net/mlx5: Correctly set gso_size when LRO is used
Subsystems (5)
PATCH 1/1 net/mlx5: Correctly set gso_size when LRO is used
From: Christoph Paasch <[email protected]>
gso_size is expected by the networking stack to be the size of the payload (thus, not including ethernet/IP/TCP-headers). However, cqe_bcnt is the full sized frame (including the headers). Dividing cqe_bcnt by lro_num_seg will then give incorrect results.
For example, running a bpftrace higher up in the TCP-stack (tcp_event_data_recv), we commonly have gso_size set to 1450 or 1451 even though in reality the payload was only 1448 bytes.
So, we need to discount the protocol headers from cqe_bcnt so we can actually divide the payload by lro_num_seg to get the real gso_size.
--- .../net/ethernet/mellanox/mlx5/core/en_rx.c | 20 +++++++++++++++---- 1 file changed, 16 insertions(+), 4 deletions(-)
drivers/net/ethernet/mellanox/mlx5/core/en_rx.c +16 -4
@@ -1154,12 +1154,14 @@ static void mlx5e_lro_update_tcp_hdr(struct mlx5_cqe64 *cqe, struct tcphdr *tcp) } } -static void mlx5e_lro_update_hdr (struct sk_buff *skb,struct mlx5_cqe64 *cqe, -u32 cqe_bcnt) +static unsigned int mlx5e_lro_update_hdr (struct sk_buff *skb, +struct mlx5_cqe64 *cqe, +u32 cqe_bcnt) {struct ethhdr *eth = (struct ethhdr *)(skb->data);struct tcphdr *tcp;int network_depth =0 ; +unsigned int hdrlen;__wsum check;__be16 proto;u16 tot_len;
@@ -1169,11 +1171,14 @@ static void mlx5e_lro_update_hdr(struct sk_buff *skb, struct mlx5_cqe64 *cqe, tot_len = cqe_bcnt - network_depth; ip_p = skb->data + network_depth; + hdrlen = network_depth;if (proto ==htons (ETH_P_IP )) {struct iphdr *ipv4 = ip_p; tcp = ip_p +sizeof (struct iphdr ); + hdrlen +=sizeof (struct iphdr ); +skb_shinfo (skb)->gso_type =SKB_GSO_TCPV4 ; ipv4->ttl = cqe->lro.min_ttl;
@@ -1193,6 +1198,8 @@ static void mlx5e_lro_update_hdr(struct sk_buff *skb, struct mlx5_cqe64 *cqe,struct ipv6hdr *ipv6 = ip_p; tcp = ip_p +sizeof (struct ipv6hdr ); + hdrlen +=sizeof (struct ipv6hdr ); +skb_shinfo (skb)->gso_type =SKB_GSO_TCPV6 ; ipv6->hop_limit = cqe->lro.min_ttl;
@@ -1205,6 +1212,10 @@ static void mlx5e_lro_update_hdr(struct sk_buff *skb, struct mlx5_cqe64 *cqe, tcp->check =tcp_v6_check (payload_len, &ipv6->saddr, &ipv6->daddr, check); } + + hdrlen += tcp->doff *4 ; + +return hdrlen; }static void *mlx5e_shampo_get_packet_hd (struct mlx5e_rq *rq,u16 header_index)
@@ -1561,8 +1572,9 @@ static inline void mlx5e_build_rx_skb(struct mlx5_cqe64 *cqe,mlx5e_macsec_offload_handle_rx_skb (netdev, skb, cqe);if (lro_num_seg >1 ) { -mlx5e_lro_update_hdr (skb, cqe, cqe_bcnt); -skb_shinfo (skb)->gso_size =DIV_ROUND_UP (cqe_bcnt, lro_num_seg); +unsigned int hdrlen =mlx5e_lro_update_hdr (skb, cqe, cqe_bcnt); + +skb_shinfo (skb)->gso_size =DIV_ROUND_UP (cqe_bcnt - hdrlen, lro_num_seg);/* Subtract one since we already counted this as one * "regular" packet in mlx5e_complete_rx_cqe() */
Signature · 1 line
2.49.0
Quoted 66 lines · On 10/07/2025 21:26, [email protected] wrote:
From: Christoph Paasch <[email protected]>
gso_sizeis expected by the networking stack to be the size of the payload (thus, not including ethernet/IP/TCP-headers). However,cqe_bcntis the full sized frame (including the headers). Dividingcqe_bcntbylro_num_segwill then give incorrect results.For example, running a bpftrace higher up in the TCP-stack (
tcp_event_data_recv), we commonly havegso_sizeset to 1450 or 1451 even though in reality the payload was only 1448 bytes.So, we need to discount the protocol headers from
cqe_bcntso we can actually divide the payload bylro_num_segto get the realgso_size.Fixes e586b3b0baee ("net/mlx5: Ethernet Datapath files")Signed-off-by Christoph Paasch <[email protected]>--- .../net/ethernet/mellanox/mlx5/core/en_rx.c | 20 +++++++++++++++---- 1 file changed, 16 insertions(+), 4 deletions(-)drivers/net/ethernet/mellanox/mlx5/core/en_rx.c +9
-2@@ -1154,12 +1154,14 @@ static void mlx5e_lro_update_tcp_hdr(struct mlx5_cqe64 *cqe, struct tcphdr *tcp) } } -static void mlx5e_lro_update_hdr (struct sk_buff *skb,struct mlx5_cqe64 *cqe, -u32 cqe_bcnt) +static unsigned int mlx5e_lro_update_hdr (struct sk_buff *skb, +struct mlx5_cqe64 *cqe, +u32 cqe_bcnt) {struct ethhdr *eth = (struct ethhdr *)(skb->data);struct tcphdr *tcp;int network_depth =0 ; +unsigned int hdrlen;__wsum check;__be16 proto;u16 tot_len;@@ -1169,11 +1171,14 @@ static void mlx5e_lro_update_hdr(struct sk_buff *skb, struct mlx5_cqe64 *cqe, tot_len = cqe_bcnt - network_depth; ip_p = skb->data + network_depth; + hdrlen = network_depth;if (proto ==htons (ETH_P_IP )) {struct iphdr *ipv4 = ip_p; tcp = ip_p +sizeof (struct iphdr ); + hdrlen +=sizeof (struct iphdr ); +skb_shinfo (skb)->gso_type =SKB_GSO_TCPV4 ; ipv4->ttl = cqe->lro.min_ttl;@@ -1193,6 +1198,8 @@ static void mlx5e_lro_update_hdr(struct sk_buff *skb, struct mlx5_cqe64 *cqe,struct ipv6hdr *ipv6 = ip_p; tcp = ip_p +sizeof (struct ipv6hdr ); + hdrlen +=sizeof (struct ipv6hdr ); +skb_shinfo (skb)->gso_type =SKB_GSO_TCPV6 ; ipv6->hop_limit = cqe->lro.min_ttl;
drivers/net/ethernet/mellanox/mlx5/core/en_rx.c static void mlx5e_lro_update_hdr(struct sk_buff *skb, struct mlx5_cqe64 *cqe, · line 1217
Quoted 6 lines
drivers/net/ethernet/mellanox/mlx5/core/en_rx.c +3
-0@@ -1205,6 +1212,10 @@ static void mlx5e_lro_update_hdr(struct sk_buff *skb, struct mlx5_cqe64 *cqe, tcp->check =tcp_v6_check (payload_len, &ipv6->saddr, &ipv6->daddr, check); } + + hdrlen += tcp->doff *4 ; +
Thanks for your patch!
Calculations seem correct.
Wouldn't it be simpler to just return the below?
(void *)tcp + tcp->doff * 4 - skb->data
Quoted 16 lines
drivers/net/ethernet/mellanox/mlx5/core/en_rx.c +4
-2+return hdrlen; }static void *mlx5e_shampo_get_packet_hd (struct mlx5e_rq *rq,u16 header_index)@@ -1561,8 +1572,9 @@ static inline void mlx5e_build_rx_skb(struct mlx5_cqe64 *cqe,mlx5e_macsec_offload_handle_rx_skb (netdev, skb, cqe);if (lro_num_seg >1 ) { -mlx5e_lro_update_hdr (skb, cqe, cqe_bcnt); -skb_shinfo (skb)->gso_size =DIV_ROUND_UP (cqe_bcnt, lro_num_seg); +unsigned int hdrlen =mlx5e_lro_update_hdr (skb, cqe, cqe_bcnt); + +skb_shinfo (skb)->gso_size =DIV_ROUND_UP (cqe_bcnt - hdrlen, lro_num_seg);/* Subtract one since we already counted this as one * "regular" packet in mlx5e_complete_rx_cqe() */
Quoted 80 lines · Tariq Toukan
Quoted 66 lines · On 10/07/2025 21:26, [email protected] wrote:
From: Christoph Paasch <[email protected]>
gso_sizeis expected by the networking stack to be the size of the payload (thus, not including ethernet/IP/TCP-headers). However,cqe_bcntis the full sized frame (including the headers). Dividingcqe_bcntbylro_num_segwill then give incorrect results.For example, running a bpftrace higher up in the TCP-stack (
tcp_event_data_recv), we commonly havegso_sizeset to 1450 or 1451 even though in reality the payload was only 1448 bytes.So, we need to discount the protocol headers from
cqe_bcntso we can actually divide the payload bylro_num_segto get the realgso_size.Fixes e586b3b0baee ("net/mlx5: Ethernet Datapath files")Signed-off-by Christoph Paasch <[email protected]>--- .../net/ethernet/mellanox/mlx5/core/en_rx.c | 20 +++++++++++++++---- 1 file changed, 16 insertions(+), 4 deletions(-)drivers/net/ethernet/mellanox/mlx5/core/en_rx.c +9
-2@@ -1154,12 +1154,14 @@ static void mlx5e_lro_update_tcp_hdr(struct mlx5_cqe64 *cqe, struct tcphdr *tcp) } } -static void mlx5e_lro_update_hdr (struct sk_buff *skb,struct mlx5_cqe64 *cqe, -u32 cqe_bcnt) +static unsigned int mlx5e_lro_update_hdr (struct sk_buff *skb, +struct mlx5_cqe64 *cqe, +u32 cqe_bcnt) {struct ethhdr *eth = (struct ethhdr *)(skb->data);struct tcphdr *tcp;int network_depth =0 ; +unsigned int hdrlen;__wsum check;__be16 proto;u16 tot_len;@@ -1169,11 +1171,14 @@ static void mlx5e_lro_update_hdr(struct sk_buff *skb, struct mlx5_cqe64 *cqe, tot_len = cqe_bcnt - network_depth; ip_p = skb->data + network_depth; + hdrlen = network_depth;if (proto ==htons (ETH_P_IP )) {struct iphdr *ipv4 = ip_p; tcp = ip_p +sizeof (struct iphdr ); + hdrlen +=sizeof (struct iphdr ); +skb_shinfo (skb)->gso_type =SKB_GSO_TCPV4 ; ipv4->ttl = cqe->lro.min_ttl;@@ -1193,6 +1198,8 @@ static void mlx5e_lro_update_hdr(struct sk_buff *skb, struct mlx5_cqe64 *cqe,struct ipv6hdr *ipv6 = ip_p; tcp = ip_p +sizeof (struct ipv6hdr ); + hdrlen +=sizeof (struct ipv6hdr ); +skb_shinfo (skb)->gso_type =SKB_GSO_TCPV6 ; ipv6->hop_limit = cqe->lro.min_ttl;drivers/net/ethernet/mellanox/mlx5/core/en_rx.c static void mlx5e_lro_update_hdr(struct sk_buff *skb, struct mlx5_cqe64 *cqe, · line 1217
Quoted 6 lines
drivers/net/ethernet/mellanox/mlx5/core/en_rx.c +3
-0@@ -1205,6 +1212,10 @@ static void mlx5e_lro_update_hdr(struct sk_buff *skb, struct mlx5_cqe64 *cqe, tcp->check =tcp_v6_check (payload_len, &ipv6->saddr, &ipv6->daddr, check); } + + hdrlen += tcp->doff *4 ; +Thanks for your patch!
Quoted 4 lines
Calculations seem correct.
Wouldn't it be simpler to just return the below?(void *)tcp +
tcp->doff* 4 -skb->data
Absolutely! I can do that!
Christoph
Quoted 19 lines
Quoted 17 lines
drivers/net/ethernet/mellanox/mlx5/core/en_rx.c +4
-2+return hdrlen; }static void *mlx5e_shampo_get_packet_hd (struct mlx5e_rq *rq,u16 header_index)@@ -1561,8 +1572,9 @@ static inline void mlx5e_build_rx_skb(struct mlx5_cqe64 *cqe,mlx5e_macsec_offload_handle_rx_skb (netdev, skb, cqe);if (lro_num_seg >1 ) { -mlx5e_lro_update_hdr (skb, cqe, cqe_bcnt); -skb_shinfo (skb)->gso_size =DIV_ROUND_UP (cqe_bcnt, lro_num_seg); +unsigned int hdrlen =mlx5e_lro_update_hdr (skb, cqe, cqe_bcnt); + +skb_shinfo (skb)->gso_size =DIV_ROUND_UP (cqe_bcnt - hdrlen, lro_num_seg);/* Subtract one since we already counted this as one * "regular" packet in mlx5e_complete_rx_cqe() */
Hi Christoph,
Quoted 10 lines · On 10/07/2025 21:26, [email protected] wrote:
From: Christoph Paasch <[email protected]>
gso_sizeis expected by the networking stack to be the size of the payload (thus, not including ethernet/IP/TCP-headers). However,cqe_bcntis the full sized frame (including the headers). Dividingcqe_bcntbylro_num_segwill then give incorrect results.For example, running a bpftrace higher up in the TCP-stack (
tcp_event_data_recv), we commonly havegso_sizeset to 1450 or 1451 even though in reality the payload was only 1448 bytes.
Other than introspecting the wrong gso_size value, is there a functional breakage that can be observed?
Quoted 11 lines · Gal Pressman
Hi Christoph,
Quoted 7 lines · On 10/07/2025 21:26, [email protected] wrote:
From: Christoph Paasch <[email protected]>
gso_sizeis expected by the networking stack to be the size of the payload (thus, not including ethernet/IP/TCP-headers). However,cqe_bcntis the full sized frame (including the headers). Dividingcqe_bcntbylro_num_segwill then give incorrect results.
Quoted 5 lines
Quoted 3 lines
For example, running a bpftrace higher up in the TCP-stack (
tcp_event_data_recv), we commonly havegso_sizeset to 1450 or 1451 even though in reality the payload was only 1448 bytes.Other than introspecting the wrong
gso_sizevalue, is there a functional breakage that can be observed?
I wouldn't call it "functional breakage", but definitely unintended consequences / lower perf :
- In tcp_measure_rcv_mss() len will be for example 1450, but. rcv_mss will be 1448 (because tp->advmss is 1448). Thus, we will always recompute scaling_ratio each time an LRO-packet is received.
- In tcp_gro_receive, it will interfere with the decision whether or not to flush and thus potentially result in less gro'ed packets.
Christoph
Quoted 18 lines · Christoph Paasch
Quoted 11 lines · Gal Pressman
Hi Christoph,
Quoted 7 lines · On 10/07/2025 21:26, [email protected] wrote:
From: Christoph Paasch <[email protected]>
gso_sizeis expected by the networking stack to be the size of the payload (thus, not including ethernet/IP/TCP-headers). However,cqe_bcntis the full sized frame (including the headers). Dividingcqe_bcntbylro_num_segwill then give incorrect results.Quoted 5 lines
Quoted 3 lines
For example, running a bpftrace higher up in the TCP-stack (
tcp_event_data_recv), we commonly havegso_sizeset to 1450 or 1451 even though in reality the payload was only 1448 bytes.Other than introspecting the wrong
gso_sizevalue, is there a functional breakage that can be observed?
Quoted 7 lines
I wouldn't call it "functional breakage", but definitely unintended consequences / lower perf :
- Intcp_measure_rcv_mss()len will be for example 1450, but.rcv_msswill be 1448 (becausetp->advmssis 1448). Thus, we will always recomputescaling_ratioeach time an LRO-packet is received.
- Intcp_gro_receive, it will interfere with the decision whether or not to flush and thus potentially result in less gro'ed packets.
Thanks!
Please put that in the commit message.