Linux Virtualization - Sep 2019 - [PATCH v2] vhost: block speculation of translated descriptors

On Wed, Sep 11, 2019 at 02:16:28PM +0200, Michal Hocko
wrote:
> On Wed 11-09-19 08:10:00, Michael S. Tsirkin wrote:
> > iovec addresses coming from vhost are assumed to be
> > pre-validated, but in fact can be speculated to a value
> > out of range.
> > 
> > Userspace address are later validated with array_index_nospec so we
can
> > be sure kernel info does not leak through these addresses, but vhost
> > must also not leak userspace info outside the allowed memory table to
> > guests.
> > 
> > Following the defence in depth principle, make sure
> > the address is not validated out of node range.
> > 
> > Signed-off-by: Michael S. Tsirkin <mst at redhat.com>
> > Acked-by: Jason Wang <jasowang at redhat.com>
> > Tested-by: Jason Wang <jasowang at redhat.com>
> 
> no need to mark fo stable? Other spectre fixes tend to be backported
> even when the security implications are not really clear. The risk
> should be low and better to be covered in case.
This is not really a fix - more a defence in depth thing,
quite similar to e.g.  commit b3bbfb3fb5d25776b8e3f361d2eedaabb0b496cd
x86: Introduce __uaccess_begin_nospec() and uaccess_try_nospec
in scope.

That one doesn't seem to be tagged for stable. Was it queued
there in practice?
> > ---
> > 
> > changes from v1: fix build on 32 bit
> > 
> >  drivers/vhost/vhost.c | 6 ++++--
> >  1 file changed, 4 insertions(+), 2 deletions(-)
> > 
> > diff --git a/drivers/vhost/vhost.c b/drivers/vhost/vhost.c
> > index 5dc174ac8cac..34ea219936e3 100644
> > --- a/drivers/vhost/vhost.c
> > +++ b/drivers/vhost/vhost.c
> > @@ -2071,8 +2071,10 @@ static int translate_desc(struct
vhost_virtqueue *vq, u64 addr, u32 len,
> >  		_iov = iov + ret;
> >  		size = node->size - addr + node->start;
> >  		_iov->iov_len = min((u64)len - s, size);
> > -		_iov->iov_base = (void __user *)(unsigned long)
> > -			(node->userspace_addr + addr - node->start);
> > +		_iov->iov_base = (void __user *)
> > +			((unsigned long)node->userspace_addr +
> > +			 array_index_nospec((unsigned long)(addr - node->start),
> > +					    node->size));
> >  		s += size;
> >  		addr += size;
> >  		++ret;
> > -- 
> > MST
> 
> -- 
> Michal Hocko
> SUSE Labs

On Wed 11-09-19 08:25:03, Michael S. Tsirkin wrote:
> On Wed, Sep 11, 2019 at 02:16:28PM +0200, Michal Hocko wrote:
> > On Wed 11-09-19 08:10:00, Michael S. Tsirkin wrote:
> > > iovec addresses coming from vhost are assumed to be
> > > pre-validated, but in fact can be speculated to a value
> > > out of range.
> > > 
> > > Userspace address are later validated with array_index_nospec so
we can
> > > be sure kernel info does not leak through these addresses, but
vhost
> > > must also not leak userspace info outside the allowed memory
table to
> > > guests.
> > > 
> > > Following the defence in depth principle, make sure
> > > the address is not validated out of node range.
> > > 
> > > Signed-off-by: Michael S. Tsirkin <mst at redhat.com>
> > > Acked-by: Jason Wang <jasowang at redhat.com>
> > > Tested-by: Jason Wang <jasowang at redhat.com>
> > 
> > no need to mark fo stable? Other spectre fixes tend to be backported
> > even when the security implications are not really clear. The risk
> > should be low and better to be covered in case.
> 
> This is not really a fix - more a defence in depth thing,
> quite similar to e.g.  commit b3bbfb3fb5d25776b8e3f361d2eedaabb0b496cd
> x86: Introduce __uaccess_begin_nospec() and uaccess_try_nospec
> in scope.
>
> That one doesn't seem to be tagged for stable. Was it queued
> there in practice?
not marked for stable but it went in. At least to 4.4.

-- 
Michal Hocko
SUSE Labs

On Wed, Sep 11, 2019 at 02:33:16PM +0200, Michal Hocko
wrote:
> On Wed 11-09-19 08:25:03, Michael S. Tsirkin wrote:
> > On Wed, Sep 11, 2019 at 02:16:28PM +0200, Michal Hocko wrote:
> > > On Wed 11-09-19 08:10:00, Michael S. Tsirkin wrote:
> > > > iovec addresses coming from vhost are assumed to be
> > > > pre-validated, but in fact can be speculated to a value
> > > > out of range.
> > > > 
> > > > Userspace address are later validated with
array_index_nospec so we can
> > > > be sure kernel info does not leak through these addresses,
but vhost
> > > > must also not leak userspace info outside the allowed memory
table to
> > > > guests.
> > > > 
> > > > Following the defence in depth principle, make sure
> > > > the address is not validated out of node range.
> > > > 
> > > > Signed-off-by: Michael S. Tsirkin <mst at redhat.com>
> > > > Acked-by: Jason Wang <jasowang at redhat.com>
> > > > Tested-by: Jason Wang <jasowang at redhat.com>
> > > 
> > > no need to mark fo stable? Other spectre fixes tend to be
backported
> > > even when the security implications are not really clear. The
risk
> > > should be low and better to be covered in case.
> > 
> > This is not really a fix - more a defence in depth thing,
> > quite similar to e.g.  commit b3bbfb3fb5d25776b8e3f361d2eedaabb0b496cd
> > x86: Introduce __uaccess_begin_nospec() and uaccess_try_nospec
> > in scope.
> >
> > That one doesn't seem to be tagged for stable. Was it queued
> > there in practice?
> 
> not marked for stable but it went in. At least to 4.4.
So I guess the answer is I don't know. If you feel it's
justified, then sure, feel free to forward.

-- 
MST