2024-01-15 19:49:11

by Andrew Melnichenko

Subject: [PATCH 1/1] vhost: Added pad cleanup if vnet_hdr is not present.

When the Qemu launched with vhost but without tap vnet_hdr,
vhost tries to copy vnet_hdr from socket iter with size 0
to the page that may contain some trash.
That trash can be interpreted as unpredictable values for
vnet_hdr.
That leads to dropping some packets and in some cases to
stalling vhost routine when the vhost_net tries to process
packets and fails in a loop.

Qemu options:
-netdev tap,vhost=on,vnet_hdr=off,...

Signed-off-by: Andrew Melnychenko <[email protected]>
---
drivers/vhost/net.c | 3 +++
1 file changed, 3 insertions(+)

diff --git a/drivers/vhost/net.c b/drivers/vhost/net.c
index f2ed7167c848..57411ac2d08b 100644
--- a/drivers/vhost/net.c
+++ b/drivers/vhost/net.c
@@ -735,6 +735,9 @@ static int vhost_net_build_xdp(struct vhost_net_virtqueue *nvq,
hdr = buf;
gso = &hdr->gso;

+ if (!sock_hlen)
+ memset(buf, 0, pad);
+
if ((gso->flags & VIRTIO_NET_HDR_F_NEEDS_CSUM) &&
vhost16_to_cpu(vq, gso->csum_start) +
vhost16_to_cpu(vq, gso->csum_offset) + 2 >
--
2.43.0

2024-01-15 21:45:10

by Yuri Benditovich

Subject: Re: [PATCH 1/1] vhost: Added pad cleanup if vnet_hdr is not present.

See also https://issues.redhat.com/browse/RHEL-1303


On Mon, Jan 15, 2024 at 9:48 PM Andrew Melnychenko <[email protected]> wrote:
>
> When the Qemu launched with vhost but without tap vnet_hdr,
> vhost tries to copy vnet_hdr from socket iter with size 0
> to the page that may contain some trash.
> That trash can be interpreted as unpredictable values for
> vnet_hdr.
> That leads to dropping some packets and in some cases to
> stalling vhost routine when the vhost_net tries to process
> packets and fails in a loop.
>
> Qemu options:
> -netdev tap,vhost=on,vnet_hdr=off,...
>
> Signed-off-by: Andrew Melnychenko <[email protected]>
> ---
> drivers/vhost/net.c | 3 +++
> 1 file changed, 3 insertions(+)
>
> diff --git a/drivers/vhost/net.c b/drivers/vhost/net.c
> index f2ed7167c848..57411ac2d08b 100644
> --- a/drivers/vhost/net.c
> +++ b/drivers/vhost/net.c
> @@ -735,6 +735,9 @@ static int vhost_net_build_xdp(struct vhost_net_virtqueue *nvq,
> hdr = buf;
> gso = &hdr->gso;
>
> + if (!sock_hlen)
> + memset(buf, 0, pad);
> +
> if ((gso->flags & VIRTIO_NET_HDR_F_NEEDS_CSUM) &&
> vhost16_to_cpu(vq, gso->csum_start) +
> vhost16_to_cpu(vq, gso->csum_offset) + 2 >
> --
> 2.43.0
>

2024-01-15 22:32:49

by Michael S. Tsirkin

Subject: Re: [PATCH 1/1] vhost: Added pad cleanup if vnet_hdr is not present.

On Mon, Jan 15, 2024 at 09:48:40PM +0200, Andrew Melnychenko wrote:
> When the Qemu launched with vhost but without tap vnet_hdr,
> vhost tries to copy vnet_hdr from socket iter with size 0
> to the page that may contain some trash.
> That trash can be interpreted as unpredictable values for
> vnet_hdr.
> That leads to dropping some packets and in some cases to
> stalling vhost routine when the vhost_net tries to process
> packets and fails in a loop.
>
> Qemu options:
> -netdev tap,vhost=on,vnet_hdr=off,...
>
> Signed-off-by: Andrew Melnychenko <[email protected]>
> ---
> drivers/vhost/net.c | 3 +++
> 1 file changed, 3 insertions(+)
>
> diff --git a/drivers/vhost/net.c b/drivers/vhost/net.c
> index f2ed7167c848..57411ac2d08b 100644
> --- a/drivers/vhost/net.c
> +++ b/drivers/vhost/net.c
> @@ -735,6 +735,9 @@ static int vhost_net_build_xdp(struct vhost_net_virtqueue *nvq,
> hdr = buf;
> gso = &hdr->gso;
>
> + if (!sock_hlen)
> + memset(buf, 0, pad);
> +
> if ((gso->flags & VIRTIO_NET_HDR_F_NEEDS_CSUM) &&
> vhost16_to_cpu(vq, gso->csum_start) +
> vhost16_to_cpu(vq, gso->csum_offset) + 2 >


Hmm need to analyse it to make sure there are no cases where we leak
some data to guest here in case where sock_hlen is set ...
> --
> 2.43.0

2024-02-08 15:35:13

by Yuri Benditovich

Subject: Re: [PATCH 1/1] vhost: Added pad cleanup if vnet_hdr is not present.

Just polite ping

On Tue, Jan 16, 2024 at 12:32 AM Michael S. Tsirkin <[email protected]> wrote:
>
> On Mon, Jan 15, 2024 at 09:48:40PM +0200, Andrew Melnychenko wrote:
> > When the Qemu launched with vhost but without tap vnet_hdr,
> > vhost tries to copy vnet_hdr from socket iter with size 0
> > to the page that may contain some trash.
> > That trash can be interpreted as unpredictable values for
> > vnet_hdr.
> > That leads to dropping some packets and in some cases to
> > stalling vhost routine when the vhost_net tries to process
> > packets and fails in a loop.
> >
> > Qemu options:
> > -netdev tap,vhost=on,vnet_hdr=off,...
> >
> > Signed-off-by: Andrew Melnychenko <[email protected]>
> > ---
> > drivers/vhost/net.c | 3 +++
> > 1 file changed, 3 insertions(+)
> >
> > diff --git a/drivers/vhost/net.c b/drivers/vhost/net.c
> > index f2ed7167c848..57411ac2d08b 100644
> > --- a/drivers/vhost/net.c
> > +++ b/drivers/vhost/net.c
> > @@ -735,6 +735,9 @@ static int vhost_net_build_xdp(struct vhost_net_virtqueue *nvq,
> > hdr = buf;
> > gso = &hdr->gso;
> >
> > + if (!sock_hlen)
> > + memset(buf, 0, pad);
> > +
> > if ((gso->flags & VIRTIO_NET_HDR_F_NEEDS_CSUM) &&
> > vhost16_to_cpu(vq, gso->csum_start) +
> > vhost16_to_cpu(vq, gso->csum_offset) + 2 >
>
>
> Hmm need to analyse it to make sure there are no cases where we leak
> some data to guest here in case where sock_hlen is set ...
> > --
> > 2.43.0
>

2024-02-22 20:03:09

by Michael S. Tsirkin

Subject: Re: [PATCH 1/1] vhost: Added pad cleanup if vnet_hdr is not present.

On Mon, Jan 15, 2024 at 05:32:25PM -0500, Michael S. Tsirkin wrote:
> On Mon, Jan 15, 2024 at 09:48:40PM +0200, Andrew Melnychenko wrote:
> > When the Qemu launched with vhost but without tap vnet_hdr,
> > vhost tries to copy vnet_hdr from socket iter with size 0
> > to the page that may contain some trash.
> > That trash can be interpreted as unpredictable values for
> > vnet_hdr.
> > That leads to dropping some packets and in some cases to
> > stalling vhost routine when the vhost_net tries to process
> > packets and fails in a loop.
> >
> > Qemu options:
> > -netdev tap,vhost=on,vnet_hdr=off,...
> >
> > Signed-off-by: Andrew Melnychenko <[email protected]>
> > ---
> > drivers/vhost/net.c | 3 +++
> > 1 file changed, 3 insertions(+)
> >
> > diff --git a/drivers/vhost/net.c b/drivers/vhost/net.c
> > index f2ed7167c848..57411ac2d08b 100644
> > --- a/drivers/vhost/net.c
> > +++ b/drivers/vhost/net.c
> > @@ -735,6 +735,9 @@ static int vhost_net_build_xdp(struct vhost_net_virtqueue *nvq,
> > hdr = buf;
> > gso = &hdr->gso;
> >
> > + if (!sock_hlen)
> > + memset(buf, 0, pad);
> > +
> > if ((gso->flags & VIRTIO_NET_HDR_F_NEEDS_CSUM) &&
> > vhost16_to_cpu(vq, gso->csum_start) +
> > vhost16_to_cpu(vq, gso->csum_offset) + 2 >
>
>
> Hmm need to analyse it to make sure there are no cases where we leak
> some data to guest here in case where sock_hlen is set ...


Could you post this analysis pls?

> > --
> > 2.43.0

2024-02-26 16:36:38

by Andrew Melnichenko

Subject: Re: [PATCH 1/1] vhost: Added pad cleanup if vnet_hdr is not present.

Hi all,
Ok, let me prepare a new patch v2, where I'll write a
description/analysis of the issue in the commit message.

On Thu, Feb 22, 2024 at 10:02 PM Michael S. Tsirkin <[email protected]> wrote:
>
> On Mon, Jan 15, 2024 at 05:32:25PM -0500, Michael S. Tsirkin wrote:
> > On Mon, Jan 15, 2024 at 09:48:40PM +0200, Andrew Melnychenko wrote:
> > > When the Qemu launched with vhost but without tap vnet_hdr,
> > > vhost tries to copy vnet_hdr from socket iter with size 0
> > > to the page that may contain some trash.
> > > That trash can be interpreted as unpredictable values for
> > > vnet_hdr.
> > > That leads to dropping some packets and in some cases to
> > > stalling vhost routine when the vhost_net tries to process
> > > packets and fails in a loop.
> > >
> > > Qemu options:
> > > -netdev tap,vhost=on,vnet_hdr=off,...
> > >
> > > Signed-off-by: Andrew Melnychenko <[email protected]>
> > > ---
> > > drivers/vhost/net.c | 3 +++
> > > 1 file changed, 3 insertions(+)
> > >
> > > diff --git a/drivers/vhost/net.c b/drivers/vhost/net.c
> > > index f2ed7167c848..57411ac2d08b 100644
> > > --- a/drivers/vhost/net.c
> > > +++ b/drivers/vhost/net.c
> > > @@ -735,6 +735,9 @@ static int vhost_net_build_xdp(struct vhost_net_virtqueue *nvq,
> > > hdr = buf;
> > > gso = &hdr->gso;
> > >
> > > + if (!sock_hlen)
> > > + memset(buf, 0, pad);
> > > +
> > > if ((gso->flags & VIRTIO_NET_HDR_F_NEEDS_CSUM) &&
> > > vhost16_to_cpu(vq, gso->csum_start) +
> > > vhost16_to_cpu(vq, gso->csum_offset) + 2 >
> >
> >
> > Hmm need to analyse it to make sure there are no cases where we leak
> > some data to guest here in case where sock_hlen is set ...
>
>
> Could you post this analysis pls?
>
> > > --
> > > 2.43.0
>