Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
MIME-Version: 1.0
References: <20210330231602.1223216-1-egranata@google.com> <YHQQL1OTOdnuOYUW@stefanha-x1.localdomain>
 <20210412094217.GA981912@infradead.org> <YHarc5gGgjyQOaA+@stefanha-x1.localdomain>
In-Reply-To: <YHarc5gGgjyQOaA+@stefanha-x1.localdomain>
From:   Enrico Granata <egranata@google.com>
Date:   Wed, 14 Apr 2021 14:12:18 -0600
Message-ID: <CAPR809uOOsULgAA647=g+GTY32KbevqytsUfQACCfuuztqdSzg@mail.gmail.com>
Subject: Re: [PATCH] virtio_blk: Add support for lifetime feature
To:     Stefan Hajnoczi <stefanha@redhat.com>
Cc:     Christoph Hellwig <hch@infradead.org>, mst@redhat.com,
        jasowang@redhat.com, pbonzini@redhat.com, axboe@kernel.dk,
        virtualization@lists.linux-foundation.org,
        linux-block@vger.kernel.org, linux-kernel@vger.kernel.org
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

First and foremost thanks for the feedback on the code. I will send
out an up-to-date patch with those comments addressed ASAP

As for the broader issue, I am definitely happy to incorporate any
feedback and work to improve the spec, but looking at embedded storage
devices both eMMC and UFS seem to me to be exposing the attributes
that are included in the virtio-blk lifetime feature, and the same
data is also used by the Android Open Source Project for Health HAL.
It does seem like this format has a lot of practical adoption in the
wider industry and that makes it fairly trivial to implement in a lot
of common embedded systems. Having this immediate path to adoption in
a variety of scenarios seems to me in and of itself valuable.

Thanks,
- Enrico

On Wed, Apr 14, 2021 at 2:44 AM Stefan Hajnoczi <stefanha@redhat.com> wrote:
>
> On Mon, Apr 12, 2021 at 10:42:17AM +0100, Christoph Hellwig wrote:
> > A note to the virtio committee:  eMMC is the worst of all the currently
> > active storage standards by a large margin.  It defines very strange
> > ad-hoc interfaces that expose very specific internals and often provides
> > very poor abstractions.  It would be great it you could reach out to the
> > wider storage community before taking bad ideas from the eMMC standard
> > and putting it into virtio.
>
> As Michael mentioned, there is still time to change the virtio-blk spec
> since this feature hasn't been released yet.
>
> Why exactly is exposing eMMC-style lifetime information problematic?
>
> Can you and Enrico discuss the use case to figure out an alternative
> interface?
>
> Thanks,
> Stefan