Re: [PATCH net v3 5/8] net: resolve uncommon partial checksums at ingress

["Michael S. Tsirkin" <mst@redhat.com>]

On Tue, Oct 20, 2026 at 08:05:29PM -0400, Willem de Bruijn wrote:
> Michael S. Tsirkin wrote:
> > On Tue, Oct 20, 2026 at 03:44:35PM -0400, Willem de Bruijn wrote:
> > > > +/* Walk the L3 header at @nhoff: where does its transport header
> > > > + * start, and which protocol is it?  Returns false if the header
> > > > + * is truncated, malformed or fragmented.
> > > > + */
> > > > +static inline bool
> > > > +virtio_net_l4_start(struct sk_buff *skb, unsigned int nhoff, __be16 proto,
> > > > +		    unsigned int *thoff, int *ip_proto)
> > > > +{
> > > > +	if (proto == htons(ETH_P_IP)) {
> > > > +		const struct iphdr *iph = virtio_net_ip4_parse(skb, nhoff, thoff);
> > > > +
> > > > +		if (!iph || ip_is_fragment(iph))
> > > > +			return false;
> > > > +		*ip_proto = iph->protocol;
> > > > +		return true;
> > > > +	}
> > > > +
> > > > +	if (proto == htons(ETH_P_IPV6)) {
> > > > +		const struct ipv6hdr *iph;
> > > > +		int flags = 0;
> > > > +
> > > > +		if (!pskb_may_pull(skb, nhoff + sizeof(*iph)))
> > > > +			return false;
> > > > +		iph = (const void *)(skb->data + nhoff);
> > > > +		if (iph->version != 6)
> > > > +			return false;
> > > > +		*thoff = nhoff + sizeof(*iph);
> > > > +		*ip_proto = iph->nexthdr;
> > > > +		if (*ip_proto != IPPROTO_TCP && *ip_proto != IPPROTO_UDP) {
> > > > +			*thoff = nhoff;
> > > > +			*ip_proto = ipv6_find_hdr(skb, thoff, -1, NULL,
> > > > +						  &flags);
> > > > +			if (*ip_proto < 0 || flags & IP6_FH_F_FRAG)
> > > > +				return false;
> > > > +		}
> > > > +		return true;
> > > > +	}
> > > > +
> > > > +	return false;
> > > > +}
> > > 
> > > for all this parsing: could we use flow dissection, like the
> > > !VIRTIO_NET_HDR_F_NEEDS_CSUM case?
> > 
> > Hmm I don't see how.  That one trusts the caller's protocol and network
> > header. We could parse L2 ourselves and use the dissector for L3+. But
> > it stops at whatever it does not know, and every miss would turn into a
> > drop or a software checksum for a packet the stack delivers fine.  Our
> > parser deliberately covers everything the stack can deliver.
> 
> You're right, the protocol and network offset need to be parsed
> manually. After that it could be passed to the flow dissector.
> 
> Totally untested:
> 
>     static inline bool virtio_net_setup_l3_input(struct sk_buff *skb, int *nhoff, __be16 *proto)
>     {
>         const struct net_device *dev = skb->dev;
> 
>         /* Devices with Ethernet framing: ARPHRD_ETHER, loopback, macvlan, etc. */
>         if (dev->header_ops == &eth_header_ops || dev->type == ARPHRD_ETHER) {
>             if (!pskb_may_pull(skb, ETH_HLEN))
>                 return false;
>             *nhoff = ETH_HLEN;
>             *proto = eth_hdr(skb)->h_proto;
>         } else if (dev->type == ARPHRD_NONE) {
>             /* Raw L3 devices (e.g. tun in IFF_TUN mode) */
>             *nhoff = 0;
>             *proto = skb->protocol;
>             if (!*proto) {
>                 if (!pskb_may_pull(skb, 1))
>                     return false;
>                 if ((skb->data[0] >> 4) == 4)
>                     *proto = htons(ETH_P_IP);
>                 else if ((skb->data[0] >> 4) == 6)
>                     *proto = htons(ETH_P_IPV6);
>                 else
>                     return false;
>             }
>         } else {
>             /* Fallback for other netdevs using caller's network header */
>             *nhoff = skb_network_offset(skb);
>             if (*nhoff < 0 || (unsigned int)*nhoff >= skb->len)
>                 return false;
>             *proto = skb->protocol;
>         }
> 
>         return true;
>     }
> 
>     static inline bool virtio_net_csum_partial_ok(struct sk_buff *skb, u8 hdr_gso_type)
>     {
>         struct flow_keys_basic keys;
>         __be16 orig_proto = skb->protocol;
>         u16 orig_net_hdr = skb->network_header;
>         __be16 proto;
>         int nhoff;
>         u16 csum_offset;
>         size_t min_thlen;
>   
>         if (!virtio_net_setup_l3_input(skb, &nhoff, &proto))
>             return false;
>   
>         skb_set_network_header(skb, nhoff);
>         skb->protocol = proto;
>   
>         if (!skb_flow_dissect_flow_keys_basic(dev_net(skb->dev), skb, &keys,
>                               NULL, 0, 0, 0, 0))
>             goto restore_fail;
>   
>         /* Disallow fragments: partial checksums cannot be retained across frags */
>         if (keys.control.flags & FLOW_DIS_IS_FRAGMENT)
>             goto restore_fail;
>   
>         /* Validate protocol matching */
>         switch (keys.basic.ip_proto) {
>         case IPPROTO_TCP:
>             if (!virtio_net_hdr_match_proto(keys.basic.n_proto, hdr_gso_type))
>                 goto restore_fail;
>             min_thlen = sizeof(struct tcphdr);
>             csum_offset = offsetof(struct tcphdr, check);
>             break;
>         case IPPROTO_UDP:
>             if (!virtio_net_hdr_match_proto(keys.basic.n_proto, hdr_gso_type))
>                 goto restore_fail;
>             min_thlen = sizeof(struct udphdr);
>             csum_offset = offsetof(struct udphdr, check);
>             break;
>         default:
>             goto restore_fail;
>         }
> 
>         /* csum_start must match the dissected transport header offset */
>         if (skb_transport_offset(skb) != keys.control.thoff)
>             goto restore_fail;
> 
>         /* csum_offset must point to the transport checksum field */
>         if (skb->csum_offset != csum_offset)
>             goto restore_fail;
> 
>         /* Ensure the transport header is linear */
>         if (!pskb_may_pull(skb, keys.control.thoff + min_thlen))
>             goto restore_fail;
> 
>         return true;
> 
>     restore_fail:
>         skb->protocol = orig_proto;
>         skb->network_header = orig_net_hdr;
>         return false;
>     }



Hmm, an interesting trick, but I don't see how this works for tunnels?
We need to dissect at inner_nh, and on success we can not leave
network_header there - the packet continues to outer IP processing.
I fear making all this work will be more code than the parsing it
replaces.


-- 
MST