Re: [Qemu-devel] [PATCH v1 2/3] intel-iommu: extend VTD emulation to allow 57-bit IOVA address width.

[Peter Xu]

On Tue, Nov 13, 2018 at 01:45:44PM +0800, Yu Zhang wrote:

[...]

> > > Since at it, another thing I thought about is making sure the IOMMU
> > > capabilities will match between host and guest IOMMU, which I think
> > > this series has ignorred so far.  E.g., when we're having assigned
> > > devices in the guest and with 5-level IOVA, we should make sure the
> > > host IOMMU supports 5-level as well before the guest starts since
> > > otherwise the shadow page synchronization could potentially fail when
> > > the requested IOVA address goes beyond 4-level.  One simple solution
> > > is just to disable device assignment for now when we're with 57bits
> > > vIOMMU but I'm not sure whether that's what you want, especially you
> > > mentioned the DPDK case (who may use assigned devices).
> > 
> > Ok I totally forgot that we don't even support any kind of check like
> > this before... So feel free to skip this comment if you want, or it
> > would be even nicer if you want to fix it as a whole. :)
> > 
> 
> Indeed. We have talked about this before. How about we focus on the 5-level
> extension for now, and solve the check issue in the future? I still do not
> have any clean solutions in mind. BTW, any suggestions on this issue? :)

I started to remember our discussions, sorry I should remember them
earlier... :)

The only thing in my mind (I think I also suggested the same thing
during that discussion, but I don't trust my memory any more...) is to
use sysfs.  Say:

  1. Scan /sys/class/iommu/dmarN for all the host IOMMUs, read cap of
     each IOMMU from /sys/class/iommu/dmar0/intel-iommu/cap,

  2. For each host iommu, scan /sys/class/iommu/dmarN/devices for all
     the devices under each host IOMMU, then we can know which IOMMU
     owns which device,

  3. For each assigned device to the guest, we lookup the previous
     information to know the mgaw for each host device, raise error
     and stop QEMU from booting if any of the host device has less
     level supported than the guest vIOMMU (possibly some more checks
     in vtd_iommu_notify_flag_changed)

(we still have some issue on vtd_iommu_notify_flag_changed since it's
 only run until the first enablement of vIOMMU, so we'll only raise
 the error during guest Linux boots with vIOMMU on. But that's another
 issue)

Regards,

-- 
Peter Xu