Re: [PATCH v6 13/18] iommu/vt-d: Handle reattach of the restored domain

[Samiullah Khawaja <skhawaja@google.com>]

On Fri, Oct 30, 2026 at 07:38:54AM +0800, Baolu Lu wrote:
>On 10/30/26 06:34, Samiullah Khawaja wrote:
>>>>diff --git a/drivers/iommu/intel/liveupdate.c 
>>>>b/drivers/iommu/intel/ liveupdate.c
>>>>index 36816d3651e6..665ebb81bd61 100644
>>>>--- a/drivers/iommu/intel/liveupdate.c
>>>>+++ b/drivers/iommu/intel/liveupdate.c
>>>>@@ -394,6 +394,137 @@ void 
>>>>intel_iommu_handle_drhd_restore_failure(struct intel_iommu 
>>>>*iommu)
>>>>               iommu->name);
>>>> }
>>>>+static void domain_detach_restored_iommu(struct dmar_domain *domain,
>>>>+                     struct intel_iommu *iommu)
>>>>+{
>>>>+    struct iommu_domain_info *info;
>>>>+
>>>>+    guard(mutex)(&iommu->did_lock);
>>>>+    info = xa_load(&domain->iommu_array, iommu->seq_id);
>>>>+    if (--info->refcnt == 0) {
>>>>+        xa_erase(&domain->iommu_array, iommu->seq_id);
>>>>+        kfree(info);
>>>>+    }
>>>
>>>This seems to be different from normal domain_detach_iommu(), which
>>>looks like this,
>>>
>>>       if (--info->refcnt == 0) {
>>>               ida_free(&iommu->domain_ida, info->did);
>>>               xa_erase(&domain->iommu_array, iommu->seq_id);
>>>               kfree(info);
>>>       }
>>>
>>>why ida_free() is skipped in the restored path?
>>
>>The restored device can be removed and probed again while the PASID and
>>context entries stay intact. So the restored did is not freed in this
>>path. Once the domain is reclaimed by the iommufd, then it takes the
>>normal detach path.
>
>Can you please add a comment to explain this?

Agreed. I will add a comment for this.
>
>Thanks,
>baolu

Thanks,
Sami