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

[Lu Baolu <baolu.lu@linux.intel.com>]

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