Re: [PATCH bpf] lwt_bpf: account for aligned neigh header length

["Emil Tsalapatis" <emil@etsalapatis.com>]

On Mon Jul 27, 2026 at 6:30 AM EDT, Junseo Lim wrote:
> ip_finish_output2() expands an skb to LL_RESERVED_SPACE(dev) before LWT
> xmit. An LWT_XMIT BPF program can then modify the skb head and still
> return BPF_OK, so bpf_xmit() rechecks the remaining headroom before the
> skb continues to neighbour output.
>
> That recheck uses dst->dev->hard_header_len. This is not enough for the
> neighbour cached-header path: neigh_hh_output() copies the cached hardware
> header using the aligned hh_cache size, HH_DATA_MOD for short headers or
> HH_DATA_ALIGN(hh_len) otherwise.
>
> On Ethernet, hard_header_len is 14 but the cached copy needs 16 bytes. If
> an LWT_XMIT BPF program calls bpf_skb_change_head(skb, 1, 0), the skb can
> still have 15 bytes of headroom after the program. xmit_check_hhlen()
> accepts that, after which neigh_hh_output() hits its headroom warning and
> drops the skb.
>
> Compare against the aligned hardware header length in xmit_check_hhlen()
> so the BPF_OK path leaves enough headroom for neigh_hh_output()'s cached
> header copy.
>
> Fixes: 3a0af8fd61f9 ("bpf: BPF for lightweight tunnel infrastructure")
> Signed-off-by: Junseo Lim <zirajs7@gmail.com>

Reviewed-by: Emil Tsalapatis <emil@etsalapatis.com>

> ---
> Tested on a veth pair with an LWT_XMIT program calling
> bpf_skb_change_head(skb, 1, 0). Before the patch, a UDP packet sent by
> the reproducer triggers the neigh_hh_output() warning and is dropped.
> With the patch, the packet is delivered without the warning.
>
> Below is an excerpt of the warning before the patch:
>
>     WARNING: ./include/net/neighbour.h:538 at ip_finish_output2+0x19f0/0x1f80, CPU#0: lwt_xmit_headro/55
>     Call Trace:
>      __ip_finish_output+0x59b/0x8b0
>      ip_finish_output+0x67/0x390
>      ip_output+0x1db/0x700
>      ip_send_skb+0x1f6/0x270
>      udp_send_skb+0x8d2/0xe00
>      udp_sendmsg+0x1717/0x2620
>      __sys_sendto+0x443/0x4e0
>
>  net/core/lwt_bpf.c | 6 ++++--
>  1 file changed, 4 insertions(+), 2 deletions(-)
>
> diff --git a/net/core/lwt_bpf.c b/net/core/lwt_bpf.c
> index bf588f508b79..2890aa59a3a0 100644
> --- a/net/core/lwt_bpf.c
> +++ b/net/core/lwt_bpf.c
> @@ -169,8 +169,10 @@ static int bpf_output(struct net *net, struct sock *sk, struct sk_buff *skb)
>  
>  static int xmit_check_hhlen(struct sk_buff *skb, int hh_len)
>  {
> -	if (skb_headroom(skb) < hh_len) {
> -		int nhead = HH_DATA_ALIGN(hh_len - skb_headroom(skb));
> +	int hh_alen = HH_DATA_ALIGN(hh_len);
> +
> +	if (skb_headroom(skb) < hh_alen) {
> +		int nhead = HH_DATA_ALIGN(hh_alen - skb_headroom(skb));
>  
>  		if (pskb_expand_head(skb, nhead, 0, GFP_ATOMIC))
>  			return -ENOMEM;