The following reply was made to PR bin/31502; it has been noted by GNATS.

From: christos@zoulas.com (Christos Zoulas)
To: gnats-bugs@netbsd.org, gnats-admin@netbsd.org,
	netbsd-bugs@netbsd.org
Cc: 
Subject: Re: bin/31502: rpc.statd doesn't save failed notifies
Date: Fri, 7 Oct 2005 08:39:39 -0400

 On Oct 7, 11:34am, xcc98be0c43465684@f4n.org (xcc98be0c43465684@f4n.org) wrote:
 -- Subject: bin/31502: rpc.statd doesn't save failed notifies
 
 Can you try this?
 
 christos
 
 Index: statd.c
 ===================================================================
 RCS file: /cvsroot/src/usr.sbin/rpc.statd/statd.c,v
 retrieving revision 1.23
 diff -u -u -r1.23 statd.c
 --- statd.c	14 Jan 2004 10:29:46 -0000	1.23
 +++ statd.c	7 Oct 2005 12:39:10 -0000
 @@ -459,54 +459,52 @@
  	time_t now = *(time_t *) ptr;
  	char *name = key->data;
  	DBT data;
 +	int error;
  
  	if (hi->notifyReqd == 0 || hi->notifyReqd > now)
  		return 0;
  
 -	if (notify_one_host(name)) {
 -give_up:
 +	/*
 +	 * If one of the initial attempts fails, we wait
 +	 * for a while and have another go.  This is necessary
 +	 * because when we have crashed, (eg. a power outage)
 +	 * it is quite possible that we won't be able to
 +	 * contact all monitored hosts immediately on restart,
 +	 * either because they crashed too and take longer
 +	 * to come up (in which case the notification isn't
 +	 * really required), or more importantly if some
 +	 * router etc. needed to reach the monitored host
 +	 * has not come back up yet.  In this case, we will
 +	 * be a bit late in re-establishing locks (after the
 +	 * grace period) but that is the best we can do.  We
 +	 * try 10 times at 5 sec intervals, 10 more times at
 +	 * 1 minute intervals, then 24 more times at hourly
 +	 * intervals, finally giving up altogether if the
 +	 * host hasn't come back to life after 24 hours.
 +	 */
 +	if (notify_one_host(name) || hi->attempts++ >= 44) {
 +		error = 0;
  		hi->notifyReqd = 0;
  		hi->attempts = 0;
 -		data.data = hi;
 -		data.size = sizeof(*hi);
 -		switch ((*db->put)(db, key, &data, 0)) {
 -		case -1:
 -			syslog(LOG_ERR, "Error storing %s (%m)", name);
 -		case 0:
 -			return 0;
 -
 -		default:
 -			abort();
 -		}
 -	}
 -	else {
 -		/*
 -		 * If one of the initial attempts fails, we wait
 -		 * for a while and have another go.  This is necessary
 -		 * because when we have crashed, (eg. a power outage)
 -		 * it is quite possible that we won't be able to
 -		 * contact all monitored hosts immediately on restart,
 -		 * either because they crashed too and take longer
 -		 * to come up (in which case the notification isn't
 -		 * really required), or more importantly if some
 -		 * router etc. needed to reach the monitored host
 -		 * has not come back up yet.  In this case, we will
 -		 * be a bit late in re-establishing locks (after the
 -		 * grace period) but that is the best we can do.  We
 -		 * try 10 times at 5 sec intervals, 10 more times at
 -		 * 1 minute intervals, then 24 more times at hourly
 -		 * intervals, finally giving up altogether if the
 -		 * host hasn't come back to life after 24 hours.
 -		 */
 -		if (hi->attempts++ >= 44)
 -			goto give_up;
 -		else if (hi->attempts < 10)
 +	} else {
 +		error = -1;
 +		if (hi->attempts < 10)
  			hi->notifyReqd += 5;
  		else if (hi->attempts < 20)
  			hi->notifyReqd += 60;
  		else
  			hi->notifyReqd += 60 * 60;
 -		return -1;
 +	}
 +	data.data = hi;
 +	data.size = sizeof(*hi);
 +	switch ((*db->put)(db, key, &data, 0)) {
 +	case -1:
 +		syslog(LOG_ERR, "Error storing %s (%m)", name);
 +	case 0:
 +		return error;
 +
 +	default:
 +		abort();
  	}
  }