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