[RFC] ip_frag: support IPv6 reassembly with extensions

[Thomas Monjalon]

11/11/2024 13:59, Konstantin Ananyev:
> 
> > 14/02/2024 04:51, Stephen Hemminger:
> > > On Tue, 13 Feb 2024 12:47:27 +0100
> > > <vignesh.purushotham.srinivas at ericsson.com> wrote:
> > >
> > > > +/*
> > > > + * Function to crawl through the extension header stack.
> > > > + * This function breaks as soon a the fragment header is
> > > > + * found and returns the total length the traversed exts
> > > > + * and the last extension before the fragment header
> > > > + */
> > > > +static inline uint32_t
> > > > +ip_frag_get_last_exthdr(struct rte_ipv6_hdr *ip_hdr, uint8_t **last_ext)
> > > > +{
> > > > +	uint32_t total_len = 0;
> > > > +	size_t ext_len = 0;
> > > > +	*last_ext = (uint8_t *)(ip_hdr + 1);
> > > > +	int next_proto = ip_hdr->proto;
> > > > +
> > > > +	while (next_proto != IPPROTO_FRAGMENT &&
> > > > +		(next_proto = rte_ipv6_get_next_ext(
> > > > +		*last_ext, next_proto, &ext_len)) >= 0) {
> > > > +
> > > > +		total_len += ext_len;
> > > > +
> > > > +		if (next_proto == IPPROTO_FRAGMENT)
> > > > +			return total_len;
> > > > +
> > > > +		*last_ext += ext_len;
> > > > +	}
> > > > +
> > > > +	return total_len;
> > > > +}
> > >
> > > Doing endless loop like this opens up DoS attacks.
> > > Better to use rte_next_skip_ip6_ext() or do similar limited loop.
> > 
> > There was no reply to this interesting comment?
> 
> I think there is a limit is the latest version for that patch:
> https://patchwork.dpdk.org/project/dpdk/patch/20241015082133.3910533-1-vignesh.purushotham.srinivas@ericsson.com/

Even if there is a new version, we should reply to comments.

> It is also has an ACK from me...
> Though looking at it once again - we'd better have an extra check here to
> make sure that total_len would not exceed mbuf->data_len. 

I suppose it can wait 25.03.