Re: [Xen-devel] [RFC v2 7/7] xen/iommu: smmu-v3: Add Xen specific code to enable the ported driver

[Julien Grall <julien.grall@xxxxxxxxxx>]

Hi Sameer,

On 22/11/17 02:17, Goel, Sameer wrote:
> On 11/20/2017 7:25 AM, Julien Grall wrote:
> > On 19/11/17 07:45, Goel, Sameer wrote:
> > > On 10/12/2017 10:36 AM, Julien Grall wrote:
> > > > Please use #if 0 rather than removing the code + comment on top. But I am not 
> > > > sure why you drop the S2 free code. Shouldn't we allocate VMID from the SMMU?
> > > I am picking up the vmid from the domain I am sharing the page tables with. 
> > > domain->arch.p2m.vmid. This seemed valid. Please let me know if we want to 
> > > generate a different vmid?
> > The processor and the SMMU may support either 8 or 16bits VMID but I don't 
> > think any thing in the spec says that both processor and SMMU must support the 
> > same value.
> >
> > So it would be possible to have a processor supporting 16-bits VMID and the 
> > SMMU only 8-bits VMID.
> >
> > Adding a check at the SMMU initialization might be a solution. But as the code 
> > for allocating the VMID is already present, then I would prefer to we stick 
> > with different the VMID.
> Sure. For this iteration I will generate an SMMU specific VMID. Do you think it 
> would be an optimization to reuse the CPU VMID if SMMU supports 16bit VMID?
The only benefit I can see is if the SMMU support broadcast TLB 
maintenance. AFAICT, the driver would need to configure the SMMU for that.
However, I think we should first focus on getting a basic SMMUv3 driver 
in Xen. Once this is done, we can look at optimizing the driver.
Cheers,

--
Julien Grall

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxxx
https://lists.xenproject.org/mailman/listinfo/xen-devel