Re: [PATCH net-next v15 15/15] quic: add packet parser base

[Kuniyuki Iwashima <kuniyu@google.com>]

From: Xin Long <lucien.xin@gmail.com>
Date: Wed, 16 Sep 2026 14:08:22 -0400
> > > +struct sock *quic_listen_sock_lookup(struct sk_buff *skb, union quic_addr *sa,
> > > +                                  union quic_addr *da, struct sock *usk,
> > > +                                  struct quic_data *alpns)
> > > +{
> >
> > [ ... ]
> >
> > > +out:
> > > +     if (sk && sk->sk_reuseport)
> > > +             sk = reuseport_select_sock(sk, quic_addr_hash(net, da), skb, 1);
> >
> > [Severity: Medium]
> > Should the hdr_len argument be 0 rather than 1 here?
> >
> > quic_udp_rcv() already did skb_pull(skb, sizeof(struct udphdr)) before
> > quic_packet_rcv() is reached, so skb->data points at the QUIC payload, and
> > reuseport_select_sock() documents hdr_len as how far the pointer must advance
> > to reach the payload.
> >
> > With a classic SO_ATTACH_REUSEPORT_CBPF program, run_bpf_filter() does:
> >
> > net/core/sock_reuseport.c:
> >         if (unlikely(!pskb_may_pull(skb, hdr_len)))
> >                 return -EINVAL;
> >         ...
> >         pskb_pull(skb, hdr_len);
> >         ...
> >         __skb_push(skb, hdr_len);
> >
> > so the filter would see the datagram starting one byte into the QUIC header:
> > all its offsets are shifted and the first byte (form/fixed/type bits) is
> > invisible, which can select the wrong socket in the reuseport group.
> >
> > [ ... ]
> >
> True:
> 
> BPF programs may need to access the QUIC header, so I will pass 0 to
> reuseport_select_sock() to ensure the full QUIC packet is visible.

Why do you use reuseport_select_sock() here ?