This patch enhances GSO segment checks by verifying the presence
of frag_list and protocol consistency, addressing low throughput
issues on IPv4 servers when used as hotspots

Specifically, it fixes a bug in GSO segmentation when forwarding
GRO packets with frag_list. The function skb_segment_list cannot
correctly process GRO skbs converted by XLAT, because XLAT only
converts the header of the head skb. As a result, skbs in the
frag_list may remain unconverted, leading to protocol
inconsistencies and reduced throughput.

To resolve this, the patch uses skb_segment to handle forwarded
packets converted by XLAT, ensuring that all fragments are
properly converted and segmented.

Signed-off-by: Jibin Zhang <jibin.zhang@mediatek.com>
---
v2: To apply the added condition to a narrower scop

  In this version, the condition (skb_has_frag_list(gso_skb) &&
(gso_skb->protocol == skb_shinfo(gso_skb)->frag_list->protocol))
is moved into inner 'if' statement to a narrower scope.

  Send out the patch again for further discussion because:

1. This issue has a significant impact and has occurred in many
countries and regions.
2. Currently, modifying BPF is not a good option, because BPF code
cannot access the header of skb on the fraglist, and the required
changes would affect a wide range of code.
3. Directly disabling GRO aggregation for XLAT flows is also not a
good solution, as this change would disable GRO even when forwarding
is not needed, and it would also require cooperation from all device
drivers.

[1]: https//patchwork.kernel.org/patch/14350844

---
 net/ipv4/tcp_offload.c | 4 +++-
 net/ipv4/udp_offload.c | 4 +++-
 2 files changed, 6 insertions(+), 2 deletions(-)

diff --git a/net/ipv4/tcp_offload.c b/net/ipv4/tcp_offload.c
index fdda18b1abda..6c2c10f37f87 100644
--- a/net/ipv4/tcp_offload.c
+++ b/net/ipv4/tcp_offload.c
@@ -107,7 +107,9 @@ static struct sk_buff *tcp4_gso_segment(struct sk_buff *skb,
 	if (skb_shinfo(skb)->gso_type & SKB_GSO_FRAGLIST) {
 		struct tcphdr *th = tcp_hdr(skb);
 
-		if (skb_pagelen(skb) - th->doff * 4 == skb_shinfo(skb)->gso_size)
+		if ((skb_pagelen(skb) - th->doff * 4 == skb_shinfo(skb)->gso_size) &&
+		    skb_has_frag_list(skb) &&
+		    (skb->protocol == skb_shinfo(skb)->frag_list->protocol))
 			return __tcp4_gso_segment_list(skb, features);
 
 		skb->ip_summed = CHECKSUM_NONE;
diff --git a/net/ipv4/udp_offload.c b/net/ipv4/udp_offload.c
index 19d0b5b09ffa..2a99f011793f 100644
--- a/net/ipv4/udp_offload.c
+++ b/net/ipv4/udp_offload.c
@@ -514,7 +514,9 @@ struct sk_buff *__udp_gso_segment(struct sk_buff *gso_skb,
 
 	if (skb_shinfo(gso_skb)->gso_type & SKB_GSO_FRAGLIST) {
 		 /* Detect modified geometry and pass those to skb_segment. */
-		if (skb_pagelen(gso_skb) - sizeof(*uh) == skb_shinfo(gso_skb)->gso_size)
+		if ((skb_pagelen(gso_skb) - sizeof(*uh) == skb_shinfo(gso_skb)->gso_size) &&
+		    skb_has_frag_list(gso_skb) &&
+		    (gso_skb->protocol == skb_shinfo(gso_skb)->frag_list->protocol))
 			return __udp_gso_segment_list(gso_skb, features, is_ipv6);
 
 		ret = __skb_linearize(gso_skb);
-- 
2.45.2

This is an AI-generated review of your patch. The human sending this
email has considered the AI review valid, or at least pausible.

For local reproduction steps see https://netdev-ai.bots.linux.dev/ai-local.html
---
net: fix segmentation of forwarding fraglist GRO

This patch adds checks for frag_list presence and protocol consistency
when handling SKB_GSO_FRAGLIST packets in GSO segmentation. This ensures
XLAT-converted packets fall back to skb_segment instead of skb_segment_list.

> diff --git a/net/ipv4/tcp_offload.c b/net/ipv4/tcp_offload.c
> index fdda18b1abda..6c2c10f37f87 100644
> --- a/net/ipv4/tcp_offload.c
> +++ b/net/ipv4/tcp_offload.c
> @@ -107,7 +107,9 @@ static struct sk_buff *tcp4_gso_segment(struct sk_buff *skb,
>  	if (skb_shinfo(skb)->gso_type & SKB_GSO_FRAGLIST) {
>  		struct tcphdr *th = tcp_hdr(skb);
>
> -		if (skb_pagelen(skb) - th->doff * 4 == skb_shinfo(skb)->gso_size)
> +		if ((skb_pagelen(skb) - th->doff * 4 == skb_shinfo(skb)->gso_size) &&
> +		    skb_has_frag_list(skb) &&
> +		    (skb->protocol == skb_shinfo(skb)->frag_list->protocol))
>  			return __tcp4_gso_segment_list(skb, features);

Should tcp6_gso_segment() in net/ipv6/tcpv6_offload.c receive the same
treatment? Looking at that function, it has the identical pattern before
this patch:

    if (skb_shinfo(skb)->gso_type & SKB_GSO_FRAGLIST) {
        struct tcphdr *th = tcp_hdr(skb);

        if (skb_pagelen(skb) - th->doff * 4 == skb_shinfo(skb)->gso_size)
            return __tcp6_gso_segment_list(skb, features);

        skb->ip_summed = CHECKSUM_NONE;
    }

If an IPv4 packet is GRO-aggregated with fraglist and then converted by
XLAT to IPv6, the same protocol mismatch could occur when it reaches
tcp6_gso_segment(). Paolo raised this point in the v1 review:
https://lore.kernel.org/all/aab6c515-12e4-48ca-8220-c0797dae781f@redhat.com/

[ ... ]
-- 
pw-bot: cr