[Dnsmasq-discuss] Re: Problem with multihomed interface

[Simon Kelley]

> Hello,
>
> Marcus Better wrote:
>> Yes, it works if I remove the other address.
>>
>> As a temporary fix, is it possible to force dnsmasq to answer requests?
>
> Your compiled binary probably lacks RTNETLINK support
> (check with strings /usr/sbin/dnsmasq |grep netlink_init).
> On OpenWRT something like the following patch is necessary.
> Than 2.25test3 works as expected in regard to multiple networks/ranges
> on one physical interface now, thanks Simon.

That's good

> I don't know if RTNETLINK support in UCLIBC has to be differenciated
> somehow, though. Please fix that in config.h appropriately.

since 2.23 (i think, certianly 2.24) RTNETLINK  should be enabled when
compiled against uclibc

>
> (Ask if you have use for dnsmasq_2.25test3-2_mipsel.ipk.)
>
> Lutz
>
> --- dnsmasq-2.25test3.orig/src/config.h      2005-12-31 15:47:57.000000000
> +0200
> +++ dnsmasq-2.25test3/src/config.h   2005-12-31 15:47:57.000000000 +0200
> @@ -246,6 +246,7 @@
>  #  else
>  #    undef HAVE_GETOPT_LONG
>  #  endif
> +#define HAVE_RTNETLINK
>  #undef HAVE_ARC4RANDOM
>  #define HAVE_RANDOM
>  #define HAVE_DEV_URANDOM
>

That doesn't look like a patch against vanilla 2.25test3 src/config.h

There's a block above there which is #if defined(__UCLIBC__) and which
does enable RTNETLINK.

Maybe you could send me the complete file, or a complete diff against the
vanilla distribution; it's possible the the openWRT patches are confusing
things.

Cheers,

Simon.