Re: [PATCH v4 2/5] iommu/vt-d: Remove the pasid present check in prq_event_thread

[Baolu Lu]

On 10/30/24 22:28, Joel Granados wrote:
> >    /*
> > diff --git a/drivers/iommu/intel/svm.c b/drivers/iommu/intel/svm.c
> > index 078d1e32a24e..ff88f31053d1 100644
> > --- a/drivers/iommu/intel/svm.c
> > +++ b/drivers/iommu/intel/svm.c
> > @@ -304,9 +304,6 @@ void intel_drain_pasid_prq(struct device *dev, u32
> > pasid)
> >           int qdep;
> >
> >           info = dev_iommu_priv_get(dev);
> > -       if (WARN_ON(!info || !dev_is_pci(dev)))
> > -               return;
> Did you mean to take out both checks?:
>    1. The info pointer check
>    2. the dev_is_pci check
>
> I can understand the dev_is_pci check, but we should definitely take
> action if info is NULL. Right?

WARN_ON(!info) is duplicate as far as I can see. Accessing
info->pri_enable when info is NULL will cause a null pointer dereference
warning. This appears irrelevant to this patch though.

> > -
> >           if (!info->pri_enabled)
> >                   return;
> >
> > Generally, intel_drain_pasid_prq() should be called if
> >
> > - a translation is removed from a pasid entry; and
> This is the path that is already mentiond
>
> > - PRI on this device is enabled.
> And this path is:
>    -> intel_iommu_enable_iopf
>      -> context_flip_pri
>        -> intel_context_flush_present
>          -> qi_flush_pasid_cache
>
> Right?

Sorry that I didn't make it clear. It should be "PRI on this device was
enabled", a.k.a. info->pri_enabled is true. I didn't meant to say in the
PRI enabling path.

--
baolu