netdev - Re: [PATCH net-next 4/4] virtio-net: remove F_GUEST

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20230619154329.GD74977@h68b04307.sqa.eu95>
Date: Mon, 19 Jun 2023 23:43:29 +0800
From: Heng Qi <hengqi@...ux.alibaba.com>
To: "Michael S. Tsirkin" <mst@...hat.com>
Cc: netdev@...r.kernel.org, bpf@...r.kernel.org,
	Jason Wang <jasowang@...hat.com>,
	Xuan Zhuo <xuanzhuo@...ux.alibaba.com>,
	"David S. Miller" <davem@...emloft.net>,
	Eric Dumazet <edumazet@...gle.com>,
	Jakub Kicinski <kuba@...nel.org>, Paolo Abeni <pabeni@...hat.com>,
	Alexei Starovoitov <ast@...nel.org>,
	Daniel Borkmann <daniel@...earbox.net>,
	Jesper Dangaard Brouer <hawk@...nel.org>,
	John Fastabend <john.fastabend@...il.com>
Subject: Re: [PATCH net-next 4/4] virtio-net: remove F_GUEST_CSUM check for
 XDP loading

On Mon, Jun 19, 2023 at 10:33:44AM -0400, Michael S. Tsirkin wrote:
> On Mon, Jun 19, 2023 at 08:41:54PM +0800, Heng Qi wrote:
> > On Mon, Jun 19, 2023 at 07:16:20AM -0400, Michael S. Tsirkin wrote:
> > > On Mon, Jun 19, 2023 at 06:57:38PM +0800, Heng Qi wrote:
> > > > Lay the foundation for the subsequent patch
> > > 
> > > which subsequent patch? this is the last one in series.
> > > 
> > > > to complete the coexistence
> > > > of XDP and virtio-net guest csum.
> > > > 
> > > > Signed-off-by: Heng Qi <hengqi@...ux.alibaba.com>
> > > > Reviewed-by: Xuan Zhuo <xuanzhuo@...ux.alibaba.com>
> > > > ---
> > > >  drivers/net/virtio_net.c | 4 +---
> > > >  1 file changed, 1 insertion(+), 3 deletions(-)
> > > > 
> > > > diff --git a/drivers/net/virtio_net.c b/drivers/net/virtio_net.c
> > > > index 25b486ab74db..79471de64b56 100644
> > > > --- a/drivers/net/virtio_net.c
> > > > +++ b/drivers/net/virtio_net.c
> > > > @@ -60,7 +60,6 @@ static const unsigned long guest_offloads[] = {
> > > >  	VIRTIO_NET_F_GUEST_TSO6,
> > > >  	VIRTIO_NET_F_GUEST_ECN,
> > > >  	VIRTIO_NET_F_GUEST_UFO,
> > > > -	VIRTIO_NET_F_GUEST_CSUM,
> > > >  	VIRTIO_NET_F_GUEST_USO4,
> > > >  	VIRTIO_NET_F_GUEST_USO6,
> > > >  	VIRTIO_NET_F_GUEST_HDRLEN
> > > 
> > > What is this doing? Drop support for VIRTIO_NET_F_GUEST_CSUM? Why?
> > 
> > guest_offloads[] is used by the VIRTIO_NET_CTRL_GUEST_OFFLOADS_SET
> > command to switch features when XDP is loaded/unloaded.
> > 
> > If the VIRTIO_NET_F_CTRL_GUEST_OFFLOADS feature is negotiated:
> > 1. When XDP is loaded, virtnet_xdp_set() uses virtnet_clear_guest_offloads()
> > to automatically turn off the features in guest_offloads[].
> > 
> > 2. when XDP is unloaded, virtnet_xdp_set() uses virtnet_restore_guest_offloads()
> > to automatically restore the features in guest_offloads[].
> > 
> > Now, this work no longer makes XDP and _F_GUEST_CSUM mutually
> > exclusive, so this patch removed the _F_GUEST_CSUM from guest_offloads[].
> > 
> > > This will disable all of guest offloads I think ..
> > 
> > No. This doesn't change the dependencies of other features on
> > _F_GUEST_CSUM. Removing _F_GUEST_CSUM here does not mean that other
> > features that depend on it will be turned off at the same time, such as
> > _F_GUEST_TSO{4,6}, F_GUEST_USO{4,6}, etc.
> > 
> > Thanks.
> 
> Hmm I don't get it.
> 
> static int virtnet_restore_guest_offloads(struct virtnet_info *vi)
> {               
>         u64 offloads = vi->guest_offloads;
>                         
>         if (!vi->guest_offloads)
>                 return 0;
>         
>         return virtnet_set_guest_offloads(vi, offloads); 
> }               
>                         
> is the bit _F_GUEST_CSUM set in vi->guest_offloads?

No, but first we doesn't clear _F_GUEST_CSUM in virtnet_clear_guest_offloads().

If VIRTIO_NET_F_CTRL_GUEST_OFFLOADS is negotiated, features that can
be dynamically controlled by the VIRTIO_NET_CTRL_GUEST_OFFLOADS_SET
command must also be negotiated. Therefore, if GRO_HW_MASK features such
as _F_GUEST_TSO exist, then _F_GUEST_CSUM must exist (according to the
dependencies defined by the spec).

Now, we only dynamically turn off/on the features contained in
guest_offloads[] through XDP loading/unloading (with this patch,
_F_GUEST_CSUM will not be controlled), and _F_GUEST_CSUM is always on.

Another point is that the virtio-net NETIF_F_RXCSUM corresponding to
_F_GUEST_CSUM is only included in dev->features, not in dev->hw_features,
which means that users cannot manually control the switch of
_F_GUEST_CSUM.

> 
> Because if it isn't then we'll try to set _F_GUEST_TSO
> without setting _F_GUEST_CSUM and that's a spec
> violation I think.

As explained above, we did not cause a specification violation.

Thanks.

> 
> 
> > > 
> > > 
> > > > @@ -3522,10 +3521,9 @@ static int virtnet_xdp_set(struct net_device *dev, struct bpf_prog *prog,
> > > >  	        virtio_has_feature(vi->vdev, VIRTIO_NET_F_GUEST_TSO6) ||
> > > >  	        virtio_has_feature(vi->vdev, VIRTIO_NET_F_GUEST_ECN) ||
> > > >  		virtio_has_feature(vi->vdev, VIRTIO_NET_F_GUEST_UFO) ||
> > > > -		virtio_has_feature(vi->vdev, VIRTIO_NET_F_GUEST_CSUM) ||
> > > >  		virtio_has_feature(vi->vdev, VIRTIO_NET_F_GUEST_USO4) ||
> > > >  		virtio_has_feature(vi->vdev, VIRTIO_NET_F_GUEST_USO6))) {
> > > > -		NL_SET_ERR_MSG_MOD(extack, "Can't set XDP while host is implementing GRO_HW/CSUM, disable GRO_HW/CSUM first");
> > > > +		NL_SET_ERR_MSG_MOD(extack, "Can't set XDP while host is implementing GRO_HW, disable GRO_HW first");
> > > >  		return -EOPNOTSUPP;
> > > >  	}
> > > >  
> > > > -- 
> > > > 2.19.1.6.gb485710b