[Dnsmasq-discuss] dnsmasq 2.61 no periodic RA's and sendto errors

[Vladislav Grishenko]

Thanks, Simon.
Now, the only noticeable thing to go is configurable RA periodical interval
values
Rfc2461 introduces:
MaxRtrAdvInterval:	valid range [4 .. 1800], default 600 seconds
MinRtrAdvInterval:	valid range [3 .. 3/4 * MaxRtrAdvInterval], default
1/3 * MaxRtrAdvInterval
AdvDefaultLifetime:	valid range 0, [MaxRtrAdvInterval .. 9000], default
3 * MaxRtrAdvInterval

So, probably enable-ra syntax needs to be expanded, something like
enable-ra,[MaxRtrAdvInterval],[MinRtrAdvInterval],[AdvDefaultLifetime]
If AdvDefaultLifetime and/or MinRtrAdvInterval are not specified - just
calculate them from specified/default MaxRtrAdvInterval.
This will make configuration more handy for sure.

Best Regards, Vladislav Grishenko


> -----Original Message-----
> From: Simon Kelley [mailto:simon at thekelleys.org.uk]
> Sent: Saturday, May 12, 2012 8:35 PM
> To: Vladislav Grishenko
> Cc: dnsmasq-discuss at lists.thekelleys.org.uk; Jim Bos
> Subject: Re: dnsmasq 2.61 no periodic RA's and sendto errors
> 
> On 11/05/12 15:16, Vladislav Grishenko wrote:
> > Hi Simon,
> >
> > dnsmasq 2.61 has couple of issues:
> > 1. No Router Advertisement and corresponding syslog sendto error when
> > replying on Router Solicitation from Windows XP and possible other
> > OSes It's caused by absence of source IPv6 address in RS packet, what
> > is allowed by RFC if host's interface doesn't acquired valid
> > link-local yet In this case RA should be sent to all nodes multicast
> > address (sources has ALL_HOSTS define, which has a bit different
> > meaning) 2. For non-slaac modes periodic RA doesn't work because next
> > RA event could be reset by empty next slaac event Actually it could
> > work only for the first time, if additional random delay is 0.
> >
> > Patches attached.
> >
> > Best Regards, Vladislav Grishenko
> >
> >
> 
> Many thanks. That solves the mystery of the sometime-missing periodic RAs.
> I was aware that could sometime happen, but hadn't reproduced the
> problem.
> 
> I've put out 2.62test1 with these fixes in place. This probably needs
another
> release, but I'll leave it a week or two in case any other regressions
come to
> light in 2.61.
> 
> Jim, I'm pretty sure this is the fix for you problem.
> 
> Cheers,
> 
> Simon.