dnsmasq is terminated after short time #510
Comments
In what way?
It should not and I cannot find a reason right now. There are a couple of siturations where we send a SIGTERM, but that either means the DHCP handler is going to be replaced or NAP was disabled.
It is supposed to. It's a short running service invoked via dbus. Anything in the output of blueman-applet when that happens? Did you have blueman-services running? |
From README.debian in both the
I have found this to be true, but did not dig into it too deeply, but rather decided to try out building from git instead.
With what? dnsmasq is the only option available on my system. The
This is not the case. The NAP service is still there and I can connect, but I don't get a DHCP address.
Got it. I just find is suspicious that
Not that I could see.
I was leaving the window open and just using the "Apply" button to turn NAP on and off. Do the services stop when you close the window? |
|
Op 04/29/2016 om 06:44 PM schreef David Lechner:
That file was last updated in
It works perfectly fine with current NM for me on Gentoo. Did you enable ~infirit |
|
I am going the other way around. The host computer is sharing it's existing Internet connection via Bluetooth and the remote device (in this case, LEGO MINDSTORMS EV3) is a client. |
Also, https://bugs.launchpad.net/ubuntu/+source/blueman/+bug/1574093 |
Ah, yes this is only through a dnsserver which we start, dnsmasq is the preferred option. If you install busybox and make a symlink udhcpd it will show up as an optiona as well.
This may be fixed with 33bebbe It was never reported here but I doubt your issues is related to this. |
|
I guess you have not read my previous comments in detail. |
Yes I did which I why I am looking to reproduce this in a VM.
Contradicts below. I gave you an alternate options to try instead of dnsmasq in case dnsmasq was causing the problems.
So it does not work. I can't reproduce the problem with my android phone under a 16.04 vm, however.. There are some changes in master that fix some bugs introduced with the move to GSettings. The best way to workaround them is to change the IP address (increment by 1) and then hit apply. This will force the daemon to be stopped and started again. The command line used for the server is below, you may need to change the paths to match debian/ubuntu. So use this to see what happens and if you can keep it alive like that. This needs root! edit: also check if ip forwarding is enabled, enable it with |
OK, to clarify,
It does work as long as I'll see if I can figure out how to make python print a stack trace in
It is already enabled.
I can try these, but I am 90% sure that the problem is not with |
|
Interesting... yesterday, I was running with |
|
OK, I've installed a fresh Ubuntu 16.04 VM so that we are working from the same place. The official Ubuntu package is working out of the box. So, as I mentioned in #513, I think the problem has to do with residual configurations from older versions. Is there any documentation that describes which settings are stored where (both in 2.0.4 and master) so that I can attempt to clean up my system? |
|
I have reverted to the official Ubuntu 16.04 package (2.0.4) on both my main machine and a (not fresh) VM that I use for development. On my main machine, everything is working as expected now. On the VM however, I am getting an error that Unfortunately, I can't pinpoint what is the difference why one is working now and one is not. Both systems were upgraded from 14.04 and already had blueman (1.23-git201403102151) installed before the upgrade. And both systems exhibited the update: |
|
OK, I think I have this reproducible now. On a fresh Ubuntu 16.04 VM: First, enable "Source code" in "Software & Updates" (can also be done by editing Check the NAP checkbox and click the "Apply" button. Seen in syslog... |
|
Did you try setting the right prefix for your system? Did you try using python3? There's no blueman package installed, right? Regarding the traceback, |
|
This is a mess, multiple version of blueman each with different issues now knowing which belong to which 😞. For blueman master: It is most likely not blueman-mechanism that stops dnsmasq but systemd.
Now dnsmasq stays alive after blueman-mechanism ends for me with my android phone. |
|
And now it dies again.. more confused now 😕. |
|
I can't reproduce this with blueman as packaged by jessie, fedora 23 or gentoo or even ubuntu 16.04. I did briefly tried master but outside of the package manager so I am not counting this. I highly doubt blueman-mechanism is stopping dnsmasq as there is nothing that I can think of that will do this. My suspicion is on systemd + cgroups cleaning things up and killing all processes left in the cgroup after blueman-mechanism ended. |
|
But it does not make sense that this happens in master only. It does not do anything different than 2.0.4 regarding dnsmasq... |
True. I would rather have a proper deb build from master to test than a |
|
Note: lintian gives me an error |
|
That systemd configuration file may cause the problem. It's not in 2-0-stable. |
|
See below, no idea why but the VM seems to be broken and can't connect with the ethernet adapter so I'll probably have to do a new one, sigh.. Is it me or is ubuntu getting more broken with each release.. |
I just used the default
I did, but I couldn't get it to work. Apparently
Yes. In my last comment this was on a VM that had not had a blueman package installed ever. Since I got the official package working now, I kind of gave up on working on this anymore. But, I did make a package for master that fixes the lintian problem and some other issues. https://github.com/ev3dev/blueman I have attached the .deb package that I created from this if you would like to try it out. |
systemd must not kill processes started by blueman-mechanism, i.e. DHCP handlers. Fixes blueman-project#510
|
Got it. #514. 😃 |
Thanks, but shouldn't it be |
No,
Full diff is here. I enabled polkit. Don't know if that is desirable, but I was just trying everything. The only other potentially useful thing is in Most of the other changes are because I was packaging |
Background:
I am trying to get NAP (Internet connection sharing via Bluetooth) working on Ubuntu 16.04. This seems broken in the official package, so I tried building/running from git master. After fixing #508, I was able to get it to work - sort of...
Problem:
After a short time, the
dnsmasqprocess started by blueman is terminated. Here is the syslog...blueman-mechanismexiting is what terminatesdnsmasq. Is this right?blueman-mechanismis not supposed to be terminating. This seems related to the timout timer. reset is only called once, so there is nothing to keep it alive.So, how do I keep
dnsmaqrunning?The text was updated successfully, but these errors were encountered: