1.9.4: TCP queries when some threads are full

Patrik Lundin patrik at sigterm.se
Tue Dec 10 14:15:57 UTC 2019

On Mon, Dec 02, 2019 at 02:55:58PM +0100, Havard Eidnes wrote:
> As you may have seen, I recently had an "encounter" with unbound
> related to handling of TCP sessions, and it looks like there was a
> bug that caused unbound with TCP client sessions which had seen an
> initial query + response not to re-arm the TCP idle timeout for that
> session.
> The patch in the pull request at
>   https://github.com/NLnetLabs/unbound/pull/122
> should be a fix for this issue if I read the code correctly.  If you
> are able to do some testing / validation of this, that would be
> greatly appreciated.

That is an interesting find, well done and thanks for sharing!

Looking at the patch I wanted to read about comm_point_start_listening()
and found some documentation at
which states:
msec    timeout in milliseconds, or -1 for no (change to the) timeout.
So seconds*1000.

The "no change to the timeout" makes me curious, as I read that to mean
"calling the function with -1 retains the previously set value", did it
not work like that in this instance?

Patrik Lundin

More information about the Unbound-users mailing list