Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
MIME-Version: 1.0
References: <20230322141031.2211141-1-viktor@daynix.com> <20230322123121-mutt-send-email-mst@kernel.org>
 <87mt44hh5f.fsf@redhat.com>
In-Reply-To: <87mt44hh5f.fsf@redhat.com>
From:   Viktor Prutyanov <viktor@daynix.com>
Date:   Wed, 22 Mar 2023 20:21:12 +0300
Message-ID: <CAPv0NP79iv=s-tdTWwb_3cWbucJSH7LTDYo0VxjdnWvpjguK3Q@mail.gmail.com>
Subject: Re: [PATCH v4] virtio: add VIRTIO_F_NOTIFICATION_DATA feature support
To:     Cornelia Huck <cohuck@redhat.com>
Cc:     "Michael S. Tsirkin" <mst@redhat.com>, jasowang@redhat.com,
        pasic@linux.ibm.com, farman@linux.ibm.com,
        linux-s390@vger.kernel.org, kvm@vger.kernel.org,
        virtualization@lists.linux-foundation.org,
        linux-kernel@vger.kernel.org, yan@daynix.com
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
Precedence: bulk

On Wed, Mar 22, 2023 at 7:42=E2=80=AFPM Cornelia Huck <cohuck@redhat.com> w=
rote:
>
> On Wed, Mar 22 2023, "Michael S. Tsirkin" <mst@redhat.com> wrote:
>
> > On Wed, Mar 22, 2023 at 05:10:31PM +0300, Viktor Prutyanov wrote:
> >> According to VirtIO spec v1.2, VIRTIO_F_NOTIFICATION_DATA feature
> >> indicates that the driver passes extra data along with the queue
> >> notifications.
> >>
> >> In a split queue case, the extra data is 16-bit available index. In a
> >> packed queue case, the extra data is 1-bit wrap counter and 15-bit
> >> available index.
> >>
> >> Add support for this feature for MMIO, channel I/O and modern PCI
> >> transports.
> >>
> >> Signed-off-by: Viktor Prutyanov <viktor@daynix.com>
> >> ---
> >>  v4: remove VP_NOTIFY macro and legacy PCI support, add
> >>     virtio_ccw_kvm_notify_with_data to virtio_ccw
> >>  v3: support feature in virtio_ccw, remove VM_NOTIFY, use avail_idx_sh=
adow,
> >>     remove byte swap, rename to vring_notification_data
> >>  v2: reject the feature in virtio_ccw, replace __le32 with u32
> >>
> >>  Tested with disabled VIRTIO_F_NOTIFICATION_DATA on qemu-system-s390x
> >>  (virtio-blk-ccw), qemu-system-riscv64 (virtio-blk-device,
> >>  virtio-rng-device), qemu-system-x86_64 (virtio-blk-pci, virtio-net-pc=
i)
> >>  to make sure nothing is broken.
> >>  Tested with enabled VIRTIO_F_NOTIFICATION_DATA on 64-bit RISC-V Linux
> >>  and my hardware implementation of virtio-rng.
> >
> > what did you test? virtio pci? mmio? guessing not ccw...
> >
> > Cornelia could you hack up something to quickly test ccw?
>
> Hm, I'm not entirely sure how notification data is supposed to be used
> in real life -- Viktor, what is your virtio-rng implementation doing;
> can this be hacked into all transports?

In hardware implementation of split VirtIO queue, the notification data fea=
ture
saves time on a memory request to avail_idx from the device side. I can
definitely say about MMIO, but most likely it is also useful for PCI.
It is also written here that this feature was added for efficiency:
https://www.mail-archive.com/virtio-dev@lists.oasis-open.org/msg02728.html
Also DPDK has support for this feature in VirtIO PCI:
https://github.com/DPDK/dpdk/commit/7e72f3ec1a8abefd9321a61e484846e16177f5b=
1

>
> (Also, if the other ccw folks have something handy, please speak up :)
>