[tor-relays] assign_to_cpuworker failed

[Ivan Markin]

diffusae:
> I nevertheless get the "clock jumped" warning once a day:
> 
> "Jan 21 18:34:53.000 [warn] Your system clock just jumped 398 seconds
> forward; assuming established circuits no longer work."
> 
> 3-6 minutes difference is quite much. I also have a ntp daemon running
> and don't know why this happens. The clock is synced?

This may not be related to this issue.
"clock jump" may not be a clock jump actually. E.g. one can just "pause"
all TCP connections for some time (e.g. on a firewall/DPI/traffic
shaper) and then resume them. In such case tor will regard this
sleep/resume in the main loop as a "clock jump" thus breaking all
circuits. This is because it produces ticks each second in the same
thread as the main loop (which is blocking awaiting for data on a socket).
There should be a ticket for this, maybe someone can arrange it.

--
Ivan Markin