Skip to content

Commit 837052d

Browse files
ogerlitzdavem330
ogerlitz
authored and
davem330
committed
net/mlx4_en: Add netdev support for TCP/IP offloads of vxlan tunneling
When the device tunneling offloads mode is vxlan do the following - call SET_PORT with the relevant setting - add DMFS steering vxlan rule for the device self and multicast mac addresses of the form: {<ETH, outer-mac> <VXLAN, ANY vnid> <ETH, ANY mac>} --> RSS QP - set relevant QPC fields in RSS context and RX ring QPs - in TX flow, set WQE fields to generate HW checksum, and handle gso skbs which are marked for encapsulation such that the HW will segment them properly. - in RX flow, read HW offloaded checksum for encapsulated packets from the CQE - advertize hw_enc_features and NETIF_F_GSO_UDP_TUNNEL to the networking stack Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com> Signed-off-by: David S. Miller <davem@davemloft.net>
1 parent 7ffdf72 commit 837052d

File tree

5 files changed

+129
-2
lines changed

5 files changed

+129
-2
lines changed

‎drivers/net/ethernet/mellanox/mlx4/en_netdev.c

Lines changed: 94 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -468,6 +468,53 @@ static void mlx4_en_u64_to_mac(unsigned char dst_mac[ETH_ALEN + 2], u64 src_mac)
468468
memset(&dst_mac[ETH_ALEN], 0, 2);
469469
}
470470

471+
472+
static int mlx4_en_tunnel_steer_add(struct mlx4_en_priv *priv, unsigned char *addr,
473+
int qpn, u64 *reg_id)
474+
{
475+
int err;
476+
struct mlx4_spec_list spec_eth_outer = { {NULL} };
477+
struct mlx4_spec_list spec_vxlan = { {NULL} };
478+
struct mlx4_spec_list spec_eth_inner = { {NULL} };
479+
480+
struct mlx4_net_trans_rule rule = {
481+
.queue_mode = MLX4_NET_TRANS_Q_FIFO,
482+
.exclusive = 0,
483+
.allow_loopback = 1,
484+
.promisc_mode = MLX4_FS_REGULAR,
485+
.priority = MLX4_DOMAIN_NIC,
486+
};
487+
488+
__be64 mac_mask = cpu_to_be64(MLX4_MAC_MASK << 16);
489+
490+
if (priv->mdev->dev->caps.tunnel_offload_mode != MLX4_TUNNEL_OFFLOAD_MODE_VXLAN)
491+
return 0; /* do nothing */
492+
493+
rule.port = priv->port;
494+
rule.qpn = qpn;
495+
INIT_LIST_HEAD(&rule.list);
496+
497+
spec_eth_outer.id = MLX4_NET_TRANS_RULE_ID_ETH;
498+
memcpy(spec_eth_outer.eth.dst_mac, addr, ETH_ALEN);
499+
memcpy(spec_eth_outer.eth.dst_mac_msk, &mac_mask, ETH_ALEN);
500+
501+
spec_vxlan.id = MLX4_NET_TRANS_RULE_ID_VXLAN; /* any vxlan header */
502+
spec_eth_inner.id = MLX4_NET_TRANS_RULE_ID_ETH; /* any inner eth header */
503+
504+
list_add_tail(&spec_eth_outer.list, &rule.list);
505+
list_add_tail(&spec_vxlan.list, &rule.list);
506+
list_add_tail(&spec_eth_inner.list, &rule.list);
507+
508+
err = mlx4_flow_attach(priv->mdev->dev, &rule, reg_id);
509+
if (err) {
510+
en_err(priv, "failed to add vxlan steering rule, err %d\n", err);
511+
return err;
512+
}
513+
en_dbg(DRV, priv, "added vxlan steering rule, mac %pM reg_id %llx\n", addr, *reg_id);
514+
return 0;
515+
}
516+
517+
471518
static int mlx4_en_uc_steer_add(struct mlx4_en_priv *priv,
472519
unsigned char *mac, int *qpn, u64 *reg_id)
473520
{
@@ -585,6 +632,10 @@ static int mlx4_en_get_qp(struct mlx4_en_priv *priv)
585632
if (err)
586633
goto steer_err;
587634

635+
if (mlx4_en_tunnel_steer_add(priv, priv->dev->dev_addr, *qpn,
636+
&priv->tunnel_reg_id))
637+
goto tunnel_err;
638+
588639
entry = kmalloc(sizeof(*entry), GFP_KERNEL);
589640
if (!entry) {
590641
err = -ENOMEM;
@@ -599,6 +650,9 @@ static int mlx4_en_get_qp(struct mlx4_en_priv *priv)
599650
return 0;
600651

601652
alloc_err:
653+
if (priv->tunnel_reg_id)
654+
mlx4_flow_detach(priv->mdev->dev, priv->tunnel_reg_id);
655+
tunnel_err:
602656
mlx4_en_uc_steer_release(priv, priv->dev->dev_addr, *qpn, reg_id);
603657

604658
steer_err:
@@ -642,6 +696,11 @@ static void mlx4_en_put_qp(struct mlx4_en_priv *priv)
642696
}
643697
}
644698

699+
if (priv->tunnel_reg_id) {
700+
mlx4_flow_detach(priv->mdev->dev, priv->tunnel_reg_id);
701+
priv->tunnel_reg_id = 0;
702+
}
703+
645704
en_dbg(DRV, priv, "Releasing qp: port %d, qpn %d\n",
646705
priv->port, qpn);
647706
mlx4_qp_release_range(dev, qpn, 1);
@@ -1044,6 +1103,12 @@ static void mlx4_en_do_multicast(struct mlx4_en_priv *priv,
10441103
if (err)
10451104
en_err(priv, "Fail to detach multicast address\n");
10461105

1106+
if (mclist->tunnel_reg_id) {
1107+
err = mlx4_flow_detach(priv->mdev->dev, mclist->tunnel_reg_id);
1108+
if (err)
1109+
en_err(priv, "Failed to detach multicast address\n");
1110+
}
1111+
10471112
/* remove from list */
10481113
list_del(&mclist->list);
10491114
kfree(mclist);
@@ -1061,6 +1126,10 @@ static void mlx4_en_do_multicast(struct mlx4_en_priv *priv,
10611126
if (err)
10621127
en_err(priv, "Fail to attach multicast address\n");
10631128

1129+
err = mlx4_en_tunnel_steer_add(priv, &mc_list[10], priv->base_qpn,
1130+
&mclist->tunnel_reg_id);
1131+
if (err)
1132+
en_err(priv, "Failed to attach multicast address\n");
10641133
}
10651134
}
10661135
}
@@ -1598,6 +1667,15 @@ int mlx4_en_start_port(struct net_device *dev)
15981667
goto tx_err;
15991668
}
16001669

1670+
if (mdev->dev->caps.tunnel_offload_mode == MLX4_TUNNEL_OFFLOAD_MODE_VXLAN) {
1671+
err = mlx4_SET_PORT_VXLAN(mdev->dev, priv->port, VXLAN_STEER_BY_OUTER_MAC);
1672+
if (err) {
1673+
en_err(priv, "Failed setting port L2 tunnel configuration, err %d\n",
1674+
err);
1675+
goto tx_err;
1676+
}
1677+
}
1678+
16011679
/* Init port */
16021680
en_dbg(HW, priv, "Initializing port\n");
16031681
err = mlx4_INIT_PORT(mdev->dev, priv->port);
@@ -2400,6 +2478,13 @@ int mlx4_en_init_netdev(struct mlx4_en_dev *mdev, int port,
24002478
if (mdev->dev->caps.steering_mode != MLX4_STEERING_MODE_A0)
24012479
dev->priv_flags |= IFF_UNICAST_FLT;
24022480

2481+
if (mdev->dev->caps.tunnel_offload_mode == MLX4_TUNNEL_OFFLOAD_MODE_VXLAN) {
2482+
dev->hw_enc_features |= NETIF_F_IP_CSUM | NETIF_F_RXCSUM |
2483+
NETIF_F_TSO | NETIF_F_GSO_UDP_TUNNEL;
2484+
dev->hw_features |= NETIF_F_GSO_UDP_TUNNEL;
2485+
dev->features |= NETIF_F_GSO_UDP_TUNNEL;
2486+
}
2487+
24032488
mdev->pndev[port] = dev;
24042489

24052490
netif_carrier_off(dev);
@@ -2429,6 +2514,15 @@ int mlx4_en_init_netdev(struct mlx4_en_dev *mdev, int port,
24292514
goto out;
24302515
}
24312516

2517+
if (mdev->dev->caps.tunnel_offload_mode == MLX4_TUNNEL_OFFLOAD_MODE_VXLAN) {
2518+
err = mlx4_SET_PORT_VXLAN(mdev->dev, priv->port, VXLAN_STEER_BY_OUTER_MAC);
2519+
if (err) {
2520+
en_err(priv, "Failed setting port L2 tunnel configuration, err %d\n",
2521+
err);
2522+
goto out;
2523+
}
2524+
}
2525+
24322526
/* Init port */
24332527
en_warn(priv, "Initializing port\n");
24342528
err = mlx4_INIT_PORT(mdev->dev, priv->port);

‎drivers/net/ethernet/mellanox/mlx4/en_resources.c

Lines changed: 6 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -68,6 +68,12 @@ void mlx4_en_fill_qp_context(struct mlx4_en_priv *priv, int size, int stride,
6868
context->db_rec_addr = cpu_to_be64(priv->res.db.dma << 2);
6969
if (!(dev->features & NETIF_F_HW_VLAN_CTAG_RX))
7070
context->param3 |= cpu_to_be32(1 << 30);
71+
72+
if (!is_tx && !rss &&
73+
(mdev->dev->caps.tunnel_offload_mode == MLX4_TUNNEL_OFFLOAD_MODE_VXLAN)) {
74+
en_dbg(HW, priv, "Setting RX qp %x tunnel mode to RX tunneled & non-tunneled\n", qpn);
75+
context->srqn = cpu_to_be32(7 << 28); /* this fills bits 30:28 */
76+
}
7177
}
7278

7379

‎drivers/net/ethernet/mellanox/mlx4/en_rx.c

Lines changed: 14 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -631,6 +631,7 @@ int mlx4_en_process_rx_cq(struct net_device *dev, struct mlx4_en_cq *cq, int bud
631631
int ip_summed;
632632
int factor = priv->cqe_factor;
633633
u64 timestamp;
634+
bool l2_tunnel;
634635

635636
if (!priv->port_up)
636637
return 0;
@@ -709,6 +710,8 @@ int mlx4_en_process_rx_cq(struct net_device *dev, struct mlx4_en_cq *cq, int bud
709710
length -= ring->fcs_del;
710711
ring->bytes += length;
711712
ring->packets++;
713+
l2_tunnel = (dev->hw_enc_features & NETIF_F_RXCSUM) &&
714+
(cqe->vlan_my_qpn & cpu_to_be32(MLX4_CQE_L2_TUNNEL));
712715

713716
if (likely(dev->features & NETIF_F_RXCSUM)) {
714717
if ((cqe->status & cpu_to_be16(MLX4_CQE_STATUS_IPOK)) &&
@@ -738,6 +741,8 @@ int mlx4_en_process_rx_cq(struct net_device *dev, struct mlx4_en_cq *cq, int bud
738741
gro_skb->data_len = length;
739742
gro_skb->ip_summed = CHECKSUM_UNNECESSARY;
740743

744+
if (l2_tunnel)
745+
gro_skb->encapsulation = 1;
741746
if ((cqe->vlan_my_qpn &
742747
cpu_to_be32(MLX4_CQE_VLAN_PRESENT_MASK)) &&
743748
(dev->features & NETIF_F_HW_VLAN_CTAG_RX)) {
@@ -790,6 +795,9 @@ int mlx4_en_process_rx_cq(struct net_device *dev, struct mlx4_en_cq *cq, int bud
790795
skb->protocol = eth_type_trans(skb, dev);
791796
skb_record_rx_queue(skb, cq->ring);
792797

798+
if (l2_tunnel)
799+
skb->encapsulation = 1;
800+
793801
if (dev->features & NETIF_F_RXHASH)
794802
skb_set_hash(skb,
795803
be32_to_cpu(cqe->immed_rss_invalid),
@@ -1057,6 +1065,12 @@ int mlx4_en_config_rss_steer(struct mlx4_en_priv *priv)
10571065
rss_mask |= MLX4_RSS_UDP_IPV4 | MLX4_RSS_UDP_IPV6;
10581066
rss_context->base_qpn_udp = rss_context->default_qpn;
10591067
}
1068+
1069+
if (mdev->dev->caps.tunnel_offload_mode == MLX4_TUNNEL_OFFLOAD_MODE_VXLAN) {
1070+
en_info(priv, "Setting RSS context tunnel type to RSS on inner headers\n");
1071+
rss_mask |= MLX4_RSS_BY_INNER_HEADERS;
1072+
}
1073+
10601074
rss_context->flags = rss_mask;
10611075
rss_context->hash_fn = MLX4_RSS_HASH_TOP;
10621076
for (i = 0; i < 10; i++)

‎drivers/net/ethernet/mellanox/mlx4/en_tx.c

Lines changed: 13 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -39,6 +39,7 @@
3939
#include <linux/if_vlan.h>
4040
#include <linux/vmalloc.h>
4141
#include <linux/tcp.h>
42+
#include <linux/ip.h>
4243
#include <linux/moduleparam.h>
4344

4445
#include "mlx4_en.h"
@@ -560,7 +561,10 @@ static int get_real_size(struct sk_buff *skb, struct net_device *dev,
560561
int real_size;
561562

562563
if (skb_is_gso(skb)) {
563-
*lso_header_size = skb_transport_offset(skb) + tcp_hdrlen(skb);
564+
if (skb->encapsulation)
565+
*lso_header_size = (skb_inner_transport_header(skb) - skb->data) + inner_tcp_hdrlen(skb);
566+
else
567+
*lso_header_size = skb_transport_offset(skb) + tcp_hdrlen(skb);
564568
real_size = CTRL_SIZE + skb_shinfo(skb)->nr_frags * DS_SIZE +
565569
ALIGN(*lso_header_size + 4, DS_SIZE);
566570
if (unlikely(*lso_header_size != skb_headlen(skb))) {
@@ -859,6 +863,14 @@ netdev_tx_t mlx4_en_xmit(struct sk_buff *skb, struct net_device *dev)
859863
tx_info->inl = 1;
860864
}
861865

866+
if (skb->encapsulation) {
867+
struct iphdr *ipv4 = (struct iphdr *)skb_inner_network_header(skb);
868+
if (ipv4->protocol == IPPROTO_TCP || ipv4->protocol == IPPROTO_UDP)
869+
op_own |= cpu_to_be32(MLX4_WQE_CTRL_IIP | MLX4_WQE_CTRL_ILP);
870+
else
871+
op_own |= cpu_to_be32(MLX4_WQE_CTRL_IIP);
872+
}
873+
862874
ring->prod += nr_txbb;
863875

864876
/* If we used a bounce buffer then copy descriptor back into place */

‎drivers/net/ethernet/mellanox/mlx4/mlx4_en.h

Lines changed: 2 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -436,6 +436,7 @@ struct mlx4_en_mc_list {
436436
enum mlx4_en_mclist_act action;
437437
u8 addr[ETH_ALEN];
438438
u64 reg_id;
439+
u64 tunnel_reg_id;
439440
};
440441

441442
struct mlx4_en_frag_info {
@@ -567,7 +568,7 @@ struct mlx4_en_priv {
567568
struct list_head filters;
568569
struct hlist_head filter_hash[1 << MLX4_EN_FILTER_HASH_SHIFT];
569570
#endif
570-
571+
u64 tunnel_reg_id;
571572
};
572573

573574
enum mlx4_en_wol {

0 commit comments

Comments
 (0)