6:40 a.m.
Hi,
Typing "ping abcd", where abcd does not exist on /etc/hosts or in the
search domains, will usually hang for a while with Connman.
I've compared with other systems and my conclusion is that libnss_dns
is retrying the DNS request to connman, since connman answer does not
include the query field.
This behaviour comes from [1], where both a invalid Query and no
upstream servers are handled in the same way.
I believe the query field must be part of the answer when the Query
message was valid, but just happens to not be possible to resolve.
Thanks,
Nuno
[1] https://git.kernel.org/pub/scm/network/connman/connman.git/tree/src/dnspr...
11:29 p.m.
Hi Nuno,
On 3/28/19 2:40 PM, Nuno Gonçalves wrote:
Hi,
Typing "ping abcd", where abcd does not exist on /etc/hosts or in the
search domains, will usually hang for a while with Connman.
I've compared with other systems and my conclusion is that libnss_dns
is retrying the DNS request to connman, since connman answer does not
include the query field.
This behaviour comes from [1], where both a invalid Query and no
upstream servers are handled in the same way.
I believe the query field must be part of the answer when the Query
message was valid, but just happens to not be possible to resolve.
Thanks for you report. Do you happen to have a patch for this? I suppose
you have the setup already running for testing.
Thanks,
Daniel
10:02 a.m.
Hi,
I did hacked this to work for me, and it will take some substantial
work to upstream. Anyway I hope I can do it soon.
I do have a question on this subsystem, when compiling with
--with-dns-backend=systemd-resolved (instead of internal), what is the
intended behaviour for tether clients?
Currently I think they get no DNS server.
Should we send them the systemd-revolved upstream server?
Should dnsproxy be split into the proxy part and the caching part,
where the proxy will continue to be used to relay requests to
systemd-resolved?
Thanks,
Nuno
On Mon, Apr 1, 2019 at 8:29 AM Daniel Wagner <wagi(a)monom.org> wrote:
Hi Nuno,
On 3/28/19 2:40 PM, Nuno Gonçalves wrote:
> Hi,
>
> Typing "ping abcd", where abcd does not exist on /etc/hosts or in the
> search domains, will usually hang for a while with Connman.
>
> I've compared with other systems and my conclusion is that libnss_dns
> is retrying the DNS request to connman, since connman answer does not
> include the query field.
>
> This behaviour comes from [1], where both a invalid Query and no
> upstream servers are handled in the same way.
>
> I believe the query field must be part of the answer when the Query
> message was valid, but just happens to not be possible to resolve.
Thanks for you report. Do you happen to have a patch for this? I suppose
you have the setup already running for testing.
Thanks,
Daniel
4:33 a.m.
Hi Nuno,
On 7/7/19 7:02 PM, Nuno Gonçalves wrote:
Hi,
I did hacked this to work for me, and it will take some substantial
work to upstream. Anyway I hope I can do it soon.
Excellent! Thanks a lot for your effort. Really appreciated.
I do have a question on this subsystem, when compiling with
--with-dns-backend=systemd-resolved (instead of internal), what is the
intended behaviour for tether clients?
Currently I think they get no DNS server.
Should we send them the systemd-revolved upstream server?
Hmm, sounds like a reasonable thing to do.
Should dnsproxy be split into the proxy part and the caching part,
where the proxy will continue to be used to relay requests to
systemd-resolved?
If this makes sense to do (also if the code allows it :)), then yes.
Thanks,
Daniel
351
days inactive
464
days old
3 comments
2 participants
participants (2)
-
Daniel Wagner -
Nuno Gonçalves