Received-SPF: pass (google.com: domain of linux-kernel+bounces-6336-linux.lists.archive=gmail.com@vger.kernel.org designates 139.178.88.99 as permitted sender) client-ip=139.178.88.99;
Precedence: bulk
MIME-Version: 1.0
References: <20231212131712.1816324-1-maxime.coquelin@redhat.com>
 <20231212131712.1816324-3-maxime.coquelin@redhat.com> <CACGkMEthp13a20TGashiFNDovK+b10mgfdX8L=3Xv05g5-eo0w@mail.gmail.com>
 <8afc44f7-cef5-4b4a-afac-85a4eb967d3e@redhat.com> <CACGkMEtNZyHDzo2e_N9SM9dw4CK8vpZHWBrk7TFFixD1T7cdjw@mail.gmail.com>
 <b793a98c-dfde-4bf6-9057-6fa359cf4723@redhat.com>
In-Reply-To: <b793a98c-dfde-4bf6-9057-6fa359cf4723@redhat.com>
From: Jason Wang <jasowang@redhat.com>
Date: Wed, 20 Dec 2023 11:50:20 +0800
Message-ID: <CACGkMEsUcVazoagKuTuFQdn55khyyDwOnYKLJuOyyt8_ib7wiA@mail.gmail.com>
Subject: Re: [PATCH v5 2/4] vduse: Temporarily disable control queue features
To: Maxime Coquelin <maxime.coquelin@redhat.com>
Cc: mst@redhat.com, xuanzhuo@linux.alibaba.com, paul@paul-moore.com, 
	jmorris@namei.org, serge@hallyn.com, stephen.smalley.work@gmail.com, 
	eparis@parisplace.org, xieyongji@bytedance.com, 
	virtualization@lists.linux-foundation.org, linux-kernel@vger.kernel.org, 
	linux-security-module@vger.kernel.org, selinux@vger.kernel.org, 
	david.marchand@redhat.com, lulu@redhat.com, casey@schaufler-ca.com
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

On Mon, Dec 18, 2023 at 5:21=E2=80=AFPM Maxime Coquelin
<maxime.coquelin@redhat.com> wrote:
>
>
>
> On 12/18/23 03:50, Jason Wang wrote:
> > On Wed, Dec 13, 2023 at 7:23=E2=80=AFPM Maxime Coquelin
> > <maxime.coquelin@redhat.com> wrote:
> >>
> >> Hi Jason,
> >>
> >> On 12/13/23 05:52, Jason Wang wrote:
> >>> On Tue, Dec 12, 2023 at 9:17=E2=80=AFPM Maxime Coquelin
> >>> <maxime.coquelin@redhat.com> wrote:
> >>>>
> >>>> Virtio-net driver control queue implementation is not safe
> >>>> when used with VDUSE. If the VDUSE application does not
> >>>> reply to control queue messages, it currently ends up
> >>>> hanging the kernel thread sending this command.
> >>>>
> >>>> Some work is on-going to make the control queue
> >>>> implementation robust with VDUSE. Until it is completed,
> >>>> let's disable control virtqueue and features that depend on
> >>>> it.
> >>>>
> >>>> Signed-off-by: Maxime Coquelin <maxime.coquelin@redhat.com>
> >>>
> >>> I wonder if it's better to fail instead of a mask as a start.
> >>
> >> I think it is better to use a mask and not fail, so that we can in the
> >> future use a recent VDUSE application with an older kernel.
> >
> > It may confuse the userspace unless userspace can do post check after
> > CREATE_DEV.
> >
> > And for blk we fail when WCE is set in feature_is_valid():
> >
> > static bool features_is_valid(u64 features)
> > {
> >          if (!(features & (1ULL << VIRTIO_F_ACCESS_PLATFORM)))
> >                  return false;
> >
> >          /* Now we only support read-only configuration space */
> >          if (features & (1ULL << VIRTIO_BLK_F_CONFIG_WCE))
> >                  return false;
> >
> >          return true;
> > }
>
> Ok, consistency with other devices types is indeed better.
>
> But should I fail if any of the feature advertised by the application is
> not listed by the VDUSE driver, or just fail if control queue is being
> advertised by the application?

Maybe it's better to fail for any other of the features that depend on
the control vq.

Thanks

>
> Thanks,
> Maxime
>
> > Thanks
> >
> >>
> >> Why would it be better to fail than negotiating?
> >>
> >> Thanks,
> >> Maxime
> >>
> >
>