Re: [PATCH v2 00/15] vfio: expose virtual Shared Virtual Addressing to VMs

[Stefan Hajnoczi <stefanha@gmail.com>]

[-- Attachment #1: Type: text/plain, Size: 5729 bytes --]

On Tue, Jun 16, 2020 at 10:00:16AM -0700, Raj, Ashok wrote:
> On Tue, Jun 16, 2020 at 04:49:28PM +0100, Stefan Hajnoczi wrote:
> > On Tue, Jun 16, 2020 at 02:26:38AM +0000, Tian, Kevin wrote:
> > > > From: Stefan Hajnoczi <stefanha@gmail.com>
> > > > Sent: Monday, June 15, 2020 6:02 PM
> > > > 
> > > > On Thu, Jun 11, 2020 at 05:15:19AM -0700, Liu Yi L wrote:
> > > > > Shared Virtual Addressing (SVA), a.k.a, Shared Virtual Memory (SVM) on
> > > > > Intel platforms allows address space sharing between device DMA and
> > > > > applications. SVA can reduce programming complexity and enhance
> > > > security.
> > > > >
> > > > > This VFIO series is intended to expose SVA usage to VMs. i.e. Sharing
> > > > > guest application address space with passthru devices. This is called
> > > > > vSVA in this series. The whole vSVA enabling requires QEMU/VFIO/IOMMU
> > > > > changes. For IOMMU and QEMU changes, they are in separate series (listed
> > > > > in the "Related series").
> > > > >
> > > > > The high-level architecture for SVA virtualization is as below, the key
> > > > > design of vSVA support is to utilize the dual-stage IOMMU translation (
> > > > > also known as IOMMU nesting translation) capability in host IOMMU.
> > > > >
> > > > >
> > > > >     .-------------.  .---------------------------.
> > > > >     |   vIOMMU    |  | Guest process CR3, FL only|
> > > > >     |             |  '---------------------------'
> > > > >     .----------------/
> > > > >     | PASID Entry |--- PASID cache flush -
> > > > >     '-------------'                       |
> > > > >     |             |                       V
> > > > >     |             |                CR3 in GPA
> > > > >     '-------------'
> > > > > Guest
> > > > > ------| Shadow |--------------------------|--------
> > > > >       v        v                          v
> > > > > Host
> > > > >     .-------------.  .----------------------.
> > > > >     |   pIOMMU    |  | Bind FL for GVA-GPA  |
> > > > >     |             |  '----------------------'
> > > > >     .----------------/  |
> > > > >     | PASID Entry |     V (Nested xlate)
> > > > >     '----------------\.------------------------------.
> > > > >     |             |   |SL for GPA-HPA, default domain|
> > > > >     |             |   '------------------------------'
> > > > >     '-------------'
> > > > > Where:
> > > > >  - FL = First level/stage one page tables
> > > > >  - SL = Second level/stage two page tables
> > > > 
> > > > Hi,
> > > > Looks like an interesting feature!
> > > > 
> > > > To check I understand this feature: can applications now pass virtual
> > > > addresses to devices instead of translating to IOVAs?
> > > > 
> > > > If yes, can guest applications restrict the vSVA address space so the
> > > > device only has access to certain regions?
> > > > 
> > > > On one hand replacing IOVA translation with virtual addresses simplifies
> > > > the application programming model, but does it give up isolation if the
> > > > device can now access all application memory?
> > > > 
> > > 
> > > with SVA each application is allocated with a unique PASID to tag its
> > > virtual address space. The device that claims SVA support must guarantee 
> > > that one application can only program the device to access its own virtual
> > > address space (i.e. all DMAs triggered by this application are tagged with
> > > the application's PASID, and are translated by IOMMU's PASID-granular
> > > page table). So, isolation is not sacrificed in SVA.
> > 
> > Isolation between applications is preserved but there is no isolation
> > between the device and the application itself. The application needs to
> > trust the device.
> 
> Right. With all convenience comes security trust. With SVA there is an
> expectation that the device has the required security boundaries properly
> implemented. FWIW, what is our guarantee today that VF's are secure from
> one another or even its own PF? They can also generate transactions with
> any of its peer id's and there is nothing an IOMMU can do today. Other than
> rely on ACS. Even BusMaster enable can be ignored and devices (malicious
> or otherwise) can generate after the BM=0. With SVM you get the benefits of
> 
> * Not having to register regions
> * Don't need to pin application space for DMA.

As along as the security model is clearly documented users can decide
whether or not SVA meets their requirements. I just wanted to clarify
what the security model is.

> 
> > 
> > Examples:
> > 
> > 1. The device can snoop secret data from readable pages in the
> >    application's virtual memory space.
> 
> Aren't there other security technologies that can address this?

Maybe the IOMMU could enforce Memory Protection Keys? Imagine each
device is assigned a subset of memory protection keys and the IOMMU
checks them on each device access. This would allow the application to
mark certain pages off-limits to the device but the IOMMU could still
walk the full process page table (no need to construct a special device
page table for the IOMMU).

> > 
> > 2. The device can gain arbitrary execution on the CPU by overwriting
> >    control flow addresses (e.g. function pointers, stack return
> >    addresses) in writable pages.
> 
> I suppose technology like CET might be able to guard. The general
> expectation is code pages and anything that needs to be protected should be
> mapped nor writable.

Function pointers are a common exception to this. They are often located
in writable heap or stack pages.

There might also be dynamic linker memory structures that are easy to
hijack.

Stefan

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 488 bytes --]