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