From mboxrd@z Thu Jan  1 00:00:00 1970
Received: from eggs.gnu.org ([2001:4830:134:3::10]:41261)
	by lists.gnu.org with esmtp (Exim 4.71)
	(envelope-from <stefanha@redhat.com>) id 1VTruF-0003mW-Bg
	for qemu-devel@nongnu.org; Wed, 09 Oct 2013 07:26:25 -0400
Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71)
	(envelope-from <stefanha@redhat.com>) id 1VTru9-0000PE-DC
	for qemu-devel@nongnu.org; Wed, 09 Oct 2013 07:26:19 -0400
Received: from mx1.redhat.com ([209.132.183.28]:14246)
	by eggs.gnu.org with esmtp (Exim 4.71)
	(envelope-from <stefanha@redhat.com>) id 1VTru9-0000P6-5L
	for qemu-devel@nongnu.org; Wed, 09 Oct 2013 07:26:13 -0400
Date: Wed, 9 Oct 2013 13:26:09 +0200
From: Stefan Hajnoczi <stefanha@redhat.com>
Message-ID: <20131009112609.GA29850@stefanha-thinkpad.muc.redhat.com>
References: <1381312531-28723-1-git-send-email-stefanha@redhat.com>
	<1381312531-28723-2-git-send-email-stefanha@redhat.com>
	<52552A77.7030504@redhat.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <52552A77.7030504@redhat.com>
Subject: Re: [Qemu-devel] [PATCH 1/2] rfifolock: add recursive FIFO lock
List-Id: <qemu-devel.nongnu.org>
List-Unsubscribe: <https://lists.nongnu.org/mailman/options/qemu-devel>,
	<mailto:qemu-devel-request@nongnu.org?subject=unsubscribe>
List-Archive: <http://lists.nongnu.org/archive/html/qemu-devel>
List-Post: <mailto:qemu-devel@nongnu.org>
List-Help: <mailto:qemu-devel-request@nongnu.org?subject=help>
List-Subscribe: <https://lists.nongnu.org/mailman/listinfo/qemu-devel>,
	<mailto:qemu-devel-request@nongnu.org?subject=subscribe>
To: Paolo Bonzini <pbonzini@redhat.com>
Cc: Kevin Wolf <kwolf@redhat.com>, xiawenc@linux.vnet.ibm.com, qemu-devel@nongnu.org, Michael Roth <mdroth@linux.vnet.ibm.com>

On Wed, Oct 09, 2013 at 12:05:43PM +0200, Paolo Bonzini wrote:
> Il 09/10/2013 11:55, Stefan Hajnoczi ha scritto:
> > +    /* Take a ticket */
> > +    unsigned int ticket = r->tail++;
> > +
> > +    if (r->nesting > 0) {
> > +        if (qemu_thread_is_self(&r->owner_thread)) {
> > +            r->tail--; /* put ticket back, we're nesting */
> > +        } else {
> 
> ticket is dead in the "nested" path, why not move it (and the increment)
> directly after the else?

The increment cannot be moved because there are 3 cases:

1. Uncontended lock: r->tail++
2. Recursive lock: do not change tail
3. Wait for contended lock: r->tail++

Case #1 needs r->tail++ too.

I find this approach clearest:

if (r->nesting == 0) {
    /* Uncontended, just take a ticket */
    r->tail++;
} else if (qemu_thread_is_self(&r->owner_thread)) {
    /* Recursive lock, no need to take a ticket */
} else {
    /* Contended case, take a ticket and wait for it */
    unsigned int ticket = r->tail++;

    while (ticket != r->head) {
        ...
    }
}

There's also a shorter approach which collapses everything into a single
if statement, but I find the if expression a harder to understand:

if (r->nesting == 0 || !qemu_thread_is_self(&r->owner_thread)) {
    unsigned int ticket = r->tail++;
    while (ticket != r->head) {
        ...
    }
}

Any preferences?

Stefan