hv: deny the launch of VM if pass-thru PIO bar isn't identical mapping

In current design, when pass-thru dev,
for the PIO bar, need to ensure the guest PIO start address
equals to host PIO start address.
Then set the VMCS io bitmap to pass-thru the corresponding
port io to guest for performance.

ACRN-DM and acrn-config should ensure the identical mapping of PIO bar.
If ACRN-DM or acrn-config failed to achieve this,
we should deny the launch of VM

Tracked-On: #6508

Signed-off-by: Liu,Junming <junming.liu@intel.com>
Reviewed-by: Zhao Yakui <yakui.zhao@intel.com>
Reviewed-by: Fei Li <fei1.li@intel.com>

hypervisor/arch/x86/guest/vm.c

@@ -599,34 +599,37 @@ int32_t create_vm(uint16_t vm_id, uint64_t pcpu_bitmap, struct acrn_vm_config *v
 		passthrough_smbios(vm, get_acrn_boot_info());
 #endif
 
-		init_vpci(vm);
+		status = init_vpci(vm);
-		enable_iommu();
-
-		/* Create virtual uart;*/
-		init_legacy_vuarts(vm, vm_config->vuart);
-
-		register_reset_port_handler(vm);
-
-		/* vpic wire_mode default is INTR */
-		vm->wire_mode = VPIC_WIRE_INTR;
-
-		/* Init full emulated vIOAPIC instance:
-		 * Present a virtual IOAPIC to guest, as a placeholder interrupt controller,
-		 * even if the guest uses PT LAPIC. This is to satisfy the guest OSes,
-		 * in some cases, though the functionality of vIOAPIC doesn't work.
-		 */
-		vioapic_init(vm);
-
-		/* Populate return VM handle */
-		*rtn_vm = vm;
-		vm->sw.io_shared_page = NULL;
-		if ((vm_config->load_order == POST_LAUNCHED_VM) && ((vm_config->guest_flags & GUEST_FLAG_IO_COMPLETION_POLLING) != 0U)) {
-			/* enable IO completion polling mode per its guest flags in vm_config. */
-			vm->sw.is_polling_ioreq = true;
-		}
-		status = set_vcpuid_entries(vm);
 		if (status == 0) {
-			vm->state = VM_CREATED;
+			enable_iommu();
+
+			/* Create virtual uart;*/
+			init_legacy_vuarts(vm, vm_config->vuart);
+
+			register_reset_port_handler(vm);
+
+			/* vpic wire_mode default is INTR */
+			vm->wire_mode = VPIC_WIRE_INTR;
+
+			/* Init full emulated vIOAPIC instance:
+			* Present a virtual IOAPIC to guest, as a placeholder interrupt controller,
+			* even if the guest uses PT LAPIC. This is to satisfy the guest OSes,
+			* in some cases, though the functionality of vIOAPIC doesn't work.
+			*/
+			vioapic_init(vm);
+
+			/* Populate return VM handle */
+			*rtn_vm = vm;
+			vm->sw.io_shared_page = NULL;
+			if ((vm_config->load_order == POST_LAUNCHED_VM)
+				&& ((vm_config->guest_flags & GUEST_FLAG_IO_COMPLETION_POLLING) != 0U)) {
+				/* enable IO completion polling mode per its guest flags in vm_config. */
+				vm->sw.is_polling_ioreq = true;
+			}
+			status = set_vcpuid_entries(vm);
+			if (status == 0) {
+				vm->state = VM_CREATED;
+			}
 		}
 	}
 

hypervisor/dm/vpci/vdev.c

@@ -149,6 +149,27 @@ static void pci_vdev_update_vbar_base(struct pci_vdev *vdev, uint32_t idx)
 	vdev->vbars[idx].base_gpa = base;
 }
 
+int32_t check_pt_dev_pio_bars(struct pci_vdev *vdev)
+{
+	int32_t ret = 0;
+	uint32_t idx;
+
+	if (vdev->pdev != NULL) {
+		for (idx = 0U; idx < vdev->nr_bars; idx++) {
+			if ((is_pci_io_bar(&vdev->vbars[idx])) && (vdev->vbars[idx].base_gpa != vdev->vbars[idx].base_hpa)) {
+				ret = -EIO;
+				pr_err("%s, PCI:%02x:%02x.%x PIO BAR%d isn't identical mapping, "
+					"host start addr is 0x%lx, while guest start addr is 0x%lx",
+					__func__, vdev->bdf.bits.b, vdev->bdf.bits.d, vdev->bdf.bits.f, idx,
+					vdev->vbars[idx].base_hpa, vdev->vbars[idx].base_gpa);
+				break;
+			}
+		}
+	}
+
+	return ret;
+}
+
 void pci_vdev_write_vbar(struct pci_vdev *vdev, uint32_t idx, uint32_t val)
 {
 	struct pci_vbar *vbar;

hypervisor/dm/vpci/vpci.c

@@ -41,7 +41,7 @@
 #include <board_info.h>
 
 
-static void vpci_init_vdevs(struct acrn_vm *vm);
+static int32_t vpci_init_vdevs(struct acrn_vm *vm);
 static int32_t vpci_read_cfg(struct acrn_vpci *vpci, union pci_bdf bdf, uint32_t offset, uint32_t bytes, uint32_t *val);
 static int32_t vpci_write_cfg(struct acrn_vpci *vpci, union pci_bdf bdf, uint32_t offset, uint32_t bytes, uint32_t val);
 static struct pci_vdev *find_available_vdev(struct acrn_vpci *vpci, union pci_bdf bdf);
@@ -205,7 +205,7 @@ static int32_t vpci_mmio_cfg_access(struct io_request *io_req, void *private_dat
  * @pre vm != NULL
  * @pre vm->vm_id < CONFIG_MAX_VM_NUM
  */
-void init_vpci(struct acrn_vm *vm)
+int32_t init_vpci(struct acrn_vm *vm)
 {
 	struct vm_io_range pci_cfgaddr_range = {
 		.base = PCI_CONFIG_ADDR,
@@ -219,6 +219,7 @@ void init_vpci(struct acrn_vm *vm)
 
 	struct acrn_vm_config *vm_config;
 	struct pci_mmcfg_region *pci_mmcfg;
+	int32_t ret = 0;
 
 	vm->iommu = create_iommu_domain(vm->vm_id, hva2hpa(vm->arch_vm.nworld_eptp), 48U);
 
@@ -242,20 +243,24 @@ void init_vpci(struct acrn_vm *vm)
 	}
 
 	/* Build up vdev list for vm */
-	vpci_init_vdevs(vm);
+	ret = vpci_init_vdevs(vm);
 
-	register_mmio_emulation_handler(vm, vpci_mmio_cfg_access, vm->vpci.pci_mmcfg.address,
+	if (ret == 0) {
-		vm->vpci.pci_mmcfg.address + get_pci_mmcfg_size(&vm->vpci.pci_mmcfg), &vm->vpci, false);
+		register_mmio_emulation_handler(vm, vpci_mmio_cfg_access, vm->vpci.pci_mmcfg.address,
+			vm->vpci.pci_mmcfg.address + get_pci_mmcfg_size(&vm->vpci.pci_mmcfg), &vm->vpci, false);
 
-	/* Intercept and handle I/O ports CF8h */
+		/* Intercept and handle I/O ports CF8h */
-	register_pio_emulation_handler(vm, PCI_CFGADDR_PIO_IDX, &pci_cfgaddr_range,
+		register_pio_emulation_handler(vm, PCI_CFGADDR_PIO_IDX, &pci_cfgaddr_range,
-		vpci_pio_cfgaddr_read, vpci_pio_cfgaddr_write);
+			vpci_pio_cfgaddr_read, vpci_pio_cfgaddr_write);
 
-	/* Intercept and handle I/O ports CFCh -- CFFh */
+		/* Intercept and handle I/O ports CFCh -- CFFh */
-	register_pio_emulation_handler(vm, PCI_CFGDATA_PIO_IDX, &pci_cfgdata_range,
+		register_pio_emulation_handler(vm, PCI_CFGDATA_PIO_IDX, &pci_cfgdata_range,
-		vpci_pio_cfgdata_read, vpci_pio_cfgdata_write);
+			vpci_pio_cfgdata_read, vpci_pio_cfgdata_write);
 
-	spinlock_init(&vm->vpci.lock);
+		spinlock_init(&vm->vpci.lock);
+	}
+
+	return ret;
 }
 
 /**
@@ -666,18 +671,25 @@ struct pci_vdev *vpci_init_vdev(struct acrn_vpci *vpci, struct acrn_vm_pci_dev_c
 /**
  * @pre vm != NULL
  */
-static void vpci_init_vdevs(struct acrn_vm *vm)
+static int32_t vpci_init_vdevs(struct acrn_vm *vm)
 {
 	uint32_t idx;
 	struct acrn_vpci *vpci = &(vm->vpci);
 	const struct acrn_vm_config *vm_config = get_vm_config(vpci2vm(vpci)->vm_id);
+	int32_t ret = 0;
 
 	for (idx = 0U; idx < vm_config->pci_dev_num; idx++) {
 		/* the vdev whose vBDF is unassigned will be created by hypercall */
 		if ((!is_postlaunched_vm(vm)) || (vm_config->pci_devs[idx].vbdf.value != UNASSIGNED_VBDF)) {
 			(void)vpci_init_vdev(vpci, &vm_config->pci_devs[idx], NULL);
+			ret = check_pt_dev_pio_bars(&vpci->pci_vdevs[idx]);
+			if (ret != 0) {
+				break;
+			}
 		}
 	}
+
+	return ret;
 }
 
 /**
@@ -733,14 +745,21 @@ int32_t vpci_assign_pcidev(struct acrn_vm *tgt_vm, struct acrn_pcidev *pcidev)
 			pci_vdev_write_vbar(vdev, idx, pcidev->bar[idx]);
 		}
 
-		vdev->flags |= pcidev->type;
+		ret = check_pt_dev_pio_bars(vdev);
-		vdev->bdf.value = pcidev->virt_bdf;
+
-		/*We should re-add the vdev to hashlist since its vbdf has changed */
+		if (ret == 0) {
-		hlist_del(&vdev->link);
+			vdev->flags |= pcidev->type;
-		hlist_add_head(&vdev->link, &vpci->vdevs_hlist_heads[hash64(vdev->bdf.value, VDEV_LIST_HASHBITS)]);
+			vdev->bdf.value = pcidev->virt_bdf;
-		vdev->parent_user = vdev_in_sos;
+			/*We should re-add the vdev to hashlist since its vbdf has changed */
+			hlist_del(&vdev->link);
+			hlist_add_head(&vdev->link, &vpci->vdevs_hlist_heads[hash64(vdev->bdf.value, VDEV_LIST_HASHBITS)]);
+			vdev->parent_user = vdev_in_sos;
+			vdev_in_sos->user = vdev;
+		} else {
+			vdev->vdev_ops->deinit_vdev(vdev);
+			vdev_in_sos->vdev_ops->init_vdev(vdev_in_sos);
+		}
 		spinlock_release(&tgt_vm->vpci.lock);
-		vdev_in_sos->user = vdev;
 	} else {
 		pr_fatal("%s, can't find PCI device %x:%x.%x for vm[%d] %x:%x.%x\n", __func__,
 			pcidev->phys_bdf >> 8U, (pcidev->phys_bdf >> 3U) & 0x1fU, pcidev->phys_bdf & 0x7U,

hypervisor/dm/vpci/vpci_priv.h

@@ -178,6 +178,8 @@ void pci_vdev_write_vbar(struct pci_vdev *vdev, uint32_t idx, uint32_t val);
 
 void vdev_pt_hide_sriov_cap(struct pci_vdev *vdev);
 
+int32_t check_pt_dev_pio_bars(struct pci_vdev *vdev);
+
 typedef void (*map_pcibar)(struct pci_vdev *vdev, uint32_t bar_idx);
 typedef void (*unmap_pcibar)(struct pci_vdev *vdev, uint32_t bar_idx);
 void vpci_update_one_vbar(struct pci_vdev *vdev, uint32_t bar_idx, uint32_t val, map_pcibar map_cb, unmap_pcibar unmap_cb);

hypervisor/include/dm/vpci.h

@@ -184,7 +184,7 @@ struct acrn_vm;
 
 extern const struct pci_vdev_ops vhostbridge_ops;
 extern const struct pci_vdev_ops vpci_bridge_ops;
-void init_vpci(struct acrn_vm *vm);
+int32_t init_vpci(struct acrn_vm *vm);
 void deinit_vpci(struct acrn_vm *vm);
 struct pci_vdev *pci_find_vdev(struct acrn_vpci *vpci, union pci_bdf vbdf);
 struct acrn_pcidev;