Message-ID: <20211008155156.626e78b5.pasic@linux.ibm.com>
Date:   Fri, 8 Oct 2021 15:51:56 +0200
From:   Halil Pasic <pasic@...ux.ibm.com>
To:     "Michael S. Tsirkin" <mst@...hat.com>
Cc:     Jason Wang <jasowang@...hat.com>,
        Xie Yongji <xieyongji@...edance.com>,
        virtualization@...ts.linux-foundation.org,
        linux-kernel@...r.kernel.org, stable@...r.kernel.org,
        markver@...ibm.com, Cornelia Huck <cohuck@...hat.com>,
        Christian Borntraeger <borntraeger@...ibm.com>,
        linux-s390@...r.kernel.org, stefanha@...hat.com,
        Raphael Norwitz <raphael.norwitz@...anix.com>,
        qemu-devel@...gnu.org, Halil Pasic <pasic@...ux.ibm.com>
Subject: Re: [PATCH v2 1/1] virtio: write back F_VERSION_1 before validate

On Fri, 8 Oct 2021 09:05:03 -0400
"Michael S. Tsirkin" <mst@...hat.com> wrote:

> On Fri, Oct 08, 2021 at 02:34:22PM +0200, Halil Pasic wrote:
> > The virtio specification virtio-v1.1-cs01 states: "Transitional devices
> > MUST detect Legacy drivers by detecting that VIRTIO_F_VERSION_1 has not
> > been acknowledged by the driver."  This is exactly what QEMU as of 6.1
> > has done relying solely on VIRTIO_F_VERSION_1 for detecting that.
> > 
> > However, the specification also says: "... the driver MAY read (but MUST
> > NOT write) the device-specific configuration fields to check that it can
> > support the device ..." before setting FEATURES_OK.
> > 
> > In that case, any transitional device relying solely on
> > VIRTIO_F_VERSION_1 for detecting legacy drivers will return data in
> > legacy format.  In particular, this implies that it is in big endian
> > format for big endian guests. This naturally confuses the driver which
> > expects little endian in the modern mode.
> > 
> > It is probably a good idea to amend the spec to clarify that
> > VIRTIO_F_VERSION_1 can only be relied on after the feature negotiation
> > is complete. However, we already have a regression so let's try to address  
> 
> actually, regressions. and we can add 
> "since originally before validate callback existed
> config space was only read after
> FEATURES_OK. See Fixes tags for relevant commits"
> 
> > it.

How about replacing the paragraph above with the following?

"It is probably a good idea to amend the spec to clarify that
VIRTIO_F_VERSION_1 can only be relied on after the feature negotiation
is complete. Before validate callback existed, config space was only
read after FEATURES_OK. However, we already have two regression, so
let's address this here as well."
> > 
> > The regressions affect the VIRTIO_NET_F_MTU feature of virtio-net and
> > the VIRTIO_BLK_F_BLK_SIZE feature of virtio-blk for BE guests when
> > virtio 1.0 is used on both sides. The latter renders virtio-blk
> > unusable with DASD backing, because things simply don't work with
> > the default.  

and add 
"See Fixes tags for relevant commits."
here.
> 
> Let's add a work around description now:
> 
> 
> For QEMU, we can work around the issue by writing out the features
> register with VIRTIO_F_VERSION_1 bit set.  We (ab) use the
s/features register/feature bits/
rationale: ccw does not have a features register, and qemu does not
really act as if its behavior was controlled by the values in a features
register. I.e. when we read the register we see VIRTIO_F_VERSION_!
because the feature is offered. In QEMU we basically read host_featues
but write the guest_features. And what drives device behavior is mostly
guest_features. 

s/(ab) use/(ab)use/

> finalize_features config op for this. It's not enough to address vhost

s/It's/This is/

> user and vhost block devices since these do not get the features until

s/vhost user and vhost block/some vhost-user and vhost-vdpa/ ?

Ratioale: I think vhost block is just a vhost-user device. On the other
hand vhost-user-fs works like charm because the config space is
implemented in qemu and not in the vhost-user device. I
didn't check vhost_net. I'm not even sure qemu offers a vhost_net
implementation. Anyway I wouldn't like to make any false statements here.

> FEATURES_OK, however it looks like these two actually never handled the
> endian-ness for legacy mode correctly, so at least that's not a
> regression.
> 
> No devices except virtio net and virtio blk seem to be affected.
> 
> Long term the right thing to do is to fix the hypervisors.
> 

Sounds good. Thanks! Are you OK with my changes proposed to your changes?

Regards,
Halil
> 
> > 
> > Cc: <stable@...r.kernel.org> #v4.11
> > Signed-off-by: Halil Pasic <pasic@...ux.ibm.com>
> > Fixes: 82e89ea077b9 ("virtio-blk: Add validation for block size in
> > config space") Fixes: fe36cbe0671e ("virtio_net: clear MTU when out
> > of range") Reported-by: markver@...ibm.com
> > ---
> >  drivers/virtio/virtio.c | 11 +++++++++++
> >  1 file changed, 11 insertions(+)
> > 
> > diff --git a/drivers/virtio/virtio.c b/drivers/virtio/virtio.c
> > index 0a5b54034d4b..236081afe9a2 100644
> > --- a/drivers/virtio/virtio.c
> > +++ b/drivers/virtio/virtio.c
> > @@ -239,6 +239,17 @@ static int virtio_dev_probe(struct device *_d)
> >  		driver_features_legacy = driver_features;
> >  	}
> >  
> > +	/*
> > +	 * Some devices detect legacy solely via F_VERSION_1. Write
> > +	 * F_VERSION_1 to force LE config space accesses before
> > FEATURES_OK for
> > +	 * these when needed.
> > +	 */
> > +	if (drv->validate && !virtio_legacy_is_little_endian()
> > +			  && device_features &
> > BIT_ULL(VIRTIO_F_VERSION_1)) {
> > +		dev->features = BIT_ULL(VIRTIO_F_VERSION_1);
> > +		dev->config->finalize_features(dev);
> > +	}
> > +
> >  	if (device_features & (1ULL << VIRTIO_F_VERSION_1))
> >  		dev->features = driver_features & device_features;
> >  	else
> > 
> > base-commit: 60a9483534ed0d99090a2ee1d4bb0b8179195f51
> > -- 
> > 2.25.1  
> 

Powered by blists - more mailing lists