1 /* 2 * QEMU Machine 3 * 4 * Copyright (C) 2014 Red Hat Inc 5 * 6 * Authors: 7 * Marcel Apfelbaum <marcel.a@redhat.com> 8 * 9 * This work is licensed under the terms of the GNU GPL, version 2 or later. 10 * See the COPYING file in the top-level directory. 11 */ 12 13 #include "qemu/osdep.h" 14 #include "qemu/option.h" 15 #include "qapi/qmp/qerror.h" 16 #include "sysemu/replay.h" 17 #include "qemu/units.h" 18 #include "hw/boards.h" 19 #include "hw/loader.h" 20 #include "qapi/error.h" 21 #include "qapi/qapi-visit-common.h" 22 #include "qapi/qapi-visit-machine.h" 23 #include "qapi/visitor.h" 24 #include "hw/sysbus.h" 25 #include "sysemu/cpus.h" 26 #include "sysemu/sysemu.h" 27 #include "sysemu/reset.h" 28 #include "sysemu/runstate.h" 29 #include "sysemu/numa.h" 30 #include "qemu/error-report.h" 31 #include "sysemu/qtest.h" 32 #include "hw/pci/pci.h" 33 #include "hw/mem/nvdimm.h" 34 #include "migration/global_state.h" 35 #include "migration/vmstate.h" 36 #include "exec/confidential-guest-support.h" 37 #include "hw/virtio/virtio.h" 38 #include "hw/virtio/virtio-pci.h" 39 40 GlobalProperty hw_compat_6_1[] = { 41 { "vhost-user-vsock-device", "seqpacket", "off" }, 42 { "nvme-ns", "shared", "off" }, 43 }; 44 const size_t hw_compat_6_1_len = G_N_ELEMENTS(hw_compat_6_1); 45 46 GlobalProperty hw_compat_6_0[] = { 47 { "gpex-pcihost", "allow-unmapped-accesses", "false" }, 48 { "i8042", "extended-state", "false"}, 49 { "nvme-ns", "eui64-default", "off"}, 50 { "e1000", "init-vet", "off" }, 51 { "e1000e", "init-vet", "off" }, 52 { "vhost-vsock-device", "seqpacket", "off" }, 53 }; 54 const size_t hw_compat_6_0_len = G_N_ELEMENTS(hw_compat_6_0); 55 56 GlobalProperty hw_compat_5_2[] = { 57 { "ICH9-LPC", "smm-compat", "on"}, 58 { "PIIX4_PM", "smm-compat", "on"}, 59 { "virtio-blk-device", "report-discard-granularity", "off" }, 60 { "virtio-net-pci-base", "vectors", "3"}, 61 }; 62 const size_t hw_compat_5_2_len = G_N_ELEMENTS(hw_compat_5_2); 63 64 GlobalProperty hw_compat_5_1[] = { 65 { "vhost-scsi", "num_queues", "1"}, 66 { "vhost-user-blk", "num-queues", "1"}, 67 { "vhost-user-scsi", "num_queues", "1"}, 68 { "virtio-blk-device", "num-queues", "1"}, 69 { "virtio-scsi-device", "num_queues", "1"}, 70 { "nvme", "use-intel-id", "on"}, 71 { "pvpanic", "events", "1"}, /* PVPANIC_PANICKED */ 72 { "pl011", "migrate-clk", "off" }, 73 { "virtio-pci", "x-ats-page-aligned", "off"}, 74 }; 75 const size_t hw_compat_5_1_len = G_N_ELEMENTS(hw_compat_5_1); 76 77 GlobalProperty hw_compat_5_0[] = { 78 { "pci-host-bridge", "x-config-reg-migration-enabled", "off" }, 79 { "virtio-balloon-device", "page-poison", "false" }, 80 { "vmport", "x-read-set-eax", "off" }, 81 { "vmport", "x-signal-unsupported-cmd", "off" }, 82 { "vmport", "x-report-vmx-type", "off" }, 83 { "vmport", "x-cmds-v2", "off" }, 84 { "virtio-device", "x-disable-legacy-check", "true" }, 85 }; 86 const size_t hw_compat_5_0_len = G_N_ELEMENTS(hw_compat_5_0); 87 88 GlobalProperty hw_compat_4_2[] = { 89 { "virtio-blk-device", "queue-size", "128"}, 90 { "virtio-scsi-device", "virtqueue_size", "128"}, 91 { "virtio-blk-device", "x-enable-wce-if-config-wce", "off" }, 92 { "virtio-blk-device", "seg-max-adjust", "off"}, 93 { "virtio-scsi-device", "seg_max_adjust", "off"}, 94 { "vhost-blk-device", "seg_max_adjust", "off"}, 95 { "usb-host", "suppress-remote-wake", "off" }, 96 { "usb-redir", "suppress-remote-wake", "off" }, 97 { "qxl", "revision", "4" }, 98 { "qxl-vga", "revision", "4" }, 99 { "fw_cfg", "acpi-mr-restore", "false" }, 100 { "virtio-device", "use-disabled-flag", "false" }, 101 }; 102 const size_t hw_compat_4_2_len = G_N_ELEMENTS(hw_compat_4_2); 103 104 GlobalProperty hw_compat_4_1[] = { 105 { "virtio-pci", "x-pcie-flr-init", "off" }, 106 }; 107 const size_t hw_compat_4_1_len = G_N_ELEMENTS(hw_compat_4_1); 108 109 GlobalProperty hw_compat_4_0[] = { 110 { "VGA", "edid", "false" }, 111 { "secondary-vga", "edid", "false" }, 112 { "bochs-display", "edid", "false" }, 113 { "virtio-vga", "edid", "false" }, 114 { "virtio-gpu-device", "edid", "false" }, 115 { "virtio-device", "use-started", "false" }, 116 { "virtio-balloon-device", "qemu-4-0-config-size", "true" }, 117 { "pl031", "migrate-tick-offset", "false" }, 118 }; 119 const size_t hw_compat_4_0_len = G_N_ELEMENTS(hw_compat_4_0); 120 121 GlobalProperty hw_compat_3_1[] = { 122 { "pcie-root-port", "x-speed", "2_5" }, 123 { "pcie-root-port", "x-width", "1" }, 124 { "memory-backend-file", "x-use-canonical-path-for-ramblock-id", "true" }, 125 { "memory-backend-memfd", "x-use-canonical-path-for-ramblock-id", "true" }, 126 { "tpm-crb", "ppi", "false" }, 127 { "tpm-tis", "ppi", "false" }, 128 { "usb-kbd", "serial", "42" }, 129 { "usb-mouse", "serial", "42" }, 130 { "usb-tablet", "serial", "42" }, 131 { "virtio-blk-device", "discard", "false" }, 132 { "virtio-blk-device", "write-zeroes", "false" }, 133 { "virtio-balloon-device", "qemu-4-0-config-size", "false" }, 134 { "pcie-root-port-base", "disable-acs", "true" }, /* Added in 4.1 */ 135 }; 136 const size_t hw_compat_3_1_len = G_N_ELEMENTS(hw_compat_3_1); 137 138 GlobalProperty hw_compat_3_0[] = {}; 139 const size_t hw_compat_3_0_len = G_N_ELEMENTS(hw_compat_3_0); 140 141 GlobalProperty hw_compat_2_12[] = { 142 { "migration", "decompress-error-check", "off" }, 143 { "hda-audio", "use-timer", "false" }, 144 { "cirrus-vga", "global-vmstate", "true" }, 145 { "VGA", "global-vmstate", "true" }, 146 { "vmware-svga", "global-vmstate", "true" }, 147 { "qxl-vga", "global-vmstate", "true" }, 148 }; 149 const size_t hw_compat_2_12_len = G_N_ELEMENTS(hw_compat_2_12); 150 151 GlobalProperty hw_compat_2_11[] = { 152 { "hpet", "hpet-offset-saved", "false" }, 153 { "virtio-blk-pci", "vectors", "2" }, 154 { "vhost-user-blk-pci", "vectors", "2" }, 155 { "e1000", "migrate_tso_props", "off" }, 156 }; 157 const size_t hw_compat_2_11_len = G_N_ELEMENTS(hw_compat_2_11); 158 159 GlobalProperty hw_compat_2_10[] = { 160 { "virtio-mouse-device", "wheel-axis", "false" }, 161 { "virtio-tablet-device", "wheel-axis", "false" }, 162 }; 163 const size_t hw_compat_2_10_len = G_N_ELEMENTS(hw_compat_2_10); 164 165 GlobalProperty hw_compat_2_9[] = { 166 { "pci-bridge", "shpc", "off" }, 167 { "intel-iommu", "pt", "off" }, 168 { "virtio-net-device", "x-mtu-bypass-backend", "off" }, 169 { "pcie-root-port", "x-migrate-msix", "false" }, 170 }; 171 const size_t hw_compat_2_9_len = G_N_ELEMENTS(hw_compat_2_9); 172 173 GlobalProperty hw_compat_2_8[] = { 174 { "fw_cfg_mem", "x-file-slots", "0x10" }, 175 { "fw_cfg_io", "x-file-slots", "0x10" }, 176 { "pflash_cfi01", "old-multiple-chip-handling", "on" }, 177 { "pci-bridge", "shpc", "on" }, 178 { TYPE_PCI_DEVICE, "x-pcie-extcap-init", "off" }, 179 { "virtio-pci", "x-pcie-deverr-init", "off" }, 180 { "virtio-pci", "x-pcie-lnkctl-init", "off" }, 181 { "virtio-pci", "x-pcie-pm-init", "off" }, 182 { "cirrus-vga", "vgamem_mb", "8" }, 183 { "isa-cirrus-vga", "vgamem_mb", "8" }, 184 }; 185 const size_t hw_compat_2_8_len = G_N_ELEMENTS(hw_compat_2_8); 186 187 GlobalProperty hw_compat_2_7[] = { 188 { "virtio-pci", "page-per-vq", "on" }, 189 { "virtio-serial-device", "emergency-write", "off" }, 190 { "ioapic", "version", "0x11" }, 191 { "intel-iommu", "x-buggy-eim", "true" }, 192 { "virtio-pci", "x-ignore-backend-features", "on" }, 193 }; 194 const size_t hw_compat_2_7_len = G_N_ELEMENTS(hw_compat_2_7); 195 196 GlobalProperty hw_compat_2_6[] = { 197 { "virtio-mmio", "format_transport_address", "off" }, 198 /* Optional because not all virtio-pci devices support legacy mode */ 199 { "virtio-pci", "disable-modern", "on", .optional = true }, 200 { "virtio-pci", "disable-legacy", "off", .optional = true }, 201 }; 202 const size_t hw_compat_2_6_len = G_N_ELEMENTS(hw_compat_2_6); 203 204 GlobalProperty hw_compat_2_5[] = { 205 { "isa-fdc", "fallback", "144" }, 206 { "pvscsi", "x-old-pci-configuration", "on" }, 207 { "pvscsi", "x-disable-pcie", "on" }, 208 { "vmxnet3", "x-old-msi-offsets", "on" }, 209 { "vmxnet3", "x-disable-pcie", "on" }, 210 }; 211 const size_t hw_compat_2_5_len = G_N_ELEMENTS(hw_compat_2_5); 212 213 GlobalProperty hw_compat_2_4[] = { 214 /* Optional because the 'scsi' property is Linux-only */ 215 { "virtio-blk-device", "scsi", "true", .optional = true }, 216 { "e1000", "extra_mac_registers", "off" }, 217 { "virtio-pci", "x-disable-pcie", "on" }, 218 { "virtio-pci", "migrate-extra", "off" }, 219 { "fw_cfg_mem", "dma_enabled", "off" }, 220 { "fw_cfg_io", "dma_enabled", "off" } 221 }; 222 const size_t hw_compat_2_4_len = G_N_ELEMENTS(hw_compat_2_4); 223 224 GlobalProperty hw_compat_2_3[] = { 225 { "virtio-blk-pci", "any_layout", "off" }, 226 { "virtio-balloon-pci", "any_layout", "off" }, 227 { "virtio-serial-pci", "any_layout", "off" }, 228 { "virtio-9p-pci", "any_layout", "off" }, 229 { "virtio-rng-pci", "any_layout", "off" }, 230 { TYPE_PCI_DEVICE, "x-pcie-lnksta-dllla", "off" }, 231 { "migration", "send-configuration", "off" }, 232 { "migration", "send-section-footer", "off" }, 233 { "migration", "store-global-state", "off" }, 234 }; 235 const size_t hw_compat_2_3_len = G_N_ELEMENTS(hw_compat_2_3); 236 237 GlobalProperty hw_compat_2_2[] = {}; 238 const size_t hw_compat_2_2_len = G_N_ELEMENTS(hw_compat_2_2); 239 240 GlobalProperty hw_compat_2_1[] = { 241 { "intel-hda", "old_msi_addr", "on" }, 242 { "VGA", "qemu-extended-regs", "off" }, 243 { "secondary-vga", "qemu-extended-regs", "off" }, 244 { "virtio-scsi-pci", "any_layout", "off" }, 245 { "usb-mouse", "usb_version", "1" }, 246 { "usb-kbd", "usb_version", "1" }, 247 { "virtio-pci", "virtio-pci-bus-master-bug-migration", "on" }, 248 }; 249 const size_t hw_compat_2_1_len = G_N_ELEMENTS(hw_compat_2_1); 250 251 MachineState *current_machine; 252 253 static char *machine_get_kernel(Object *obj, Error **errp) 254 { 255 MachineState *ms = MACHINE(obj); 256 257 return g_strdup(ms->kernel_filename); 258 } 259 260 static void machine_set_kernel(Object *obj, const char *value, Error **errp) 261 { 262 MachineState *ms = MACHINE(obj); 263 264 g_free(ms->kernel_filename); 265 ms->kernel_filename = g_strdup(value); 266 } 267 268 static char *machine_get_initrd(Object *obj, Error **errp) 269 { 270 MachineState *ms = MACHINE(obj); 271 272 return g_strdup(ms->initrd_filename); 273 } 274 275 static void machine_set_initrd(Object *obj, const char *value, Error **errp) 276 { 277 MachineState *ms = MACHINE(obj); 278 279 g_free(ms->initrd_filename); 280 ms->initrd_filename = g_strdup(value); 281 } 282 283 static char *machine_get_append(Object *obj, Error **errp) 284 { 285 MachineState *ms = MACHINE(obj); 286 287 return g_strdup(ms->kernel_cmdline); 288 } 289 290 static void machine_set_append(Object *obj, const char *value, Error **errp) 291 { 292 MachineState *ms = MACHINE(obj); 293 294 g_free(ms->kernel_cmdline); 295 ms->kernel_cmdline = g_strdup(value); 296 } 297 298 static char *machine_get_dtb(Object *obj, Error **errp) 299 { 300 MachineState *ms = MACHINE(obj); 301 302 return g_strdup(ms->dtb); 303 } 304 305 static void machine_set_dtb(Object *obj, const char *value, Error **errp) 306 { 307 MachineState *ms = MACHINE(obj); 308 309 g_free(ms->dtb); 310 ms->dtb = g_strdup(value); 311 } 312 313 static char *machine_get_dumpdtb(Object *obj, Error **errp) 314 { 315 MachineState *ms = MACHINE(obj); 316 317 return g_strdup(ms->dumpdtb); 318 } 319 320 static void machine_set_dumpdtb(Object *obj, const char *value, Error **errp) 321 { 322 MachineState *ms = MACHINE(obj); 323 324 g_free(ms->dumpdtb); 325 ms->dumpdtb = g_strdup(value); 326 } 327 328 static void machine_get_phandle_start(Object *obj, Visitor *v, 329 const char *name, void *opaque, 330 Error **errp) 331 { 332 MachineState *ms = MACHINE(obj); 333 int64_t value = ms->phandle_start; 334 335 visit_type_int(v, name, &value, errp); 336 } 337 338 static void machine_set_phandle_start(Object *obj, Visitor *v, 339 const char *name, void *opaque, 340 Error **errp) 341 { 342 MachineState *ms = MACHINE(obj); 343 int64_t value; 344 345 if (!visit_type_int(v, name, &value, errp)) { 346 return; 347 } 348 349 ms->phandle_start = value; 350 } 351 352 static char *machine_get_dt_compatible(Object *obj, Error **errp) 353 { 354 MachineState *ms = MACHINE(obj); 355 356 return g_strdup(ms->dt_compatible); 357 } 358 359 static void machine_set_dt_compatible(Object *obj, const char *value, Error **errp) 360 { 361 MachineState *ms = MACHINE(obj); 362 363 g_free(ms->dt_compatible); 364 ms->dt_compatible = g_strdup(value); 365 } 366 367 static bool machine_get_dump_guest_core(Object *obj, Error **errp) 368 { 369 MachineState *ms = MACHINE(obj); 370 371 return ms->dump_guest_core; 372 } 373 374 static void machine_set_dump_guest_core(Object *obj, bool value, Error **errp) 375 { 376 MachineState *ms = MACHINE(obj); 377 378 ms->dump_guest_core = value; 379 } 380 381 static bool machine_get_mem_merge(Object *obj, Error **errp) 382 { 383 MachineState *ms = MACHINE(obj); 384 385 return ms->mem_merge; 386 } 387 388 static void machine_set_mem_merge(Object *obj, bool value, Error **errp) 389 { 390 MachineState *ms = MACHINE(obj); 391 392 ms->mem_merge = value; 393 } 394 395 static bool machine_get_usb(Object *obj, Error **errp) 396 { 397 MachineState *ms = MACHINE(obj); 398 399 return ms->usb; 400 } 401 402 static void machine_set_usb(Object *obj, bool value, Error **errp) 403 { 404 MachineState *ms = MACHINE(obj); 405 406 ms->usb = value; 407 ms->usb_disabled = !value; 408 } 409 410 static bool machine_get_graphics(Object *obj, Error **errp) 411 { 412 MachineState *ms = MACHINE(obj); 413 414 return ms->enable_graphics; 415 } 416 417 static void machine_set_graphics(Object *obj, bool value, Error **errp) 418 { 419 MachineState *ms = MACHINE(obj); 420 421 ms->enable_graphics = value; 422 } 423 424 static char *machine_get_firmware(Object *obj, Error **errp) 425 { 426 MachineState *ms = MACHINE(obj); 427 428 return g_strdup(ms->firmware); 429 } 430 431 static void machine_set_firmware(Object *obj, const char *value, Error **errp) 432 { 433 MachineState *ms = MACHINE(obj); 434 435 g_free(ms->firmware); 436 ms->firmware = g_strdup(value); 437 } 438 439 static void machine_set_suppress_vmdesc(Object *obj, bool value, Error **errp) 440 { 441 MachineState *ms = MACHINE(obj); 442 443 ms->suppress_vmdesc = value; 444 } 445 446 static bool machine_get_suppress_vmdesc(Object *obj, Error **errp) 447 { 448 MachineState *ms = MACHINE(obj); 449 450 return ms->suppress_vmdesc; 451 } 452 453 static char *machine_get_memory_encryption(Object *obj, Error **errp) 454 { 455 MachineState *ms = MACHINE(obj); 456 457 if (ms->cgs) { 458 return g_strdup(object_get_canonical_path_component(OBJECT(ms->cgs))); 459 } 460 461 return NULL; 462 } 463 464 static void machine_set_memory_encryption(Object *obj, const char *value, 465 Error **errp) 466 { 467 Object *cgs = 468 object_resolve_path_component(object_get_objects_root(), value); 469 470 if (!cgs) { 471 error_setg(errp, "No such memory encryption object '%s'", value); 472 return; 473 } 474 475 object_property_set_link(obj, "confidential-guest-support", cgs, errp); 476 } 477 478 static void machine_check_confidential_guest_support(const Object *obj, 479 const char *name, 480 Object *new_target, 481 Error **errp) 482 { 483 /* 484 * So far the only constraint is that the target has the 485 * TYPE_CONFIDENTIAL_GUEST_SUPPORT interface, and that's checked 486 * by the QOM core 487 */ 488 } 489 490 static bool machine_get_nvdimm(Object *obj, Error **errp) 491 { 492 MachineState *ms = MACHINE(obj); 493 494 return ms->nvdimms_state->is_enabled; 495 } 496 497 static void machine_set_nvdimm(Object *obj, bool value, Error **errp) 498 { 499 MachineState *ms = MACHINE(obj); 500 501 ms->nvdimms_state->is_enabled = value; 502 } 503 504 static bool machine_get_hmat(Object *obj, Error **errp) 505 { 506 MachineState *ms = MACHINE(obj); 507 508 return ms->numa_state->hmat_enabled; 509 } 510 511 static void machine_set_hmat(Object *obj, bool value, Error **errp) 512 { 513 MachineState *ms = MACHINE(obj); 514 515 ms->numa_state->hmat_enabled = value; 516 } 517 518 static char *machine_get_nvdimm_persistence(Object *obj, Error **errp) 519 { 520 MachineState *ms = MACHINE(obj); 521 522 return g_strdup(ms->nvdimms_state->persistence_string); 523 } 524 525 static void machine_set_nvdimm_persistence(Object *obj, const char *value, 526 Error **errp) 527 { 528 MachineState *ms = MACHINE(obj); 529 NVDIMMState *nvdimms_state = ms->nvdimms_state; 530 531 if (strcmp(value, "cpu") == 0) { 532 nvdimms_state->persistence = 3; 533 } else if (strcmp(value, "mem-ctrl") == 0) { 534 nvdimms_state->persistence = 2; 535 } else { 536 error_setg(errp, "-machine nvdimm-persistence=%s: unsupported option", 537 value); 538 return; 539 } 540 541 g_free(nvdimms_state->persistence_string); 542 nvdimms_state->persistence_string = g_strdup(value); 543 } 544 545 void machine_class_allow_dynamic_sysbus_dev(MachineClass *mc, const char *type) 546 { 547 QAPI_LIST_PREPEND(mc->allowed_dynamic_sysbus_devices, g_strdup(type)); 548 } 549 550 bool device_is_dynamic_sysbus(MachineClass *mc, DeviceState *dev) 551 { 552 Object *obj = OBJECT(dev); 553 554 if (!object_dynamic_cast(obj, TYPE_SYS_BUS_DEVICE)) { 555 return false; 556 } 557 558 return device_type_is_dynamic_sysbus(mc, object_get_typename(obj)); 559 } 560 561 bool device_type_is_dynamic_sysbus(MachineClass *mc, const char *type) 562 { 563 bool allowed = false; 564 strList *wl; 565 ObjectClass *klass = object_class_by_name(type); 566 567 for (wl = mc->allowed_dynamic_sysbus_devices; 568 !allowed && wl; 569 wl = wl->next) { 570 allowed |= !!object_class_dynamic_cast(klass, wl->value); 571 } 572 573 return allowed; 574 } 575 576 static char *machine_get_memdev(Object *obj, Error **errp) 577 { 578 MachineState *ms = MACHINE(obj); 579 580 return g_strdup(ms->ram_memdev_id); 581 } 582 583 static void machine_set_memdev(Object *obj, const char *value, Error **errp) 584 { 585 MachineState *ms = MACHINE(obj); 586 587 g_free(ms->ram_memdev_id); 588 ms->ram_memdev_id = g_strdup(value); 589 } 590 591 HotpluggableCPUList *machine_query_hotpluggable_cpus(MachineState *machine) 592 { 593 int i; 594 HotpluggableCPUList *head = NULL; 595 MachineClass *mc = MACHINE_GET_CLASS(machine); 596 597 /* force board to initialize possible_cpus if it hasn't been done yet */ 598 mc->possible_cpu_arch_ids(machine); 599 600 for (i = 0; i < machine->possible_cpus->len; i++) { 601 Object *cpu; 602 HotpluggableCPU *cpu_item = g_new0(typeof(*cpu_item), 1); 603 604 cpu_item->type = g_strdup(machine->possible_cpus->cpus[i].type); 605 cpu_item->vcpus_count = machine->possible_cpus->cpus[i].vcpus_count; 606 cpu_item->props = g_memdup(&machine->possible_cpus->cpus[i].props, 607 sizeof(*cpu_item->props)); 608 609 cpu = machine->possible_cpus->cpus[i].cpu; 610 if (cpu) { 611 cpu_item->has_qom_path = true; 612 cpu_item->qom_path = object_get_canonical_path(cpu); 613 } 614 QAPI_LIST_PREPEND(head, cpu_item); 615 } 616 return head; 617 } 618 619 /** 620 * machine_set_cpu_numa_node: 621 * @machine: machine object to modify 622 * @props: specifies which cpu objects to assign to 623 * numa node specified by @props.node_id 624 * @errp: if an error occurs, a pointer to an area to store the error 625 * 626 * Associate NUMA node specified by @props.node_id with cpu slots that 627 * match socket/core/thread-ids specified by @props. It's recommended to use 628 * query-hotpluggable-cpus.props values to specify affected cpu slots, 629 * which would lead to exact 1:1 mapping of cpu slots to NUMA node. 630 * 631 * However for CLI convenience it's possible to pass in subset of properties, 632 * which would affect all cpu slots that match it. 633 * Ex for pc machine: 634 * -smp 4,cores=2,sockets=2 -numa node,nodeid=0 -numa node,nodeid=1 \ 635 * -numa cpu,node-id=0,socket_id=0 \ 636 * -numa cpu,node-id=1,socket_id=1 637 * will assign all child cores of socket 0 to node 0 and 638 * of socket 1 to node 1. 639 * 640 * On attempt of reassigning (already assigned) cpu slot to another NUMA node, 641 * return error. 642 * Empty subset is disallowed and function will return with error in this case. 643 */ 644 void machine_set_cpu_numa_node(MachineState *machine, 645 const CpuInstanceProperties *props, Error **errp) 646 { 647 MachineClass *mc = MACHINE_GET_CLASS(machine); 648 NodeInfo *numa_info = machine->numa_state->nodes; 649 bool match = false; 650 int i; 651 652 if (!mc->possible_cpu_arch_ids) { 653 error_setg(errp, "mapping of CPUs to NUMA node is not supported"); 654 return; 655 } 656 657 /* disabling node mapping is not supported, forbid it */ 658 assert(props->has_node_id); 659 660 /* force board to initialize possible_cpus if it hasn't been done yet */ 661 mc->possible_cpu_arch_ids(machine); 662 663 for (i = 0; i < machine->possible_cpus->len; i++) { 664 CPUArchId *slot = &machine->possible_cpus->cpus[i]; 665 666 /* reject unsupported by board properties */ 667 if (props->has_thread_id && !slot->props.has_thread_id) { 668 error_setg(errp, "thread-id is not supported"); 669 return; 670 } 671 672 if (props->has_core_id && !slot->props.has_core_id) { 673 error_setg(errp, "core-id is not supported"); 674 return; 675 } 676 677 if (props->has_socket_id && !slot->props.has_socket_id) { 678 error_setg(errp, "socket-id is not supported"); 679 return; 680 } 681 682 if (props->has_die_id && !slot->props.has_die_id) { 683 error_setg(errp, "die-id is not supported"); 684 return; 685 } 686 687 /* skip slots with explicit mismatch */ 688 if (props->has_thread_id && props->thread_id != slot->props.thread_id) { 689 continue; 690 } 691 692 if (props->has_core_id && props->core_id != slot->props.core_id) { 693 continue; 694 } 695 696 if (props->has_die_id && props->die_id != slot->props.die_id) { 697 continue; 698 } 699 700 if (props->has_socket_id && props->socket_id != slot->props.socket_id) { 701 continue; 702 } 703 704 /* reject assignment if slot is already assigned, for compatibility 705 * of legacy cpu_index mapping with SPAPR core based mapping do not 706 * error out if cpu thread and matched core have the same node-id */ 707 if (slot->props.has_node_id && 708 slot->props.node_id != props->node_id) { 709 error_setg(errp, "CPU is already assigned to node-id: %" PRId64, 710 slot->props.node_id); 711 return; 712 } 713 714 /* assign slot to node as it's matched '-numa cpu' key */ 715 match = true; 716 slot->props.node_id = props->node_id; 717 slot->props.has_node_id = props->has_node_id; 718 719 if (machine->numa_state->hmat_enabled) { 720 if ((numa_info[props->node_id].initiator < MAX_NODES) && 721 (props->node_id != numa_info[props->node_id].initiator)) { 722 error_setg(errp, "The initiator of CPU NUMA node %" PRId64 723 " should be itself (got %" PRIu16 ")", 724 props->node_id, numa_info[props->node_id].initiator); 725 return; 726 } 727 numa_info[props->node_id].has_cpu = true; 728 numa_info[props->node_id].initiator = props->node_id; 729 } 730 } 731 732 if (!match) { 733 error_setg(errp, "no match found"); 734 } 735 } 736 737 static void machine_get_smp(Object *obj, Visitor *v, const char *name, 738 void *opaque, Error **errp) 739 { 740 MachineState *ms = MACHINE(obj); 741 SMPConfiguration *config = &(SMPConfiguration){ 742 .has_cpus = true, .cpus = ms->smp.cpus, 743 .has_sockets = true, .sockets = ms->smp.sockets, 744 .has_dies = true, .dies = ms->smp.dies, 745 .has_clusters = true, .clusters = ms->smp.clusters, 746 .has_cores = true, .cores = ms->smp.cores, 747 .has_threads = true, .threads = ms->smp.threads, 748 .has_maxcpus = true, .maxcpus = ms->smp.max_cpus, 749 }; 750 751 if (!visit_type_SMPConfiguration(v, name, &config, &error_abort)) { 752 return; 753 } 754 } 755 756 static void machine_set_smp(Object *obj, Visitor *v, const char *name, 757 void *opaque, Error **errp) 758 { 759 MachineState *ms = MACHINE(obj); 760 g_autoptr(SMPConfiguration) config = NULL; 761 762 if (!visit_type_SMPConfiguration(v, name, &config, errp)) { 763 return; 764 } 765 766 machine_parse_smp_config(ms, config, errp); 767 } 768 769 static void machine_class_init(ObjectClass *oc, void *data) 770 { 771 MachineClass *mc = MACHINE_CLASS(oc); 772 773 /* Default 128 MB as guest ram size */ 774 mc->default_ram_size = 128 * MiB; 775 mc->rom_file_has_mr = true; 776 777 /* numa node memory size aligned on 8MB by default. 778 * On Linux, each node's border has to be 8MB aligned 779 */ 780 mc->numa_mem_align_shift = 23; 781 782 object_class_property_add_str(oc, "kernel", 783 machine_get_kernel, machine_set_kernel); 784 object_class_property_set_description(oc, "kernel", 785 "Linux kernel image file"); 786 787 object_class_property_add_str(oc, "initrd", 788 machine_get_initrd, machine_set_initrd); 789 object_class_property_set_description(oc, "initrd", 790 "Linux initial ramdisk file"); 791 792 object_class_property_add_str(oc, "append", 793 machine_get_append, machine_set_append); 794 object_class_property_set_description(oc, "append", 795 "Linux kernel command line"); 796 797 object_class_property_add_str(oc, "dtb", 798 machine_get_dtb, machine_set_dtb); 799 object_class_property_set_description(oc, "dtb", 800 "Linux kernel device tree file"); 801 802 object_class_property_add_str(oc, "dumpdtb", 803 machine_get_dumpdtb, machine_set_dumpdtb); 804 object_class_property_set_description(oc, "dumpdtb", 805 "Dump current dtb to a file and quit"); 806 807 object_class_property_add(oc, "smp", "SMPConfiguration", 808 machine_get_smp, machine_set_smp, 809 NULL, NULL); 810 object_class_property_set_description(oc, "smp", 811 "CPU topology"); 812 813 object_class_property_add(oc, "phandle-start", "int", 814 machine_get_phandle_start, machine_set_phandle_start, 815 NULL, NULL); 816 object_class_property_set_description(oc, "phandle-start", 817 "The first phandle ID we may generate dynamically"); 818 819 object_class_property_add_str(oc, "dt-compatible", 820 machine_get_dt_compatible, machine_set_dt_compatible); 821 object_class_property_set_description(oc, "dt-compatible", 822 "Overrides the \"compatible\" property of the dt root node"); 823 824 object_class_property_add_bool(oc, "dump-guest-core", 825 machine_get_dump_guest_core, machine_set_dump_guest_core); 826 object_class_property_set_description(oc, "dump-guest-core", 827 "Include guest memory in a core dump"); 828 829 object_class_property_add_bool(oc, "mem-merge", 830 machine_get_mem_merge, machine_set_mem_merge); 831 object_class_property_set_description(oc, "mem-merge", 832 "Enable/disable memory merge support"); 833 834 object_class_property_add_bool(oc, "usb", 835 machine_get_usb, machine_set_usb); 836 object_class_property_set_description(oc, "usb", 837 "Set on/off to enable/disable usb"); 838 839 object_class_property_add_bool(oc, "graphics", 840 machine_get_graphics, machine_set_graphics); 841 object_class_property_set_description(oc, "graphics", 842 "Set on/off to enable/disable graphics emulation"); 843 844 object_class_property_add_str(oc, "firmware", 845 machine_get_firmware, machine_set_firmware); 846 object_class_property_set_description(oc, "firmware", 847 "Firmware image"); 848 849 object_class_property_add_bool(oc, "suppress-vmdesc", 850 machine_get_suppress_vmdesc, machine_set_suppress_vmdesc); 851 object_class_property_set_description(oc, "suppress-vmdesc", 852 "Set on to disable self-describing migration"); 853 854 object_class_property_add_link(oc, "confidential-guest-support", 855 TYPE_CONFIDENTIAL_GUEST_SUPPORT, 856 offsetof(MachineState, cgs), 857 machine_check_confidential_guest_support, 858 OBJ_PROP_LINK_STRONG); 859 object_class_property_set_description(oc, "confidential-guest-support", 860 "Set confidential guest scheme to support"); 861 862 /* For compatibility */ 863 object_class_property_add_str(oc, "memory-encryption", 864 machine_get_memory_encryption, machine_set_memory_encryption); 865 object_class_property_set_description(oc, "memory-encryption", 866 "Set memory encryption object to use"); 867 868 object_class_property_add_str(oc, "memory-backend", 869 machine_get_memdev, machine_set_memdev); 870 object_class_property_set_description(oc, "memory-backend", 871 "Set RAM backend" 872 "Valid value is ID of hostmem based backend"); 873 } 874 875 static void machine_class_base_init(ObjectClass *oc, void *data) 876 { 877 MachineClass *mc = MACHINE_CLASS(oc); 878 mc->max_cpus = mc->max_cpus ?: 1; 879 mc->min_cpus = mc->min_cpus ?: 1; 880 mc->default_cpus = mc->default_cpus ?: 1; 881 882 if (!object_class_is_abstract(oc)) { 883 const char *cname = object_class_get_name(oc); 884 assert(g_str_has_suffix(cname, TYPE_MACHINE_SUFFIX)); 885 mc->name = g_strndup(cname, 886 strlen(cname) - strlen(TYPE_MACHINE_SUFFIX)); 887 mc->compat_props = g_ptr_array_new(); 888 } 889 } 890 891 static void machine_initfn(Object *obj) 892 { 893 MachineState *ms = MACHINE(obj); 894 MachineClass *mc = MACHINE_GET_CLASS(obj); 895 896 container_get(obj, "/peripheral"); 897 container_get(obj, "/peripheral-anon"); 898 899 ms->dump_guest_core = true; 900 ms->mem_merge = true; 901 ms->enable_graphics = true; 902 ms->kernel_cmdline = g_strdup(""); 903 904 if (mc->nvdimm_supported) { 905 Object *obj = OBJECT(ms); 906 907 ms->nvdimms_state = g_new0(NVDIMMState, 1); 908 object_property_add_bool(obj, "nvdimm", 909 machine_get_nvdimm, machine_set_nvdimm); 910 object_property_set_description(obj, "nvdimm", 911 "Set on/off to enable/disable " 912 "NVDIMM instantiation"); 913 914 object_property_add_str(obj, "nvdimm-persistence", 915 machine_get_nvdimm_persistence, 916 machine_set_nvdimm_persistence); 917 object_property_set_description(obj, "nvdimm-persistence", 918 "Set NVDIMM persistence" 919 "Valid values are cpu, mem-ctrl"); 920 } 921 922 if (mc->cpu_index_to_instance_props && mc->get_default_cpu_node_id) { 923 ms->numa_state = g_new0(NumaState, 1); 924 object_property_add_bool(obj, "hmat", 925 machine_get_hmat, machine_set_hmat); 926 object_property_set_description(obj, "hmat", 927 "Set on/off to enable/disable " 928 "ACPI Heterogeneous Memory Attribute " 929 "Table (HMAT)"); 930 } 931 932 /* default to mc->default_cpus */ 933 ms->smp.cpus = mc->default_cpus; 934 ms->smp.max_cpus = mc->default_cpus; 935 ms->smp.sockets = 1; 936 ms->smp.dies = 1; 937 ms->smp.clusters = 1; 938 ms->smp.cores = 1; 939 ms->smp.threads = 1; 940 } 941 942 static void machine_finalize(Object *obj) 943 { 944 MachineState *ms = MACHINE(obj); 945 946 g_free(ms->kernel_filename); 947 g_free(ms->initrd_filename); 948 g_free(ms->kernel_cmdline); 949 g_free(ms->dtb); 950 g_free(ms->dumpdtb); 951 g_free(ms->dt_compatible); 952 g_free(ms->firmware); 953 g_free(ms->device_memory); 954 g_free(ms->nvdimms_state); 955 g_free(ms->numa_state); 956 } 957 958 bool machine_usb(MachineState *machine) 959 { 960 return machine->usb; 961 } 962 963 int machine_phandle_start(MachineState *machine) 964 { 965 return machine->phandle_start; 966 } 967 968 bool machine_dump_guest_core(MachineState *machine) 969 { 970 return machine->dump_guest_core; 971 } 972 973 bool machine_mem_merge(MachineState *machine) 974 { 975 return machine->mem_merge; 976 } 977 978 static char *cpu_slot_to_string(const CPUArchId *cpu) 979 { 980 GString *s = g_string_new(NULL); 981 if (cpu->props.has_socket_id) { 982 g_string_append_printf(s, "socket-id: %"PRId64, cpu->props.socket_id); 983 } 984 if (cpu->props.has_die_id) { 985 if (s->len) { 986 g_string_append_printf(s, ", "); 987 } 988 g_string_append_printf(s, "die-id: %"PRId64, cpu->props.die_id); 989 } 990 if (cpu->props.has_core_id) { 991 if (s->len) { 992 g_string_append_printf(s, ", "); 993 } 994 g_string_append_printf(s, "core-id: %"PRId64, cpu->props.core_id); 995 } 996 if (cpu->props.has_thread_id) { 997 if (s->len) { 998 g_string_append_printf(s, ", "); 999 } 1000 g_string_append_printf(s, "thread-id: %"PRId64, cpu->props.thread_id); 1001 } 1002 return g_string_free(s, false); 1003 } 1004 1005 static void numa_validate_initiator(NumaState *numa_state) 1006 { 1007 int i; 1008 NodeInfo *numa_info = numa_state->nodes; 1009 1010 for (i = 0; i < numa_state->num_nodes; i++) { 1011 if (numa_info[i].initiator == MAX_NODES) { 1012 error_report("The initiator of NUMA node %d is missing, use " 1013 "'-numa node,initiator' option to declare it", i); 1014 exit(1); 1015 } 1016 1017 if (!numa_info[numa_info[i].initiator].present) { 1018 error_report("NUMA node %" PRIu16 " is missing, use " 1019 "'-numa node' option to declare it first", 1020 numa_info[i].initiator); 1021 exit(1); 1022 } 1023 1024 if (!numa_info[numa_info[i].initiator].has_cpu) { 1025 error_report("The initiator of NUMA node %d is invalid", i); 1026 exit(1); 1027 } 1028 } 1029 } 1030 1031 static void machine_numa_finish_cpu_init(MachineState *machine) 1032 { 1033 int i; 1034 bool default_mapping; 1035 GString *s = g_string_new(NULL); 1036 MachineClass *mc = MACHINE_GET_CLASS(machine); 1037 const CPUArchIdList *possible_cpus = mc->possible_cpu_arch_ids(machine); 1038 1039 assert(machine->numa_state->num_nodes); 1040 for (i = 0; i < possible_cpus->len; i++) { 1041 if (possible_cpus->cpus[i].props.has_node_id) { 1042 break; 1043 } 1044 } 1045 default_mapping = (i == possible_cpus->len); 1046 1047 for (i = 0; i < possible_cpus->len; i++) { 1048 const CPUArchId *cpu_slot = &possible_cpus->cpus[i]; 1049 1050 if (!cpu_slot->props.has_node_id) { 1051 /* fetch default mapping from board and enable it */ 1052 CpuInstanceProperties props = cpu_slot->props; 1053 1054 props.node_id = mc->get_default_cpu_node_id(machine, i); 1055 if (!default_mapping) { 1056 /* record slots with not set mapping, 1057 * TODO: make it hard error in future */ 1058 char *cpu_str = cpu_slot_to_string(cpu_slot); 1059 g_string_append_printf(s, "%sCPU %d [%s]", 1060 s->len ? ", " : "", i, cpu_str); 1061 g_free(cpu_str); 1062 1063 /* non mapped cpus used to fallback to node 0 */ 1064 props.node_id = 0; 1065 } 1066 1067 props.has_node_id = true; 1068 machine_set_cpu_numa_node(machine, &props, &error_fatal); 1069 } 1070 } 1071 1072 if (machine->numa_state->hmat_enabled) { 1073 numa_validate_initiator(machine->numa_state); 1074 } 1075 1076 if (s->len && !qtest_enabled()) { 1077 warn_report("CPU(s) not present in any NUMA nodes: %s", 1078 s->str); 1079 warn_report("All CPU(s) up to maxcpus should be described " 1080 "in NUMA config, ability to start up with partial NUMA " 1081 "mappings is obsoleted and will be removed in future"); 1082 } 1083 g_string_free(s, true); 1084 } 1085 1086 MemoryRegion *machine_consume_memdev(MachineState *machine, 1087 HostMemoryBackend *backend) 1088 { 1089 MemoryRegion *ret = host_memory_backend_get_memory(backend); 1090 1091 if (memory_region_is_mapped(ret)) { 1092 error_report("memory backend %s can't be used multiple times.", 1093 object_get_canonical_path_component(OBJECT(backend))); 1094 exit(EXIT_FAILURE); 1095 } 1096 host_memory_backend_set_mapped(backend, true); 1097 vmstate_register_ram_global(ret); 1098 return ret; 1099 } 1100 1101 void machine_run_board_init(MachineState *machine) 1102 { 1103 MachineClass *machine_class = MACHINE_GET_CLASS(machine); 1104 ObjectClass *oc = object_class_by_name(machine->cpu_type); 1105 CPUClass *cc; 1106 1107 /* This checkpoint is required by replay to separate prior clock 1108 reading from the other reads, because timer polling functions query 1109 clock values from the log. */ 1110 replay_checkpoint(CHECKPOINT_INIT); 1111 1112 if (machine->ram_memdev_id) { 1113 Object *o; 1114 o = object_resolve_path_type(machine->ram_memdev_id, 1115 TYPE_MEMORY_BACKEND, NULL); 1116 machine->ram = machine_consume_memdev(machine, MEMORY_BACKEND(o)); 1117 } 1118 1119 if (machine->numa_state) { 1120 numa_complete_configuration(machine); 1121 if (machine->numa_state->num_nodes) { 1122 machine_numa_finish_cpu_init(machine); 1123 } 1124 } 1125 1126 /* If the machine supports the valid_cpu_types check and the user 1127 * specified a CPU with -cpu check here that the user CPU is supported. 1128 */ 1129 if (machine_class->valid_cpu_types && machine->cpu_type) { 1130 int i; 1131 1132 for (i = 0; machine_class->valid_cpu_types[i]; i++) { 1133 if (object_class_dynamic_cast(oc, 1134 machine_class->valid_cpu_types[i])) { 1135 /* The user specificed CPU is in the valid field, we are 1136 * good to go. 1137 */ 1138 break; 1139 } 1140 } 1141 1142 if (!machine_class->valid_cpu_types[i]) { 1143 /* The user specified CPU is not valid */ 1144 error_report("Invalid CPU type: %s", machine->cpu_type); 1145 error_printf("The valid types are: %s", 1146 machine_class->valid_cpu_types[0]); 1147 for (i = 1; machine_class->valid_cpu_types[i]; i++) { 1148 error_printf(", %s", machine_class->valid_cpu_types[i]); 1149 } 1150 error_printf("\n"); 1151 1152 exit(1); 1153 } 1154 } 1155 1156 /* Check if CPU type is deprecated and warn if so */ 1157 cc = CPU_CLASS(oc); 1158 if (cc && cc->deprecation_note) { 1159 warn_report("CPU model %s is deprecated -- %s", machine->cpu_type, 1160 cc->deprecation_note); 1161 } 1162 1163 if (machine->cgs) { 1164 /* 1165 * With confidential guests, the host can't see the real 1166 * contents of RAM, so there's no point in it trying to merge 1167 * areas. 1168 */ 1169 machine_set_mem_merge(OBJECT(machine), false, &error_abort); 1170 1171 /* 1172 * Virtio devices can't count on directly accessing guest 1173 * memory, so they need iommu_platform=on to use normal DMA 1174 * mechanisms. That requires also disabling legacy virtio 1175 * support for those virtio pci devices which allow it. 1176 */ 1177 object_register_sugar_prop(TYPE_VIRTIO_PCI, "disable-legacy", 1178 "on", true); 1179 object_register_sugar_prop(TYPE_VIRTIO_DEVICE, "iommu_platform", 1180 "on", false); 1181 } 1182 1183 accel_init_interfaces(ACCEL_GET_CLASS(machine->accelerator)); 1184 machine_class->init(machine); 1185 phase_advance(PHASE_MACHINE_INITIALIZED); 1186 } 1187 1188 static NotifierList machine_init_done_notifiers = 1189 NOTIFIER_LIST_INITIALIZER(machine_init_done_notifiers); 1190 1191 void qemu_add_machine_init_done_notifier(Notifier *notify) 1192 { 1193 notifier_list_add(&machine_init_done_notifiers, notify); 1194 if (phase_check(PHASE_MACHINE_READY)) { 1195 notify->notify(notify, NULL); 1196 } 1197 } 1198 1199 void qemu_remove_machine_init_done_notifier(Notifier *notify) 1200 { 1201 notifier_remove(notify); 1202 } 1203 1204 void qdev_machine_creation_done(void) 1205 { 1206 cpu_synchronize_all_post_init(); 1207 1208 if (current_machine->boot_once) { 1209 qemu_boot_set(current_machine->boot_once, &error_fatal); 1210 qemu_register_reset(restore_boot_order, g_strdup(current_machine->boot_order)); 1211 } 1212 1213 /* 1214 * ok, initial machine setup is done, starting from now we can 1215 * only create hotpluggable devices 1216 */ 1217 phase_advance(PHASE_MACHINE_READY); 1218 qdev_assert_realized_properly(); 1219 1220 /* TODO: once all bus devices are qdevified, this should be done 1221 * when bus is created by qdev.c */ 1222 /* 1223 * TODO: If we had a main 'reset container' that the whole system 1224 * lived in, we could reset that using the multi-phase reset 1225 * APIs. For the moment, we just reset the sysbus, which will cause 1226 * all devices hanging off it (and all their child buses, recursively) 1227 * to be reset. Note that this will *not* reset any Device objects 1228 * which are not attached to some part of the qbus tree! 1229 */ 1230 qemu_register_reset(resettable_cold_reset_fn, sysbus_get_default()); 1231 1232 notifier_list_notify(&machine_init_done_notifiers, NULL); 1233 1234 if (rom_check_and_register_reset() != 0) { 1235 exit(1); 1236 } 1237 1238 replay_start(); 1239 1240 /* This checkpoint is required by replay to separate prior clock 1241 reading from the other reads, because timer polling functions query 1242 clock values from the log. */ 1243 replay_checkpoint(CHECKPOINT_RESET); 1244 qemu_system_reset(SHUTDOWN_CAUSE_NONE); 1245 register_global_state(); 1246 } 1247 1248 static const TypeInfo machine_info = { 1249 .name = TYPE_MACHINE, 1250 .parent = TYPE_OBJECT, 1251 .abstract = true, 1252 .class_size = sizeof(MachineClass), 1253 .class_init = machine_class_init, 1254 .class_base_init = machine_class_base_init, 1255 .instance_size = sizeof(MachineState), 1256 .instance_init = machine_initfn, 1257 .instance_finalize = machine_finalize, 1258 }; 1259 1260 static void machine_register_types(void) 1261 { 1262 type_register_static(&machine_info); 1263 } 1264 1265 type_init(machine_register_types) 1266