From owner-freebsd-current  Sat Aug 23 00:19:59 1997
Return-Path: <owner-freebsd-current>
Received: (from root@localhost)
          by hub.freebsd.org (8.8.7/8.8.7) id AAA27377
          for current-outgoing; Sat, 23 Aug 1997 00:19:59 -0700 (PDT)
Received: from scanner.worldgate.com (scanner.worldgate.com [198.161.84.3])
          by hub.freebsd.org (8.8.7/8.8.7) with ESMTP id AAA27371;
          Sat, 23 Aug 1997 00:19:53 -0700 (PDT)
Received: from znep.com (uucp@localhost)
	by scanner.worldgate.com (8.8.5/8.8.5) with UUCP id BAA14729;
	Sat, 23 Aug 1997 01:19:44 -0600 (MDT)
Received: from localhost (marcs@localhost) by alive.znep.com (8.7.5/8.7.3) with SMTP id BAA07684; Sat, 23 Aug 1997 01:19:46 -0600 (MDT)
Date: Sat, 23 Aug 1997 01:19:45 -0600 (MDT)
From: Marc Slemko <marcs@znep.com>
To: Sean Eric Fagan <sef@Kithrup.COM>
cc: julian@FreeBSD.ORG, current@FreeBSD.ORG
Subject: Re: I've broken ping
In-Reply-To: <199707120707.AAA23391@kithrup.com>
Message-ID: <Pine.BSF.3.95.970823010850.6677D-100000@alive.znep.com>
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII
Sender: owner-freebsd-current@FreeBSD.ORG
X-Loop: FreeBSD.org
Precedence: bulk

On Sat, 12 Jul 1997, Sean Eric Fagan wrote:

> >(or if sean can post the patch someone else can check it in)
> >(to 2.2 and 3.0) (don;t you have commit privs sean?)
> 
> Yeah, I have patches, but I was hoping you'd be able to test it more
> extensively than I could :).
> 
> I can check it in tomorrow, otherwise.
> 
> For the curious, the patches are below.  I'm going to bed now :).

Running under 2.2-stable from a week or two ago, both the 2.2 and -current
ping are still broken. I am seeing it hang in pr_addr if the reverse isn't
resolvable; ^C wont kill it because it appears like a function call is
being restarted.  I am confused about why this is happening.  ISTR some
changes in FreeBSD WRT this that I haven't had time to follow.

Changing the sigaction() back to a signal() fixes this particular case.
The signal() man page says function calls will be restarted, but the
sigaction() man page implies they won't without SA_RESTART.  That follows
with what I am aware that recent BSD code does.  It doesn't seem to fit
with what I am seeing.

> 
> Index: ping.c
> ===================================================================
> RCS file: /home/ncvs/src/sbin/ping/ping.c,v
> retrieving revision 1.23
> diff -u -r1.23 ping.c
> --- ping.c	1997/07/09 20:33:58	1.23
> +++ ping.c	1997/07/12 03:55:41
> @@ -423,8 +423,19 @@
>  	else
>  		(void)printf("PING %s: %d data bytes\n", hostname, datalen);
>  
> -	(void)signal(SIGINT, stopit);
> -	(void)signal(SIGALRM, catcher);
> +	si_sa.sa_handler = stopit;
> +	sigemptyset(&si_sa.sa_mask);
> +	si_sa.sa_flags = 0;
> +	if (sigaction(SIGINT, &si_sa, 0) == -1) {
> +		err(EX_OSERR, "sigaction SIGINT");
> +	}
> +
> +	si_sa.sa_handler = catcher;
> +	sigemptyset(&si_sa.sa_mask);
> +	si_sa.sa_flags = 0;
> +	if (sigaction(SIGALRM, &si_sa, 0) == -1) {
> +		err(EX_OSERR, "sigaction SIGALRM");
> +	}
>  
>  	/*
>  	 * Use sigaction instead of signal() to get unambiguous semantics
> @@ -508,9 +519,17 @@
>  catcher(int sig)
>  {
>  	int waittime;
> +	struct sigaction si_sa;
>  
>  	pinger();
> -	(void)signal(SIGALRM, catcher);
> +
> +	si_sa.sa_handler = catcher;
> +	sigemptyset(&si_sa.sa_mask);
> +	si_sa.sa_flags = 0;
> +	if (sigaction(SIGALRM, &si_sa, 0) == -1) {
> +		err(EX_OSERR, "sigaction");
> +	}
> +
>  	if (!npackets || ntransmitted < npackets)
>  		alarm((u_int)interval);
>  	else {
> @@ -520,7 +539,9 @@
>  				waittime = 1;
>  		} else
>  			waittime = MAXWAIT;
> -		(void)signal(SIGALRM, stopit);
> +		finish_up = 1;
> +		si_sa.sa_handler = stopit;
> +		(void)sigaction(SIGALRM, &si_sa, 0);
>  		(void)alarm((u_int)waittime);
>  	}
>  }
>