netdev - Re: [net-next,v9,15/15] quic: add packet parser base

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <CADvbK_e8Sbm8Q3KVR=uZZDYuQLTdprRskN=jgompaSeYDkmhFw@mail.gmail.com>
Date: Thu, 5 Feb 2026 14:02:58 -0500
From: Xin Long <lucien.xin@...il.com>
To: Simon Horman <horms@...nel.org>
Cc: steved@...hat.com, marcelo.leitner@...il.com, aahringo@...hat.com, 
	alibuda@...ux.alibaba.com, jbaron@...mai.com, hare@...e.de, kuba@...nel.org, 
	tom@...pey.com, linux-cifs@...r.kernel.org, daniel@...x.se, 
	quic@...ts.linux.dev, jlayton@...nel.org, tfanelli@...hat.com, 
	dhowells@...hat.com, linkinjeon@...nel.org, hepengtao@...omi.com, 
	pc@...guebit.com, kernel-tls-handshake@...ts.linux.dev, 
	illiliti@...tonmail.com, xiyou.wangcong@...il.com, 
	andrew.gospodarek@...adcom.com, mail@...nericson.me, edumazet@...gle.com, 
	pabeni@...hat.com, matttbe@...nel.org, netdev@...r.kernel.org, 
	smfrench@...il.com, metze@...ba.org, mbuhl@...nbsd.org, 
	chuck.lever@...cle.com, dreibh@...ula.no, davem@...emloft.net, 
	sd@...asysnail.net
Subject: Re: [net-next,v9,15/15] quic: add packet parser base

On Thu, Feb 5, 2026 at 6:55 AM Simon Horman <horms@...nel.org> wrote:
>
> This is an AI-generated review of your patch. The human sending this
> email has considered the AI review valid, or at least plausible.
>
> For local reproduction steps see https://netdev-ai.bots.linux.dev/ai-local.html
> ---
> quic: add packet parser base
>
> This patch uses 'quic_packet' to handle packing of QUIC packets on the
> receive (RX) path. It introduces mechanisms to parse the ALPN from client
> Initial packets to determine the correct listener socket, routes received
> packets accordingly, and adds support for ICMP-based MTU updates by
> locating the relevant socket and updating the stored PMTU.
>
> > diff --git a/net/quic/packet.c b/net/quic/packet.c
> > index cf64a6aaa795..16a7db6ccd87 100644
> > --- a/net/quic/packet.c
> > +++ b/net/quic/packet.c
>
> [ ... ]
>
> > @@ -14,6 +14,637 @@
> >
> >  #define QUIC_HLEN            1
> >
> > +/* Handle ICMP Toobig packet and update QUIC socket path MTU. */
> > +static int quic_packet_rcv_err(struct sock *sk, struct sk_buff *skb)
> > +{
> > +     union quic_addr daddr, saddr;
> > +     u32 info;
> > +
> > +     /* All we can do is lookup the matching QUIC socket by addresses. */
> > +     quic_get_msg_addrs(skb, &saddr, &daddr);
> > +     sk = quic_sock_lookup(skb, &daddr, &saddr, sk, NULL);
>
> Are the address arguments swapped in quic_packet_rcv_err()? Looking at
> other call sites in this file, quic_get_msg_addrs is consistently called
> with &daddr first, then &saddr:
>
>   Line 476 in quic_packet_get_listen_sock():
>     quic_get_msg_addrs(skb, &daddr, &saddr);
>
>   Line 513 in quic_packet_get_sock():
>     quic_get_msg_addrs(skb, &daddr, &saddr);
>
>   Line 536 in quic_packet_get_sock():
>     quic_get_msg_addrs(skb, &daddr, &saddr);
>
> But here quic_packet_rcv_err() passes &saddr first, then &daddr.
>
> Looking at the implementation of quic_v4_get_msg_addrs() in
> net/quic/family.c, the function extracts addresses from the packet as:
>
>   sa->v4.sin_port = uh->source;
>   sa->v4.sin_addr.s_addr = ip_hdr(skb)->saddr;
>   da->v4.sin_port = uh->dest;
>   da->v4.sin_addr.s_addr = ip_hdr(skb)->daddr;
>
> So sa gets the packet's source address (remote), and da gets the packet's
> destination address (local).
>
> Then quic_sock_lookup() in net/quic/socket.c expects:
>
>   if (quic_cmp_sk_addr(tmp, quic_path_saddr(paths, 0), sa) &&
>       quic_cmp_sk_addr(tmp, quic_path_daddr(paths, 0), da) &&
>
> This compares the socket's local address with sa and the socket's remote
> address with da, meaning sa should be the local (packet dest) address and
> da should be the remote (packet source) address.
>
> With the swapped arguments, quic_packet_rcv_err() would pass the remote
> address where the local is expected and vice versa, causing the socket
> lookup to fail. This would prevent ICMP Packet Too Big messages from being
> processed correctly, breaking Path MTU Discovery.
>
The code is correct, I may add a comment for the explanation:


        /* ICMP embeds the original outgoing QUIC packet, so
saddr/daddr are reversed when
         * parsed. Only address-based socket lookup is possible in this case.
         */

Thanks.