VT-d: Separate VT-d page table from P2M table

Signed-off-by: Weidong Han <weidong.han@intel.com>
Signed-off-by: Anthony Xu <anthony.xu@intel.com>

tools/python/xen/xend/XendDomainInfo.py

@@ -2135,8 +2135,13 @@ class XendDomainInfo:
             # set memory limit
             xc.domain_setmaxmem(self.domid, maxmem)
 
+            # Reserve 1 page per MiB of RAM for separate VT-d page table.
+            vtd_mem = 4 * (self.info['memory_static_max'] / 1024 / 1024)
+            # Round vtd_mem up to a multiple of a MiB.
+            vtd_mem = ((vtd_mem + 1023) / 1024) * 1024
+
             # Make sure there's enough RAM available for the domain
-            balloon.free(memory + shadow)
+            balloon.free(memory + shadow + vtd_mem)
 
             # Set up the shadow memory
             shadow_cur = xc.shadow_mem_control(self.domid, shadow / 1024)

xen/arch/x86/mm/p2m.c

@@ -259,18 +259,10 @@ p2m_set_entry(struct domain *d, unsigned long gfn, mfn_t mfn, p2m_type_t p2mt)
 
     if ( iommu_enabled && is_hvm_domain(d) )
     {
-        if ( boot_cpu_data.x86_vendor == X86_VENDOR_INTEL )
-        {
-            if ( (p2mt == p2m_mmio_direct) )
-                iommu_flush(d, gfn, (u64*)p2m_entry);
-        }
-        else if ( boot_cpu_data.x86_vendor == X86_VENDOR_AMD )
-        {
-            if ( p2mt == p2m_ram_rw )
-                iommu_map_page(d, gfn, mfn_x(mfn));
-            else
-                iommu_unmap_page(d, gfn);
-        }
+        if ( p2mt == p2m_ram_rw )
+            iommu_map_page(d, gfn, mfn_x(mfn));
+        else
+            iommu_unmap_page(d, gfn);
     }
 
     /* Success */
@@ -529,11 +521,6 @@ int p2m_alloc_table(struct domain *d,
             goto error;
     }
 
-#if CONFIG_PAGING_LEVELS >= 3
-    if (vtd_enabled && is_hvm_domain(d))
-        iommu_set_pgd(d);
-#endif
-
     P2M_PRINTK("p2m table initialised (%u pages)\n", page_count);
     p2m_unlock(p2m);
     return 0;

xen/drivers/passthrough/vtd/iommu.c

@@ -671,13 +671,26 @@ void dma_pte_free_pagetable(struct domain *domain, u64 start, u64 end)
     }
 
     /* free pgd */
-    if ( start == 0 && end == ((((u64)1) << addr_width) - 1) )
+    if ( start == 0 && end >= ((((u64)1) << addr_width) - 1) )
     {
         free_pgtable_maddr(hd->pgd_maddr);
         hd->pgd_maddr = 0;
     }
 }
 
+ /* free all VT-d page tables when shut down or destroy domain. */
+static void iommu_free_pagetable(struct domain *domain)
+{
+    struct hvm_iommu *hd = domain_hvm_iommu(domain);
+    int addr_width = agaw_to_width(hd->agaw);
+    u64 start, end;
+
+    start = 0;
+    end = (((u64)1) << addr_width) - 1;
+
+    dma_pte_free_pagetable(domain, start, end);
+}
+
 static int iommu_set_root_entry(struct iommu *iommu)
 {
     u32 cmd, sts;
@@ -1472,9 +1485,9 @@ void iommu_domain_teardown(struct domain *d)
     if ( list_empty(&acpi_drhd_units) )
         return;
 
-    iommu_domid_release(d);
-    iommu_free_pgd(d);
+    iommu_free_pagetable(d);
     return_devices_to_dom0(d);
+    iommu_domid_release(d);
 }
 
 static int domain_context_mapped(struct pci_dev *pdev)