[Olsr-users] INFINITE cost in all links

[Henning Rogge]

On Mon, Jul 28, 2014 at 3:23 PM, Eric Malkowski <(spam-protected)> wrote:
> It is interesting that the gateway node stops transmitting OLSR packets all
> together.
> In an older version (we're talking 0.5.6 I think) there was an old timer bug
> where the system call times() return value goes negative and then the timers
> wouldn't work anymore until you reboot.  I find it interesting that this
> problem happens and upon restarting OLSRd it still won't work and you have
> to resort to reboot so uptime of the OS is zeroed.  Does it seem to be
> uptime related?

We moved away from times() because of this bug to "gettimeofday()" and
added a special wrapper that fix time-zone jumping and stuff like
this.

If gettimeofday() returns an error, OLSR will immediately end with the
error message "OS clock is not working, have to shut down OLSR" and a
return code of 1. It cannot work without a clock.

(Olsrd2 use clock_gettime() (if available) to ask the OS for the
raw_monotinic or monotonic clock... but I would guess this could also
fail.)

So running it from the command line instead of forking into background
should reveal this.

On Mon, Jul 28, 2014 at 7:37 PM, Antonio Anselmi <(spam-protected)> wrote:
> Hi Henning,
>
> I think it's not a network error since the the adHoc beacons work (all node
> are associated at layer2). I think that the Eric Malkowski reply in
> mailing-list could be a good hint (olsrd timers go off so scheduler stops to
> send packets).

I have seen cases where the interface changed in a subtile way to that
every attempt to send a packet to the OLSR socket returned an error.

In both cases (time related and network related) you should see an
output in the debug log.

If possible please run the routing agent with debug level 1 and look
at the last lines.

Henning Rogge