Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
MIME-Version: 1.0
References: <20190207205730.199332-1-evgreen@chromium.org> <20190207205730.199332-3-evgreen@chromium.org>
 <yq1d0nvqeix.fsf@oracle.com>
In-Reply-To: <yq1d0nvqeix.fsf@oracle.com>
From:   Evan Green <evgreen@chromium.org>
Date:   Thu, 14 Feb 2019 10:00:05 -0800
Message-ID: <CAE=gft6bdM2dLc9RhGYMWp8bz6_c-W8Yy-CUTM9wXe+UiYkXEw@mail.gmail.com>
Subject: Re: [PATCH v2 2/2] loop: Better discard support for block devices
To:     "Martin K. Petersen" <martin.petersen@oracle.com>
Cc:     Jens Axboe <axboe@kernel.dk>, Bart Van Assche <bvanassche@acm.org>,
        Gwendal Grignou <gwendal@chromium.org>,
        Alexis Savery <asavery@chromium.org>,
        Ming Lei <ming.lei@redhat.com>,
        linux-block <linux-block@vger.kernel.org>,
        LKML <linux-kernel@vger.kernel.org>
Content-Type: text/plain; charset="UTF-8"
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

On Wed, Feb 13, 2019 at 6:40 PM Martin K. Petersen
<martin.petersen@oracle.com> wrote:
>
>
> Evan,
>
> > If the backing device for a loop device is a block device, then mirror
> > the discard properties of the underlying block device into the loop
> > device. While in there, differentiate between REQ_OP_DISCARD and
> > REQ_OP_WRITE_ZEROES, which are different for block devices, but which
> > the loop device had just been lumping together.
>
> Bubbling up the queue limits from the backing device is fine. However,
> I'm not sure why you are requiring a filesystem to be on a
> discard-capable device for REQ_OP_DISCARD to have an effect? Punching a
> hole in a file is semantically the same as discarding.
>

Hi Martin,
Thanks so much for taking a look at this patch, I was getting nervous
it was languishing again. I got confused by this comment though. My
intention was to not change behavior for loop devices backed by a
regular file system file. The changes in loop_config_discard() should
result in QUEUE_FLAG_DISCARD being set for backings of regular files
that support f_op->fallocate(), same as before my patch. The change in
lo_discard() to call blk_queue_discard() on the loopback queue itself
was just shorthand to avoid duplicating all those if statements from
loop_config_discard again. Am I missing a spot where I've implicitly
changed the behavior for file-backed loop devices?
-Evan