diff --git a/xen/drivers/passthrough/amd/iommu_cmd.c b/xen/drivers/passthrough/amd/iommu_cmd.c
index fdf089e28b..f3a6a5ed86 100644
--- a/xen/drivers/passthrough/amd/iommu_cmd.c
+++ b/xen/drivers/passthrough/amd/iommu_cmd.c
@@ -277,6 +277,20 @@ static void invalidate_interrupt_table(struct amd_iommu *iommu, u16 device_id)
     send_iommu_command(iommu, cmd);
 }
 
+void invalidate_iommu_all(struct amd_iommu *iommu)
+{
+    u32 cmd[4], entry;
+
+    cmd[3] = cmd[2] = cmd[0] = 0;
+
+    set_field_in_reg_u32(IOMMU_CMD_INVALIDATE_IOMMU_ALL, 0,
+                         IOMMU_CMD_OPCODE_MASK, IOMMU_CMD_OPCODE_SHIFT,
+                         &entry);
+    cmd[1] = entry;
+
+    send_iommu_command(iommu, cmd);
+}
+
 void amd_iommu_flush_iotlb(struct pci_dev *pdev,
                            uint64_t gaddr, unsigned int order)
 {
@@ -380,3 +394,11 @@ void amd_iommu_flush_intremap(struct amd_iommu *iommu, uint16_t bdf)
     invalidate_interrupt_table(iommu, bdf);
     flush_command_buffer(iommu);
 }
+
+void amd_iommu_flush_all_caches(struct amd_iommu *iommu)
+{
+    ASSERT( spin_is_locked(&iommu->lock) );
+
+    invalidate_iommu_all(iommu);
+    flush_command_buffer(iommu);
+}
diff --git a/xen/drivers/passthrough/amd/iommu_init.c b/xen/drivers/passthrough/amd/iommu_init.c
index 12fe3e40e4..404638d194 100644
--- a/xen/drivers/passthrough/amd/iommu_init.c
+++ b/xen/drivers/passthrough/amd/iommu_init.c
@@ -598,6 +598,9 @@ static void enable_iommu(struct amd_iommu *iommu)
     set_iommu_event_log_control(iommu, IOMMU_CONTROL_ENABLED);
     set_iommu_translation_control(iommu, IOMMU_CONTROL_ENABLED);
 
+    if ( iommu_has_feature(iommu, IOMMU_EXT_FEATURE_IASUP_SHIFT) )
+        amd_iommu_flush_all_caches(iommu);
+
     iommu->enabled = 1;
     spin_unlock_irqrestore(&iommu->lock, flags);
 
@@ -970,6 +973,9 @@ void amd_iommu_resume(void)
     }
 
     /* flush all cache entries after iommu re-enabled */
-    invalidate_all_devices();
-    invalidate_all_domain_pages();
+    if ( !iommu_has_feature(iommu, IOMMU_EXT_FEATURE_IASUP_SHIFT) )
+    {
+        invalidate_all_devices();
+        invalidate_all_domain_pages();
+    }
 }
diff --git a/xen/include/asm-x86/hvm/svm/amd-iommu-defs.h b/xen/include/asm-x86/hvm/svm/amd-iommu-defs.h
index c6bf6cd8ec..a9325cb724 100644
--- a/xen/include/asm-x86/hvm/svm/amd-iommu-defs.h
+++ b/xen/include/asm-x86/hvm/svm/amd-iommu-defs.h
@@ -192,6 +192,7 @@
 #define IOMMU_CMD_INVALIDATE_IOMMU_PAGES	0x3
 #define IOMMU_CMD_INVALIDATE_IOTLB_PAGES	0x4
 #define IOMMU_CMD_INVALIDATE_INT_TABLE		0x5
+#define IOMMU_CMD_INVALIDATE_IOMMU_ALL      0x8
 
 /* COMPLETION_WAIT command */
 #define IOMMU_COMP_WAIT_DATA_BUFFER_SIZE	8
diff --git a/xen/include/asm-x86/hvm/svm/amd-iommu-proto.h b/xen/include/asm-x86/hvm/svm/amd-iommu-proto.h
index 5a327bfad0..960e9ff783 100644
--- a/xen/include/asm-x86/hvm/svm/amd-iommu-proto.h
+++ b/xen/include/asm-x86/hvm/svm/amd-iommu-proto.h
@@ -78,6 +78,7 @@ void amd_iommu_flush_iotlb(struct pci_dev *pdev, uint64_t gaddr,
                            unsigned int order);
 void amd_iommu_flush_device(struct amd_iommu *iommu, uint16_t bdf);
 void amd_iommu_flush_intremap(struct amd_iommu *iommu, uint16_t bdf);
+void amd_iommu_flush_all_caches(struct amd_iommu *iommu);
 
 /* find iommu for bdf */
 struct amd_iommu *find_iommu_for_device(int seg, int bdf);