netdev - Re: IP oversized ip oacket from - header size should be skipped?

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]

Date: Fri, 28 Jun 2024 15:35:34 +0200
From: Ian Kumlien <ian.kumlien@...il.com>
To: Florian Westphal <fw@...len.de>
Cc: Linux Kernel Network Developers <netdev@...r.kernel.org>
Subject: Re: IP oversized ip oacket from - header size should be skipped?

So this bug predates 2.6.12-rc2, been digging a bit now... Unless
gp->len has been pointing to something else weird.

On Fri, Jun 28, 2024 at 1:44 PM Ian Kumlien <ian.kumlien@...il.com> wrote:
> On Fri, Jun 28, 2024 at 1:28 PM Ian Kumlien <ian.kumlien@...il.com> wrote:
> > On Fri, Jun 28, 2024 at 12:55 PM Ian Kumlien <ian.kumlien@...il.com> wrote:
> > > On Fri, Jun 28, 2024 at 12:53 PM Florian Westphal <fw@...len.de> wrote:
> > > > Ian Kumlien <ian.kumlien@...il.com> wrote:
> > > > > Hi,
> > > > >
> > > > > In net/ipv4/ip_fragment.c line 412:
> > > > > static int ip_frag_reasm(struct ipq *qp, struct sk_buff *skb,
> > > > >                          struct sk_buff *prev_tail, struct net_device *dev)
> > > > > {
> > > > > ...
> > > > >         len = ip_hdrlen(skb) + qp->q.len;
> > > > >         err = -E2BIG;
> > > > >         if (len > 65535)
> > > > >                 goto out_oversize;
> > > > > ....
> > > > >
> > > > > We can expand the expression to:
> > > > > len = (ip_hdr(skb)->ihl * 4) + qp->q.len;
> > > > >
> > > > > But it's still weird since the definition of q->len is: "total length
> > > > > of the original datagram"
> > > >
> > > > AFAICS datagram == l4 payload, so adding ihl is correct.
> > >
> > > But then it should be added and multiplied by the count of fragments?
> > > which doesn't make sense to me...
> > >
> > > I have a security scanner that generates big packets (looking for
> > > overflows using nmap nasl) that causes this to happen on send....
> >
> > So my thinking is that the packet is 65535 or thereabouts which would
> > mean 44 segments, 43 would be 1500 bytes while the last one would be
> > 1035
> >
> > To me it seems extremely unlikely that we would hit the limit in the
> > case of all packets being l4 - but I'll do some more testing
>
> So, I realize that i'm not the best at this but I can't get this to fit.
>
> The 65535 comes from the 16 bit ip total length field, which includes
> header and data.
> The minimum length is 20 which would be just the IP header.
>
> Now, IF we are comparing to 65535 then it HAS to be the full packet (ie l3)
>
> If we are making this comparison with l4 data, then we are not RFC
> compliant IMHO