Re: [PATCH v2 2/3] iommu/sva: Stop using ioasid_set for SVA

[Jacob Pan]

Hi Kevin,

On Wed, 15 Feb 2023 03:24:18 +0000, "Tian, Kevin" <kevin.tian@xxxxxxxxx>
wrote:

> > From: Jacob Pan <jacob.jun.pan@xxxxxxxxxxxxxxx>
> > Sent: Tuesday, February 14, 2023 2:44 PM
> > 
> > From: Jason Gunthorpe <jgg@xxxxxxxxxx>
> > 
> > Instead SVA drivers can use a simple global IDA to allocate PASIDs for
> > each mm_struct.
> > 
> > Future work would be to allow drivers using the SVA APIs to reserve
> > global PASIDs from this IDA for their internal use, eg with the DMA API
> > PASID support.  
> 
> I think DMA API PASID will need new API around this IDA. SVA APIs should
> be just for SVA.
> 
> > +void mm_pasid_drop(struct mm_struct *mm)
> >  {
> > -	return mmget_not_zero(mm);
> > -}
> > +	pr_alert("%s %d", __func__, mm->pasid);
> > +	if (likely(!pasid_valid(mm->pasid)))
> > +		return;
> > 
> > -/**
> > - * iommu_sva_find() - Find mm associated to the given PASID
> > - * @pasid: Process Address Space ID assigned to the mm
> > - *
> > - * On success a reference to the mm is taken, and must be released with
> > mmput().
> > - *
> > - * Returns the mm corresponding to this PASID, or an error if not
> > found.
> > - */
> > -struct mm_struct *iommu_sva_find(ioasid_t pasid)
> > -{
> > -	return ioasid_find(&iommu_sva_pasid, pasid, __mmget_not_zero);
> > +	ida_free(&iommu_global_pasid_ida, mm->pasid);
> >  }  
> 
> when moving this function following line is missed:
> 
> 	mm->pasid = INVALID_IOASID;
> 
not needed, mm is gone by now.

> btw the current placement looks asymmetric. We now have
> mm_pasid_init() and mm_pasid_set() in mm.h while mm_pasid_drop()
> in iommu-sva.c.
> Is it cleaner to introduce a iommu_sva_free_pasid() helper and then
> call it from mm_pasid_drop() in mm.h?
That is what we had before, it seems logical to me. Jason?

Thanks,

Jacob