From mboxrd@z Thu Jan  1 00:00:00 1970
Received: from eggs.gnu.org ([2001:4830:134:3::10]:55996)
	by lists.gnu.org with esmtp (Exim 4.71)
	(envelope-from <pbonzini@redhat.com>) id 1ZAJYb-0003v8-75
	for qemu-devel@nongnu.org; Wed, 01 Jul 2015 11:04:19 -0400
Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71)
	(envelope-from <pbonzini@redhat.com>) id 1ZAJYV-0000rO-Lu
	for qemu-devel@nongnu.org; Wed, 01 Jul 2015 11:04:13 -0400
References: <1435761950-26714-1-git-send-email-stefanha@redhat.com>
	<5593FD66.5020002@redhat.com>
	<CAJSP0QVk+uKK=NPx8R2naGMAqnQfNLsuJJM2iH8xfyaer3_+rw@mail.gmail.com>
From: Paolo Bonzini <pbonzini@redhat.com>
Message-ID: <5594015C.20104@redhat.com>
Date: Wed, 1 Jul 2015 17:03:56 +0200
MIME-Version: 1.0
In-Reply-To: <CAJSP0QVk+uKK=NPx8R2naGMAqnQfNLsuJJM2iH8xfyaer3_+rw@mail.gmail.com>
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: 7bit
Subject: Re: [Qemu-devel] [Qemu-block] [PATCH] block/mirror: limit qiov to
	IOV_MAX elements
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: Stefan Hajnoczi <stefanha@gmail.com>
Cc: qemu block <qemu-block@nongnu.org>, qemu-devel <qemu-devel@nongnu.org>, Stefan Hajnoczi <stefanha@redhat.com>



On 01/07/2015 16:59, Stefan Hajnoczi wrote:
> I found it annoying to write it backwards too, but it's for consistency:
> 
>   if (s->buf_free_count < nb_chunks + added_chunks) {
>       trace_mirror_break_buf_busy(s, nb_chunks, s->in_flight);
>       break;
>   }
>   if (IOV_MAX < nb_chunks + added_chunks) {
>       trace_mirror_break_iov_max(s, nb_chunks, added_chunks);
>       break;
>   }
> 
> It's the same type of check as s->buf_free_count (which isn't modified
> by this loop either so it's a yoda conditional).

Hmm, right.  The problem goes back to:

        while (nb_chunks == 0 && s->buf_free_count < added_chunks) {
            trace_mirror_yield_buf_busy(s, nb_chunks, s->in_flight);
            qemu_coroutine_yield();
        }

where s->buf_free_count _is_ modified by the loop.  The if below:

        if (s->buf_free_count < nb_chunks + added_chunks) {
            trace_mirror_break_buf_busy(s, nb_chunks, s->in_flight);
            break;
        }

is written as a < check for consistency, and the one you add exacerbates
the problem.  If you want you can change the < to > in the "while" loop
as well; otherwise the patch is okay as is.

Paolo