RE: [PATCH 3/6] vfio: remove the unused mdev iommu hook

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



> From: Jason Gunthorpe <jgg@xxxxxxxx>
> Sent: Monday, May 10, 2021 11:55 PM
> 
> On Mon, May 10, 2021 at 08:54:02AM +0200, Christoph Hellwig wrote:
> > The iommu_device field in struct mdev_device has never been used
> > since it was added more than 2 years ago.
> >
> > Signed-off-by: Christoph Hellwig <hch@xxxxxx>
> > ---
> >  drivers/vfio/vfio_iommu_type1.c | 132 ++++++--------------------------
> >  include/linux/mdev.h            |  20 -----
> >  2 files changed, 25 insertions(+), 127 deletions(-)
> 
> I asked Intel folks to deal with this a month ago:
> 
> https://lore.kernel.org/kvm/20210406200030.GA425310@xxxxxxxxxx/
> 
> So lets just remove it, it is clearly a bad idea
> 
> Reviewed-by: Jason Gunthorpe <jgg@xxxxxxxxxx>
> 

Want to get a clearer picture on what needs to be redesigned after this
removal.

Are you specially concerned about this iommu_device hack which 
directly connects mdev_device to iommu layer or the entire removed
logic including the aux domain concept? For the former we are now 
following up the referred thread to find a clean way. But for the latter 
we feel it's still necessary regardless of how iommu interface is redesigned 
to support device connection from the upper level driver. The reason is 
that with mdev or subdevice one physical device could be attached to 
multiple domains now. there could be a primary domain with DOMAIN_
DMA type for DMA_API use by parent driver itself, and multiple auxiliary 
domains with DOMAIN_UNMANAGED types for subdevices assigned to 
different VMs. In this case It's a different model from existing policy 
which allows only one domain per device. In removed code we followed 
Joerg's suggestion to keep existing iommu_attach_device for connecting 
device to the primary domain and then add new iommu_aux_attach_
device for connecting device to auxiliary domains. Lots of removed iommu 
code deal with the management of auxiliary domains in the iommu core 
layer and intel-iommu drivers, which imho is largely generic and could
be leveraged w/o intrusive refactoring to support redesigned iommu 
interface.

Does it sound a reasonable position?

Thanks
Kevin




[Index of Archives]     [KVM ARM]     [KVM ia64]     [KVM ppc]     [Virtualization Tools]     [Spice Development]     [Libvirt]     [Libvirt Users]     [Linux USB Devel]     [Linux Audio Users]     [Yosemite Questions]     [Linux Kernel]     [Linux SCSI]     [XFree86]

  Powered by Linux