytes
[ring
] = rx_bytes
;
1425 priv
->last_moder_jiffies
= jiffies
;
1428 static void mlx4_en_do_get_stats(struct work_struct
*work
)
1430 struct delayed_work
*delay
= to_delayed_work(work
);
1431 struct mlx4_en_priv
*priv
= container_of(delay
, struct mlx4_en_priv
,
1433 struct mlx4_en_dev
*mdev
= priv
->mdev
;
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);
1441 en_dbg(HW
, priv
, "Could not update stats\n");
1443 mlx4_en_auto_moderation(priv
);
1446 queue_delayed_work(mdev
->workqueue
, &priv
->stats_task
, STATS_DELAY
);
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;
1452 mutex_unlock(&mdev
->state_lock
);
1455 /* mlx4_en_service_task - Run service task for tasks that needed to be done
1458 static void mlx4_en_service_task(struct work_struct
*work
)
1460 struct delayed_work
*delay
= to_delayed_work(work
);
1461 struct mlx4_en_priv
*priv
= container_of(delay
, struct mlx4_en_priv
,
1463 struct mlx4_en_dev
*mdev
= priv
->mdev
;
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
);
1470 mlx4_en_recover_from_oom(priv
);
1471 queue_delayed_work(mdev
->workqueue
, &priv
->service_task
,
1472 SERVICE_TASK_DELAY
);
1474 mutex_unlock(&mdev
->state_lock
);
1477 static void mlx4_en_linkstate(struct work_struct
*work
)
1479 struct mlx4_en_priv
*priv
= container_of(work
, struct mlx4_en_priv
,
1481 struct mlx4_en_dev
*mdev
= priv
->mdev
;
1482 int linkstate
= priv
->link_state
;
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
);
1492 en_info(priv
, "Link Up\n");
1493 netif_carrier_on(priv
->dev
);
1496 priv
->last_link_state
= linkstate
;
1497 mutex_unlock(&mdev
->state_lock
);
1500 static int mlx4_en_init_affinity_hint(struct mlx4_en_priv
*priv
, int ring_idx
)
1502 struct mlx4_en_rx_ring
*ring
= priv
->rx_ring
[ring_idx
];
1503 int numa_node
= priv
->mdev
->dev
->numa_node
;
1506 if (!zalloc_cpumask_var(&ring
->affinity_mask
, GFP_KERNEL
))
1509 ret
= cpumask_set_cpu_local_first(ring_idx
, numa_node
,
1510 ring
->affinity_mask
);
1512 free_cpumask_var(ring
->affinity_mask
);
1517 static void mlx4_en_free_affinity_hint(struct mlx4_en_priv
*priv
, int ring_idx
)
1519 free_cpumask_var(priv
->rx_ring
[ring_idx
]->affinity_mask
);
1522 int mlx4_en_start_port(struct net_device
*dev
)
1524 struct mlx4_en_priv
*priv
= netdev_priv(dev
);
1525 struct mlx4_en_dev
*mdev
= priv
->mdev
;
1526 struct mlx4_en_cq
*cq
;
1527 struct mlx4_en_tx_ring
*tx_ring
;
1533 u8 mc_list
[16] = {0};
1535 if (priv
->port_up
) {
1536 en_dbg(DRV
, priv
, "start port called while port already up\n");
1540 INIT_LIST_HEAD(&priv
->mc_list
);
1541 INIT_LIST_HEAD(&priv
->curr_list
);
1542 INIT_LIST_HEAD(&priv
->ethtool_list
);
1543 memset(&priv
->ethtool_rules
[0], 0,
1544 sizeof(struct ethtool_flow_id
) * MAX_NUM_OF_FS_RULES
);
1546 /* Calculate Rx buf size */
1547 dev
->mtu
= min(dev
->mtu
, priv
->max_mtu
);
1548 mlx4_en_calc_rx_buf(dev
);
1549 en_dbg(DRV
, priv
, "Rx buf size:%d\n", priv
->rx_skb_size
);
1551 /* Configure rx cq's and rings */
1552 err
= mlx4_en_activate_rx_rings(priv
);
1554 en_err(priv
, "Failed to activate RX rings\n");
1557 for (i
= 0; i
< priv
->rx_ring_num
; i
++) {
1558 cq
= priv
->rx_cq
[i
];
1560 mlx4_en_cq_init_lock(cq
);
1562 err
= mlx4_en_init_affinity_hint(priv
, i
);
1564 en_err(priv
, "Failed preparing IRQ affinity hint\n");
1568 err
= mlx4_en_activate_cq(priv
, cq
, i
);
1570 en_err(priv
, "Failed activating Rx CQ\n");
1571 mlx4_en_free_affinity_hint(priv
, i
);
1575 for (j
= 0; j
< cq
->size
; j
++) {
1576 struct mlx4_cqe
*cqe
= NULL
;
1578 cqe
= mlx4_en_get_cqe(cq
->buf
, j
, priv
->cqe_size
) +
1580 cqe
->owner_sr_opcode
= MLX4_CQE_OWNER_MASK
;
1583 err
= mlx4_en_set_cq_moder(priv
, cq
);
1585 en_err(priv
, "Failed setting cq moderation parameters\n");
1586 mlx4_en_deactivate_cq(priv
, cq
);
1587 mlx4_en_free_affinity_hint(priv
, i
);
1590 mlx4_en_arm_cq(priv
, cq
);
1591 priv
->rx_ring
[i
]->cqn
= cq
->mcq
.cqn
;
1596 en_dbg(DRV
, priv
, "Getting qp number for port %d\n", priv
->port
);
1597 err
= mlx4_en_get_qp(priv
);
1599 en_err(priv
, "Failed getting eth qp\n");
1602 mdev
->mac_removed
[priv
->port
] = 0;
1604 err
= mlx4_en_config_rss_steer(priv
);
1606 en_err(priv
, "Failed configuring rss steering\n");
1610 err
= mlx4_en_create_drop_qp(priv
);
1614 /* Configure tx cq's and rings */
1615 for (i
= 0; i
< priv
->tx_ring_num
; i
++) {
1617 cq
= priv
->tx_cq
[i
];
1618 err
= mlx4_en_activate_cq(priv
, cq
, i
);
1620 en_err(priv
, "Failed allocating Tx CQ\n");
1623 err
= mlx4_en_set_cq_moder(priv
, cq
);
1625 en_err(priv
, "Failed setting cq moderation parameters\n");
1626 mlx4_en_deactivate_cq(priv
, cq
);
1629 en_dbg(DRV
, priv
, "Resetting index of collapsed CQ:%d to -1\n", i
);
1630 cq
->buf
->wqe_index
= cpu_to_be16(0xffff);
1632 /* Configure ring */
1633 tx_ring
= priv
->tx_ring
[i
];
1634 err
= mlx4_en_activate_tx_ring(priv
, tx_ring
, cq
->mcq
.cqn
,
1635 i
/ priv
->num_tx_rings_p_up
);
1637 en_err(priv
, "Failed allocating Tx ring\n");
1638 mlx4_en_deactivate_cq(priv
, cq
);
1641 tx_ring
->tx_queue
= netdev_get_tx_queue(dev
, i
);
1643 /* Arm CQ for TX completions */
1644 mlx4_en_arm_cq(priv
, cq
);
1646 /* Set initial ownership of all Tx TXBBs to SW (1) */
1647 for (j
= 0; j
< tx_ring
->buf_size
; j
+= STAMP_STRIDE
)
1648 *((u32
*) (tx_ring
->buf
+ j
)) = 0xffffffff;
1652 /* Configure port */
1653 err
= mlx4_SET_PORT_general(mdev
->dev
, priv
->port
,
1654 priv
->rx_skb_size
+ ETH_FCS_LEN
,
1655 priv
->prof
->tx_pause
,
1657 priv
->prof
->rx_pause
,
1658 priv
->prof
->rx_ppp
);
1660 en_err(priv
, "Failed setting port general configurations for port %d, with error %d\n",
1664 /* Set default qp number */
1665 err
= mlx4_SET_PORT_qpn_calc(mdev
->dev
, priv
->port
, priv
->base_qpn
, 0);
1667 en_err(priv
, "Failed setting default qp numbers\n");
1671 if (mdev
->dev
->caps
.tunnel_offload_mode
== MLX4_TUNNEL_OFFLOAD_MODE_VXLAN
) {
1672 err
= mlx4_SET_PORT_VXLAN(mdev
->dev
, priv
->port
, VXLAN_STEER_BY_OUTER_MAC
, 1);
1674 en_err(priv
, "Failed setting port L2 tunnel configuration, err %d\n",
1681 en_dbg(HW
, priv
, "Initializing port\n");
1682 err
= mlx4_INIT_PORT(mdev
->dev
, priv
->port
);
1684 en_err(priv
, "Failed Initializing port\n");
1688 /* Attach rx QP to bradcast address */
1689 eth_broadcast_addr(&mc_list
[10]);
1690 mc_list
[5] = priv
->port
; /* needed for B0 steering support */
1691 if (mlx4_multicast_attach(mdev
->dev
, &priv
->rss_map
.indir_qp
, mc_list
,
1692 priv
->port
, 0, MLX4_PROT_ETH
,
1693 &priv
->broadcast_id
))
1694 mlx4_warn(mdev
, "Failed Attaching Broadcast\n");
1696 /* Must redo promiscuous mode setup. */
1697 priv
->flags
&= ~(MLX4_EN_FLAG_PROMISC
| MLX4_EN_FLAG_MC_PROMISC
);
1699 /* Schedule multicast task to populate multicast list */
1700 queue_work(mdev
->workqueue
, &priv
->rx_mode_task
);
1702 #ifdef CONFIG_MLX4_EN_VXLAN
1703 if (priv
->mdev
->dev
->caps
.tunnel_offload_mode
== MLX4_TUNNEL_OFFLOAD_MODE_VXLAN
)
1704 vxlan_get_rx_port(dev
);
1706 priv
->port_up
= true;
1707 netif_tx_start_all_queues(dev
);
1708 netif_device_attach(dev
);
1713 while (tx_index
--) {
1714 mlx4_en_deactivate_tx_ring(priv
, priv
->tx_ring
[tx_index
]);
1715 mlx4_en_deactivate_cq(priv
, priv
->tx_cq
[tx_index
]);
1717 mlx4_en_destroy_drop_qp(priv
);
1719 mlx4_en_release_rss_steer(priv
);
1721 mlx4_en_put_qp(priv
);
1723 while (rx_index
--) {
1724 mlx4_en_deactivate_cq(priv
, priv
->rx_cq
[rx_index
]);
1725 mlx4_en_free_affinity_hint(priv
, rx_index
);
1727 for (i
= 0; i
< priv
->rx_ring_num
; i
++)
1728 mlx4_en_deactivate_rx_ring(priv
, priv
->rx_ring
[i
]);
1730 return err
; /* need to close devices */
1734 void mlx4_en_stop_port(struct net_device
*dev
, int detach
)
1736 struct mlx4_en_priv
*priv
= netdev_priv(dev
);
1737 struct mlx4_en_dev
*mdev
= priv
->mdev
;
1738 struct mlx4_en_mc_list
*mclist
, *tmp
;
1739 struct ethtool_flow_id
*flow
, *tmp_flow
;
1741 u8 mc_list
[16] = {0};
1743 if (!priv
->port_up
) {
1744 en_dbg(DRV
, priv
, "stop port called while port already down\n");
1749 mlx4_CLOSE_PORT(mdev
->dev
, priv
->port
);
1751 /* Synchronize with tx routine */
1752 netif_tx_lock_bh(dev
);
1754 netif_device_detach(dev
);
1755 netif_tx_stop_all_queues(dev
);
1756 netif_tx_unlock_bh(dev
);
1758 netif_tx_disable(dev
);
1760 /* Set port as not active */
1761 priv
->port_up
= false;
1763 /* Promsicuous mode */
1764 if (mdev
->dev
->caps
.steering_mode
==
1765 MLX4_STEERING_MODE_DEVICE_MANAGED
) {
1766 priv
->flags
&= ~(MLX4_EN_FLAG_PROMISC
|
1767 MLX4_EN_FLAG_MC_PROMISC
);
1768 mlx4_flow_steer_promisc_remove(mdev
->dev
,
1770 MLX4_FS_ALL_DEFAULT
);
1771 mlx4_flow_steer_promisc_remove(mdev
->dev
,
1773 MLX4_FS_MC_DEFAULT
);
1774 } else if (priv
->flags
& MLX4_EN_FLAG_PROMISC
) {
1775 priv
->flags
&= ~MLX4_EN_FLAG_PROMISC
;
1777 /* Disable promiscouos mode */
1778 mlx4_unicast_promisc_remove(mdev
->dev
, priv
->base_qpn
,
1781 /* Disable Multicast promisc */
1782 if (priv
->flags
& MLX4_EN_FLAG_MC_PROMISC
) {
1783 mlx4_multicast_promisc_remove(mdev
->dev
, priv
->base_qpn
,
1785 priv
->flags
&= ~MLX4_EN_FLAG_MC_PROMISC
;
1789 /* Detach All multicasts */
1790 eth_broadcast_addr(&mc_list
[10]);
1791 mc_list
[5] = priv
->port
; /* needed for B0 steering support */
1792 mlx4_multicast_detach(mdev
->dev
, &priv
->rss_map
.indir_qp
, mc_list
,
1793 MLX4_PROT_ETH
, priv
->broadcast_id
);
1794 list_for_each_entry(mclist
, &priv
->curr_list
, list
) {
1795 memcpy(&mc_list
[10], mclist
->addr
, ETH_ALEN
);
1796 mc_list
[5] = priv
->port
;
1797 mlx4_multicast_detach(mdev
->dev
, &priv
->rss_map
.indir_qp
,
1798 mc_list
, MLX4_PROT_ETH
, mclist
->reg_id
);
1799 if (mclist
->tunnel_reg_id
)
1800 mlx4_flow_detach(mdev
->dev
, mclist
->tunnel_reg_id
);
1802 mlx4_en_clear_list(dev
);
1803 list_for_each_entry_safe(mclist
, tmp
, &priv
->curr_list
, list
) {
1804 list_del(&mclist
->list
);
1808 /* Flush multicast filter */
1809 mlx4_SET_MCAST_FLTR(mdev
->dev
, priv
->port
, 0, 1, MLX4_MCAST_CONFIG
);
1811 /* Remove flow steering rules for the port*/
1812 if (mdev
->dev
->caps
.steering_mode
==
1813 MLX4_STEERING_MODE_DEVICE_MANAGED
) {
1815 list_for_each_entry_safe(flow
, tmp_flow
,
1816 &priv
->ethtool_list
, list
) {
1817 mlx4_flow_detach(mdev
->dev
, flow
->id
);
1818 list_del(&flow
->list
);
1822 mlx4_en_destroy_drop_qp(priv
);
1825 for (i
= 0; i
< priv
->tx_ring_num
; i
++) {
1826 mlx4_en_deactivate_tx_ring(priv
, priv
->tx_ring
[i
]);
1827 mlx4_en_deactivate_cq(priv
, priv
->tx_cq
[i
]);
1831 for (i
= 0; i
< priv
->tx_ring_num
; i
++)
1832 mlx4_en_free_tx_buf(dev
, priv
->tx_ring
[i
]);
1835 mlx4_en_release_rss_steer(priv
);
1837 /* Unregister Mac address for the port */
1838 mlx4_en_put_qp(priv
);
1839 if (!(mdev
->dev
->caps
.flags2
& MLX4_DEV_CAP_FLAG2_REASSIGN_MAC_EN
))
1840 mdev
->mac_removed
[priv
->port
] = 1;
1843 for (i
= 0; i
< priv
->rx_ring_num
; i
++) {
1844 struct mlx4_en_cq
*cq
= priv
->rx_cq
[i
];
1847 while (!mlx4_en_cq_lock_napi(cq
)) {
1848 pr_info("CQ %d locked\n", i
);
1853 napi_synchronize(&cq
->napi
);
1854 mlx4_en_deactivate_rx_ring(priv
, priv
->rx_ring
[i
]);
1855 mlx4_en_deactivate_cq(priv
, cq
);
1857 mlx4_en_free_affinity_hint(priv
, i
);
1861 static void mlx4_en_restart(struct work_struct
*work
)
1863 struct mlx4_en_priv
*priv
= container_of(work
, struct mlx4_en_priv
,
1865 struct mlx4_en_dev
*mdev
= priv
->mdev
;
1866 struct net_device
*dev
= priv
->dev
;
1868 en_dbg(DRV
, priv
, "Watchdog task called for port %d\n", priv
->port
);
1870 mutex_lock(&mdev
->state_lock
);
1871 if (priv
->port_up
) {
1872 mlx4_en_stop_port(dev
, 1);
1873 if (mlx4_en_start_port(dev
))
1874 en_err(priv
, "Failed restarting port %d\n", priv
->port
);
1876 mutex_unlock(&mdev
->state_lock
);
1879 static void mlx4_en_clear_stats(struct net_device
*dev
)
1881 struct mlx4_en_priv
*priv
= netdev_priv(dev
);
1882 struct mlx4_en_dev
*mdev
= priv
->mdev
;
1885 if (mlx4_en_DUMP_ETH_STATS(mdev
, priv
->port
, 1))
1886 en_dbg(HW
, priv
, "Failed dumping statistics\n");
1888 memset(&priv
->stats
, 0, sizeof(priv
->stats
));
1889 memset(&priv
->pstats
, 0, sizeof(priv
->pstats
));
1890 memset(&priv
->pkstats
, 0, sizeof(priv
->pkstats
));
1891 memset(&priv
->port_stats
, 0, sizeof(priv
->port_stats
));
1892 memset(&priv
->rx_flowstats
, 0, sizeof(priv
->rx_flowstats
));
1893 memset(&priv
->tx_flowstats
, 0, sizeof(priv
->tx_flowstats
));
1894 memset(&priv
->rx_priority_flowstats
, 0,
1895 sizeof(priv
->rx_priority_flowstats
));
1896 memset(&priv
->tx_priority_flowstats
, 0,
1897 sizeof(priv
->tx_priority_flowstats
));
1899 for (i
= 0; i
< priv
->tx_ring_num
; i
++) {
1900 priv
->tx_ring
[i
]->bytes
= 0;
1901 priv
->tx_ring
[i
]->packets
= 0;
1902 priv
->tx_ring
[i
]->tx_csum
= 0;
1904 for (i
= 0; i
< priv
->rx_ring_num
; i
++) {
1905 priv
->rx_ring
[i
]->bytes
= 0;
1906 priv
->rx_ring
[i
]->packets
= 0;
1907 priv
->rx_ring
[i
]->csum_ok
= 0;
1908 priv
->rx_ring
[i
]->csum_none
= 0;
1909 priv
->rx_ring
[i
]->csum_complete
= 0;
1913 static int mlx4_en_open(struct net_device
*dev
)
1915 struct mlx4_en_priv
*priv
= netdev_priv(dev
);
1916 struct mlx4_en_dev
*mdev
= priv
->mdev
;
1919 mutex_lock(&mdev
->state_lock
);
1921 if (!mdev
->device_up
) {
1922 en_err(priv
, "Cannot open - device down/disabled\n");
1927 /* Reset HW statistics and SW counters */
1928 mlx4_en_clear_stats(dev
);
1930 err
= mlx4_en_start_port(dev
);
1932 en_err(priv
, "Failed starting port:%d\n", priv
->port
);
1935 mutex_unlock(&mdev
->state_lock
);
1940 static int mlx4_en_close(struct net_device
*dev
)
1942 struct mlx4_en_priv
*priv
= netdev_priv(dev
);
1943 struct mlx4_en_dev
*mdev
= priv
->mdev
;
1945 en_dbg(IFDOWN
, priv
, "Close port called\n");
1947 mutex_lock(&mdev
->state_lock
);
1949 mlx4_en_stop_port(dev
, 0);
1950 netif_carrier_off(dev
);
1952 mutex_unlock(&mdev
->state_lock
);
1956 void mlx4_en_free_resources(struct mlx4_en_priv
*priv
)
1960 #ifdef CONFIG_RFS_ACCEL
1961 free_irq_cpu_rmap(priv
->dev
->rx_cpu_rmap
);
1962 priv
->dev
->rx_cpu_rmap
= NULL
;
1965 for (i
= 0; i
< priv
->tx_ring_num
; i
++) {
1966 if (priv
->tx_ring
&& priv
->tx_ring
[i
])
1967 mlx4_en_destroy_tx_ring(priv
, &priv
->tx_ring
[i
]);
1968 if (priv
->tx_cq
&& priv
->tx_cq
[i
])
1969 mlx4_en_destroy_cq(priv
, &priv
->tx_cq
[i
]);
1972 for (i
= 0; i
< priv
->rx_ring_num
; i
++) {
1973 if (priv
->rx_ring
[i
])
1974 mlx4_en_destroy_rx_ring(priv
, &priv
->rx_ring
[i
],
1975 priv
->prof
->rx_ring_size
, priv
->stride
);
1977 mlx4_en_destroy_cq(priv
, &priv
->rx_cq
[i
]);
1980 if (priv
->base_tx_qpn
) {
1981 mlx4_qp_release_range(priv
->mdev
->dev
, priv
->base_tx_qpn
, priv
->tx_ring_num
);
1982 priv
->base_tx_qpn
= 0;
1986 int mlx4_en_alloc_resources(struct mlx4_en_priv
*priv
)
1988 struct mlx4_en_port_profile
*prof
= priv
->prof
;
1992 /* Create tx Rings */
1993 for (i
= 0; i
< priv
->tx_ring_num
; i
++) {
1994 node
= cpu_to_node(i
% num_online_cpus());
1995 if (mlx4_en_create_cq(priv
, &priv
->tx_cq
[i
],
1996 prof
->tx_ring_size
, i
, TX
, node
))
1999 if (mlx4_en_create_tx_ring(priv
, &priv
->tx_ring
[i
],
2000 prof
->tx_ring_size
, TXBB_SIZE
,
2005 /* Create rx Rings */
2006 for (i
= 0; i
< priv
->rx_ring_num
; i
++) {
2007 node
= cpu_to_node(i
% num_online_cpus());
2008 if (mlx4_en_create_cq(priv
, &priv
->rx_cq
[i
],
2009 prof
->rx_ring_size
, i
, RX
, node
))
2012 if (mlx4_en_create_rx_ring(priv
, &priv
->rx_ring
[i
],
2013 prof
->rx_ring_size
, priv
->stride
,
2018 #ifdef CONFIG_RFS_ACCEL
2019 if (priv
->mdev
->dev
->caps
.comp_pool
) {
2020 priv
->dev
->rx_cpu_rmap
= alloc_irq_cpu_rmap(priv
->mdev
->dev
->caps
.comp_pool
);
2021 if (!priv
->dev
->rx_cpu_rmap
)
2029 en_err(priv
, "Failed to allocate NIC resources\n");
2030 for (i
= 0; i
< priv
->rx_ring_num
; i
++) {
2031 if (priv
->rx_ring
[i
])
2032 mlx4_en_destroy_rx_ring(priv
, &priv
->rx_ring
[i
],
2036 mlx4_en_destroy_cq(priv
, &priv
->rx_cq
[i
]);
2038 for (i
= 0; i
< priv
->tx_ring_num
; i
++) {
2039 if (priv
->tx_ring
[i
])
2040 mlx4_en_destroy_tx_ring(priv
, &priv
->tx_ring
[i
]);
2042 mlx4_en_destroy_cq(priv
, &priv
->tx_cq
[i
]);
2048 void mlx4_en_destroy_netdev(struct net_device
*dev
)
2050 struct mlx4_en_priv
*priv
= netdev_priv(dev
);
2051 struct mlx4_en_dev
*mdev
= priv
->mdev
;
2053 en_dbg(DRV
, priv
, "Destroying netdev on port:%d\n", priv
->port
);
2055 /* Unregister device - this will close the port if it was up */
2056 if (priv
->registered
)
2057 unregister_netdev(dev
);
2059 if (priv
->allocated
)
2060 mlx4_free_hwq_res(mdev
->dev
, &priv
->res
, MLX4_EN_PAGE_SIZE
);
2062 cancel_delayed_work(&priv
->stats_task
);
2063 cancel_delayed_work(&priv
->service_task
);
2064 /* flush any pending task for this netdev */
2065 flush_workqueue(mdev
->workqueue
);
2067 /* Detach the netdev so tasks would not attempt to access it */
2068 mutex_lock(&mdev
->state_lock
);
2069 mdev
->pndev
[priv
->port
] = NULL
;
2070 mdev
->upper
[priv
->port
] = NULL
;
2071 mutex_unlock(&mdev
->state_lock
);
2073 mlx4_en_free_resources(priv
);
2075 kfree(priv
->tx_ring
);
2081 static int mlx4_en_change_mtu(struct net_device
*dev
, int new_mtu
)
2083 struct mlx4_en_priv
*priv
= netdev_priv(dev
);
2084 struct mlx4_en_dev
*mdev
= priv
->mdev
;
2087 en_dbg(DRV
, priv
, "Change MTU called - current:%d new:%d\n",
2090 if ((new_mtu
< MLX4_EN_MIN_MTU
) || (new_mtu
> priv
->max_mtu
)) {
2091 en_err(priv
, "Bad MTU size:%d.\n", new_mtu
);
2096 if (netif_running(dev
)) {
2097 mutex_lock(&mdev
->state_lock
);
2098 if (!mdev
->device_up
) {
2099 /* NIC is probably restarting - let watchdog task reset
2101 en_dbg(DRV
, priv
, "Change MTU called with card down!?\n");
2103 mlx4_en_stop_port(dev
, 1);
2104 err
= mlx4_en_start_port(dev
);
2106 en_err(priv
, "Failed restarting port:%d\n",
2108 queue_work(mdev
->workqueue
, &priv
->watchdog_task
);
2111 mutex_unlock(&mdev
->state_lock
);
2116 static int mlx4_en_hwtstamp_set(struct net_device
*dev
, struct ifreq
*ifr
)
2118 struct mlx4_en_priv
*priv
= netdev_priv(dev
);
2119 struct mlx4_en_dev
*mdev
= priv
->mdev
;
2120 struct hwtstamp_config config
;
2122 if (copy_from_user(&config
, ifr
->ifr_data
, sizeof(config
)))
2125 /* reserved for future extensions */
2129 /* device doesn't support time stamping */
2130 if (!(mdev
->dev
->caps
.flags2
& MLX4_DEV_CAP_FLAG2_TS
))
2133 /* TX HW timestamp */
2134 switch (config
.tx_type
) {
2135 case HWTSTAMP_TX_OFF
:
2136 case HWTSTAMP_TX_ON
:
2142 /* RX HW timestamp */
2143 switch (config
.rx_filter
) {
2144 case HWTSTAMP_FILTER_NONE
:
2146 case HWTSTAMP_FILTER_ALL
:
2147 case HWTSTAMP_FILTER_SOME
:
2148 case HWTSTAMP_FILTER_PTP_V1_L4_EVENT
:
2149 case HWTSTAMP_FILTER_PTP_V1_L4_SYNC
:
2150 case HWTSTAMP_FILTER_PTP_V1_L4_DELAY_REQ
:
2151 case HWTSTAMP_FILTER_PTP_V2_L4_EVENT
:
2152 case HWTSTAMP_FILTER_PTP_V2_L4_SYNC
:
2153 case HWTSTAMP_FILTER_PTP_V2_L4_DELAY_REQ
:
2154 case HWTSTAMP_FILTER_PTP_V2_L2_EVENT
:
2155 case HWTSTAMP_FILTER_PTP_V2_L2_SYNC
:
2156 case HWTSTAMP_FILTER_PTP_V2_L2_DELAY_REQ
:
2157 case HWTSTAMP_FILTER_PTP_V2_EVENT
:
2158 case HWTSTAMP_FILTER_PTP_V2_SYNC
:
2159 case HWTSTAMP_FILTER_PTP_V2_DELAY_REQ
:
2160 config
.rx_filter
= HWTSTAMP_FILTER_ALL
;
2166 if (mlx4_en_reset_config(dev
, config
, dev
->features
)) {
2167 config
.tx_type
= HWTSTAMP_TX_OFF
;
2168 config
.rx_filter
= HWTSTAMP_FILTER_NONE
;
2171 return copy_to_user(ifr
->ifr_data
, &config
,
2172 sizeof(config
)) ? -EFAULT
: 0;
2175 static int mlx4_en_hwtstamp_get(struct net_device
*dev
, struct ifreq
*ifr
)
2177 struct mlx4_en_priv
*priv
= netdev_priv(dev
);
2179 return copy_to_user(ifr
->ifr_data
, &priv
->hwtstamp_config
,
2180 sizeof(priv
->hwtstamp_config
)) ? -EFAULT
: 0;
2183 static int mlx4_en_ioctl(struct net_device
*dev
, struct ifreq
*ifr
, int cmd
)
2187 return mlx4_en_hwtstamp_set(dev
, ifr
);
2189 return mlx4_en_hwtstamp_get(dev
, ifr
);
2195 static int mlx4_en_set_features(struct net_device
*netdev
,
2196 netdev_features_t features
)
2198 struct mlx4_en_priv
*priv
= netdev_priv(netdev
);
2202 if (DEV_FEATURE_CHANGED(netdev
, features
, NETIF_F_RXFCS
)) {
2203 en_info(priv
, "Turn %s RX-FCS\n",
2204 (features
& NETIF_F_RXFCS
) ? "ON" : "OFF");
2208 if (DEV_FEATURE_CHANGED(netdev
, features
, NETIF_F_RXALL
)) {
2209 u8 ignore_fcs_value
= (features
& NETIF_F_RXALL
) ? 1 : 0;
2211 en_info(priv
, "Turn %s RX-ALL\n",
2212 ignore_fcs_value
? "ON" : "OFF");
2213 ret
= mlx4_SET_PORT_fcs_check(priv
->mdev
->dev
,
2214 priv
->port
, ignore_fcs_value
);
2219 if (DEV_FEATURE_CHANGED(netdev
, features
, NETIF_F_HW_VLAN_CTAG_RX
)) {
2220 en_info(priv
, "Turn %s RX vlan strip offload\n",
2221 (features
& NETIF_F_HW_VLAN_CTAG_RX
) ? "ON" : "OFF");
2225 if (DEV_FEATURE_CHANGED(netdev
, features
, NETIF_F_HW_VLAN_CTAG_TX
))
2226 en_info(priv
, "Turn %s TX vlan strip offload\n",
2227 (features
& NETIF_F_HW_VLAN_CTAG_TX
) ? "ON" : "OFF");
2229 if (DEV_FEATURE_CHANGED(netdev
, features
, NETIF_F_LOOPBACK
)) {
2230 en_info(priv
, "Turn %s loopback\n",
2231 (features
& NETIF_F_LOOPBACK
) ? "ON" : "OFF");
2232 mlx4_en_update_loopback_state(netdev
, features
);
2236 ret
= mlx4_en_reset_config(netdev
, priv
->hwtstamp_config
,
2245 static int mlx4_en_set_vf_mac(struct net_device
*dev
, int queue
, u8
*mac
)
2247 struct mlx4_en_priv
*en_priv
= netdev_priv(dev
);
2248 struct mlx4_en_dev
*mdev
= en_priv
->mdev
;
2249 u64 mac_u64
= mlx4_mac_to_u64(mac
);
2251 if (!is_valid_ether_addr(mac
))
2254 return mlx4_set_vf_mac(mdev
->dev
, en_priv
->port
, queue
, mac_u64
);
2257 static int mlx4_en_set_vf_vlan(struct net_device
*dev
, int vf
, u16 vlan
, u8 qos
)
2259 struct mlx4_en_priv
*en_priv
= netdev_priv(dev
);
2260 struct mlx4_en_dev
*mdev
= en_priv
->mdev
;
2262 return mlx4_set_vf_vlan(mdev
->dev
, en_priv
->port
, vf
, vlan
, qos
);
2265 static int mlx4_en_set_vf_rate(struct net_device
*dev
, int vf
, int min_tx_rate
,
2268 struct mlx4_en_priv
*en_priv
= netdev_priv(dev
);
2269 struct mlx4_en_dev
*mdev
= en_priv
->mdev
;
2271 return mlx4_set_vf_rate(mdev
->dev
, en_priv
->port
, vf
, min_tx_rate
,
2275 static int mlx4_en_set_vf_spoofchk(struct net_device
*dev
, int vf
, bool setting
)
2277 struct mlx4_en_priv
*en_priv
= netdev_priv(dev
);
2278 struct mlx4_en_dev
*mdev
= en_priv
->mdev
;
2280 return mlx4_set_vf_spoofchk(mdev
->dev
, en_priv
->port
, vf
, setting
);
2283 static int mlx4_en_get_vf_config(struct net_device
*dev
, int vf
, struct ifla_vf_info
*ivf
)
2285 struct mlx4_en_priv
*en_priv
= netdev_priv(dev
);
2286 struct mlx4_en_dev
*mdev
= en_priv
->mdev
;
2288 return mlx4_get_vf_config(mdev
->dev
, en_priv
->port
, vf
, ivf
);
2291 static int mlx4_en_set_vf_link_state(struct net_device
*dev
, int vf
, int link_state
)
2293 struct mlx4_en_priv
*en_priv
= netdev_priv(dev
);
2294 struct mlx4_en_dev
*mdev
= en_priv
->mdev
;
2296 return mlx4_set_vf_link_state(mdev
->dev
, en_priv
->port
, vf
, link_state
);
2299 #define PORT_ID_BYTE_LEN 8
2300 static int mlx4_en_get_phys_port_id(struct net_device
*dev
,
2301 struct netdev_phys_item_id
*ppid
)
2303 struct mlx4_en_priv
*priv
= netdev_priv(dev
);
2304 struct mlx4_dev
*mdev
= priv
->mdev
->dev
;
2306 u64 phys_port_id
= mdev
->caps
.phys_port_id
[priv
->port
];
2311 ppid
->id_len
= sizeof(phys_port_id
);
2312 for (i
= PORT_ID_BYTE_LEN
- 1; i
>= 0; --i
) {
2313 ppid
->id
[i
] = phys_port_id
& 0xff;
2319 #ifdef CONFIG_MLX4_EN_VXLAN
2320 static void mlx4_en_add_vxlan_offloads(struct work_struct
*work
)
2323 struct mlx4_en_priv
*priv
= container_of(work
, struct mlx4_en_priv
,
2326 ret
= mlx4_config_vxlan_port(priv
->mdev
->dev
, priv
->vxlan_port
);
2330 ret
= mlx4_SET_PORT_VXLAN(priv
->mdev
->dev
, priv
->port
,
2331 VXLAN_STEER_BY_OUTER_MAC
, 1);
2334 en_err(priv
, "failed setting L2 tunnel configuration ret %d\n", ret
);
2339 priv
->dev
->hw_enc_features
|= NETIF_F_IP_CSUM
| NETIF_F_RXCSUM
|
2340 NETIF_F_TSO
| NETIF_F_GSO_UDP_TUNNEL
;
2341 priv
->dev
->hw_features
|= NETIF_F_GSO_UDP_TUNNEL
;
2342 priv
->dev
->features
|= NETIF_F_GSO_UDP_TUNNEL
;
2345 static void mlx4_en_del_vxlan_offloads(struct work_struct
*work
)
2348 struct mlx4_en_priv
*priv
= container_of(work
, struct mlx4_en_priv
,
2350 /* unset offloads */
2351 priv
->dev
->hw_enc_features
&= ~(NETIF_F_IP_CSUM
| NETIF_F_RXCSUM
|
2352 NETIF_F_TSO
| NETIF_F_GSO_UDP_TUNNEL
);
2353 priv
->dev
->hw_features
&= ~NETIF_F_GSO_UDP_TUNNEL
;
2354 priv
->dev
->features
&= ~NETIF_F_GSO_UDP_TUNNEL
;
2356 ret
= mlx4_SET_PORT_VXLAN(priv
->mdev
->dev
, priv
->port
,
2357 VXLAN_STEER_BY_OUTER_MAC
, 0);
2359 en_err(priv
, "failed setting L2 tunnel configuration ret %d\n", ret
);
2361 priv
->vxlan_port
= 0;
2364 static void mlx4_en_add_vxlan_port(struct net_device
*dev
,
2365 sa_family_t sa_family
, __be16 port
)
2367 struct mlx4_en_priv
*priv
= netdev_priv(dev
);
2368 __be16 current_port
;
2370 if (priv
->mdev
->dev
->caps
.tunnel_offload_mode
!= MLX4_TUNNEL_OFFLOAD_MODE_VXLAN
)
2373 if (sa_family
== AF_INET6
)
2376 current_port
= priv
->vxlan_port
;
2377 if (current_port
&& current_port
!= port
) {
2378 en_warn(priv
, "vxlan port %d configured, can't add port %d\n",
2379 ntohs(current_port
), ntohs(port
));
2383 priv
->vxlan_port
= port
;
2384 queue_work(priv
->mdev
->workqueue
, &priv
->vxlan_add_task
);
2387 static void mlx4_en_del_vxlan_port(struct net_device
*dev
,
2388 sa_family_t sa_family
, __be16 port
)
2390 struct mlx4_en_priv
*priv
= netdev_priv(dev
);
2391 __be16 current_port
;
2393 if (priv
->mdev
->dev
->caps
.tunnel_offload_mode
!= MLX4_TUNNEL_OFFLOAD_MODE_VXLAN
)
2396 if (sa_family
== AF_INET6
)
2399 current_port
= priv
->vxlan_port
;
2400 if (current_port
!= port
) {
2401 en_dbg(DRV
, priv
, "vxlan port %d isn't configured, ignoring\n", ntohs(port
));
2405 queue_work(priv
->mdev
->workqueue
, &priv
->vxlan_del_task
);
2408 static netdev_features_t
mlx4_en_features_check(struct sk_buff
*skb
,
2409 struct net_device
*dev
,
2410 netdev_features_t features
)
2412 features
= vlan_features_check(skb
, features
);
2413 return vxlan_features_check(skb
, features
);
2417 static int mlx4_en_set_tx_maxrate(struct net_device
*dev
, int queue_index
, u32 maxrate
)
2419 struct mlx4_en_priv
*priv
= netdev_priv(dev
);
2420 struct mlx4_en_tx_ring
*tx_ring
= priv
->tx_ring
[queue_index
];
2421 struct mlx4_update_qp_params params
;
2424 if (!(priv
->mdev
->dev
->caps
.flags2
& MLX4_DEV_CAP_FLAG2_QP_RATE_LIMIT
))
2427 /* rate provided to us in Mbs, check if it fits into 12 bits, if not use Gbs */
2428 if (maxrate
>> 12) {
2429 params
.rate_unit
= MLX4_QP_RATE_LIMIT_GBS
;
2430 params
.rate_val
= maxrate
/ 1000;
2431 } else if (maxrate
) {
2432 params
.rate_unit
= MLX4_QP_RATE_LIMIT_MBS
;
2433 params
.rate_val
= maxrate
;
2434 } else { /* zero serves to revoke the QP rate-limitation */
2435 params
.rate_unit
= 0;
2436 params
.rate_val
= 0;
2439 err
= mlx4_update_qp(priv
->mdev
->dev
, tx_ring
->qpn
, MLX4_UPDATE_QP_RATE_LIMIT
,
2444 static const struct net_device_ops mlx4_netdev_ops
= {
2445 .ndo_open
= mlx4_en_open
,
2446 .ndo_stop
= mlx4_en_close
,
2447 .ndo_start_xmit
= mlx4_en_xmit
,
2448 .ndo_select_queue
= mlx4_en_select_queue
,
2449 .ndo_get_stats
= mlx4_en_get_stats
,
2450 .ndo_set_rx_mode
= mlx4_en_set_rx_mode
,
2451 .ndo_set_mac_address
= mlx4_en_set_mac
,
2452 .ndo_validate_addr
= eth_validate_addr
,
2453 .ndo_change_mtu
= mlx4_en_change_mtu
,
2454 .ndo_do_ioctl
= mlx4_en_ioctl
,
2455 .ndo_tx_timeout
= mlx4_en_tx_timeout
,
2456 .ndo_vlan_rx_add_vid
= mlx4_en_vlan_rx_add_vid
,
2457 .ndo_vlan_rx_kill_vid
= mlx4_en_vlan_rx_kill_vid
,
2458 #ifdef CONFIG_NET_POLL_CONTROLLER
2459 .ndo_poll_controller
= mlx4_en_netpoll
,
2461 .ndo_set_features
= mlx4_en_set_features
,
2462 .ndo_setup_tc
= mlx4_en_setup_tc
,
2463 #ifdef CONFIG_RFS_ACCEL
2464 .ndo_rx_flow_steer
= mlx4_en_filter_rfs
,
2466 #ifdef CONFIG_NET_RX_BUSY_POLL
2467 .ndo_busy_poll
= mlx4_en_low_latency_recv
,
2469 .ndo_get_phys_port_id
= mlx4_en_get_phys_port_id
,
2470 #ifdef CONFIG_MLX4_EN_VXLAN
2471 .ndo_add_vxlan_port
= mlx4_en_add_vxlan_port
,
2472 .ndo_del_vxlan_port
= mlx4_en_del_vxlan_port
,
2473 .ndo_features_check
= mlx4_en_features_check
,
2475 .ndo_set_tx_maxrate
= mlx4_en_set_tx_maxrate
,
2478 static const struct net_device_ops mlx4_netdev_ops_master
= {
2479 .ndo_open
= mlx4_en_open
,
2480 .ndo_stop
= mlx4_en_close
,
2481 .ndo_start_xmit
= mlx4_en_xmit
,
2482 .ndo_select_queue
= mlx4_en_select_queue
,
2483 .ndo_get_stats
= mlx4_en_get_stats
,
2484 .ndo_set_rx_mode
= mlx4_en_set_rx_mode
,
2485 .ndo_set_mac_address
= mlx4_en_set_mac
,
2486 .ndo_validate_addr
= eth_validate_addr
,
2487 .ndo_change_mtu
= mlx4_en_change_mtu
,
2488 .ndo_tx_timeout
= mlx4_en_tx_timeout
,
2489 .ndo_vlan_rx_add_vid
= mlx4_en_vlan_rx_add_vid
,
2490 .ndo_vlan_rx_kill_vid
= mlx4_en_vlan_rx_kill_vid
,
2491 .ndo_set_vf_mac
= mlx4_en_set_vf_mac
,
2492 .ndo_set_vf_vlan
= mlx4_en_set_vf_vlan
,
2493 .ndo_set_vf_rate
= mlx4_en_set_vf_rate
,
2494 .ndo_set_vf_spoofchk
= mlx4_en_set_vf_spoofchk
,
2495 .ndo_set_vf_link_state
= mlx4_en_set_vf_link_state
,
2496 .ndo_get_vf_config
= mlx4_en_get_vf_config
,
2497 #ifdef CONFIG_NET_POLL_CONTROLLER
2498 .ndo_poll_controller
= mlx4_en_netpoll
,
2500 .ndo_set_features
= mlx4_en_set_features
,
2501 .ndo_setup_tc
= mlx4_en_setup_tc
,
2502 #ifdef CONFIG_RFS_ACCEL
2503 .ndo_rx_flow_steer
= mlx4_en_filter_rfs
,
2505 .ndo_get_phys_port_id
= mlx4_en_get_phys_port_id
,
2506 #ifdef CONFIG_MLX4_EN_VXLAN
2507 .ndo_add_vxlan_port
= mlx4_en_add_vxlan_port
,
2508 .ndo_del_vxlan_port
= mlx4_en_del_vxlan_port
,
2509 .ndo_features_check
= mlx4_en_features_check
,
2511 .ndo_set_tx_maxrate
= mlx4_en_set_tx_maxrate
,
2514 struct mlx4_en_bond
{
2515 struct work_struct work
;
2516 struct mlx4_en_priv
*priv
;
2518 struct mlx4_port_map port_map
;
2521 static void mlx4_en_bond_work(struct work_struct
*work
)
2523 struct mlx4_en_bond
*bond
= container_of(work
,
2524 struct mlx4_en_bond
,
2527 struct mlx4_dev
*dev
= bond
->priv
->mdev
->dev
;
2529 if (bond
->is_bonded
) {
2530 if (!mlx4_is_bonded(dev
)) {
2531 err
= mlx4_bond(dev
);
2533 en_err(bond
->priv
, "Fail to bond device\n");
2536 err
= mlx4_port_map_set(dev
, &bond
->port_map
);
2538 en_err(bond
->priv
, "Fail to set port map [%d][%d]: %d\n",
2539 bond
->port_map
.port1
,
2540 bond
->port_map
.port2
,
2543 } else if (mlx4_is_bonded(dev
)) {
2544 err
= mlx4_unbond(dev
);
2546 en_err(bond
->priv
, "Fail to unbond device\n");
2548 dev_put(bond
->priv
->dev
);
2552 static int mlx4_en_queue_bond_work(struct mlx4_en_priv
*priv
, int is_bonded
,
2553 u8 v2p_p1
, u8 v2p_p2
)
2555 struct mlx4_en_bond
*bond
= NULL
;
2557 bond
= kzalloc(sizeof(*bond
), GFP_ATOMIC
);
2561 INIT_WORK(&bond
->work
, mlx4_en_bond_work
);
2563 bond
->is_bonded
= is_bonded
;
2564 bond
->port_map
.port1
= v2p_p1
;
2565 bond
->port_map
.port2
= v2p_p2
;
2566 dev_hold(priv
->dev
);
2567 queue_work(priv
->mdev
->workqueue
, &bond
->work
);
2571 int mlx4_en_netdev_event(struct notifier_block
*this,
2572 unsigned long event
, void *ptr
)
2574 struct net_device
*ndev
= netdev_notifier_info_to_dev(ptr
);
2576 struct mlx4_en_dev
*mdev
;
2577 struct mlx4_dev
*dev
;
2578 int i
, num_eth_ports
= 0;
2579 bool do_bond
= true;
2580 struct mlx4_en_priv
*priv
;
2584 if (!net_eq(dev_net(ndev
), &init_net
))
2587 mdev
= container_of(this, struct mlx4_en_dev
, nb
);
2590 /* Go into this mode only when two network devices set on two ports
2591 * of the same mlx4 device are slaves of the same bonding master
2593 mlx4_foreach_port(i
, dev
, MLX4_PORT_TYPE_ETH
) {
2595 if (!port
&& (mdev
->pndev
[i
] == ndev
))
2597 mdev
->upper
[i
] = mdev
->pndev
[i
] ?
2598 netdev_master_upper_dev_get(mdev
->pndev
[i
]) : NULL
;
2599 /* condition not met: network device is a slave */
2600 if (!mdev
->upper
[i
])
2602 if (num_eth_ports
< 2)
2604 /* condition not met: same master */
2605 if (mdev
->upper
[i
] != mdev
->upper
[i
-1])
2608 /* condition not met: 2 salves */
2609 do_bond
= (num_eth_ports
== 2) ? do_bond
: false;
2611 /* handle only events that come with enough info */
2612 if ((do_bond
&& (event
!= NETDEV_BONDING_INFO
)) || !port
)
2615 priv
= netdev_priv(ndev
);
2617 struct netdev_notifier_bonding_info
*notifier_info
= ptr
;
2618 struct netdev_bonding_info
*bonding_info
=
2619 ¬ifier_info
->bonding_info
;
2621 /* required mode 1, 2 or 4 */
2622 if ((bonding_info
->master
.bond_mode
!= BOND_MODE_ACTIVEBACKUP
) &&
2623 (bonding_info
->master
.bond_mode
!= BOND_MODE_XOR
) &&
2624 (bonding_info
->master
.bond_mode
!= BOND_MODE_8023AD
))
2627 /* require exactly 2 slaves */
2628 if (bonding_info
->master
.num_slaves
!= 2)
2633 if (bonding_info
->master
.bond_mode
==
2634 BOND_MODE_ACTIVEBACKUP
) {
2635 /* in active-backup mode virtual ports are
2636 * mapped to the physical port of the active
2638 if (bonding_info
->slave
.state
==
2639 BOND_STATE_BACKUP
) {
2647 } else { /* BOND_STATE_ACTIVE */
2656 } else { /* Active-Active */
2657 /* in active-active mode a virtual port is
2658 * mapped to the native physical port if and only
2659 * if the physical port is up */
2660 __s8 link
= bonding_info
->slave
.link
;
2666 if ((link
== BOND_LINK_UP
) ||
2667 (link
== BOND_LINK_FAIL
)) {
2672 } else { /* BOND_LINK_DOWN || BOND_LINK_BACK */
2682 mlx4_en_queue_bond_work(priv
, do_bond
,
2683 v2p_port1
, v2p_port2
);
2688 void mlx4_en_update_pfc_stats_bitmap(struct mlx4_dev
*dev
,
2689 struct mlx4_en_stats_bitmap
*stats_bitmap
,
2690 u8 rx_ppp
, u8 rx_pause
,
2691 u8 tx_ppp
, u8 tx_pause
)
2693 int last_i
= NUM_MAIN_STATS
+ NUM_PORT_STATS
;
2695 if (!mlx4_is_slave(dev
) &&
2696 (dev
->caps
.flags2
& MLX4_DEV_CAP_FLAG2_FLOWSTATS_EN
)) {
2697 mutex_lock(&stats_bitmap
->mutex
);
2698 bitmap_clear(stats_bitmap
->bitmap
, last_i
, NUM_FLOW_STATS
);
2701 bitmap_set(stats_bitmap
->bitmap
, last_i
,
2702 NUM_FLOW_PRIORITY_STATS_RX
);
2703 last_i
+= NUM_FLOW_PRIORITY_STATS_RX
;
2705 if (rx_pause
&& !(rx_ppp
))
2706 bitmap_set(stats_bitmap
->bitmap
, last_i
,
2708 last_i
+= NUM_FLOW_STATS_RX
;
2711 bitmap_set(stats_bitmap
->bitmap
, last_i
,
2712 NUM_FLOW_PRIORITY_STATS_TX
);
2713 last_i
+= NUM_FLOW_PRIORITY_STATS_TX
;
2715 if (tx_pause
&& !(tx_ppp
))
2716 bitmap_set(stats_bitmap
->bitmap
, last_i
,
2718 last_i
+= NUM_FLOW_STATS_TX
;
2720 mutex_unlock(&stats_bitmap
->mutex
);
2724 void mlx4_en_set_stats_bitmap(struct mlx4_dev
*dev
,
2725 struct mlx4_en_stats_bitmap
*stats_bitmap
,
2726 u8 rx_ppp
, u8 rx_pause
,
2727 u8 tx_ppp
, u8 tx_pause
)
2731 mutex_init(&stats_bitmap
->mutex
);
2732 bitmap_zero(stats_bitmap
->bitmap
, NUM_ALL_STATS
);
2734 if (mlx4_is_slave(dev
)) {
2735 bitmap_set(stats_bitmap
->bitmap
, last_i
+
2736 MLX4_FIND_NETDEV_STAT(rx_packets
), 1);
2737 bitmap_set(stats_bitmap
->bitmap
, last_i
+
2738 MLX4_FIND_NETDEV_STAT(tx_packets
), 1);
2739 bitmap_set(stats_bitmap
->bitmap
, last_i
+
2740 MLX4_FIND_NETDEV_STAT(rx_bytes
), 1);
2741 bitmap_set(stats_bitmap
->bitmap
, last_i
+
2742 MLX4_FIND_NETDEV_STAT(tx_bytes
), 1);
2743 bitmap_set(stats_bitmap
->bitmap
, last_i
+
2744 MLX4_FIND_NETDEV_STAT(rx_dropped
), 1);
2745 bitmap_set(stats_bitmap
->bitmap
, last_i
+
2746 MLX4_FIND_NETDEV_STAT(tx_dropped
), 1);
2748 bitmap_set(stats_bitmap
->bitmap
, last_i
, NUM_MAIN_STATS
);
2750 last_i
+= NUM_MAIN_STATS
;
2752 bitmap_set(stats_bitmap
->bitmap
, last_i
, NUM_PORT_STATS
);
2753 last_i
+= NUM_PORT_STATS
;
2755 mlx4_en_update_pfc_stats_bitmap(dev
, stats_bitmap
,
2758 last_i
+= NUM_FLOW_STATS
;
2760 if (!mlx4_is_slave(dev
))
2761 bitmap_set(stats_bitmap
->bitmap
, last_i
, NUM_PKT_STATS
);
2764 int mlx4_en_init_netdev(struct mlx4_en_dev
*mdev
, int port
,
2765 struct mlx4_en_port_profile
*prof
)
2767 struct net_device
*dev
;
2768 struct mlx4_en_priv
*priv
;
2773 dev
= alloc_etherdev_mqs(sizeof(struct mlx4_en_priv
),
2774 MAX_TX_RINGS
, MAX_RX_RINGS
);
2778 netif_set_real_num_tx_queues(dev
, prof
->tx_ring_num
);
2779 netif_set_real_num_rx_queues(dev
, prof
->rx_ring_num
);
2781 SET_NETDEV_DEV(dev
, &mdev
->dev
->persist
->pdev
->dev
);
2782 dev
->dev_port
= port
- 1;
2785 * Initialize driver private data
2788 priv
= netdev_priv(dev
);
2789 memset(priv
, 0, sizeof(struct mlx4_en_priv
));
2790 spin_lock_init(&priv
->stats_lock
);
2791 INIT_WORK(&priv
->rx_mode_task
, mlx4_en_do_set_rx_mode
);
2792 INIT_WORK(&priv
->watchdog_task
, mlx4_en_restart
);
2793 INIT_WORK(&priv
->linkstate_task
, mlx4_en_linkstate
);
2794 INIT_DELAYED_WORK(&priv
->stats_task
, mlx4_en_do_get_stats
);
2795 INIT_DELAYED_WORK(&priv
->service_task
, mlx4_en_service_task
);
2796 #ifdef CONFIG_MLX4_EN_VXLAN
2797 INIT_WORK(&priv
->vxlan_add_task
, mlx4_en_add_vxlan_offloads
);
2798 INIT_WORK(&priv
->vxlan_del_task
, mlx4_en_del_vxlan_offloads
);
2800 #ifdef CONFIG_RFS_ACCEL
2801 INIT_LIST_HEAD(&priv
->filters
);
2802 spin_lock_init(&priv
->filters_lock
);
2807 priv
->ddev
= &mdev
->pdev
->dev
;
2810 priv
->port_up
= false;
2811 priv
->flags
= prof
->flags
;
2812 priv
->pflags
= MLX4_EN_PRIV_FLAGS_BLUEFLAME
;
2813 priv
->ctrl_flags
= cpu_to_be32(MLX4_WQE_CTRL_CQ_UPDATE
|
2814 MLX4_WQE_CTRL_SOLICITED
);
2815 priv
->num_tx_rings_p_up
= mdev
->profile
.num_tx_rings_p_up
;
2816 priv
->tx_ring_num
= prof
->tx_ring_num
;
2817 priv
->tx_work_limit
= MLX4_EN_DEFAULT_TX_WORK
;
2818 netdev_rss_key_fill(priv
->rss_key
, sizeof(priv
->rss_key
));
2820 priv
->tx_ring
= kzalloc(sizeof(struct mlx4_en_tx_ring
*) * MAX_TX_RINGS
,
2822 if (!priv
->tx_ring
) {
2826 priv
->tx_cq
= kzalloc(sizeof(struct mlx4_en_cq
*) * MAX_TX_RINGS
,
2832 priv
->rx_ring_num
= prof
->rx_ring_num
;
2833 priv
->cqe_factor
= (mdev
->dev
->caps
.cqe_size
== 64) ? 1 : 0;
2834 priv
->cqe_size
= mdev
->dev
->caps
.cqe_size
;
2835 priv
->mac_index
= -1;
2836 priv
->msg_enable
= MLX4_EN_MSG_LEVEL
;
2837 #ifdef CONFIG_MLX4_EN_DCB
2838 if (!mlx4_is_slave(priv
->mdev
->dev
)) {
2839 if (mdev
->dev
->caps
.flags2
& MLX4_DEV_CAP_FLAG2_ETS_CFG
) {
2840 dev
->dcbnl_ops
= &mlx4_en_dcbnl_ops
;
2842 en_info(priv
, "enabling only PFC DCB ops\n");
2843 dev
->dcbnl_ops
= &mlx4_en_dcbnl_pfc_ops
;
2848 for (i
= 0; i
< MLX4_EN_MAC_HASH_SIZE
; ++i
)
2849 INIT_HLIST_HEAD(&priv
->mac_hash
[i
]);
2851 /* Query for default mac and max mtu */
2852 priv
->max_mtu
= mdev
->dev
->caps
.eth_mtu_cap
[priv
->port
];
2854 if (mdev
->dev
->caps
.rx_checksum_flags_port
[priv
->port
] &
2855 MLX4_RX_CSUM_MODE_VAL_NON_TCP_UDP
)
2856 priv
->flags
|= MLX4_EN_FLAG_RX_CSUM_NON_TCP_UDP
;
2858 /* Set default MAC */
2859 dev
->addr_len
= ETH_ALEN
;
2860 mlx4_en_u64_to_mac(dev
->dev_addr
, mdev
->dev
->caps
.def_mac
[priv
->port
]);
2861 if (!is_valid_ether_addr(dev
->dev_addr
)) {
2862 if (mlx4_is_slave(priv
->mdev
->dev
)) {
2863 eth_hw_addr_random(dev
);
2864 en_warn(priv
, "Assigned random MAC address %pM\n", dev
->dev_addr
);
2865 mac_u64
= mlx4_mac_to_u64(dev
->dev_addr
);
2866 mdev
->dev
->caps
.def_mac
[priv
->port
] = mac_u64
;
2868 en_err(priv
, "Port: %d, invalid mac burned: %pM, quiting\n",
2869 priv
->port
, dev
->dev_addr
);
2875 memcpy(priv
->current_mac
, dev
->dev_addr
, sizeof(priv
->current_mac
));
2877 priv
->stride
= roundup_pow_of_two(sizeof(struct mlx4_en_rx_desc
) +
2878 DS_SIZE
* MLX4_EN_MAX_RX_FRAGS
);
2879 err
= mlx4_en_alloc_resources(priv
);
2883 /* Initialize time stamping config */
2884 priv
->hwtstamp_config
.flags
= 0;
2885 priv
->hwtstamp_config
.tx_type
= HWTSTAMP_TX_OFF
;
2886 priv
->hwtstamp_config
.rx_filter
= HWTSTAMP_FILTER_NONE
;
2888 /* Allocate page for receive rings */
2889 err
= mlx4_alloc_hwq_res(mdev
->dev
, &priv
->res
,
2890 MLX4_EN_PAGE_SIZE
, MLX4_EN_PAGE_SIZE
);
2892 en_err(priv
, "Failed to allocate page for rx qps\n");
2895 priv
->allocated
= 1;
2898 * Initialize netdev entry points
2900 if (mlx4_is_master(priv
->mdev
->dev
))
2901 dev
->netdev_ops
= &mlx4_netdev_ops_master
;
2903 dev
->netdev_ops
= &mlx4_netdev_ops
;
2904 dev
->watchdog_timeo
= MLX4_EN_WATCHDOG_TIMEOUT
;
2905 netif_set_real_num_tx_queues(dev
, priv
->tx_ring_num
);
2906 netif_set_real_num_rx_queues(dev
, priv
->rx_ring_num
);
2908 dev
->ethtool_ops
= &mlx4_en_ethtool_ops
;
2911 * Set driver features
2913 dev
->hw_features
= NETIF_F_SG
| NETIF_F_IP_CSUM
| NETIF_F_IPV6_CSUM
;
2914 if (mdev
->LSO_support
)
2915 dev
->hw_features
|= NETIF_F_TSO
| NETIF_F_TSO6
;
2917 dev
->vlan_features
= dev
->hw_features
;
2919 dev
->hw_features
|= NETIF_F_RXCSUM
| NETIF_F_RXHASH
;
2920 dev
->features
= dev
->hw_features
| NETIF_F_HIGHDMA
|
2921 NETIF_F_HW_VLAN_CTAG_TX
| NETIF_F_HW_VLAN_CTAG_RX
|
2922 NETIF_F_HW_VLAN_CTAG_FILTER
;
2923 dev
->hw_features
|= NETIF_F_LOOPBACK
|
2924 NETIF_F_HW_VLAN_CTAG_TX
| NETIF_F_HW_VLAN_CTAG_RX
;
2926 if (mdev
->dev
->caps
.flags
& MLX4_DEV_CAP_FLAG_FCS_KEEP
)
2927 dev
->hw_features
|= NETIF_F_RXFCS
;
2929 if (mdev
->dev
->caps
.flags2
& MLX4_DEV_CAP_FLAG2_IGNORE_FCS
)
2930 dev
->hw_features
|= NETIF_F_RXALL
;
2932 if (mdev
->dev
->caps
.steering_mode
==
2933 MLX4_STEERING_MODE_DEVICE_MANAGED
&&
2934 mdev
->dev
->caps
.dmfs_high_steer_mode
!= MLX4_STEERING_DMFS_A0_STATIC
)
2935 dev
->hw_features
|= NETIF_F_NTUPLE
;
2937 if (mdev
->dev
->caps
.steering_mode
!= MLX4_STEERING_MODE_A0
)
2938 dev
->priv_flags
|= IFF_UNICAST_FLT
;
2940 /* Setting a default hash function value */
2941 if (mdev
->dev
->caps
.flags2
& MLX4_DEV_CAP_FLAG2_RSS_TOP
) {
2942 priv
->rss_hash_fn
= ETH_RSS_HASH_TOP
;
2943 } else if (mdev
->dev
->caps
.flags2
& MLX4_DEV_CAP_FLAG2_RSS_XOR
) {
2944 priv
->rss_hash_fn
= ETH_RSS_HASH_XOR
;
2947 "No RSS hash capabilities exposed, using Toeplitz\n");
2948 priv
->rss_hash_fn
= ETH_RSS_HASH_TOP
;
2951 mdev
->pndev
[port
] = dev
;
2952 mdev
->upper
[port
] = NULL
;
2954 netif_carrier_off(dev
);
2955 mlx4_en_set_default_moderation(priv
);
2957 en_warn(priv
, "Using %d TX rings\n", prof
->tx_ring_num
);
2958 en_warn(priv
, "Using %d RX rings\n", prof
->rx_ring_num
);
2960 mlx4_en_update_loopback_state(priv
->dev
, priv
->dev
->features
);
2962 /* Configure port */
2963 mlx4_en_calc_rx_buf(dev
);
2964 err
= mlx4_SET_PORT_general(mdev
->dev
, priv
->port
,
2965 priv
->rx_skb_size
+ ETH_FCS_LEN
,
2966 prof
->tx_pause
, prof
->tx_ppp
,
2967 prof
->rx_pause
, prof
->rx_ppp
);
2969 en_err(priv
, "Failed setting port general configurations for port %d, with error %d\n",
2974 if (mdev
->dev
->caps
.tunnel_offload_mode
== MLX4_TUNNEL_OFFLOAD_MODE_VXLAN
) {
2975 err
= mlx4_SET_PORT_VXLAN(mdev
->dev
, priv
->port
, VXLAN_STEER_BY_OUTER_MAC
, 1);
2977 en_err(priv
, "Failed setting port L2 tunnel configuration, err %d\n",
2984 en_warn(priv
, "Initializing port\n");
2985 err
= mlx4_INIT_PORT(mdev
->dev
, priv
->port
);
2987 en_err(priv
, "Failed Initializing port\n");
2990 queue_delayed_work(mdev
->workqueue
, &priv
->stats_task
, STATS_DELAY
);
2992 if (mdev
->dev
->caps
.flags2
& MLX4_DEV_CAP_FLAG2_TS
)
2993 queue_delayed_work(mdev
->workqueue
, &priv
->service_task
,
2994 SERVICE_TASK_DELAY
);
2996 mlx4_en_set_stats_bitmap(mdev
->dev
, &priv
->stats_bitmap
,
2997 mdev
->profile
.prof
[priv
->port
].rx_ppp
,
2998 mdev
->profile
.prof
[priv
->port
].rx_pause
,
2999 mdev
->profile
.prof
[priv
->port
].tx_ppp
,
3000 mdev
->profile
.prof
[priv
->port
].tx_pause
);
3002 err
= register_netdev(dev
);
3004 en_err(priv
, "Netdev registration failed for port %d\n", port
);
3008 priv
->registered
= 1;
3013 mlx4_en_destroy_netdev(dev
);
3017 int mlx4_en_reset_config(struct net_device
*dev
,
3018 struct hwtstamp_config ts_config
,
3019 netdev_features_t features
)
3021 struct mlx4_en_priv
*priv
= netdev_priv(dev
);
3022 struct mlx4_en_dev
*mdev
= priv
->mdev
;
3026 if (priv
->hwtstamp_config
.tx_type
== ts_config
.tx_type
&&
3027 priv
->hwtstamp_config
.rx_filter
== ts_config
.rx_filter
&&
3028 !DEV_FEATURE_CHANGED(dev
, features
, NETIF_F_HW_VLAN_CTAG_RX
) &&
3029 !DEV_FEATURE_CHANGED(dev
, features
, NETIF_F_RXFCS
))
3030 return 0; /* Nothing to change */
3032 if (DEV_FEATURE_CHANGED(dev
, features
, NETIF_F_HW_VLAN_CTAG_RX
) &&
3033 (features
& NETIF_F_HW_VLAN_CTAG_RX
) &&
3034 (priv
->hwtstamp_config
.rx_filter
!= HWTSTAMP_FILTER_NONE
)) {
3035 en_warn(priv
, "Can't turn ON rx vlan offload while time-stamping rx filter is ON\n");
3039 mutex_lock(&mdev
->state_lock
);
3040 if (priv
->port_up
) {
3042 mlx4_en_stop_port(dev
, 1);
3045 mlx4_en_free_resources(priv
);
3047 en_warn(priv
, "Changing device configuration rx filter(%x) rx vlan(%x)\n",
3048 ts_config
.rx_filter
, !!(features
& NETIF_F_HW_VLAN_CTAG_RX
));
3050 priv
->hwtstamp_config
.tx_type
= ts_config
.tx_type
;
3051 priv
->hwtstamp_config
.rx_filter
= ts_config
.rx_filter
;
3053 if (DEV_FEATURE_CHANGED(dev
, features
, NETIF_F_HW_VLAN_CTAG_RX
)) {
3054 if (features
& NETIF_F_HW_VLAN_CTAG_RX
)
3055 dev
->features
|= NETIF_F_HW_VLAN_CTAG_RX
;
3057 dev
->features
&= ~NETIF_F_HW_VLAN_CTAG_RX
;
3058 } else if (ts_config
.rx_filter
== HWTSTAMP_FILTER_NONE
) {
3059 /* RX time-stamping is OFF, update the RX vlan offload
3060 * to the latest wanted state
3062 if (dev
->wanted_features
& NETIF_F_HW_VLAN_CTAG_RX
)
3063 dev
->features
|= NETIF_F_HW_VLAN_CTAG_RX
;
3065 dev
->features
&= ~NETIF_F_HW_VLAN_CTAG_RX
;
3068 if (DEV_FEATURE_CHANGED(dev
, features
, NETIF_F_RXFCS
)) {
3069 if (features
& NETIF_F_RXFCS
)
3070 dev
->features
|= NETIF_F_RXFCS
;
3072 dev
->features
&= ~NETIF_F_RXFCS
;
3075 /* RX vlan offload and RX time-stamping can't co-exist !
3076 * Regardless of the caller's choice,
3077 * Turn Off RX vlan offload in case of time-stamping is ON
3079 if (ts_config
.rx_filter
!= HWTSTAMP_FILTER_NONE
) {
3080 if (dev
->features
& NETIF_F_HW_VLAN_CTAG_RX
)
3081 en_warn(priv
, "Turning off RX vlan offload since RX time-stamping is ON\n");
3082 dev
->features
&= ~NETIF_F_HW_VLAN_CTAG_RX
;
3085 err
= mlx4_en_alloc_resources(priv
);
3087 en_err(priv
, "Failed reallocating port resources\n");
3091 err
= mlx4_en_start_port(dev
);
3093 en_err(priv
, "Failed starting port\n");
3097 mutex_unlock(&mdev
->state_lock
);
3098 netdev_features_change(dev
);