Re: [RFC 2/7] iommu/vt-d: Remove iova handling code from non-dma ops path

[Lu Baolu]

Hi,

On 5/5/19 9:19 AM, Lu Baolu wrote:
> Hi,
>
> On 5/4/19 9:23 PM, Tom Murphy via iommu wrote:
> > @@ -4181,58 +4168,37 @@ static int intel_iommu_memory_notifier(struct 
> > notifier_block *nb,
> > ÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂ unsigned long val, void *v)
> > Â {
> > ÂÂÂÂÂ struct memory_notify *mhp = v;
> > -ÂÂÂ unsigned long long start, end;
> > -ÂÂÂ unsigned long start_vpfn, last_vpfn;
> > +ÂÂÂ unsigned long start_vpfn = mm_to_dma_pfn(mhp->start_pfn);
> > +ÂÂÂ unsigned long last_vpfn = mm_to_dma_pfn(mhp->start_pfn +
> > +ÂÂÂÂÂÂÂÂÂÂÂ mhp->nr_pages - 1);
> > ÂÂÂÂÂ switch (val) {
> > ÂÂÂÂÂ case MEM_GOING_ONLINE:
> > -ÂÂÂÂÂÂÂ start = mhp->start_pfn << PAGE_SHIFT;
> > -ÂÂÂÂÂÂÂ end = ((mhp->start_pfn + mhp->nr_pages) << PAGE_SHIFT) - 1;
> > -ÂÂÂÂÂÂÂ if (iommu_domain_identity_map(si_domain, start, end)) {
> > -ÂÂÂÂÂÂÂÂÂÂÂ pr_warn("Failed to build identity map for [%llx-%llx]\n",
> > -ÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂ start, end);
> > +ÂÂÂÂÂÂÂ if (iommu_domain_identity_map(si_domain, start_vpfn,
> > +ÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂ last_vpfn)) {
> > +ÂÂÂÂÂÂÂÂÂÂÂ pr_warn("Failed to build identity map for [%lx-%lx]\n",
> > +ÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂ start_vpfn, last_vpfn);
> > ÂÂÂÂÂÂÂÂÂÂÂÂÂ return NOTIFY_BAD;
> > ÂÂÂÂÂÂÂÂÂ }
> > ÂÂÂÂÂÂÂÂÂ break;
>
> Actually we don't need to update the si_domain if iommu hardware
> supports pass-through mode. This should be made in a separated patch
> anyway.

Oh! please ignore it.

This callback is only registered when hardware doesn't support pass
through mode.

        if (si_domain && !hw_pass_through)
                register_memory_notifier(&intel_iommu_memory_nb);

Best regards,
Lu Baolu