[Swan-dev] [Swan-commit] Changes to ref refs/heads/master

[Andrew Cagney]

On Thu, 28 Feb 2019 at 11:48, Paul Wouters <paul at nohats.ca> wrote:
>
> On Thu, 28 Feb 2019, Andrew Cagney wrote:
>
> >> New commits:
> >> commit c46b7d010ba30670a768b1651070a666211e648c
> >> Author: Paul Wouters <pwouters at redhat.com>
> >> Date:   Wed Feb 27 23:24:24 2019 -0500
> >>
> >>     IKEv1: Another follow up on aa6b8949
> >>
> >>     It also broke msgid handling for DPD and DELETE routines.
> >
> > I suspect this doesn't fix the bug (I went through and tracked down
> > and fixed all the cases that caused a test fail, this one didn't come
> > up):
> >
> > - the host ordered msgid and stored in the header, will be switched to
> > network order on the way out
> >
> > - so the hash code should be fed the hton(msgid) value (in fact, there
> > are so many cases of the same code it should be moved to a function)
>
> interop tests show it did fix it :)
>
> One issue is that the msgid is random, and if you stuff it in the
> ike header using the struct, it is assumed host order and will be
> flipped by out_struct()
>
> The hash calculation therefor needs the flipped version to do its hash
> calculation.
>
> I agree we could use a single function for this.
>
> Note there is one case where this must not be done. I left a comment.
>
> See c46b7d010ba30670a7

That's the case I'm looking at.  Not doing it is demonstrably wrong,
yet you're suggesting it shouldn't be done?  Is there a test case?

Andrew