Re: [PATCH v4 3/5] iommu: Add verisilicon IOMMU driver

[Benjamin Gaignard]

Le 04/07/2025 à 19:54, Jason Gunthorpe a écrit :
> On Mon, Jun 23, 2025 at 05:39:27PM +0200, Benjamin Gaignard wrote:
> > The Verisilicon IOMMU hardware block can be found in combination
> > with Verisilicon hardware video codecs (encoders or decoders) on
> > different SoCs.
> > Enable it will allow us to use non contiguous memory allocators
> > for Verisilicon video codecs.
> >
> > Signed-off-by: Benjamin Gaignard <benjamin.gaignard@xxxxxxxxxxxxx>
> > ---
> > change in version 4:
> > - Kconfig dependencies
> > - Fix the remarks done by Jason and Robin: locking, clocks, macros
> >    probing, pm_runtime, atomic allocation.
> It broadly seems OK to me
>
> Though I did notice this:
>
> > +static struct iommu_domain *vsi_iommu_domain_alloc_paging(struct device *dev)
> > +{
> > +	struct vsi_iommu *iommu = dev_iommu_priv_get(dev);
> > +	struct vsi_iommu_domain *vsi_domain;
> > +
> > +	vsi_domain = kzalloc(sizeof(*vsi_domain), GFP_KERNEL);
> > +	if (!vsi_domain)
> > +		return NULL;
> > +
> > +	vsi_domain->iommu = iommu;
> So we store the iommu in the domain? And use the iommu->lock all over
> the place
>
> > +static int vsi_iommu_attach_device(struct iommu_domain *domain,
> > +				   struct device *dev)
> > +{
> > +	struct vsi_iommu *iommu = dev_iommu_priv_get(dev);
> > +	struct vsi_iommu_domain *vsi_domain = to_vsi_domain(domain);
> > +	unsigned long flags;
> > +	int ret = 0;
> > +
> > +	ret = pm_runtime_resume_and_get(iommu->dev);
> > +	if (ret < 0)
> > +		return ret;
> > +
> > +	spin_lock_irqsave(&iommu->lock, flags);
> > +	/* iommu already attached */
> > +	if (iommu->domain == domain)
> > +		goto unlock;
> But here we don't check that the domain matches the iommu of dev.
>
> This seems a bit weird to me, I didn't quite get why the domain uses
> iommu->lock instead of just having its own per-domain lock?
>
> But if it does use iommu->lock then this does need to prevent using
> domains with the wrong iommu because the also use the wrong lock and
> then this:

I would like to avoid that but maybe a static spinlock can solve that problem ?

Regards,
Benjamin

> > +
> > +	vsi_iommu_enable(iommu, domain);
> > +	list_add_tail(&iommu->node, &vsi_domain->iommus);
> Is not safe?
>
> Jason