Skip to content

Commit 4946218

Browse files
author
Mohammad Kabat
committed
mlx4: support BIG TCP packets
Bugzilla: https://bugzilla.redhat.com/2112965 Upstream-status: v5.19-rc1 Conflicts: - drivers/net/ethernet/mellanox/mlx4/en_tx.c Context diff due to upstream commit already ported in contrast to upstream ordering, 504148f ("net: add skb_[inner_]tcp_all_headers helpers") commit 1169a64 Author: Eric Dumazet <edumazet@google.com> Date: Fri May 13 11:34:07 2022 -0700 mlx4: support BIG TCP packets mlx4 supports LSOv2 just fine. IPv6 stack inserts a temporary Hop-by-Hop header with JUMBO TLV for big packets. We need to ignore the HBH header when populating TX descriptor. Tested: Before: (not enabling bigger TSO/GRO packets) ip link set dev eth0 gso_max_size 65536 gro_max_size 65536 netperf -H lpaa18 -t TCP_RR -T2,2 -l 10 -Cc -- -r 70000,70000 MIGRATED TCP REQUEST/RESPONSE TEST from ::0 (::) port 0 AF_INET6 to lpaa18.prod.google.com () port 0 AF_INET6 : first burst 0 : cpu bind Local /Remote Socket Size Request Resp. Elapsed Trans. CPU CPU S.dem S.dem Send Recv Size Size Time Rate local remote local remote bytes bytes bytes bytes secs. per sec % S % S us/Tr us/Tr 262144 540000 70000 70000 10.00 6591.45 0.86 1.34 62.490 97.446 262144 540000 After: (enabling bigger TSO/GRO packets) ip link set dev eth0 gso_max_size 185000 gro_max_size 185000 netperf -H lpaa18 -t TCP_RR -T2,2 -l 10 -Cc -- -r 70000,70000 MIGRATED TCP REQUEST/RESPONSE TEST from ::0 (::) port 0 AF_INET6 to lpaa18.prod.google.com () port 0 AF_INET6 : first burst 0 : cpu bind Local /Remote Socket Size Request Resp. Elapsed Trans. CPU CPU S.dem S.dem Send Recv Size Size Time Rate local remote local remote bytes bytes bytes bytes secs. per sec % S % S us/Tr us/Tr 262144 540000 70000 70000 10.00 8383.95 0.95 1.01 54.432 57.584 262144 540000 Signed-off-by: Eric Dumazet <edumazet@google.com> Reviewed-by: Tariq Toukan <tariqt@nvidia.com> Acked-by: Alexander Duyck <alexanderduyck@fb.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Mohammad Kabat <mkabat@redhat.com>
1 parent 51c7ac8 commit 4946218

File tree

2 files changed

+41
-9
lines changed

2 files changed

+41
-9
lines changed

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

Lines changed: 3 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -3417,6 +3417,9 @@ int mlx4_en_init_netdev(struct mlx4_en_dev *mdev, int port,
34173417
dev->min_mtu = ETH_MIN_MTU;
34183418
dev->max_mtu = priv->max_mtu;
34193419

3420+
/* supports LSOv2 packets. */
3421+
netif_set_tso_max_size(dev, GSO_MAX_SIZE);
3422+
34203423
mdev->pndev[port] = dev;
34213424
mdev->upper[port] = NULL;
34223425

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

Lines changed: 38 additions & 9 deletions
Original file line numberDiff line numberDiff line change
@@ -43,6 +43,7 @@
4343
#include <linux/ip.h>
4444
#include <linux/ipv6.h>
4545
#include <linux/indirect_call_wrapper.h>
46+
#include <net/ipv6.h>
4647

4748
#include "mlx4_en.h"
4849

@@ -634,19 +635,28 @@ static int get_real_size(const struct sk_buff *skb,
634635
struct net_device *dev,
635636
int *lso_header_size,
636637
bool *inline_ok,
637-
void **pfrag)
638+
void **pfrag,
639+
int *hopbyhop)
638640
{
639641
struct mlx4_en_priv *priv = netdev_priv(dev);
640642
int real_size;
641643

642644
if (shinfo->gso_size) {
643645
*inline_ok = false;
644-
if (skb->encapsulation)
646+
*hopbyhop = 0;
647+
if (skb->encapsulation) {
645648
*lso_header_size = skb_inner_tcp_all_headers(skb);
646-
else
649+
} else {
650+
/* Detects large IPV6 TCP packets and prepares for removal of
651+
* HBH header that has been pushed by ip6_xmit(),
652+
* mainly so that tcpdump can dissect them.
653+
*/
654+
if (ipv6_has_hopopt_jumbo(skb))
655+
*hopbyhop = sizeof(struct hop_jumbo_hdr);
647656
*lso_header_size = skb_tcp_all_headers(skb);
657+
}
648658
real_size = CTRL_SIZE + shinfo->nr_frags * DS_SIZE +
649-
ALIGN(*lso_header_size + 4, DS_SIZE);
659+
ALIGN(*lso_header_size - *hopbyhop + 4, DS_SIZE);
650660
if (unlikely(*lso_header_size != skb_headlen(skb))) {
651661
/* We add a segment for the skb linear buffer only if
652662
* it contains data */
@@ -873,6 +883,7 @@ netdev_tx_t mlx4_en_xmit(struct sk_buff *skb, struct net_device *dev)
873883
int desc_size;
874884
int real_size;
875885
u32 index, bf_index;
886+
struct ipv6hdr *h6;
876887
__be32 op_own;
877888
int lso_header_size;
878889
void *fragptr = NULL;
@@ -881,6 +892,7 @@ netdev_tx_t mlx4_en_xmit(struct sk_buff *skb, struct net_device *dev)
881892
bool stop_queue;
882893
bool inline_ok;
883894
u8 data_offset;
895+
int hopbyhop;
884896
bool bf_ok;
885897

886898
tx_ind = skb_get_queue_mapping(skb);
@@ -890,7 +902,7 @@ netdev_tx_t mlx4_en_xmit(struct sk_buff *skb, struct net_device *dev)
890902
goto tx_drop;
891903

892904
real_size = get_real_size(skb, shinfo, dev, &lso_header_size,
893-
&inline_ok, &fragptr);
905+
&inline_ok, &fragptr, &hopbyhop);
894906
if (unlikely(!real_size))
895907
goto tx_drop_count;
896908

@@ -943,7 +955,7 @@ netdev_tx_t mlx4_en_xmit(struct sk_buff *skb, struct net_device *dev)
943955
data = &tx_desc->data;
944956
data_offset = offsetof(struct mlx4_en_tx_desc, data);
945957
} else {
946-
int lso_align = ALIGN(lso_header_size + 4, DS_SIZE);
958+
int lso_align = ALIGN(lso_header_size - hopbyhop + 4, DS_SIZE);
947959

948960
data = (void *)&tx_desc->lso + lso_align;
949961
data_offset = offsetof(struct mlx4_en_tx_desc, lso) + lso_align;
@@ -1008,14 +1020,31 @@ netdev_tx_t mlx4_en_xmit(struct sk_buff *skb, struct net_device *dev)
10081020
((ring->prod & ring->size) ?
10091021
cpu_to_be32(MLX4_EN_BIT_DESC_OWN) : 0);
10101022

1023+
lso_header_size -= hopbyhop;
10111024
/* Fill in the LSO prefix */
10121025
tx_desc->lso.mss_hdr_size = cpu_to_be32(
10131026
shinfo->gso_size << 16 | lso_header_size);
10141027

1015-
/* Copy headers;
1016-
* note that we already verified that it is linear */
1017-
memcpy(tx_desc->lso.header, skb->data, lso_header_size);
10181028

1029+
if (unlikely(hopbyhop)) {
1030+
/* remove the HBH header.
1031+
* Layout: [Ethernet header][IPv6 header][HBH][TCP header]
1032+
*/
1033+
memcpy(tx_desc->lso.header, skb->data, ETH_HLEN + sizeof(*h6));
1034+
h6 = (struct ipv6hdr *)((char *)tx_desc->lso.header + ETH_HLEN);
1035+
h6->nexthdr = IPPROTO_TCP;
1036+
/* Copy the TCP header after the IPv6 one */
1037+
memcpy(h6 + 1,
1038+
skb->data + ETH_HLEN + sizeof(*h6) +
1039+
sizeof(struct hop_jumbo_hdr),
1040+
tcp_hdrlen(skb));
1041+
/* Leave ipv6 payload_len set to 0, as LSO v2 specs request. */
1042+
} else {
1043+
/* Copy headers;
1044+
* note that we already verified that it is linear
1045+
*/
1046+
memcpy(tx_desc->lso.header, skb->data, lso_header_size);
1047+
}
10191048
ring->tso_packets++;
10201049

10211050
i = shinfo->gso_segs;

0 commit comments

Comments
 (0)