Re: [PATCH 1/3] iommu/vt-d: Remove redundant IOTLB flush

[Lu Baolu]

On 2020/3/21 0:20, Jacob Pan wrote:
> On Fri, 20 Mar 2020 21:45:26 +0800
> Lu Baolu <baolu.lu@xxxxxxxxxxxxxxx> wrote:
>
> > On 2020/3/20 12:32, Jacob Pan wrote:
> > > IOTLB flush already included in the PASID tear down process. There
> > > is no need to flush again.
> >
> > It seems that intel_pasid_tear_down_entry() doesn't flush the pasid
> > based device TLB?
> I saw this code in intel_pasid_tear_down_entry(). Isn't the last line
> flush the devtlb? Not in guest of course since the passdown tlb flush
> is inclusive.
>
> 	pasid_cache_invalidation_with_pasid(iommu, did, pasid);
> 	iotlb_invalidation_with_pasid(iommu, did, pasid);
>
> 	/* Device IOTLB doesn't need to be flushed in caching mode. */
> 	if (!cap_caching_mode(iommu->cap))
> 		devtlb_invalidation_with_pasid(iommu, dev, pasid);

But devtlb_invalidation_with_pasid() doesn't do the right thing, it
flushes the device tlb, instead of pasid-based device tlb.

static void
devtlb_invalidation_with_pasid(struct intel_iommu *iommu,
                               struct device *dev, int pasid)
{
        struct device_domain_info *info;
        u16 sid, qdep, pfsid;

        info = dev->archdata.iommu;
        if (!info || !info->ats_enabled)
                return;

        sid = info->bus << 8 | info->devfn;
        qdep = info->ats_qdep;
        pfsid = info->pfsid;

        qi_flush_dev_iotlb(iommu, sid, pfsid, qdep, 0, 64 - 
VTD_PAGE_SHIFT);
}

Best regards,
baolu

> > Best regards,
> > baolu
> >
> > > Cc: Lu Baolu <baolu.lu@xxxxxxxxxxxxxxx>
> > > Signed-off-by: Jacob Pan <jacob.jun.pan@xxxxxxxxxxxxxxx>
> > > ---
> > >    drivers/iommu/intel-svm.c | 6 ++----
> > >    1 file changed, 2 insertions(+), 4 deletions(-)
> > >
> > > diff --git a/drivers/iommu/intel-svm.c b/drivers/iommu/intel-svm.c
> > > index 8f42d717d8d7..1483f1845762 100644
> > > --- a/drivers/iommu/intel-svm.c
> > > +++ b/drivers/iommu/intel-svm.c
> > > @@ -268,10 +268,9 @@ static void intel_mm_release(struct
> > > mmu_notifier *mn, struct mm_struct *mm)
> > >    	 * *has* to handle gracefully without affecting other
> > > processes. */
> > >    	rcu_read_lock();
> > > -	list_for_each_entry_rcu(sdev, &svm->devs, list) {
> > > +	list_for_each_entry_rcu(sdev, &svm->devs, list)
> > >    		intel_pasid_tear_down_entry(svm->iommu,
> > > sdev->dev, svm->pasid);
> > > -		intel_flush_svm_range_dev(svm, sdev, 0, -1, 0);
> > > -	}
> > > +
> > >    	rcu_read_unlock();
> > >    
> > >    }
> > > @@ -731,7 +730,6 @@ int intel_svm_unbind_mm(struct device *dev, int
> > > pasid)
> > >    			 * large and has to be physically
> > > contiguous. So it's
> > >    			 * hard to be as defensive as we might
> > > like. */ intel_pasid_tear_down_entry(iommu, dev, svm->pasid);
> > > -			intel_flush_svm_range_dev(svm, sdev, 0,
> > > -1, 0); kfree_rcu(sdev, rcu);
> > >    
> > >    			if (list_empty(&svm->devs)) {
> > >    
>
> [Jacob Pan]