[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: xfs / xfree86 ipv6



Hi Alistar,
        i did cc to debian-ipv6 mailing list since you might not be the
only one having this problem and other people might look at it as well.

On Sat, 8 Mar 2003, Alistair Phipps wrote:

> First of all, xfs.  As I mentioned, the ipv4 (sid) version listens on
> ipv4 TCP, but the ipv6 version doesn't listen on TCP at all (ipv4 or
> ipv6 sockets), using exactly the same config file (the default
> /etc/X11/fs/config , except with no-listen = tcp commented out - I
> have attached the file as xfs-config.txt).

I did check this one carefully and it looks like we get 2 different
behaviours. Not sure why... it can be everything.. kernel, libc, pkgs, a
mix of woody and sid as well me not understanding 100% your setup.

Anyway if i use the default configuration for xfs with:

no-listen = tcp

i get xfs listening on IPv6

lsof | grep xfs | grep IPv6 shows:
xfs 10339 root 3u IPv6 74383 TCP *:font-service (LISTEN)

and netstat -Ainet6 -a
tcp        0      0 *:font-service          *:*                     LISTEN

if i comment no-listen = tcp so that in the config looks like:

#no-listen = tcp

xfs will only work on unix sockets. so no network code get initialized.
(like if you specify -nolisten tcp,tcp6 to the server)

I can also specify no-listen = tcp6 in the config:

no-listen = tcp6

at this point xfs will listen only on IPv4.

> I tried running the ipv6 xfs manually and it gave me an error:
>
> _FontTransSocketINET6CreateListener: ...SocketCreateListener() failed
> _FontTransMakeAllCOTSServerListeners: server already running

I did use only the /etc/init.d/xfs script to handle the start and stop
of the daemon and i did not get this error at all.

> There was definitely no other instance running.  Could this error be
> because it's trying to bind to both the ipv6 mapped ipv4 address and the
> ipv4 address itself?

If you specify no-listen = tcp the daemon binds to IPv6 then it will
listen also on IPv4, this if you have a vanilla kernel or a usagi
kernel with cat /proc/sys/net/ipv6/conf/all/bindv6only = 0
If you are running usagi with that options turned to 1 then there should
be no error. In the latter case let me know and i will have to reproduce
it locally (i don't run usagi atm).

> The other problem was the problem with lack of authorisation with an xdm
> session connection - at least that seemed the likely cause.
>
> "AUDIT: Fri Mar 7 23:03:39 2003: 1535 X: client 1 rejected from IPv6
> 2001:618:4:15a2::1 port 53952" is an example of one of the errors,

As we discussed on IRC I will just report some info here for completness.
The AUDIT error you get is due to a missing permission for the client to
connect to the server.

> which
> looks just like the error when the host isn't in .Xauthority.  But I
> think xdm should be adding the host itself.

xauth does not support ipv6. I have been starting looking at it but i have
a very little knowlegde about its internal that makes very complex for me
to understand what should go where.

>  Maybe it's not adding it
> properly?  As I said, the xdm login screen gets displayed, but as soon
> as I log in correctly, I just appear to get logged out again, as no
> clients can connect.

It could easily be.

> My /home directory is shared by NFS between the client and server, so I
> think the mit magic cookie sharing should work ok (at least I think
> that's how it works!).

afaik but feel free to correct me also MMC are associated to ip addresses
so if xauth does not know about ipv6 then it is normal that it will never
be there.

>
> I've attached the contents of my /etc/X11/xdm directory as a .bz2 file
> so you can see the config.

I don't see anything special in there that could cause the problem.

> BTW to get it working again, I just had to downgrade xfree86-server on
> the connecting machine to the non-ipv6 version.

can you be a bit more specific here? connecting machine is the client?
did you downgrade all the pkgs or only one? in case which pkgs?

> Let me know if there's anything else I can do to help, or if you have
> any ideas for me!

Thanks a lot for your report first of all. If you can provide more
information that would reduce a lot the area that i will have to look.

Thanks
Fabio

--
drac (1.11-7) unstable; urgency=low
  * added IPv6 patch from the great IPv6 Team

 -- Noel Koethe <noel@debian.org>  Sun,  9 Feb 2003 19:33:00 +0100



Reply to: