Re: [PATCH] iommu/vt-d: do not use flush-queue when caching-mode is on

[Lu Baolu]

On 1/27/21 8:26 AM, Lu Baolu wrote:
> > +{
> > +    struct dmar_domain *dmar_domain = to_dmar_domain(domain);
> > +    struct intel_iommu *iommu = domain_get_iommu(dmar_domain);
> > +
> > +    if (intel_iommu_strict)
> > +        return 0;
> > +
> > +    /*
> > +     * The flush queue implementation does not perform page-selective
> > +     * invalidations that are required for efficient TLB flushes in 
> > virtual
> > +     * environments. The benefit of batching is likely to be much 
> > lower than
> > +     * the overhead of synchronizing the virtual and physical IOMMU
> > +     * page-tables.
> > +     */
> > +    if (iommu && cap_caching_mode(iommu->cap)) {
> > +        pr_warn_once("IOMMU batching is partially disabled due to 
> > virtualization");
> > +        return 0;
> > +    }
>
> domain_get_iommu() only returns the first iommu, and could return NULL
> when this is called before domain attaching to any device. A better
> choice could be check caching mode globally and return false if caching
> mode is supported on any iommu.
>
>         struct dmar_drhd_unit *drhd;
>         struct intel_iommu *iommu;
>
>         rcu_read_lock();
>         for_each_active_iommu(iommu, drhd) {
>                  if (cap_caching_mode(iommu->cap))
>                          return false;

We should unlock rcu before return here. Sorry!

>          }
>          rcu_read_unlock();

Best regards,
baolu