Use conditional intervals in heartbeat_callback
In ticket:19454#comment:4 i suggested some rationale about the (now removed) comment in the heartbeat callback function.
The attached patch uses the return value of log_heartbeat to choose between calling the callback after HeartbeatPeriod number of seconds or on the next second.
Activity
changed milestone to %Tor: 0.3.2.x-final
The current behavior is to always call the heartbeat callback every
HeartbeatPeriodseconds even if logging of the heartbeat has failed.The patch ensures the heartbeat callback is only called every
HeartbeatPeriodseconds if logging of the heartbeat message was successful. Otherwise it will report to the scheduler that the callback should be called again in the next second.The only case where
log_heartbeatreturns != 0 is103 if (public_server_mode(options) && !hibernating) { 104 /* Let's check if we are in the current cached consensus. */ 105 if (!(me = router_get_my_routerinfo())) 106 return -1; /* Something stinks, we won't even attempt this. */ 107 else 108 if (!node_get_by_id(me->cache_info.identity_digest)) 109 log_fn(LOG_NOTICE, LD_HEARTBEAT, "Heartbeat: It seems like we are not " 110 "in the cached consensus."); 111 }So in the case when 'something stinks' the patch reschedules calling the callback in the next second instead of after
HeartbeartPeriodseconds.Hi! I applied cypherpunks' patch in my
bug19476branch. It mostly LGTM.I made the following small changes in that branch:
- Instead of rescheduling for the next second—if the writing the heartbeat log message failed, retry
MIN_HEARTBEAT_PERIODseconds later. (The documentation forMIN_HEARTBEAT_PERIODstates that it is the "Lowest allowable value for HeartbeatPeriod; if this is too low, we might expose more information than we're comfortable with.") - Documented this behaviour in the
heartbeat_callback()docstring. - Added a changes file.
- Instead of rescheduling for the next second—if the writing the heartbeat log message failed, retry
moved to tpo/core/tor#19476 (closed)
