[Xen-devel] [PATCH v2] AMD/IOMMU: don't "add" IOMMUs

["Jan Beulich" <JBeulich@xxxxxxxx>]

For find_iommu_for_device() to consistently (independent of ACPI tables)
return NULL for the PCI devices corresponding to IOMMUs, make sure
IOMMUs don't get mapped to themselves by ivrs_mappings[].

While amd_iommu_add_device() won't be called for IOMMUs from
pci_add_device(), as IOMMUs have got marked r/o,
_setup_hwdom_pci_devices() calls there nevertheless. Avoid issuing the
bogus debugging only "No iommu for ...; cannot be handed to ..." log
message as well as the non-debugging "setup ... for ... failed (-19)"
one.

Signed-off-by: Jan Beulich <jbeulich@xxxxxxxx>
---
v2: Also change add_ivrs_mapping_entry() and find_iommu_for_device()
    (I should have checked why I didn't recall seeing the log messages
    on the older AMD system before sending v1). Only return success
    from amd_iommu_add_device() for Dom0.
---
The situation for host bridges is similar: Behavior currently depends
on whether they're within the device ranges reported by IVRS entries.
On the older system 0000:00:00.0 is included, while on the newer one
it's not. Whether it gets its device table entry point to Dom0's page
tables depends on this.

--- a/xen/drivers/passthrough/amd/iommu_acpi.c
+++ b/xen/drivers/passthrough/amd/iommu_acpi.c
@@ -81,8 +81,8 @@ static void __init add_ivrs_mapping_entr
              ivrs_mappings[alias_id].intremap_inuse = shared_intremap_inuse;
          }
     }
-    /* assgin iommu hardware */
-    ivrs_mappings[bdf].iommu = iommu;
+    /* Assign IOMMU hardware, but don't map an IOMMU by itself. */
+    ivrs_mappings[bdf].iommu = iommu->bdf != bdf ? iommu : NULL;
 }
 
 static struct amd_iommu * __init find_iommu_from_bdf_cap(
--- a/xen/drivers/passthrough/amd/pci_amd_iommu.c
+++ b/xen/drivers/passthrough/amd/pci_amd_iommu.c
@@ -43,7 +43,7 @@ struct amd_iommu *find_iommu_for_device(
     {
         unsigned int bd0 = bdf & ~PCI_FUNC(~0);
 
-        if ( ivrs_mappings[bd0].iommu )
+        if ( ivrs_mappings[bd0].iommu && ivrs_mappings[bd0].iommu->bdf != bdf )
         {
             struct ivrs_mappings tmp = ivrs_mappings[bd0];
 
@@ -424,6 +424,11 @@ static int amd_iommu_add_device(u8 devfn
         return -EINVAL;
 
     bdf = PCI_BDF2(pdev->bus, pdev->devfn);
+
+    for_each_amd_iommu(iommu)
+        if ( pdev->seg == iommu->seg && bdf == iommu->bdf )
+            return is_hardware_domain(pdev->domain) ? 0 : -ENODEV;
+
     iommu = find_iommu_for_device(pdev->seg, bdf);
     if ( unlikely(!iommu) )
     {





_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxxx
https://lists.xenproject.org/mailman/listinfo/xen-devel