Re: [PATCH v1 5/5] bpf: clear decap tunnel GSO state in skb_adjust_room

[bot+bpf-ci@kernel.org]

[-- Attachment #1: Type: text/plain, Size: 2457 bytes --]

> diff --git a/net/core/filter.c b/net/core/filter.c
> --- a/net/core/filter.c
> +++ b/net/core/filter.c
> @@ -3665,6 +3665,37 @@ static int bpf_skb_net_shrink(struct sk_buff *skb, u32 off, u32 len_diff,
>  		if (!(flags & BPF_F_ADJ_ROOM_FIXED_GSO))
>  			skb_increase_gso_size(shinfo, len_diff);
>
> +		/* Selective GSO flag clearing based on decap type.
> +		 * Only clear the flags for the tunnel layer being removed.
> +		 */
> +		if ((flags & BPF_F_ADJ_ROOM_DECAP_L4_UDP) &&
> +		    (shinfo->gso_type & (SKB_GSO_UDP_TUNNEL |
> +					 SKB_GSO_UDP_TUNNEL_CSUM)))
> +			shinfo->gso_type &= ~(SKB_GSO_UDP_TUNNEL |
> +					      SKB_GSO_UDP_TUNNEL_CSUM);

Should SKB_GSO_TUNNEL_REMCSUM also be cleared here when decapping
a UDP tunnel?

VXLAN and Geneve set SKB_GSO_TUNNEL_REMCSUM alongside
SKB_GSO_UDP_TUNNEL during encapsulation:

net/ipv4/udp_offload.c:udp_gro_complete():
    skb_shinfo(skb)->gso_type |= SKB_GSO_TUNNEL_REMCSUM;

drivers/net/vxlan/vxlan_core.c:vxlan_build_skb():
    type |= SKB_GSO_TUNNEL_REMCSUM;

After clearing SKB_GSO_UDP_TUNNEL{,_CSUM} but not TUNNEL_REMCSUM,
the gso_type has an orphaned tunnel flag with no associated tunnel
type.

> +
> +		/* Clear encapsulation flag only when no tunnel GSO flags remain */
> +		if (flags & BPF_F_ADJ_ROOM_DECAP_MASK) {
> +			if (!(shinfo->gso_type & (SKB_GSO_UDP_TUNNEL |
> +						  SKB_GSO_UDP_TUNNEL_CSUM |
> +						  SKB_GSO_GRE |
> +						  SKB_GSO_GRE_CSUM |
> +						  SKB_GSO_IPXIP4 |
> +						  SKB_GSO_IPXIP6)))

This check does not include SKB_GSO_TUNNEL_REMCSUM. If a packet
had both SKB_GSO_UDP_TUNNEL and SKB_GSO_TUNNEL_REMCSUM set, the
L4_UDP clearing above removes UDP_TUNNEL but leaves REMCSUM. This
check then concludes no tunnel GSO flags remain and clears
skb->encapsulation, even though a tunnel-related GSO bit is still
present.

The kernel's own UDP tunnel segmentation in
__skb_udp_tunnel_segment() explicitly handles TUNNEL_REMCSUM as
part of tunnel processing:

    remcsum = !!(skb_shinfo(skb)->gso_type &
                 SKB_GSO_TUNNEL_REMCSUM);

Would it make sense to include TUNNEL_REMCSUM in both the L4_UDP
clearing and this encapsulation check?

> +				if (skb->encapsulation)
> +					skb->encapsulation = 0;
> +		}


---
AI reviewed your patch. Please fix the bug or email reply why it's not a bug.
See: https://github.com/kernel-patches/vmtest/blob/master/ci/claude/README.md

CI run summary: https://github.com/kernel-patches/bpf/actions/runs/23194329334