Re: [PATCH v8 10/10] iommu/vt-d: Add iotlb flush for nested domain

[Yi Liu]

On 2023/12/28 15:11, Tian, Kevin wrote:
> > From: Liu, Yi L <yi.l.liu@xxxxxxxxx>
> > Sent: Thursday, December 28, 2023 12:14 AM
> >
> >
> > +static void nested_flush_pasid_iotlb(struct intel_iommu *iommu,
> > +				     struct dmar_domain *domain, u64 addr,
> > +				     unsigned long npages, bool ih)
> > +{
> > +	u16 did = domain_id_iommu(domain, iommu);
> > +	unsigned long flags;
> > +
> > +	spin_lock_irqsave(&domain->lock, flags);
> > +	if (!list_empty(&domain->devices))
> > +		qi_flush_piotlb(iommu, did, IOMMU_NO_PASID, addr,
> > +				npages, ih, NULL);
>
> Not sure the check of domain->devices makes sense here. The outer
> loop is xa_for_each(&domain->iommu_array, i, info) which, if found,
> then implies certain devices behind this iommu must be attached to
> this domain.

yes.

> I wonder whether it's clearer to remove this function and just call
> qi_flush_piotlb() from intel_nested_flush_cache().

yeah, with above remark, it would be more clear to call qi_flush_piotlb()
from intel_nested_flush_cache().

> > +
> > +	/*
> > +	 * Invalidation queue error (i.e. IQE) will not be reported to user
> > +	 * as it's caused only by driver internal bug.
> > +	 */
>
> here lack of:
>
> 	*error = 0;

yes.

> > +	if (fault & DMA_FSTS_ICE)
> > +		*error |= IOMMU_HWPT_INVALIDATE_VTD_S1_ICE;
> > +	if (fault & DMA_FSTS_ITE)
> > +		*error |= IOMMU_HWPT_INVALIDATE_VTD_S1_ITE;
> > +}
> > +

--
Regards,
Yi Liu