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/accel.h" 15 #include "sysemu/replay.h" 16 #include "hw/boards.h" 17 #include "hw/loader.h" 18 #include "qapi/error.h" 19 #include "qapi/qapi-visit-machine.h" 20 #include "qom/object_interfaces.h" 21 #include "sysemu/cpus.h" 22 #include "sysemu/sysemu.h" 23 #include "sysemu/reset.h" 24 #include "sysemu/runstate.h" 25 #include "sysemu/xen.h" 26 #include "sysemu/qtest.h" 27 #include "hw/pci/pci_bridge.h" 28 #include "hw/mem/nvdimm.h" 29 #include "migration/global_state.h" 30 #include "exec/confidential-guest-support.h" 31 #include "hw/virtio/virtio-pci.h" 32 #include "hw/virtio/virtio-net.h" 33 #include "audio/audio.h" 34 35 GlobalProperty hw_compat_8_1[] = { 36 { TYPE_PCI_BRIDGE, "x-pci-express-writeable-slt-bug", "true" }, 37 { "ramfb", "x-migrate", "off" }, 38 }; 39 const size_t hw_compat_8_1_len = G_N_ELEMENTS(hw_compat_8_1); 40 41 GlobalProperty hw_compat_8_0[] = { 42 { "migration", "multifd-flush-after-each-section", "on"}, 43 { TYPE_PCI_DEVICE, "x-pcie-ari-nextfn-1", "on" }, 44 { TYPE_VIRTIO_NET, "host_uso", "off"}, 45 { TYPE_VIRTIO_NET, "guest_uso4", "off"}, 46 { TYPE_VIRTIO_NET, "guest_uso6", "off"}, 47 }; 48 const size_t hw_compat_8_0_len = G_N_ELEMENTS(hw_compat_8_0); 49 50 GlobalProperty hw_compat_7_2[] = { 51 { "e1000e", "migrate-timadj", "off" }, 52 { "virtio-mem", "x-early-migration", "false" }, 53 { "migration", "x-preempt-pre-7-2", "true" }, 54 { TYPE_PCI_DEVICE, "x-pcie-err-unc-mask", "off" }, 55 }; 56 const size_t hw_compat_7_2_len = G_N_ELEMENTS(hw_compat_7_2); 57 58 GlobalProperty hw_compat_7_1[] = { 59 { "virtio-device", "queue_reset", "false" }, 60 { "virtio-rng-pci", "vectors", "0" }, 61 { "virtio-rng-pci-transitional", "vectors", "0" }, 62 { "virtio-rng-pci-non-transitional", "vectors", "0" }, 63 }; 64 const size_t hw_compat_7_1_len = G_N_ELEMENTS(hw_compat_7_1); 65 66 GlobalProperty hw_compat_7_0[] = { 67 { "arm-gicv3-common", "force-8-bit-prio", "on" }, 68 { "nvme-ns", "eui64-default", "on"}, 69 }; 70 const size_t hw_compat_7_0_len = G_N_ELEMENTS(hw_compat_7_0); 71 72 GlobalProperty hw_compat_6_2[] = { 73 { "PIIX4_PM", "x-not-migrate-acpi-index", "on"}, 74 }; 75 const size_t hw_compat_6_2_len = G_N_ELEMENTS(hw_compat_6_2); 76 77 GlobalProperty hw_compat_6_1[] = { 78 { "vhost-user-vsock-device", "seqpacket", "off" }, 79 { "nvme-ns", "shared", "off" }, 80 }; 81 const size_t hw_compat_6_1_len = G_N_ELEMENTS(hw_compat_6_1); 82 83 GlobalProperty hw_compat_6_0[] = { 84 { "gpex-pcihost", "allow-unmapped-accesses", "false" }, 85 { "i8042", "extended-state", "false"}, 86 { "nvme-ns", "eui64-default", "off"}, 87 { "e1000", "init-vet", "off" }, 88 { "e1000e", "init-vet", "off" }, 89 { "vhost-vsock-device", "seqpacket", "off" }, 90 }; 91 const size_t hw_compat_6_0_len = G_N_ELEMENTS(hw_compat_6_0); 92 93 GlobalProperty hw_compat_5_2[] = { 94 { "ICH9-LPC", "smm-compat", "on"}, 95 { "PIIX4_PM", "smm-compat", "on"}, 96 { "virtio-blk-device", "report-discard-granularity", "off" }, 97 { "virtio-net-pci-base", "vectors", "3"}, 98 }; 99 const size_t hw_compat_5_2_len = G_N_ELEMENTS(hw_compat_5_2); 100 101 GlobalProperty hw_compat_5_1[] = { 102 { "vhost-scsi", "num_queues", "1"}, 103 { "vhost-user-blk", "num-queues", "1"}, 104 { "vhost-user-scsi", "num_queues", "1"}, 105 { "virtio-blk-device", "num-queues", "1"}, 106 { "virtio-scsi-device", "num_queues", "1"}, 107 { "nvme", "use-intel-id", "on"}, 108 { "pvpanic", "events", "1"}, /* PVPANIC_PANICKED */ 109 { "pl011", "migrate-clk", "off" }, 110 { "virtio-pci", "x-ats-page-aligned", "off"}, 111 }; 112 const size_t hw_compat_5_1_len = G_N_ELEMENTS(hw_compat_5_1); 113 114 GlobalProperty hw_compat_5_0[] = { 115 { "pci-host-bridge", "x-config-reg-migration-enabled", "off" }, 116 { "virtio-balloon-device", "page-poison", "false" }, 117 { "vmport", "x-read-set-eax", "off" }, 118 { "vmport", "x-signal-unsupported-cmd", "off" }, 119 { "vmport", "x-report-vmx-type", "off" }, 120 { "vmport", "x-cmds-v2", "off" }, 121 { "virtio-device", "x-disable-legacy-check", "true" }, 122 }; 123 const size_t hw_compat_5_0_len = G_N_ELEMENTS(hw_compat_5_0); 124 125 GlobalProperty hw_compat_4_2[] = { 126 { "virtio-blk-device", "queue-size", "128"}, 127 { "virtio-scsi-device", "virtqueue_size", "128"}, 128 { "virtio-blk-device", "x-enable-wce-if-config-wce", "off" }, 129 { "virtio-blk-device", "seg-max-adjust", "off"}, 130 { "virtio-scsi-device", "seg_max_adjust", "off"}, 131 { "vhost-blk-device", "seg_max_adjust", "off"}, 132 { "usb-host", "suppress-remote-wake", "off" }, 133 { "usb-redir", "suppress-remote-wake", "off" }, 134 { "qxl", "revision", "4" }, 135 { "qxl-vga", "revision", "4" }, 136 { "fw_cfg", "acpi-mr-restore", "false" }, 137 { "virtio-device", "use-disabled-flag", "false" }, 138 }; 139 const size_t hw_compat_4_2_len = G_N_ELEMENTS(hw_compat_4_2); 140 141 GlobalProperty hw_compat_4_1[] = { 142 { "virtio-pci", "x-pcie-flr-init", "off" }, 143 }; 144 const size_t hw_compat_4_1_len = G_N_ELEMENTS(hw_compat_4_1); 145 146 GlobalProperty hw_compat_4_0[] = { 147 { "VGA", "edid", "false" }, 148 { "secondary-vga", "edid", "false" }, 149 { "bochs-display", "edid", "false" }, 150 { "virtio-vga", "edid", "false" }, 151 { "virtio-gpu-device", "edid", "false" }, 152 { "virtio-device", "use-started", "false" }, 153 { "virtio-balloon-device", "qemu-4-0-config-size", "true" }, 154 { "pl031", "migrate-tick-offset", "false" }, 155 }; 156 const size_t hw_compat_4_0_len = G_N_ELEMENTS(hw_compat_4_0); 157 158 GlobalProperty hw_compat_3_1[] = { 159 { "pcie-root-port", "x-speed", "2_5" }, 160 { "pcie-root-port", "x-width", "1" }, 161 { "memory-backend-file", "x-use-canonical-path-for-ramblock-id", "true" }, 162 { "memory-backend-memfd", "x-use-canonical-path-for-ramblock-id", "true" }, 163 { "tpm-crb", "ppi", "false" }, 164 { "tpm-tis", "ppi", "false" }, 165 { "usb-kbd", "serial", "42" }, 166 { "usb-mouse", "serial", "42" }, 167 { "usb-tablet", "serial", "42" }, 168 { "virtio-blk-device", "discard", "false" }, 169 { "virtio-blk-device", "write-zeroes", "false" }, 170 { "virtio-balloon-device", "qemu-4-0-config-size", "false" }, 171 { "pcie-root-port-base", "disable-acs", "true" }, /* Added in 4.1 */ 172 }; 173 const size_t hw_compat_3_1_len = G_N_ELEMENTS(hw_compat_3_1); 174 175 GlobalProperty hw_compat_3_0[] = {}; 176 const size_t hw_compat_3_0_len = G_N_ELEMENTS(hw_compat_3_0); 177 178 GlobalProperty hw_compat_2_12[] = { 179 { "migration", "decompress-error-check", "off" }, 180 { "hda-audio", "use-timer", "false" }, 181 { "cirrus-vga", "global-vmstate", "true" }, 182 { "VGA", "global-vmstate", "true" }, 183 { "vmware-svga", "global-vmstate", "true" }, 184 { "qxl-vga", "global-vmstate", "true" }, 185 }; 186 const size_t hw_compat_2_12_len = G_N_ELEMENTS(hw_compat_2_12); 187 188 GlobalProperty hw_compat_2_11[] = { 189 { "hpet", "hpet-offset-saved", "false" }, 190 { "virtio-blk-pci", "vectors", "2" }, 191 { "vhost-user-blk-pci", "vectors", "2" }, 192 { "e1000", "migrate_tso_props", "off" }, 193 }; 194 const size_t hw_compat_2_11_len = G_N_ELEMENTS(hw_compat_2_11); 195 196 GlobalProperty hw_compat_2_10[] = { 197 { "virtio-mouse-device", "wheel-axis", "false" }, 198 { "virtio-tablet-device", "wheel-axis", "false" }, 199 }; 200 const size_t hw_compat_2_10_len = G_N_ELEMENTS(hw_compat_2_10); 201 202 GlobalProperty hw_compat_2_9[] = { 203 { "pci-bridge", "shpc", "off" }, 204 { "intel-iommu", "pt", "off" }, 205 { "virtio-net-device", "x-mtu-bypass-backend", "off" }, 206 { "pcie-root-port", "x-migrate-msix", "false" }, 207 }; 208 const size_t hw_compat_2_9_len = G_N_ELEMENTS(hw_compat_2_9); 209 210 GlobalProperty hw_compat_2_8[] = { 211 { "fw_cfg_mem", "x-file-slots", "0x10" }, 212 { "fw_cfg_io", "x-file-slots", "0x10" }, 213 { "pflash_cfi01", "old-multiple-chip-handling", "on" }, 214 { "pci-bridge", "shpc", "on" }, 215 { TYPE_PCI_DEVICE, "x-pcie-extcap-init", "off" }, 216 { "virtio-pci", "x-pcie-deverr-init", "off" }, 217 { "virtio-pci", "x-pcie-lnkctl-init", "off" }, 218 { "virtio-pci", "x-pcie-pm-init", "off" }, 219 { "cirrus-vga", "vgamem_mb", "8" }, 220 { "isa-cirrus-vga", "vgamem_mb", "8" }, 221 }; 222 const size_t hw_compat_2_8_len = G_N_ELEMENTS(hw_compat_2_8); 223 224 GlobalProperty hw_compat_2_7[] = { 225 { "virtio-pci", "page-per-vq", "on" }, 226 { "virtio-serial-device", "emergency-write", "off" }, 227 { "ioapic", "version", "0x11" }, 228 { "intel-iommu", "x-buggy-eim", "true" }, 229 { "virtio-pci", "x-ignore-backend-features", "on" }, 230 }; 231 const size_t hw_compat_2_7_len = G_N_ELEMENTS(hw_compat_2_7); 232 233 GlobalProperty hw_compat_2_6[] = { 234 { "virtio-mmio", "format_transport_address", "off" }, 235 /* Optional because not all virtio-pci devices support legacy mode */ 236 { "virtio-pci", "disable-modern", "on", .optional = true }, 237 { "virtio-pci", "disable-legacy", "off", .optional = true }, 238 }; 239 const size_t hw_compat_2_6_len = G_N_ELEMENTS(hw_compat_2_6); 240 241 GlobalProperty hw_compat_2_5[] = { 242 { "isa-fdc", "fallback", "144" }, 243 { "pvscsi", "x-old-pci-configuration", "on" }, 244 { "pvscsi", "x-disable-pcie", "on" }, 245 { "vmxnet3", "x-old-msi-offsets", "on" }, 246 { "vmxnet3", "x-disable-pcie", "on" }, 247 }; 248 const size_t hw_compat_2_5_len = G_N_ELEMENTS(hw_compat_2_5); 249 250 GlobalProperty hw_compat_2_4[] = { 251 /* Optional because the 'scsi' property is Linux-only */ 252 { "virtio-blk-device", "scsi", "true", .optional = true }, 253 { "e1000", "extra_mac_registers", "off" }, 254 { "virtio-pci", "x-disable-pcie", "on" }, 255 { "virtio-pci", "migrate-extra", "off" }, 256 { "fw_cfg_mem", "dma_enabled", "off" }, 257 { "fw_cfg_io", "dma_enabled", "off" } 258 }; 259 const size_t hw_compat_2_4_len = G_N_ELEMENTS(hw_compat_2_4); 260 261 GlobalProperty hw_compat_2_3[] = { 262 { "virtio-blk-pci", "any_layout", "off" }, 263 { "virtio-balloon-pci", "any_layout", "off" }, 264 { "virtio-serial-pci", "any_layout", "off" }, 265 { "virtio-9p-pci", "any_layout", "off" }, 266 { "virtio-rng-pci", "any_layout", "off" }, 267 { TYPE_PCI_DEVICE, "x-pcie-lnksta-dllla", "off" }, 268 { "migration", "send-configuration", "off" }, 269 { "migration", "send-section-footer", "off" }, 270 { "migration", "store-global-state", "off" }, 271 }; 272 const size_t hw_compat_2_3_len = G_N_ELEMENTS(hw_compat_2_3); 273 274 GlobalProperty hw_compat_2_2[] = {}; 275 const size_t hw_compat_2_2_len = G_N_ELEMENTS(hw_compat_2_2); 276 277 GlobalProperty hw_compat_2_1[] = { 278 { "intel-hda", "old_msi_addr", "on" }, 279 { "VGA", "qemu-extended-regs", "off" }, 280 { "secondary-vga", "qemu-extended-regs", "off" }, 281 { "virtio-scsi-pci", "any_layout", "off" }, 282 { "usb-mouse", "usb_version", "1" }, 283 { "usb-kbd", "usb_version", "1" }, 284 { "virtio-pci", "virtio-pci-bus-master-bug-migration", "on" }, 285 }; 286 const size_t hw_compat_2_1_len = G_N_ELEMENTS(hw_compat_2_1); 287 288 MachineState *current_machine; 289 290 static char *machine_get_kernel(Object *obj, Error **errp) 291 { 292 MachineState *ms = MACHINE(obj); 293 294 return g_strdup(ms->kernel_filename); 295 } 296 297 static void machine_set_kernel(Object *obj, const char *value, Error **errp) 298 { 299 MachineState *ms = MACHINE(obj); 300 301 g_free(ms->kernel_filename); 302 ms->kernel_filename = g_strdup(value); 303 } 304 305 static char *machine_get_initrd(Object *obj, Error **errp) 306 { 307 MachineState *ms = MACHINE(obj); 308 309 return g_strdup(ms->initrd_filename); 310 } 311 312 static void machine_set_initrd(Object *obj, const char *value, Error **errp) 313 { 314 MachineState *ms = MACHINE(obj); 315 316 g_free(ms->initrd_filename); 317 ms->initrd_filename = g_strdup(value); 318 } 319 320 static char *machine_get_append(Object *obj, Error **errp) 321 { 322 MachineState *ms = MACHINE(obj); 323 324 return g_strdup(ms->kernel_cmdline); 325 } 326 327 static void machine_set_append(Object *obj, const char *value, Error **errp) 328 { 329 MachineState *ms = MACHINE(obj); 330 331 g_free(ms->kernel_cmdline); 332 ms->kernel_cmdline = g_strdup(value); 333 } 334 335 static char *machine_get_dtb(Object *obj, Error **errp) 336 { 337 MachineState *ms = MACHINE(obj); 338 339 return g_strdup(ms->dtb); 340 } 341 342 static void machine_set_dtb(Object *obj, const char *value, Error **errp) 343 { 344 MachineState *ms = MACHINE(obj); 345 346 g_free(ms->dtb); 347 ms->dtb = g_strdup(value); 348 } 349 350 static char *machine_get_dumpdtb(Object *obj, Error **errp) 351 { 352 MachineState *ms = MACHINE(obj); 353 354 return g_strdup(ms->dumpdtb); 355 } 356 357 static void machine_set_dumpdtb(Object *obj, const char *value, Error **errp) 358 { 359 MachineState *ms = MACHINE(obj); 360 361 g_free(ms->dumpdtb); 362 ms->dumpdtb = g_strdup(value); 363 } 364 365 static void machine_get_phandle_start(Object *obj, Visitor *v, 366 const char *name, void *opaque, 367 Error **errp) 368 { 369 MachineState *ms = MACHINE(obj); 370 int64_t value = ms->phandle_start; 371 372 visit_type_int(v, name, &value, errp); 373 } 374 375 static void machine_set_phandle_start(Object *obj, Visitor *v, 376 const char *name, void *opaque, 377 Error **errp) 378 { 379 MachineState *ms = MACHINE(obj); 380 int64_t value; 381 382 if (!visit_type_int(v, name, &value, errp)) { 383 return; 384 } 385 386 ms->phandle_start = value; 387 } 388 389 static char *machine_get_dt_compatible(Object *obj, Error **errp) 390 { 391 MachineState *ms = MACHINE(obj); 392 393 return g_strdup(ms->dt_compatible); 394 } 395 396 static void machine_set_dt_compatible(Object *obj, const char *value, Error **errp) 397 { 398 MachineState *ms = MACHINE(obj); 399 400 g_free(ms->dt_compatible); 401 ms->dt_compatible = g_strdup(value); 402 } 403 404 static bool machine_get_dump_guest_core(Object *obj, Error **errp) 405 { 406 MachineState *ms = MACHINE(obj); 407 408 return ms->dump_guest_core; 409 } 410 411 static void machine_set_dump_guest_core(Object *obj, bool value, Error **errp) 412 { 413 MachineState *ms = MACHINE(obj); 414 415 ms->dump_guest_core = value; 416 } 417 418 static bool machine_get_mem_merge(Object *obj, Error **errp) 419 { 420 MachineState *ms = MACHINE(obj); 421 422 return ms->mem_merge; 423 } 424 425 static void machine_set_mem_merge(Object *obj, bool value, Error **errp) 426 { 427 MachineState *ms = MACHINE(obj); 428 429 ms->mem_merge = value; 430 } 431 432 static bool machine_get_usb(Object *obj, Error **errp) 433 { 434 MachineState *ms = MACHINE(obj); 435 436 return ms->usb; 437 } 438 439 static void machine_set_usb(Object *obj, bool value, Error **errp) 440 { 441 MachineState *ms = MACHINE(obj); 442 443 ms->usb = value; 444 ms->usb_disabled = !value; 445 } 446 447 static bool machine_get_graphics(Object *obj, Error **errp) 448 { 449 MachineState *ms = MACHINE(obj); 450 451 return ms->enable_graphics; 452 } 453 454 static void machine_set_graphics(Object *obj, bool value, Error **errp) 455 { 456 MachineState *ms = MACHINE(obj); 457 458 ms->enable_graphics = value; 459 } 460 461 static char *machine_get_firmware(Object *obj, Error **errp) 462 { 463 MachineState *ms = MACHINE(obj); 464 465 return g_strdup(ms->firmware); 466 } 467 468 static void machine_set_firmware(Object *obj, const char *value, Error **errp) 469 { 470 MachineState *ms = MACHINE(obj); 471 472 g_free(ms->firmware); 473 ms->firmware = g_strdup(value); 474 } 475 476 static void machine_set_suppress_vmdesc(Object *obj, bool value, Error **errp) 477 { 478 MachineState *ms = MACHINE(obj); 479 480 ms->suppress_vmdesc = value; 481 } 482 483 static bool machine_get_suppress_vmdesc(Object *obj, Error **errp) 484 { 485 MachineState *ms = MACHINE(obj); 486 487 return ms->suppress_vmdesc; 488 } 489 490 static char *machine_get_memory_encryption(Object *obj, Error **errp) 491 { 492 MachineState *ms = MACHINE(obj); 493 494 if (ms->cgs) { 495 return g_strdup(object_get_canonical_path_component(OBJECT(ms->cgs))); 496 } 497 498 return NULL; 499 } 500 501 static void machine_set_memory_encryption(Object *obj, const char *value, 502 Error **errp) 503 { 504 Object *cgs = 505 object_resolve_path_component(object_get_objects_root(), value); 506 507 if (!cgs) { 508 error_setg(errp, "No such memory encryption object '%s'", value); 509 return; 510 } 511 512 object_property_set_link(obj, "confidential-guest-support", cgs, errp); 513 } 514 515 static void machine_check_confidential_guest_support(const Object *obj, 516 const char *name, 517 Object *new_target, 518 Error **errp) 519 { 520 /* 521 * So far the only constraint is that the target has the 522 * TYPE_CONFIDENTIAL_GUEST_SUPPORT interface, and that's checked 523 * by the QOM core 524 */ 525 } 526 527 static bool machine_get_nvdimm(Object *obj, Error **errp) 528 { 529 MachineState *ms = MACHINE(obj); 530 531 return ms->nvdimms_state->is_enabled; 532 } 533 534 static void machine_set_nvdimm(Object *obj, bool value, Error **errp) 535 { 536 MachineState *ms = MACHINE(obj); 537 538 ms->nvdimms_state->is_enabled = value; 539 } 540 541 static bool machine_get_hmat(Object *obj, Error **errp) 542 { 543 MachineState *ms = MACHINE(obj); 544 545 return ms->numa_state->hmat_enabled; 546 } 547 548 static void machine_set_hmat(Object *obj, bool value, Error **errp) 549 { 550 MachineState *ms = MACHINE(obj); 551 552 ms->numa_state->hmat_enabled = value; 553 } 554 555 static void machine_get_mem(Object *obj, Visitor *v, const char *name, 556 void *opaque, Error **errp) 557 { 558 MachineState *ms = MACHINE(obj); 559 MemorySizeConfiguration mem = { 560 .has_size = true, 561 .size = ms->ram_size, 562 .has_max_size = !!ms->ram_slots, 563 .max_size = ms->maxram_size, 564 .has_slots = !!ms->ram_slots, 565 .slots = ms->ram_slots, 566 }; 567 MemorySizeConfiguration *p_mem = &mem; 568 569 visit_type_MemorySizeConfiguration(v, name, &p_mem, &error_abort); 570 } 571 572 static void machine_set_mem(Object *obj, Visitor *v, const char *name, 573 void *opaque, Error **errp) 574 { 575 ERRP_GUARD(); 576 MachineState *ms = MACHINE(obj); 577 MachineClass *mc = MACHINE_GET_CLASS(obj); 578 MemorySizeConfiguration *mem; 579 580 if (!visit_type_MemorySizeConfiguration(v, name, &mem, errp)) { 581 return; 582 } 583 584 if (!mem->has_size) { 585 mem->has_size = true; 586 mem->size = mc->default_ram_size; 587 } 588 mem->size = QEMU_ALIGN_UP(mem->size, 8192); 589 if (mc->fixup_ram_size) { 590 mem->size = mc->fixup_ram_size(mem->size); 591 } 592 if ((ram_addr_t)mem->size != mem->size) { 593 error_setg(errp, "ram size too large"); 594 goto out_free; 595 } 596 597 if (mem->has_max_size) { 598 if (mem->max_size < mem->size) { 599 error_setg(errp, "invalid value of maxmem: " 600 "maximum memory size (0x%" PRIx64 ") must be at least " 601 "the initial memory size (0x%" PRIx64 ")", 602 mem->max_size, mem->size); 603 goto out_free; 604 } 605 if (mem->has_slots && mem->slots && mem->max_size == mem->size) { 606 error_setg(errp, "invalid value of maxmem: " 607 "memory slots were specified but maximum memory size " 608 "(0x%" PRIx64 ") is equal to the initial memory size " 609 "(0x%" PRIx64 ")", mem->max_size, mem->size); 610 goto out_free; 611 } 612 ms->maxram_size = mem->max_size; 613 } else { 614 if (mem->has_slots) { 615 error_setg(errp, "slots specified but no max-size"); 616 goto out_free; 617 } 618 ms->maxram_size = mem->size; 619 } 620 ms->ram_size = mem->size; 621 ms->ram_slots = mem->has_slots ? mem->slots : 0; 622 out_free: 623 qapi_free_MemorySizeConfiguration(mem); 624 } 625 626 static char *machine_get_nvdimm_persistence(Object *obj, Error **errp) 627 { 628 MachineState *ms = MACHINE(obj); 629 630 return g_strdup(ms->nvdimms_state->persistence_string); 631 } 632 633 static void machine_set_nvdimm_persistence(Object *obj, const char *value, 634 Error **errp) 635 { 636 MachineState *ms = MACHINE(obj); 637 NVDIMMState *nvdimms_state = ms->nvdimms_state; 638 639 if (strcmp(value, "cpu") == 0) { 640 nvdimms_state->persistence = 3; 641 } else if (strcmp(value, "mem-ctrl") == 0) { 642 nvdimms_state->persistence = 2; 643 } else { 644 error_setg(errp, "-machine nvdimm-persistence=%s: unsupported option", 645 value); 646 return; 647 } 648 649 g_free(nvdimms_state->persistence_string); 650 nvdimms_state->persistence_string = g_strdup(value); 651 } 652 653 void machine_class_allow_dynamic_sysbus_dev(MachineClass *mc, const char *type) 654 { 655 QAPI_LIST_PREPEND(mc->allowed_dynamic_sysbus_devices, g_strdup(type)); 656 } 657 658 bool device_is_dynamic_sysbus(MachineClass *mc, DeviceState *dev) 659 { 660 Object *obj = OBJECT(dev); 661 662 if (!object_dynamic_cast(obj, TYPE_SYS_BUS_DEVICE)) { 663 return false; 664 } 665 666 return device_type_is_dynamic_sysbus(mc, object_get_typename(obj)); 667 } 668 669 bool device_type_is_dynamic_sysbus(MachineClass *mc, const char *type) 670 { 671 bool allowed = false; 672 strList *wl; 673 ObjectClass *klass = object_class_by_name(type); 674 675 for (wl = mc->allowed_dynamic_sysbus_devices; 676 !allowed && wl; 677 wl = wl->next) { 678 allowed |= !!object_class_dynamic_cast(klass, wl->value); 679 } 680 681 return allowed; 682 } 683 684 static char *machine_get_audiodev(Object *obj, Error **errp) 685 { 686 MachineState *ms = MACHINE(obj); 687 688 return g_strdup(ms->audiodev); 689 } 690 691 static void machine_set_audiodev(Object *obj, const char *value, 692 Error **errp) 693 { 694 MachineState *ms = MACHINE(obj); 695 696 if (!audio_state_by_name(value, errp)) { 697 return; 698 } 699 700 g_free(ms->audiodev); 701 ms->audiodev = g_strdup(value); 702 } 703 704 HotpluggableCPUList *machine_query_hotpluggable_cpus(MachineState *machine) 705 { 706 int i; 707 HotpluggableCPUList *head = NULL; 708 MachineClass *mc = MACHINE_GET_CLASS(machine); 709 710 /* force board to initialize possible_cpus if it hasn't been done yet */ 711 mc->possible_cpu_arch_ids(machine); 712 713 for (i = 0; i < machine->possible_cpus->len; i++) { 714 Object *cpu; 715 HotpluggableCPU *cpu_item = g_new0(typeof(*cpu_item), 1); 716 717 cpu_item->type = g_strdup(machine->possible_cpus->cpus[i].type); 718 cpu_item->vcpus_count = machine->possible_cpus->cpus[i].vcpus_count; 719 cpu_item->props = g_memdup(&machine->possible_cpus->cpus[i].props, 720 sizeof(*cpu_item->props)); 721 722 cpu = machine->possible_cpus->cpus[i].cpu; 723 if (cpu) { 724 cpu_item->qom_path = object_get_canonical_path(cpu); 725 } 726 QAPI_LIST_PREPEND(head, cpu_item); 727 } 728 return head; 729 } 730 731 /** 732 * machine_set_cpu_numa_node: 733 * @machine: machine object to modify 734 * @props: specifies which cpu objects to assign to 735 * numa node specified by @props.node_id 736 * @errp: if an error occurs, a pointer to an area to store the error 737 * 738 * Associate NUMA node specified by @props.node_id with cpu slots that 739 * match socket/core/thread-ids specified by @props. It's recommended to use 740 * query-hotpluggable-cpus.props values to specify affected cpu slots, 741 * which would lead to exact 1:1 mapping of cpu slots to NUMA node. 742 * 743 * However for CLI convenience it's possible to pass in subset of properties, 744 * which would affect all cpu slots that match it. 745 * Ex for pc machine: 746 * -smp 4,cores=2,sockets=2 -numa node,nodeid=0 -numa node,nodeid=1 \ 747 * -numa cpu,node-id=0,socket_id=0 \ 748 * -numa cpu,node-id=1,socket_id=1 749 * will assign all child cores of socket 0 to node 0 and 750 * of socket 1 to node 1. 751 * 752 * On attempt of reassigning (already assigned) cpu slot to another NUMA node, 753 * return error. 754 * Empty subset is disallowed and function will return with error in this case. 755 */ 756 void machine_set_cpu_numa_node(MachineState *machine, 757 const CpuInstanceProperties *props, Error **errp) 758 { 759 MachineClass *mc = MACHINE_GET_CLASS(machine); 760 NodeInfo *numa_info = machine->numa_state->nodes; 761 bool match = false; 762 int i; 763 764 if (!mc->possible_cpu_arch_ids) { 765 error_setg(errp, "mapping of CPUs to NUMA node is not supported"); 766 return; 767 } 768 769 /* disabling node mapping is not supported, forbid it */ 770 assert(props->has_node_id); 771 772 /* force board to initialize possible_cpus if it hasn't been done yet */ 773 mc->possible_cpu_arch_ids(machine); 774 775 for (i = 0; i < machine->possible_cpus->len; i++) { 776 CPUArchId *slot = &machine->possible_cpus->cpus[i]; 777 778 /* reject unsupported by board properties */ 779 if (props->has_thread_id && !slot->props.has_thread_id) { 780 error_setg(errp, "thread-id is not supported"); 781 return; 782 } 783 784 if (props->has_core_id && !slot->props.has_core_id) { 785 error_setg(errp, "core-id is not supported"); 786 return; 787 } 788 789 if (props->has_cluster_id && !slot->props.has_cluster_id) { 790 error_setg(errp, "cluster-id is not supported"); 791 return; 792 } 793 794 if (props->has_socket_id && !slot->props.has_socket_id) { 795 error_setg(errp, "socket-id is not supported"); 796 return; 797 } 798 799 if (props->has_die_id && !slot->props.has_die_id) { 800 error_setg(errp, "die-id is not supported"); 801 return; 802 } 803 804 /* skip slots with explicit mismatch */ 805 if (props->has_thread_id && props->thread_id != slot->props.thread_id) { 806 continue; 807 } 808 809 if (props->has_core_id && props->core_id != slot->props.core_id) { 810 continue; 811 } 812 813 if (props->has_cluster_id && 814 props->cluster_id != slot->props.cluster_id) { 815 continue; 816 } 817 818 if (props->has_die_id && props->die_id != slot->props.die_id) { 819 continue; 820 } 821 822 if (props->has_socket_id && props->socket_id != slot->props.socket_id) { 823 continue; 824 } 825 826 /* reject assignment if slot is already assigned, for compatibility 827 * of legacy cpu_index mapping with SPAPR core based mapping do not 828 * error out if cpu thread and matched core have the same node-id */ 829 if (slot->props.has_node_id && 830 slot->props.node_id != props->node_id) { 831 error_setg(errp, "CPU is already assigned to node-id: %" PRId64, 832 slot->props.node_id); 833 return; 834 } 835 836 /* assign slot to node as it's matched '-numa cpu' key */ 837 match = true; 838 slot->props.node_id = props->node_id; 839 slot->props.has_node_id = props->has_node_id; 840 841 if (machine->numa_state->hmat_enabled) { 842 if ((numa_info[props->node_id].initiator < MAX_NODES) && 843 (props->node_id != numa_info[props->node_id].initiator)) { 844 error_setg(errp, "The initiator of CPU NUMA node %" PRId64 845 " should be itself (got %" PRIu16 ")", 846 props->node_id, numa_info[props->node_id].initiator); 847 return; 848 } 849 numa_info[props->node_id].has_cpu = true; 850 numa_info[props->node_id].initiator = props->node_id; 851 } 852 } 853 854 if (!match) { 855 error_setg(errp, "no match found"); 856 } 857 } 858 859 static void machine_get_smp(Object *obj, Visitor *v, const char *name, 860 void *opaque, Error **errp) 861 { 862 MachineState *ms = MACHINE(obj); 863 SMPConfiguration *config = &(SMPConfiguration){ 864 .has_cpus = true, .cpus = ms->smp.cpus, 865 .has_sockets = true, .sockets = ms->smp.sockets, 866 .has_dies = true, .dies = ms->smp.dies, 867 .has_clusters = true, .clusters = ms->smp.clusters, 868 .has_cores = true, .cores = ms->smp.cores, 869 .has_threads = true, .threads = ms->smp.threads, 870 .has_maxcpus = true, .maxcpus = ms->smp.max_cpus, 871 }; 872 873 if (!visit_type_SMPConfiguration(v, name, &config, &error_abort)) { 874 return; 875 } 876 } 877 878 static void machine_set_smp(Object *obj, Visitor *v, const char *name, 879 void *opaque, Error **errp) 880 { 881 MachineState *ms = MACHINE(obj); 882 g_autoptr(SMPConfiguration) config = NULL; 883 884 if (!visit_type_SMPConfiguration(v, name, &config, errp)) { 885 return; 886 } 887 888 machine_parse_smp_config(ms, config, errp); 889 } 890 891 static void machine_get_boot(Object *obj, Visitor *v, const char *name, 892 void *opaque, Error **errp) 893 { 894 MachineState *ms = MACHINE(obj); 895 BootConfiguration *config = &ms->boot_config; 896 visit_type_BootConfiguration(v, name, &config, &error_abort); 897 } 898 899 static void machine_free_boot_config(MachineState *ms) 900 { 901 g_free(ms->boot_config.order); 902 g_free(ms->boot_config.once); 903 g_free(ms->boot_config.splash); 904 } 905 906 static void machine_copy_boot_config(MachineState *ms, BootConfiguration *config) 907 { 908 MachineClass *machine_class = MACHINE_GET_CLASS(ms); 909 910 machine_free_boot_config(ms); 911 ms->boot_config = *config; 912 if (!config->order) { 913 ms->boot_config.order = g_strdup(machine_class->default_boot_order); 914 } 915 } 916 917 static void machine_set_boot(Object *obj, Visitor *v, const char *name, 918 void *opaque, Error **errp) 919 { 920 ERRP_GUARD(); 921 MachineState *ms = MACHINE(obj); 922 BootConfiguration *config = NULL; 923 924 if (!visit_type_BootConfiguration(v, name, &config, errp)) { 925 return; 926 } 927 if (config->order) { 928 validate_bootdevices(config->order, errp); 929 if (*errp) { 930 goto out_free; 931 } 932 } 933 if (config->once) { 934 validate_bootdevices(config->once, errp); 935 if (*errp) { 936 goto out_free; 937 } 938 } 939 940 machine_copy_boot_config(ms, config); 941 /* Strings live in ms->boot_config. */ 942 free(config); 943 return; 944 945 out_free: 946 qapi_free_BootConfiguration(config); 947 } 948 949 void machine_add_audiodev_property(MachineClass *mc) 950 { 951 ObjectClass *oc = OBJECT_CLASS(mc); 952 953 object_class_property_add_str(oc, "audiodev", 954 machine_get_audiodev, 955 machine_set_audiodev); 956 object_class_property_set_description(oc, "audiodev", 957 "Audiodev to use for default machine devices"); 958 } 959 960 static void machine_class_init(ObjectClass *oc, void *data) 961 { 962 MachineClass *mc = MACHINE_CLASS(oc); 963 964 /* Default 128 MB as guest ram size */ 965 mc->default_ram_size = 128 * MiB; 966 mc->rom_file_has_mr = true; 967 968 /* numa node memory size aligned on 8MB by default. 969 * On Linux, each node's border has to be 8MB aligned 970 */ 971 mc->numa_mem_align_shift = 23; 972 973 object_class_property_add_str(oc, "kernel", 974 machine_get_kernel, machine_set_kernel); 975 object_class_property_set_description(oc, "kernel", 976 "Linux kernel image file"); 977 978 object_class_property_add_str(oc, "initrd", 979 machine_get_initrd, machine_set_initrd); 980 object_class_property_set_description(oc, "initrd", 981 "Linux initial ramdisk file"); 982 983 object_class_property_add_str(oc, "append", 984 machine_get_append, machine_set_append); 985 object_class_property_set_description(oc, "append", 986 "Linux kernel command line"); 987 988 object_class_property_add_str(oc, "dtb", 989 machine_get_dtb, machine_set_dtb); 990 object_class_property_set_description(oc, "dtb", 991 "Linux kernel device tree file"); 992 993 object_class_property_add_str(oc, "dumpdtb", 994 machine_get_dumpdtb, machine_set_dumpdtb); 995 object_class_property_set_description(oc, "dumpdtb", 996 "Dump current dtb to a file and quit"); 997 998 object_class_property_add(oc, "boot", "BootConfiguration", 999 machine_get_boot, machine_set_boot, 1000 NULL, NULL); 1001 object_class_property_set_description(oc, "boot", 1002 "Boot configuration"); 1003 1004 object_class_property_add(oc, "smp", "SMPConfiguration", 1005 machine_get_smp, machine_set_smp, 1006 NULL, NULL); 1007 object_class_property_set_description(oc, "smp", 1008 "CPU topology"); 1009 1010 object_class_property_add(oc, "phandle-start", "int", 1011 machine_get_phandle_start, machine_set_phandle_start, 1012 NULL, NULL); 1013 object_class_property_set_description(oc, "phandle-start", 1014 "The first phandle ID we may generate dynamically"); 1015 1016 object_class_property_add_str(oc, "dt-compatible", 1017 machine_get_dt_compatible, machine_set_dt_compatible); 1018 object_class_property_set_description(oc, "dt-compatible", 1019 "Overrides the \"compatible\" property of the dt root node"); 1020 1021 object_class_property_add_bool(oc, "dump-guest-core", 1022 machine_get_dump_guest_core, machine_set_dump_guest_core); 1023 object_class_property_set_description(oc, "dump-guest-core", 1024 "Include guest memory in a core dump"); 1025 1026 object_class_property_add_bool(oc, "mem-merge", 1027 machine_get_mem_merge, machine_set_mem_merge); 1028 object_class_property_set_description(oc, "mem-merge", 1029 "Enable/disable memory merge support"); 1030 1031 object_class_property_add_bool(oc, "usb", 1032 machine_get_usb, machine_set_usb); 1033 object_class_property_set_description(oc, "usb", 1034 "Set on/off to enable/disable usb"); 1035 1036 object_class_property_add_bool(oc, "graphics", 1037 machine_get_graphics, machine_set_graphics); 1038 object_class_property_set_description(oc, "graphics", 1039 "Set on/off to enable/disable graphics emulation"); 1040 1041 object_class_property_add_str(oc, "firmware", 1042 machine_get_firmware, machine_set_firmware); 1043 object_class_property_set_description(oc, "firmware", 1044 "Firmware image"); 1045 1046 object_class_property_add_bool(oc, "suppress-vmdesc", 1047 machine_get_suppress_vmdesc, machine_set_suppress_vmdesc); 1048 object_class_property_set_description(oc, "suppress-vmdesc", 1049 "Set on to disable self-describing migration"); 1050 1051 object_class_property_add_link(oc, "confidential-guest-support", 1052 TYPE_CONFIDENTIAL_GUEST_SUPPORT, 1053 offsetof(MachineState, cgs), 1054 machine_check_confidential_guest_support, 1055 OBJ_PROP_LINK_STRONG); 1056 object_class_property_set_description(oc, "confidential-guest-support", 1057 "Set confidential guest scheme to support"); 1058 1059 /* For compatibility */ 1060 object_class_property_add_str(oc, "memory-encryption", 1061 machine_get_memory_encryption, machine_set_memory_encryption); 1062 object_class_property_set_description(oc, "memory-encryption", 1063 "Set memory encryption object to use"); 1064 1065 object_class_property_add_link(oc, "memory-backend", TYPE_MEMORY_BACKEND, 1066 offsetof(MachineState, memdev), object_property_allow_set_link, 1067 OBJ_PROP_LINK_STRONG); 1068 object_class_property_set_description(oc, "memory-backend", 1069 "Set RAM backend" 1070 "Valid value is ID of hostmem based backend"); 1071 1072 object_class_property_add(oc, "memory", "MemorySizeConfiguration", 1073 machine_get_mem, machine_set_mem, 1074 NULL, NULL); 1075 object_class_property_set_description(oc, "memory", 1076 "Memory size configuration"); 1077 } 1078 1079 static void machine_class_base_init(ObjectClass *oc, void *data) 1080 { 1081 MachineClass *mc = MACHINE_CLASS(oc); 1082 mc->max_cpus = mc->max_cpus ?: 1; 1083 mc->min_cpus = mc->min_cpus ?: 1; 1084 mc->default_cpus = mc->default_cpus ?: 1; 1085 1086 if (!object_class_is_abstract(oc)) { 1087 const char *cname = object_class_get_name(oc); 1088 assert(g_str_has_suffix(cname, TYPE_MACHINE_SUFFIX)); 1089 mc->name = g_strndup(cname, 1090 strlen(cname) - strlen(TYPE_MACHINE_SUFFIX)); 1091 mc->compat_props = g_ptr_array_new(); 1092 } 1093 } 1094 1095 static void machine_initfn(Object *obj) 1096 { 1097 MachineState *ms = MACHINE(obj); 1098 MachineClass *mc = MACHINE_GET_CLASS(obj); 1099 1100 container_get(obj, "/peripheral"); 1101 container_get(obj, "/peripheral-anon"); 1102 1103 ms->dump_guest_core = true; 1104 ms->mem_merge = true; 1105 ms->enable_graphics = true; 1106 ms->kernel_cmdline = g_strdup(""); 1107 ms->ram_size = mc->default_ram_size; 1108 ms->maxram_size = mc->default_ram_size; 1109 1110 if (mc->nvdimm_supported) { 1111 ms->nvdimms_state = g_new0(NVDIMMState, 1); 1112 object_property_add_bool(obj, "nvdimm", 1113 machine_get_nvdimm, machine_set_nvdimm); 1114 object_property_set_description(obj, "nvdimm", 1115 "Set on/off to enable/disable " 1116 "NVDIMM instantiation"); 1117 1118 object_property_add_str(obj, "nvdimm-persistence", 1119 machine_get_nvdimm_persistence, 1120 machine_set_nvdimm_persistence); 1121 object_property_set_description(obj, "nvdimm-persistence", 1122 "Set NVDIMM persistence" 1123 "Valid values are cpu, mem-ctrl"); 1124 } 1125 1126 if (mc->cpu_index_to_instance_props && mc->get_default_cpu_node_id) { 1127 ms->numa_state = g_new0(NumaState, 1); 1128 object_property_add_bool(obj, "hmat", 1129 machine_get_hmat, machine_set_hmat); 1130 object_property_set_description(obj, "hmat", 1131 "Set on/off to enable/disable " 1132 "ACPI Heterogeneous Memory Attribute " 1133 "Table (HMAT)"); 1134 } 1135 1136 /* default to mc->default_cpus */ 1137 ms->smp.cpus = mc->default_cpus; 1138 ms->smp.max_cpus = mc->default_cpus; 1139 ms->smp.sockets = 1; 1140 ms->smp.dies = 1; 1141 ms->smp.clusters = 1; 1142 ms->smp.cores = 1; 1143 ms->smp.threads = 1; 1144 1145 machine_copy_boot_config(ms, &(BootConfiguration){ 0 }); 1146 } 1147 1148 static void machine_finalize(Object *obj) 1149 { 1150 MachineState *ms = MACHINE(obj); 1151 1152 machine_free_boot_config(ms); 1153 g_free(ms->kernel_filename); 1154 g_free(ms->initrd_filename); 1155 g_free(ms->kernel_cmdline); 1156 g_free(ms->dtb); 1157 g_free(ms->dumpdtb); 1158 g_free(ms->dt_compatible); 1159 g_free(ms->firmware); 1160 g_free(ms->device_memory); 1161 g_free(ms->nvdimms_state); 1162 g_free(ms->numa_state); 1163 g_free(ms->audiodev); 1164 } 1165 1166 bool machine_usb(MachineState *machine) 1167 { 1168 return machine->usb; 1169 } 1170 1171 int machine_phandle_start(MachineState *machine) 1172 { 1173 return machine->phandle_start; 1174 } 1175 1176 bool machine_dump_guest_core(MachineState *machine) 1177 { 1178 return machine->dump_guest_core; 1179 } 1180 1181 bool machine_mem_merge(MachineState *machine) 1182 { 1183 return machine->mem_merge; 1184 } 1185 1186 static char *cpu_slot_to_string(const CPUArchId *cpu) 1187 { 1188 GString *s = g_string_new(NULL); 1189 if (cpu->props.has_socket_id) { 1190 g_string_append_printf(s, "socket-id: %"PRId64, cpu->props.socket_id); 1191 } 1192 if (cpu->props.has_die_id) { 1193 if (s->len) { 1194 g_string_append_printf(s, ", "); 1195 } 1196 g_string_append_printf(s, "die-id: %"PRId64, cpu->props.die_id); 1197 } 1198 if (cpu->props.has_cluster_id) { 1199 if (s->len) { 1200 g_string_append_printf(s, ", "); 1201 } 1202 g_string_append_printf(s, "cluster-id: %"PRId64, cpu->props.cluster_id); 1203 } 1204 if (cpu->props.has_core_id) { 1205 if (s->len) { 1206 g_string_append_printf(s, ", "); 1207 } 1208 g_string_append_printf(s, "core-id: %"PRId64, cpu->props.core_id); 1209 } 1210 if (cpu->props.has_thread_id) { 1211 if (s->len) { 1212 g_string_append_printf(s, ", "); 1213 } 1214 g_string_append_printf(s, "thread-id: %"PRId64, cpu->props.thread_id); 1215 } 1216 return g_string_free(s, false); 1217 } 1218 1219 static void numa_validate_initiator(NumaState *numa_state) 1220 { 1221 int i; 1222 NodeInfo *numa_info = numa_state->nodes; 1223 1224 for (i = 0; i < numa_state->num_nodes; i++) { 1225 if (numa_info[i].initiator == MAX_NODES) { 1226 continue; 1227 } 1228 1229 if (!numa_info[numa_info[i].initiator].present) { 1230 error_report("NUMA node %" PRIu16 " is missing, use " 1231 "'-numa node' option to declare it first", 1232 numa_info[i].initiator); 1233 exit(1); 1234 } 1235 1236 if (!numa_info[numa_info[i].initiator].has_cpu) { 1237 error_report("The initiator of NUMA node %d is invalid", i); 1238 exit(1); 1239 } 1240 } 1241 } 1242 1243 static void machine_numa_finish_cpu_init(MachineState *machine) 1244 { 1245 int i; 1246 bool default_mapping; 1247 GString *s = g_string_new(NULL); 1248 MachineClass *mc = MACHINE_GET_CLASS(machine); 1249 const CPUArchIdList *possible_cpus = mc->possible_cpu_arch_ids(machine); 1250 1251 assert(machine->numa_state->num_nodes); 1252 for (i = 0; i < possible_cpus->len; i++) { 1253 if (possible_cpus->cpus[i].props.has_node_id) { 1254 break; 1255 } 1256 } 1257 default_mapping = (i == possible_cpus->len); 1258 1259 for (i = 0; i < possible_cpus->len; i++) { 1260 const CPUArchId *cpu_slot = &possible_cpus->cpus[i]; 1261 1262 if (!cpu_slot->props.has_node_id) { 1263 /* fetch default mapping from board and enable it */ 1264 CpuInstanceProperties props = cpu_slot->props; 1265 1266 props.node_id = mc->get_default_cpu_node_id(machine, i); 1267 if (!default_mapping) { 1268 /* record slots with not set mapping, 1269 * TODO: make it hard error in future */ 1270 char *cpu_str = cpu_slot_to_string(cpu_slot); 1271 g_string_append_printf(s, "%sCPU %d [%s]", 1272 s->len ? ", " : "", i, cpu_str); 1273 g_free(cpu_str); 1274 1275 /* non mapped cpus used to fallback to node 0 */ 1276 props.node_id = 0; 1277 } 1278 1279 props.has_node_id = true; 1280 machine_set_cpu_numa_node(machine, &props, &error_fatal); 1281 } 1282 } 1283 1284 if (machine->numa_state->hmat_enabled) { 1285 numa_validate_initiator(machine->numa_state); 1286 } 1287 1288 if (s->len && !qtest_enabled()) { 1289 warn_report("CPU(s) not present in any NUMA nodes: %s", 1290 s->str); 1291 warn_report("All CPU(s) up to maxcpus should be described " 1292 "in NUMA config, ability to start up with partial NUMA " 1293 "mappings is obsoleted and will be removed in future"); 1294 } 1295 g_string_free(s, true); 1296 } 1297 1298 static void validate_cpu_cluster_to_numa_boundary(MachineState *ms) 1299 { 1300 MachineClass *mc = MACHINE_GET_CLASS(ms); 1301 NumaState *state = ms->numa_state; 1302 const CPUArchIdList *possible_cpus = mc->possible_cpu_arch_ids(ms); 1303 const CPUArchId *cpus = possible_cpus->cpus; 1304 int i, j; 1305 1306 if (state->num_nodes <= 1 || possible_cpus->len <= 1) { 1307 return; 1308 } 1309 1310 /* 1311 * The Linux scheduling domain can't be parsed when the multiple CPUs 1312 * in one cluster have been associated with different NUMA nodes. However, 1313 * it's fine to associate one NUMA node with CPUs in different clusters. 1314 */ 1315 for (i = 0; i < possible_cpus->len; i++) { 1316 for (j = i + 1; j < possible_cpus->len; j++) { 1317 if (cpus[i].props.has_socket_id && 1318 cpus[i].props.has_cluster_id && 1319 cpus[i].props.has_node_id && 1320 cpus[j].props.has_socket_id && 1321 cpus[j].props.has_cluster_id && 1322 cpus[j].props.has_node_id && 1323 cpus[i].props.socket_id == cpus[j].props.socket_id && 1324 cpus[i].props.cluster_id == cpus[j].props.cluster_id && 1325 cpus[i].props.node_id != cpus[j].props.node_id) { 1326 warn_report("CPU-%d and CPU-%d in socket-%" PRId64 "-cluster-%" PRId64 1327 " have been associated with node-%" PRId64 " and node-%" PRId64 1328 " respectively. It can cause OSes like Linux to" 1329 " misbehave", i, j, cpus[i].props.socket_id, 1330 cpus[i].props.cluster_id, cpus[i].props.node_id, 1331 cpus[j].props.node_id); 1332 } 1333 } 1334 } 1335 } 1336 1337 MemoryRegion *machine_consume_memdev(MachineState *machine, 1338 HostMemoryBackend *backend) 1339 { 1340 MemoryRegion *ret = host_memory_backend_get_memory(backend); 1341 1342 if (host_memory_backend_is_mapped(backend)) { 1343 error_report("memory backend %s can't be used multiple times.", 1344 object_get_canonical_path_component(OBJECT(backend))); 1345 exit(EXIT_FAILURE); 1346 } 1347 host_memory_backend_set_mapped(backend, true); 1348 vmstate_register_ram_global(ret); 1349 return ret; 1350 } 1351 1352 static bool create_default_memdev(MachineState *ms, const char *path, Error **errp) 1353 { 1354 Object *obj; 1355 MachineClass *mc = MACHINE_GET_CLASS(ms); 1356 bool r = false; 1357 1358 obj = object_new(path ? TYPE_MEMORY_BACKEND_FILE : TYPE_MEMORY_BACKEND_RAM); 1359 if (path) { 1360 if (!object_property_set_str(obj, "mem-path", path, errp)) { 1361 goto out; 1362 } 1363 } 1364 if (!object_property_set_int(obj, "size", ms->ram_size, errp)) { 1365 goto out; 1366 } 1367 object_property_add_child(object_get_objects_root(), mc->default_ram_id, 1368 obj); 1369 /* Ensure backend's memory region name is equal to mc->default_ram_id */ 1370 if (!object_property_set_bool(obj, "x-use-canonical-path-for-ramblock-id", 1371 false, errp)) { 1372 goto out; 1373 } 1374 if (!user_creatable_complete(USER_CREATABLE(obj), errp)) { 1375 goto out; 1376 } 1377 r = object_property_set_link(OBJECT(ms), "memory-backend", obj, errp); 1378 1379 out: 1380 object_unref(obj); 1381 return r; 1382 } 1383 1384 1385 void machine_run_board_init(MachineState *machine, const char *mem_path, Error **errp) 1386 { 1387 ERRP_GUARD(); 1388 MachineClass *machine_class = MACHINE_GET_CLASS(machine); 1389 ObjectClass *oc = object_class_by_name(machine->cpu_type); 1390 CPUClass *cc; 1391 1392 /* This checkpoint is required by replay to separate prior clock 1393 reading from the other reads, because timer polling functions query 1394 clock values from the log. */ 1395 replay_checkpoint(CHECKPOINT_INIT); 1396 1397 if (!xen_enabled()) { 1398 /* On 32-bit hosts, QEMU is limited by virtual address space */ 1399 if (machine->ram_size > (2047 << 20) && HOST_LONG_BITS == 32) { 1400 error_setg(errp, "at most 2047 MB RAM can be simulated"); 1401 return; 1402 } 1403 } 1404 1405 if (machine->memdev) { 1406 ram_addr_t backend_size = object_property_get_uint(OBJECT(machine->memdev), 1407 "size", &error_abort); 1408 if (backend_size != machine->ram_size) { 1409 error_setg(errp, "Machine memory size does not match the size of the memory backend"); 1410 return; 1411 } 1412 } else if (machine_class->default_ram_id && machine->ram_size && 1413 numa_uses_legacy_mem()) { 1414 if (object_property_find(object_get_objects_root(), 1415 machine_class->default_ram_id)) { 1416 error_setg(errp, "object's id '%s' is reserved for the default" 1417 " RAM backend, it can't be used for any other purposes", 1418 machine_class->default_ram_id); 1419 error_append_hint(errp, 1420 "Change the object's 'id' to something else or disable" 1421 " automatic creation of the default RAM backend by setting" 1422 " 'memory-backend=%s' with '-machine'.\n", 1423 machine_class->default_ram_id); 1424 return; 1425 } 1426 if (!create_default_memdev(current_machine, mem_path, errp)) { 1427 return; 1428 } 1429 } 1430 1431 if (machine->numa_state) { 1432 numa_complete_configuration(machine); 1433 if (machine->numa_state->num_nodes) { 1434 machine_numa_finish_cpu_init(machine); 1435 if (machine_class->cpu_cluster_has_numa_boundary) { 1436 validate_cpu_cluster_to_numa_boundary(machine); 1437 } 1438 } 1439 } 1440 1441 if (!machine->ram && machine->memdev) { 1442 machine->ram = machine_consume_memdev(machine, machine->memdev); 1443 } 1444 1445 /* If the machine supports the valid_cpu_types check and the user 1446 * specified a CPU with -cpu check here that the user CPU is supported. 1447 */ 1448 if (machine_class->valid_cpu_types && machine->cpu_type) { 1449 int i; 1450 1451 for (i = 0; machine_class->valid_cpu_types[i]; i++) { 1452 if (object_class_dynamic_cast(oc, 1453 machine_class->valid_cpu_types[i])) { 1454 /* The user specified CPU is in the valid field, we are 1455 * good to go. 1456 */ 1457 break; 1458 } 1459 } 1460 1461 if (!machine_class->valid_cpu_types[i]) { 1462 /* The user specified CPU is not valid */ 1463 error_report("Invalid CPU type: %s", machine->cpu_type); 1464 error_printf("The valid types are: %s", 1465 machine_class->valid_cpu_types[0]); 1466 for (i = 1; machine_class->valid_cpu_types[i]; i++) { 1467 error_printf(", %s", machine_class->valid_cpu_types[i]); 1468 } 1469 error_printf("\n"); 1470 1471 exit(1); 1472 } 1473 } 1474 1475 /* Check if CPU type is deprecated and warn if so */ 1476 cc = CPU_CLASS(oc); 1477 if (cc && cc->deprecation_note) { 1478 warn_report("CPU model %s is deprecated -- %s", machine->cpu_type, 1479 cc->deprecation_note); 1480 } 1481 1482 if (machine->cgs) { 1483 /* 1484 * With confidential guests, the host can't see the real 1485 * contents of RAM, so there's no point in it trying to merge 1486 * areas. 1487 */ 1488 machine_set_mem_merge(OBJECT(machine), false, &error_abort); 1489 1490 /* 1491 * Virtio devices can't count on directly accessing guest 1492 * memory, so they need iommu_platform=on to use normal DMA 1493 * mechanisms. That requires also disabling legacy virtio 1494 * support for those virtio pci devices which allow it. 1495 */ 1496 object_register_sugar_prop(TYPE_VIRTIO_PCI, "disable-legacy", 1497 "on", true); 1498 object_register_sugar_prop(TYPE_VIRTIO_DEVICE, "iommu_platform", 1499 "on", false); 1500 } 1501 1502 accel_init_interfaces(ACCEL_GET_CLASS(machine->accelerator)); 1503 machine_class->init(machine); 1504 phase_advance(PHASE_MACHINE_INITIALIZED); 1505 } 1506 1507 static NotifierList machine_init_done_notifiers = 1508 NOTIFIER_LIST_INITIALIZER(machine_init_done_notifiers); 1509 1510 void qemu_add_machine_init_done_notifier(Notifier *notify) 1511 { 1512 notifier_list_add(&machine_init_done_notifiers, notify); 1513 if (phase_check(PHASE_MACHINE_READY)) { 1514 notify->notify(notify, NULL); 1515 } 1516 } 1517 1518 void qemu_remove_machine_init_done_notifier(Notifier *notify) 1519 { 1520 notifier_remove(notify); 1521 } 1522 1523 void qdev_machine_creation_done(void) 1524 { 1525 cpu_synchronize_all_post_init(); 1526 1527 if (current_machine->boot_config.once) { 1528 qemu_boot_set(current_machine->boot_config.once, &error_fatal); 1529 qemu_register_reset(restore_boot_order, g_strdup(current_machine->boot_config.order)); 1530 } 1531 1532 /* 1533 * ok, initial machine setup is done, starting from now we can 1534 * only create hotpluggable devices 1535 */ 1536 phase_advance(PHASE_MACHINE_READY); 1537 qdev_assert_realized_properly(); 1538 1539 /* TODO: once all bus devices are qdevified, this should be done 1540 * when bus is created by qdev.c */ 1541 /* 1542 * TODO: If we had a main 'reset container' that the whole system 1543 * lived in, we could reset that using the multi-phase reset 1544 * APIs. For the moment, we just reset the sysbus, which will cause 1545 * all devices hanging off it (and all their child buses, recursively) 1546 * to be reset. Note that this will *not* reset any Device objects 1547 * which are not attached to some part of the qbus tree! 1548 */ 1549 qemu_register_reset(resettable_cold_reset_fn, sysbus_get_default()); 1550 1551 notifier_list_notify(&machine_init_done_notifiers, NULL); 1552 1553 if (rom_check_and_register_reset() != 0) { 1554 exit(1); 1555 } 1556 1557 replay_start(); 1558 1559 /* This checkpoint is required by replay to separate prior clock 1560 reading from the other reads, because timer polling functions query 1561 clock values from the log. */ 1562 replay_checkpoint(CHECKPOINT_RESET); 1563 qemu_system_reset(SHUTDOWN_CAUSE_NONE); 1564 register_global_state(); 1565 } 1566 1567 static const TypeInfo machine_info = { 1568 .name = TYPE_MACHINE, 1569 .parent = TYPE_OBJECT, 1570 .abstract = true, 1571 .class_size = sizeof(MachineClass), 1572 .class_init = machine_class_init, 1573 .class_base_init = machine_class_base_init, 1574 .instance_size = sizeof(MachineState), 1575 .instance_init = machine_initfn, 1576 .instance_finalize = machine_finalize, 1577 }; 1578 1579 static void machine_register_types(void) 1580 { 1581 type_register_static(&machine_info); 1582 } 1583 1584 type_init(machine_register_types) 1585