Re: [PATCH v4 3/6] iommu/sva: Stop using ioasid_set for SVA

[Jason Gunthorpe]

On Tue, Mar 07, 2023 at 02:32:09PM -0800, Jacob Pan wrote:
> Hi Jason,
> 
> On Fri, 3 Mar 2023 09:15:45 -0400, Jason Gunthorpe <jgg@xxxxxxxxxx> wrote:
> 
> > On Fri, Mar 03, 2023 at 05:57:41PM +0800, Baolu Lu wrote:
> > > On 2023/3/3 17:32, Jean-Philippe Brucker wrote:  
> > > > > I suppose the common thing is reserving some kind of special
> > > > > PASIDs.  
> > > > Are you planning to use RID_PASID != 0 in VT-d?  Otherwise we could
> > > > just communicate min_pasid from the IOMMU driver the same way we do
> > > > max_pasid.
> > > > 
> > > > Otherwise I guess re-introduce a lighter ioasid_alloc() that the IOMMU
> > > > driver calls to reserve PASID0/RID_PASID.  
> > > 
> > > Yes. We probably will use a non-zero RID_PASID in the future. An
> > > interface to reserve (or allocate) a PASID from iommu_global_pasid_ida
> > > should work then.  
> > 
> > Just allowing the driver to store XA_ZERO_ENTRY would be fine
> > 
> It looks like there are incoming users of iommu_sva_find()
> https://lore.kernel.org/lkml/20230306163138.587484-1-fenghua.yu@xxxxxxxxx/T/#m1fc97725a0e56ea269c8bdabacee447070d51846
> Should we keep the xa here instead of the global ida?

I'm not sure this should be in the iommu core, it is really gross.

I would expect IDXD to keep track of the PASID's and mms it is using
and do this kind of stuff itself.

And why is this using access_remote_vm anyhow? If you know you are in
a kthread then kthread_use_mm() is probably better anyhow.

In any event we don't need a iommu_sva_find() function to wrapper
xa_load for another function inside the same .c file.

Jason