On Thu, Jan 19, 2017, at 01:14 AM, teor wrote:
On 19 Jan 2017, at 12:00, Geoff Down geoffdown@fastmail.net wrote: Jan 19 00:47:42.000 [notice] Tor 0.2.9.8 opening log file. Jan 19 00:47:55.000 [warn] Server managed proxy encountered a method error. (obfs3 Could not set up listener (0.0.0.0:xxxxx) for 'obfs3' (Address already in use).) Jan 19 00:47:55.000 [warn] Managed proxy at '/usr/bin/obfsproxy' failed the configuration protocol and will be destroyed.
This could be a bug in tor in pt_configure_remaining_proxies(), or it could be a bug in obfsproxy, or it could not be a bug.
What are the ServerTransport* lines in your torrc, default torrc, and command-line?
In the torrc I have been using for at least a year: ServerTransportPlugin obfs3 exec /usr/bin/obfsproxy managed -none elsewhere. I should have said that previous (stable) versions of tor didn't show this behaviour.
What are the info-level log lines around the lines you quoted?
For example, do you see something like: Preparing managed proxy '%s' for restart. (If so, this could be a bug in proxy_prepare_for_restart().) Nothing changed for managed proxy '%s' after HUP: not restarting.
I had to stop and restart tor to change log level: the same thing happens on startup.
Jan 19 01:48:07.000 [info] launch_managed_proxy(): Managed proxy at '/usr/bin/obfsproxy' has spawned with PID '16433'. Jan 19 02:14:22.000 [info] notify_waitpid_callback_by_pid(): Child process 16433 has exited; running callback. Jan 19 02:14:23.000 [info] handle_proxy_line(): Got a line from managed proxy '/usr/bin/obfsproxy': (VERSION 1) Jan 19 02:14:23.000 [info] handle_proxy_line(): Got a line from managed proxy '/usr/bin/obfsproxy': (SMETHOD-ERROR obfs3 Could not set up listener (0.0.0.0:xxxxx) for 'obfs3' (Address already in use).) Jan 19 02:14:23.000 [warn] Server managed proxy encountered a method error. (obfs3 Could not set up listener (0.0.0.0:xxxxx) for 'obfs3' (Address already in use).) Jan 19 02:14:23.000 [warn] Managed proxy at '/usr/bin/obfsproxy' failed the configuration protocol and will be destroyed. Jan 19 02:14:23.000 [info] tor_process_handle_destroy(): Terminated process with PID '16433'.
Then a HUP produced the same sequence (but a different PID), with no sign of the lines you mention above regarding 'restart'. The original obsfproxy process 17356 is still in the same state.