Virtualization - Oct 2021 - [PATCH 1/1] virtio: write back F_VERSION_1 before validate

On Wed, Oct 06 2021, Halil Pasic <pasic at linux.ibm.com> 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: driver MAY read (but MUST NOT
> write) the device-specific configuration fields to check that it can
> support the device before setting FEATURES_OK.
Suggest to put the citations from the spec into quotes, so that they are
distinguishable from the rest of the text.
>
> 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 regression so let's try to
address
s/regression/a regression/
> it.
Maybe mention what the regression is?

Also mention that we use this workaround for modern on BE only?
>
> Signed-off-by: Halil Pasic <pasic at linux.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 at us.ibm.com
> ---
>  drivers/virtio/virtio.c | 10 ++++++++++
>  1 file changed, 10 insertions(+)
>
> diff --git a/drivers/virtio/virtio.c b/drivers/virtio/virtio.c
> index 0a5b54034d4b..494cfecd3376 100644
> --- a/drivers/virtio/virtio.c
> +++ b/drivers/virtio/virtio.c
> @@ -239,6 +239,16 @@ 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 for these when needed.
"...to force LE config space accesses before FEATURES_OK for these when
needed (BE)."

?
> +	 */
> +	if (drv->validate && !virtio_legacy_is_little_endian()
> +			  && BIT_ULL(VIRTIO_F_VERSION_1) & device_features) {
Nit: putting device_features first would read more naturally to me.
> +		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
Patch LGTM.

On Thu, 07 Oct 2021 13:52:24 +0200
Cornelia Huck <cohuck at redhat.com> wrote:
> On Wed, Oct 06 2021, Halil Pasic <pasic at linux.ibm.com> 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: driver MAY read (but MUST NOT
> > write) the device-specific configuration fields to check that it can
> > support the device before setting FEATURES_OK.  
> 
> Suggest to put the citations from the spec into quotes, so that they are
> distinguishable from the rest of the text.
For the record: I basically took Michael's description, the one which you
said you prefer, with some minor changes.

This is one of the changes, which renders this a paraphrase and not a
quote. Michael didn't use quotation marks so I was not sure it is was
a word by word quote anyway. It was. But the spec depends on "During this
step" which does not make any sense without the context. That is why I made
the end of step explicit.

I think we are fine without quotation marks. Those who care can read the
spec.
> 
> >
> > 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 regression so let's try to
address
> 
> s/regression/a regression/
> 
Yes. Was like this in the original. Will change
> > it.  
> 
> Maybe mention what the regression is?
How about the following?

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.
> 
> Also mention that we use this workaround for modern on BE only?
We have that already, don't we. The sentence that starts with "In
particular". The regression description should reinforce that
sufficiently IMHO.
> 
> >
> > Signed-off-by: Halil Pasic <pasic at linux.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 at us.ibm.com
> > ---
> >  drivers/virtio/virtio.c | 10 ++++++++++
> >  1 file changed, 10 insertions(+)
> >
> > diff --git a/drivers/virtio/virtio.c b/drivers/virtio/virtio.c
> > index 0a5b54034d4b..494cfecd3376 100644
> > --- a/drivers/virtio/virtio.c
> > +++ b/drivers/virtio/virtio.c
> > @@ -239,6 +239,16 @@ 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 for these when needed.  
> 
> "...to force LE config space accesses before FEATURES_OK for these
when
> needed (BE)."
> 
> ?
Can do, but I would rather omit the (BE) at the end. All the conditions
are necessary:
* have validate callback
* device offered VERSION_1
* virtio legacy is be
> 
> > +	 */
> > +	if (drv->validate && !virtio_legacy_is_little_endian()
> > +			  && BIT_ULL(VIRTIO_F_VERSION_1) & device_features) {
> 
> Nit: putting device_features first would read more naturally to me.
> 
Can do.
> > +		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  
> 
> Patch LGTM.
> 
> 
Thanks for having a look. If you are fine with the proposed solution
please tell me, so I can send out a v2.

If not let us work towards an acceptable solution.

Regards,
Halil