Virtualization - Feb 2023 - [PATCH v2] vhost/vdpa: Add MSI translation tables to iommu for software-managed MSI

On Fri, Feb 17, 2023 at 01:35:59PM +0800, Jason Wang
wrote:
> On Fri, Feb 17, 2023 at 8:15 AM Jason Gunthorpe <jgg at nvidia.com>
wrote:
> >
> > On Tue, Feb 07, 2023 at 08:08:43PM +0800, Nanyong Sun wrote:
> > > From: Rong Wang <wangrong68 at huawei.com>
> > >
> > > Once enable iommu domain for one device, the MSI
> > > translation tables have to be there for software-managed MSI.
> > > Otherwise, platform with software-managed MSI without an
> > > irq bypass function, can not get a correct memory write event
> > > from pcie, will not get irqs.
> > > The solution is to obtain the MSI phy base address from
> > > iommu reserved region, and set it to iommu MSI cookie,
> > > then translation tables will be created while request irq.
> >
> > Probably not what anyone wants to hear, but I would prefer we not add
> > more uses of this stuff. It looks like we have to get rid of
> > iommu_get_msi_cookie() :\
> >
> > I'd like it if vdpa could move to iommufd not keep copying stuff
from
> > it..
> 
> Yes, but we probably need a patch for -stable.
Hmm do we? this looks like it's enabling new platforms is not a bugfix...
> >
> > Also the iommu_group_has_isolated_msi() check is missing on the vdpa
> > path, and it is missing the iommu ownership mechanism.
> 
> Ok.
> 
> >
> > Also which in-tree VDPA driver that uses the iommu runs on ARM? Please
> 
> ifcvf and vp_vpda are two drivers that use platform IOMMU.
> 
> Thanks
> 
> > don't propose core changes for unmerged drivers. :(
> >
> > Jason
> >

On Fri, Feb 17, 2023 at 6:11 PM Michael S. Tsirkin <mst at redhat.com>
wrote:
>
> On Fri, Feb 17, 2023 at 01:35:59PM +0800, Jason Wang wrote:
> > On Fri, Feb 17, 2023 at 8:15 AM Jason Gunthorpe <jgg at
nvidia.com> wrote:
> > >
> > > On Tue, Feb 07, 2023 at 08:08:43PM +0800, Nanyong Sun wrote:
> > > > From: Rong Wang <wangrong68 at huawei.com>
> > > >
> > > > Once enable iommu domain for one device, the MSI
> > > > translation tables have to be there for software-managed
MSI.
> > > > Otherwise, platform with software-managed MSI without an
> > > > irq bypass function, can not get a correct memory write
event
> > > > from pcie, will not get irqs.
> > > > The solution is to obtain the MSI phy base address from
> > > > iommu reserved region, and set it to iommu MSI cookie,
> > > > then translation tables will be created while request irq.
> > >
> > > Probably not what anyone wants to hear, but I would prefer we not
add
> > > more uses of this stuff. It looks like we have to get rid of
> > > iommu_get_msi_cookie() :\
> > >
> > > I'd like it if vdpa could move to iommufd not keep copying
stuff from
> > > it..
> >
> > Yes, but we probably need a patch for -stable.
>
> Hmm do we? this looks like it's enabling new platforms is not a
bugfix...
I think we haven't limited vDPA to any specific arch in the past?

Thanks
>
> > >
> > > Also the iommu_group_has_isolated_msi() check is missing on the
vdpa
> > > path, and it is missing the iommu ownership mechanism.
> >
> > Ok.
> >
> > >
> > > Also which in-tree VDPA driver that uses the iommu runs on ARM?
Please
> >
> > ifcvf and vp_vpda are two drivers that use platform IOMMU.
> >
> > Thanks
> >
> > > don't propose core changes for unmerged drivers. :(
> > >
> > > Jason
> > >
>