From owner-freebsd-arch@FreeBSD.ORG  Fri Nov 21 01:27:45 2003
Return-Path: <owner-freebsd-arch@FreeBSD.ORG>
Delivered-To: freebsd-arch@freebsd.org
Received: from mx1.FreeBSD.org (mx1.freebsd.org [216.136.204.125])
	by hub.freebsd.org (Postfix) with ESMTP
	id 822E716A4CE; Fri, 21 Nov 2003 01:27:45 -0800 (PST)
Received: from eth0.b.smtp.sonic.net (eth0.b.smtp.sonic.net [64.142.19.4])
	by mx1.FreeBSD.org (Postfix) with ESMTP
	id B520543FE3; Fri, 21 Nov 2003 01:27:44 -0800 (PST)
	(envelope-from gharris@sonic.net)
Received: from quadrajet.sonic.net (adsl-209-204-185-249.sonic.net
	[209.204.185.249])hAL9Rgni008972;	Fri, 21 Nov 2003 01:27:43 -0800
Received: (from guy@localhost)
	by quadrajet.sonic.net (8.9.3/8.9.3) id BAA52399;
	Fri, 21 Nov 2003 01:27:41 -0800 (PST)
	(envelope-from gharris)
Date: Fri, 21 Nov 2003 01:27:41 -0800
From: Guy Harris <guy@alum.mit.edu>
To: Bruce Evans <bde@zeta.org.au>
Message-ID: <20031121012741.G329@quadrajet.sonic.net>
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
User-Agent: Mutt/1.2.4i
cc: arch@FreeBSD.org
cc: fenner@FreeBSD.org
Subject: Re: bpf/pcap are weird
X-BeenThere: freebsd-arch@freebsd.org
X-Mailman-Version: 2.1.1
Precedence: list
List-Id: Discussion related to FreeBSD architecture
	<freebsd-arch.freebsd.org>
List-Unsubscribe: <http://lists.freebsd.org/mailman/listinfo/freebsd-arch>,
	<mailto:freebsd-arch-request@freebsd.org?subject=unsubscribe>
List-Archive: <http://lists.freebsd.org/pipermail/freebsd-arch>
List-Post: <mailto:freebsd-arch@freebsd.org>
List-Help: <mailto:freebsd-arch-request@freebsd.org?subject=help>
List-Subscribe: <http://lists.freebsd.org/mailman/listinfo/freebsd-arch>,
	<mailto:freebsd-arch-request@freebsd.org?subject=subscribe>
X-List-Received-Date: Fri, 21 Nov 2003 09:27:45 -0000

> > bpfpoll() is reported to be broken; see PR 36219.
> 
> Yes, that's the PR that indicated that "select()"/"poll()" don't, in
> fact, work correctly with timeouts.
> 
> That was fixed in 4.5...
> 
> > Rev.1.113 of bpf.c may have disturbed this.
> 
> ...but might have been re-broken.
> 
> > It removed the comment which said that
> > bpf_ready() doesn't acually imitate resultof(FIONREAD) != 0.
> 
> ...and it also removed the check for "d->bd_state == BPF_TIMED_OUT" that
> made "select()"/"poll()" work with timeouts.

I didn't read the code fully - no, it didn't.  Revision 1.113 changed

	if (events & (POLLIN | POLLRDNORM)) {
		/*
		 * An imitation of the FIONREAD ioctl code.
		 * XXX not quite.  An exact imitation:
		 *      if (d->b_slen != 0 ||
		 *          (d->bd_hbuf != NULL && d->bd_hlen != 0)
		 */
		if (d->bd_hlen != 0 ||
		    ((d->bd_immediate || d->bd_state == BPF_TIMED_OUT) &&
		    d->bd_slen != 0))
			revents |= events & (POLLIN | POLLRDNORM);
		else {
			selrecord(td, &d->bd_sel);
			/* Start the read timeout if necessary. */
			if (d->bd_rtout > 0 && d->bd_state == BPF_IDLE) {
				callout_reset(&d->bd_callout, d->bd_rtout,
				    bpf_timed_out, d);
				d->bd_state = BPF_WAITING;
			}
		}
	}

to

	if (events & (POLLIN | POLLRDNORM)) {
		if (bpf_ready(d))
			revents |= events & (POLLIN | POLLRDNORM);
		else {
			selrecord(td, &d->bd_sel);
			/* Start the read timeout if necessary. */
			if (d->bd_rtout > 0 && d->bd_state == BPF_IDLE) {
				callout_reset(&d->bd_callout, d->bd_rtout,
				    bpf_timed_out, d);
				d->bd_state = BPF_WAITING;
			}
		}
	}

but revision 1.23 of "bpfdesc.h" introduce a macro "bpf_ready(bd)" that
does

	#define bpf_ready(bd)                                            \
		((bd)->bd_hlen != 0 ||                                   \
		 (((bd)->bd_immediate || (bd)->bd_state == BPF_TIMED_OUT) && \
		  (bd)->bd_slen != 0))

so the actual "bpf_poll()" code doesn't change.

Revision 1.23 was MFC'ed to the 4.x branch in revision 1.14.2.3, which
is in 4.9.

(Also, the problem I mentioned with FreeBSD 4.4 not supporting the
workaround for "select()" not working is in FreeBSD 4.3 as well.)