Some testing and review have pointed out another location where the STOP_RESET is not being handled exactly right. The case of a restart_on='refresh' should also use a STOP_RESET. This has been added to this fix as well.
http://cr.opensolaris.org/~swilcox/6920199_webrev.1/ http://cr.opensolaris.org/~swilcox/6920199_webrev_inc.1/ Simple change -- add the check for refresh at the same point the check for restart is being done, and send the appropriate STOP event on either case as opposed to just the RESTART case. Tony Nguyen wrote: > Sean, > > Looks good. > > -tn > > On 01/27/10 03:38 PM, Sean Wilcox wrote: >> This webrev includes the fix for 6919271, which has been reviewed but >> I'm hoping to put this fix back with that one. >> >> In any case this is a simple fix, where inetd was missing an update to >> handle the new RESTARTER_EVENT_TYPE_STOP_RESET event that is effectively >> a pass-thru to the RESTARTER_EVENT_TYPE_STOP event, but in the bugs case >> was handled as an >> unknown and therefor not offlining the dependent services therefor not >> letting the final service to offline. When the final service was then >> onlined it was already onlined and therefor did not transition the >> non-inetd controlled services back online that had been offlined via the >> disable of rpc/bind. >> >> http://cr.opensolaris.org/~swilcox/6920119_webrev.0/ >> >> Thanks, >> > > > -- Sean Wilcox 303.272.9711 x79711
