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 { "vfio-pci-nohotplug", "x-ramfb-migrate", "off" } 39 }; 40 const size_t hw_compat_8_1_len = G_N_ELEMENTS(hw_compat_8_1); 41 42 GlobalProperty hw_compat_8_0[] = { 43 { "migration", "multifd-flush-after-each-section", "on"}, 44 { TYPE_PCI_DEVICE, "x-pcie-ari-nextfn-1", "on" }, 45 { TYPE_VIRTIO_NET, "host_uso", "off"}, 46 { TYPE_VIRTIO_NET, "guest_uso4", "off"}, 47 { TYPE_VIRTIO_NET, "guest_uso6", "off"}, 48 }; 49 const size_t hw_compat_8_0_len = G_N_ELEMENTS(hw_compat_8_0); 50 51 GlobalProperty hw_compat_7_2[] = { 52 { "e1000e", "migrate-timadj", "off" }, 53 { "virtio-mem", "x-early-migration", "false" }, 54 { "migration", "x-preempt-pre-7-2", "true" }, 55 { TYPE_PCI_DEVICE, "x-pcie-err-unc-mask", "off" }, 56 }; 57 const size_t hw_compat_7_2_len = G_N_ELEMENTS(hw_compat_7_2); 58 59 GlobalProperty hw_compat_7_1[] = { 60 { "virtio-device", "queue_reset", "false" }, 61 { "virtio-rng-pci", "vectors", "0" }, 62 { "virtio-rng-pci-transitional", "vectors", "0" }, 63 { "virtio-rng-pci-non-transitional", "vectors", "0" }, 64 }; 65 const size_t hw_compat_7_1_len = G_N_ELEMENTS(hw_compat_7_1); 66 67 GlobalProperty hw_compat_7_0[] = { 68 { "arm-gicv3-common", "force-8-bit-prio", "on" }, 69 { "nvme-ns", "eui64-default", "on"}, 70 }; 71 const size_t hw_compat_7_0_len = G_N_ELEMENTS(hw_compat_7_0); 72 73 GlobalProperty hw_compat_6_2[] = { 74 { "PIIX4_PM", "x-not-migrate-acpi-index", "on"}, 75 }; 76 const size_t hw_compat_6_2_len = G_N_ELEMENTS(hw_compat_6_2); 77 78 GlobalProperty hw_compat_6_1[] = { 79 { "vhost-user-vsock-device", "seqpacket", "off" }, 80 { "nvme-ns", "shared", "off" }, 81 }; 82 const size_t hw_compat_6_1_len = G_N_ELEMENTS(hw_compat_6_1); 83 84 GlobalProperty hw_compat_6_0[] = { 85 { "gpex-pcihost", "allow-unmapped-accesses", "false" }, 86 { "i8042", "extended-state", "false"}, 87 { "nvme-ns", "eui64-default", "off"}, 88 { "e1000", "init-vet", "off" }, 89 { "e1000e", "init-vet", "off" }, 90 { "vhost-vsock-device", "seqpacket", "off" }, 91 }; 92 const size_t hw_compat_6_0_len = G_N_ELEMENTS(hw_compat_6_0); 93 94 GlobalProperty hw_compat_5_2[] = { 95 { "ICH9-LPC", "smm-compat", "on"}, 96 { "PIIX4_PM", "smm-compat", "on"}, 97 { "virtio-blk-device", "report-discard-granularity", "off" }, 98 { "virtio-net-pci-base", "vectors", "3"}, 99 }; 100 const size_t hw_compat_5_2_len = G_N_ELEMENTS(hw_compat_5_2); 101 102 GlobalProperty hw_compat_5_1[] = { 103 { "vhost-scsi", "num_queues", "1"}, 104 { "vhost-user-blk", "num-queues", "1"}, 105 { "vhost-user-scsi", "num_queues", "1"}, 106 { "virtio-blk-device", "num-queues", "1"}, 107 { "virtio-scsi-device", "num_queues", "1"}, 108 { "nvme", "use-intel-id", "on"}, 109 { "pvpanic", "events", "1"}, /* PVPANIC_PANICKED */ 110 { "pl011", "migrate-clk", "off" }, 111 { "virtio-pci", "x-ats-page-aligned", "off"}, 112 }; 113 const size_t hw_compat_5_1_len = G_N_ELEMENTS(hw_compat_5_1); 114 115 GlobalProperty hw_compat_5_0[] = { 116 { "pci-host-bridge", "x-config-reg-migration-enabled", "off" }, 117 { "virtio-balloon-device", "page-poison", "false" }, 118 { "vmport", "x-read-set-eax", "off" }, 119 { "vmport", "x-signal-unsupported-cmd", "off" }, 120 { "vmport", "x-report-vmx-type", "off" }, 121 { "vmport", "x-cmds-v2", "off" }, 122 { "virtio-device", "x-disable-legacy-check", "true" }, 123 }; 124 const size_t hw_compat_5_0_len = G_N_ELEMENTS(hw_compat_5_0); 125 126 GlobalProperty hw_compat_4_2[] = { 127 { "virtio-blk-device", "queue-size", "128"}, 128 { "virtio-scsi-device", "virtqueue_size", "128"}, 129 { "virtio-blk-device", "x-enable-wce-if-config-wce", "off" }, 130 { "virtio-blk-device", "seg-max-adjust", "off"}, 131 { "virtio-scsi-device", "seg_max_adjust", "off"}, 132 { "vhost-blk-device", "seg_max_adjust", "off"}, 133 { "usb-host", "suppress-remote-wake", "off" }, 134 { "usb-redir", "suppress-remote-wake", "off" }, 135 { "qxl", "revision", "4" }, 136 { "qxl-vga", "revision", "4" }, 137 { "fw_cfg", "acpi-mr-restore", "false" }, 138 { "virtio-device", "use-disabled-flag", "false" }, 139 }; 140 const size_t hw_compat_4_2_len = G_N_ELEMENTS(hw_compat_4_2); 141 142 GlobalProperty hw_compat_4_1[] = { 143 { "virtio-pci", "x-pcie-flr-init", "off" }, 144 }; 145 const size_t hw_compat_4_1_len = G_N_ELEMENTS(hw_compat_4_1); 146 147 GlobalProperty hw_compat_4_0[] = { 148 { "VGA", "edid", "false" }, 149 { "secondary-vga", "edid", "false" }, 150 { "bochs-display", "edid", "false" }, 151 { "virtio-vga", "edid", "false" }, 152 { "virtio-gpu-device", "edid", "false" }, 153 { "virtio-device", "use-started", "false" }, 154 { "virtio-balloon-device", "qemu-4-0-config-size", "true" }, 155 { "pl031", "migrate-tick-offset", "false" }, 156 }; 157 const size_t hw_compat_4_0_len = G_N_ELEMENTS(hw_compat_4_0); 158 159 GlobalProperty hw_compat_3_1[] = { 160 { "pcie-root-port", "x-speed", "2_5" }, 161 { "pcie-root-port", "x-width", "1" }, 162 { "memory-backend-file", "x-use-canonical-path-for-ramblock-id", "true" }, 163 { "memory-backend-memfd", "x-use-canonical-path-for-ramblock-id", "true" }, 164 { "tpm-crb", "ppi", "false" }, 165 { "tpm-tis", "ppi", "false" }, 166 { "usb-kbd", "serial", "42" }, 167 { "usb-mouse", "serial", "42" }, 168 { "usb-tablet", "serial", "42" }, 169 { "virtio-blk-device", "discard", "false" }, 170 { "virtio-blk-device", "write-zeroes", "false" }, 171 { "virtio-balloon-device", "qemu-4-0-config-size", "false" }, 172 { "pcie-root-port-base", "disable-acs", "true" }, /* Added in 4.1 */ 173 }; 174 const size_t hw_compat_3_1_len = G_N_ELEMENTS(hw_compat_3_1); 175 176 GlobalProperty hw_compat_3_0[] = {}; 177 const size_t hw_compat_3_0_len = G_N_ELEMENTS(hw_compat_3_0); 178 179 GlobalProperty hw_compat_2_12[] = { 180 { "migration", "decompress-error-check", "off" }, 181 { "hda-audio", "use-timer", "false" }, 182 { "cirrus-vga", "global-vmstate", "true" }, 183 { "VGA", "global-vmstate", "true" }, 184 { "vmware-svga", "global-vmstate", "true" }, 185 { "qxl-vga", "global-vmstate", "true" }, 186 }; 187 const size_t hw_compat_2_12_len = G_N_ELEMENTS(hw_compat_2_12); 188 189 GlobalProperty hw_compat_2_11[] = { 190 { "hpet", "hpet-offset-saved", "false" }, 191 { "virtio-blk-pci", "vectors", "2" }, 192 { "vhost-user-blk-pci", "vectors", "2" }, 193 { "e1000", "migrate_tso_props", "off" }, 194 }; 195 const size_t hw_compat_2_11_len = G_N_ELEMENTS(hw_compat_2_11); 196 197 GlobalProperty hw_compat_2_10[] = { 198 { "virtio-mouse-device", "wheel-axis", "false" }, 199 { "virtio-tablet-device", "wheel-axis", "false" }, 200 }; 201 const size_t hw_compat_2_10_len = G_N_ELEMENTS(hw_compat_2_10); 202 203 GlobalProperty hw_compat_2_9[] = { 204 { "pci-bridge", "shpc", "off" }, 205 { "intel-iommu", "pt", "off" }, 206 { "virtio-net-device", "x-mtu-bypass-backend", "off" }, 207 { "pcie-root-port", "x-migrate-msix", "false" }, 208 }; 209 const size_t hw_compat_2_9_len = G_N_ELEMENTS(hw_compat_2_9); 210 211 GlobalProperty hw_compat_2_8[] = { 212 { "fw_cfg_mem", "x-file-slots", "0x10" }, 213 { "fw_cfg_io", "x-file-slots", "0x10" }, 214 { "pflash_cfi01", "old-multiple-chip-handling", "on" }, 215 { "pci-bridge", "shpc", "on" }, 216 { TYPE_PCI_DEVICE, "x-pcie-extcap-init", "off" }, 217 { "virtio-pci", "x-pcie-deverr-init", "off" }, 218 { "virtio-pci", "x-pcie-lnkctl-init", "off" }, 219 { "virtio-pci", "x-pcie-pm-init", "off" }, 220 { "cirrus-vga", "vgamem_mb", "8" }, 221 { "isa-cirrus-vga", "vgamem_mb", "8" }, 222 }; 223 const size_t hw_compat_2_8_len = G_N_ELEMENTS(hw_compat_2_8); 224 225 GlobalProperty hw_compat_2_7[] = { 226 { "virtio-pci", "page-per-vq", "on" }, 227 { "virtio-serial-device", "emergency-write", "off" }, 228 { "ioapic", "version", "0x11" }, 229 { "intel-iommu", "x-buggy-eim", "true" }, 230 { "virtio-pci", "x-ignore-backend-features", "on" }, 231 }; 232 const size_t hw_compat_2_7_len = G_N_ELEMENTS(hw_compat_2_7); 233 234 GlobalProperty hw_compat_2_6[] = { 235 { "virtio-mmio", "format_transport_address", "off" }, 236 /* Optional because not all virtio-pci devices support legacy mode */ 237 { "virtio-pci", "disable-modern", "on", .optional = true }, 238 { "virtio-pci", "disable-legacy", "off", .optional = true }, 239 }; 240 const size_t hw_compat_2_6_len = G_N_ELEMENTS(hw_compat_2_6); 241 242 GlobalProperty hw_compat_2_5[] = { 243 { "isa-fdc", "fallback", "144" }, 244 { "pvscsi", "x-old-pci-configuration", "on" }, 245 { "pvscsi", "x-disable-pcie", "on" }, 246 { "vmxnet3", "x-old-msi-offsets", "on" }, 247 { "vmxnet3", "x-disable-pcie", "on" }, 248 }; 249 const size_t hw_compat_2_5_len = G_N_ELEMENTS(hw_compat_2_5); 250 251 GlobalProperty hw_compat_2_4[] = { 252 /* Optional because the 'scsi' property is Linux-only */ 253 { "virtio-blk-device", "scsi", "true", .optional = true }, 254 { "e1000", "extra_mac_registers", "off" }, 255 { "virtio-pci", "x-disable-pcie", "on" }, 256 { "virtio-pci", "migrate-extra", "off" }, 257 { "fw_cfg_mem", "dma_enabled", "off" }, 258 { "fw_cfg_io", "dma_enabled", "off" } 259 }; 260 const size_t hw_compat_2_4_len = G_N_ELEMENTS(hw_compat_2_4); 261 262 GlobalProperty hw_compat_2_3[] = { 263 { "virtio-blk-pci", "any_layout", "off" }, 264 { "virtio-balloon-pci", "any_layout", "off" }, 265 { "virtio-serial-pci", "any_layout", "off" }, 266 { "virtio-9p-pci", "any_layout", "off" }, 267 { "virtio-rng-pci", "any_layout", "off" }, 268 { TYPE_PCI_DEVICE, "x-pcie-lnksta-dllla", "off" }, 269 { "migration", "send-configuration", "off" }, 270 { "migration", "send-section-footer", "off" }, 271 { "migration", "store-global-state", "off" }, 272 }; 273 const size_t hw_compat_2_3_len = G_N_ELEMENTS(hw_compat_2_3); 274 275 GlobalProperty hw_compat_2_2[] = {}; 276 const size_t hw_compat_2_2_len = G_N_ELEMENTS(hw_compat_2_2); 277 278 GlobalProperty hw_compat_2_1[] = { 279 { "intel-hda", "old_msi_addr", "on" }, 280 { "VGA", "qemu-extended-regs", "off" }, 281 { "secondary-vga", "qemu-extended-regs", "off" }, 282 { "virtio-scsi-pci", "any_layout", "off" }, 283 { "usb-mouse", "usb_version", "1" }, 284 { "usb-kbd", "usb_version", "1" }, 285 { "virtio-pci", "virtio-pci-bus-master-bug-migration", "on" }, 286 }; 287 const size_t hw_compat_2_1_len = G_N_ELEMENTS(hw_compat_2_1); 288 289 MachineState *current_machine; 290 291 static char *machine_get_kernel(Object *obj, Error **errp) 292 { 293 MachineState *ms = MACHINE(obj); 294 295 return g_strdup(ms->kernel_filename); 296 } 297 298 static void machine_set_kernel(Object *obj, const char *value, Error **errp) 299 { 300 MachineState *ms = MACHINE(obj); 301 302 g_free(ms->kernel_filename); 303 ms->kernel_filename = g_strdup(value); 304 } 305 306 static char *machine_get_initrd(Object *obj, Error **errp) 307 { 308 MachineState *ms = MACHINE(obj); 309 310 return g_strdup(ms->initrd_filename); 311 } 312 313 static void machine_set_initrd(Object *obj, const char *value, Error **errp) 314 { 315 MachineState *ms = MACHINE(obj); 316 317 g_free(ms->initrd_filename); 318 ms->initrd_filename = g_strdup(value); 319 } 320 321 static char *machine_get_append(Object *obj, Error **errp) 322 { 323 MachineState *ms = MACHINE(obj); 324 325 return g_strdup(ms->kernel_cmdline); 326 } 327 328 static void machine_set_append(Object *obj, const char *value, Error **errp) 329 { 330 MachineState *ms = MACHINE(obj); 331 332 g_free(ms->kernel_cmdline); 333 ms->kernel_cmdline = g_strdup(value); 334 } 335 336 static char *machine_get_dtb(Object *obj, Error **errp) 337 { 338 MachineState *ms = MACHINE(obj); 339 340 return g_strdup(ms->dtb); 341 } 342 343 static void machine_set_dtb(Object *obj, const char *value, Error **errp) 344 { 345 MachineState *ms = MACHINE(obj); 346 347 g_free(ms->dtb); 348 ms->dtb = g_strdup(value); 349 } 350 351 static char *machine_get_dumpdtb(Object *obj, Error **errp) 352 { 353 MachineState *ms = MACHINE(obj); 354 355 return g_strdup(ms->dumpdtb); 356 } 357 358 static void machine_set_dumpdtb(Object *obj, const char *value, Error **errp) 359 { 360 MachineState *ms = MACHINE(obj); 361 362 g_free(ms->dumpdtb); 363 ms->dumpdtb = g_strdup(value); 364 } 365 366 static void machine_get_phandle_start(Object *obj, Visitor *v, 367 const char *name, void *opaque, 368 Error **errp) 369 { 370 MachineState *ms = MACHINE(obj); 371 int64_t value = ms->phandle_start; 372 373 visit_type_int(v, name, &value, errp); 374 } 375 376 static void machine_set_phandle_start(Object *obj, Visitor *v, 377 const char *name, void *opaque, 378 Error **errp) 379 { 380 MachineState *ms = MACHINE(obj); 381 int64_t value; 382 383 if (!visit_type_int(v, name, &value, errp)) { 384 return; 385 } 386 387 ms->phandle_start = value; 388 } 389 390 static char *machine_get_dt_compatible(Object *obj, Error **errp) 391 { 392 MachineState *ms = MACHINE(obj); 393 394 return g_strdup(ms->dt_compatible); 395 } 396 397 static void machine_set_dt_compatible(Object *obj, const char *value, Error **errp) 398 { 399 MachineState *ms = MACHINE(obj); 400 401 g_free(ms->dt_compatible); 402 ms->dt_compatible = g_strdup(value); 403 } 404 405 static bool machine_get_dump_guest_core(Object *obj, Error **errp) 406 { 407 MachineState *ms = MACHINE(obj); 408 409 return ms->dump_guest_core; 410 } 411 412 static void machine_set_dump_guest_core(Object *obj, bool value, Error **errp) 413 { 414 MachineState *ms = MACHINE(obj); 415 416 ms->dump_guest_core = value; 417 } 418 419 static bool machine_get_mem_merge(Object *obj, Error **errp) 420 { 421 MachineState *ms = MACHINE(obj); 422 423 return ms->mem_merge; 424 } 425 426 static void machine_set_mem_merge(Object *obj, bool value, Error **errp) 427 { 428 MachineState *ms = MACHINE(obj); 429 430 ms->mem_merge = value; 431 } 432 433 static bool machine_get_usb(Object *obj, Error **errp) 434 { 435 MachineState *ms = MACHINE(obj); 436 437 return ms->usb; 438 } 439 440 static void machine_set_usb(Object *obj, bool value, Error **errp) 441 { 442 MachineState *ms = MACHINE(obj); 443 444 ms->usb = value; 445 ms->usb_disabled = !value; 446 } 447 448 static bool machine_get_graphics(Object *obj, Error **errp) 449 { 450 MachineState *ms = MACHINE(obj); 451 452 return ms->enable_graphics; 453 } 454 455 static void machine_set_graphics(Object *obj, bool value, Error **errp) 456 { 457 MachineState *ms = MACHINE(obj); 458 459 ms->enable_graphics = value; 460 } 461 462 static char *machine_get_firmware(Object *obj, Error **errp) 463 { 464 MachineState *ms = MACHINE(obj); 465 466 return g_strdup(ms->firmware); 467 } 468 469 static void machine_set_firmware(Object *obj, const char *value, Error **errp) 470 { 471 MachineState *ms = MACHINE(obj); 472 473 g_free(ms->firmware); 474 ms->firmware = g_strdup(value); 475 } 476 477 static void machine_set_suppress_vmdesc(Object *obj, bool value, Error **errp) 478 { 479 MachineState *ms = MACHINE(obj); 480 481 ms->suppress_vmdesc = value; 482 } 483 484 static bool machine_get_suppress_vmdesc(Object *obj, Error **errp) 485 { 486 MachineState *ms = MACHINE(obj); 487 488 return ms->suppress_vmdesc; 489 } 490 491 static char *machine_get_memory_encryption(Object *obj, Error **errp) 492 { 493 MachineState *ms = MACHINE(obj); 494 495 if (ms->cgs) { 496 return g_strdup(object_get_canonical_path_component(OBJECT(ms->cgs))); 497 } 498 499 return NULL; 500 } 501 502 static void machine_set_memory_encryption(Object *obj, const char *value, 503 Error **errp) 504 { 505 Object *cgs = 506 object_resolve_path_component(object_get_objects_root(), value); 507 508 if (!cgs) { 509 error_setg(errp, "No such memory encryption object '%s'", value); 510 return; 511 } 512 513 object_property_set_link(obj, "confidential-guest-support", cgs, errp); 514 } 515 516 static void machine_check_confidential_guest_support(const Object *obj, 517 const char *name, 518 Object *new_target, 519 Error **errp) 520 { 521 /* 522 * So far the only constraint is that the target has the 523 * TYPE_CONFIDENTIAL_GUEST_SUPPORT interface, and that's checked 524 * by the QOM core 525 */ 526 } 527 528 static bool machine_get_nvdimm(Object *obj, Error **errp) 529 { 530 MachineState *ms = MACHINE(obj); 531 532 return ms->nvdimms_state->is_enabled; 533 } 534 535 static void machine_set_nvdimm(Object *obj, bool value, Error **errp) 536 { 537 MachineState *ms = MACHINE(obj); 538 539 ms->nvdimms_state->is_enabled = value; 540 } 541 542 static bool machine_get_hmat(Object *obj, Error **errp) 543 { 544 MachineState *ms = MACHINE(obj); 545 546 return ms->numa_state->hmat_enabled; 547 } 548 549 static void machine_set_hmat(Object *obj, bool value, Error **errp) 550 { 551 MachineState *ms = MACHINE(obj); 552 553 ms->numa_state->hmat_enabled = value; 554 } 555 556 static void machine_get_mem(Object *obj, Visitor *v, const char *name, 557 void *opaque, Error **errp) 558 { 559 MachineState *ms = MACHINE(obj); 560 MemorySizeConfiguration mem = { 561 .has_size = true, 562 .size = ms->ram_size, 563 .has_max_size = !!ms->ram_slots, 564 .max_size = ms->maxram_size, 565 .has_slots = !!ms->ram_slots, 566 .slots = ms->ram_slots, 567 }; 568 MemorySizeConfiguration *p_mem = &mem; 569 570 visit_type_MemorySizeConfiguration(v, name, &p_mem, &error_abort); 571 } 572 573 static void machine_set_mem(Object *obj, Visitor *v, const char *name, 574 void *opaque, Error **errp) 575 { 576 ERRP_GUARD(); 577 MachineState *ms = MACHINE(obj); 578 MachineClass *mc = MACHINE_GET_CLASS(obj); 579 MemorySizeConfiguration *mem; 580 581 if (!visit_type_MemorySizeConfiguration(v, name, &mem, errp)) { 582 return; 583 } 584 585 if (!mem->has_size) { 586 mem->has_size = true; 587 mem->size = mc->default_ram_size; 588 } 589 mem->size = QEMU_ALIGN_UP(mem->size, 8192); 590 if (mc->fixup_ram_size) { 591 mem->size = mc->fixup_ram_size(mem->size); 592 } 593 if ((ram_addr_t)mem->size != mem->size) { 594 error_setg(errp, "ram size too large"); 595 goto out_free; 596 } 597 598 if (mem->has_max_size) { 599 if (mem->max_size < mem->size) { 600 error_setg(errp, "invalid value of maxmem: " 601 "maximum memory size (0x%" PRIx64 ") must be at least " 602 "the initial memory size (0x%" PRIx64 ")", 603 mem->max_size, mem->size); 604 goto out_free; 605 } 606 if (mem->has_slots && mem->slots && mem->max_size == mem->size) { 607 error_setg(errp, "invalid value of maxmem: " 608 "memory slots were specified but maximum memory size " 609 "(0x%" PRIx64 ") is equal to the initial memory size " 610 "(0x%" PRIx64 ")", mem->max_size, mem->size); 611 goto out_free; 612 } 613 ms->maxram_size = mem->max_size; 614 } else { 615 if (mem->has_slots) { 616 error_setg(errp, "slots specified but no max-size"); 617 goto out_free; 618 } 619 ms->maxram_size = mem->size; 620 } 621 ms->ram_size = mem->size; 622 ms->ram_slots = mem->has_slots ? mem->slots : 0; 623 out_free: 624 qapi_free_MemorySizeConfiguration(mem); 625 } 626 627 static char *machine_get_nvdimm_persistence(Object *obj, Error **errp) 628 { 629 MachineState *ms = MACHINE(obj); 630 631 return g_strdup(ms->nvdimms_state->persistence_string); 632 } 633 634 static void machine_set_nvdimm_persistence(Object *obj, const char *value, 635 Error **errp) 636 { 637 MachineState *ms = MACHINE(obj); 638 NVDIMMState *nvdimms_state = ms->nvdimms_state; 639 640 if (strcmp(value, "cpu") == 0) { 641 nvdimms_state->persistence = 3; 642 } else if (strcmp(value, "mem-ctrl") == 0) { 643 nvdimms_state->persistence = 2; 644 } else { 645 error_setg(errp, "-machine nvdimm-persistence=%s: unsupported option", 646 value); 647 return; 648 } 649 650 g_free(nvdimms_state->persistence_string); 651 nvdimms_state->persistence_string = g_strdup(value); 652 } 653 654 void machine_class_allow_dynamic_sysbus_dev(MachineClass *mc, const char *type) 655 { 656 QAPI_LIST_PREPEND(mc->allowed_dynamic_sysbus_devices, g_strdup(type)); 657 } 658 659 bool device_is_dynamic_sysbus(MachineClass *mc, DeviceState *dev) 660 { 661 Object *obj = OBJECT(dev); 662 663 if (!object_dynamic_cast(obj, TYPE_SYS_BUS_DEVICE)) { 664 return false; 665 } 666 667 return device_type_is_dynamic_sysbus(mc, object_get_typename(obj)); 668 } 669 670 bool device_type_is_dynamic_sysbus(MachineClass *mc, const char *type) 671 { 672 bool allowed = false; 673 strList *wl; 674 ObjectClass *klass = object_class_by_name(type); 675 676 for (wl = mc->allowed_dynamic_sysbus_devices; 677 !allowed && wl; 678 wl = wl->next) { 679 allowed |= !!object_class_dynamic_cast(klass, wl->value); 680 } 681 682 return allowed; 683 } 684 685 static char *machine_get_audiodev(Object *obj, Error **errp) 686 { 687 MachineState *ms = MACHINE(obj); 688 689 return g_strdup(ms->audiodev); 690 } 691 692 static void machine_set_audiodev(Object *obj, const char *value, 693 Error **errp) 694 { 695 MachineState *ms = MACHINE(obj); 696 697 if (!audio_state_by_name(value, errp)) { 698 return; 699 } 700 701 g_free(ms->audiodev); 702 ms->audiodev = g_strdup(value); 703 } 704 705 HotpluggableCPUList *machine_query_hotpluggable_cpus(MachineState *machine) 706 { 707 int i; 708 HotpluggableCPUList *head = NULL; 709 MachineClass *mc = MACHINE_GET_CLASS(machine); 710 711 /* force board to initialize possible_cpus if it hasn't been done yet */ 712 mc->possible_cpu_arch_ids(machine); 713 714 for (i = 0; i < machine->possible_cpus->len; i++) { 715 Object *cpu; 716 HotpluggableCPU *cpu_item = g_new0(typeof(*cpu_item), 1); 717 718 cpu_item->type = g_strdup(machine->possible_cpus->cpus[i].type); 719 cpu_item->vcpus_count = machine->possible_cpus->cpus[i].vcpus_count; 720 cpu_item->props = g_memdup(&machine->possible_cpus->cpus[i].props, 721 sizeof(*cpu_item->props)); 722 723 cpu = machine->possible_cpus->cpus[i].cpu; 724 if (cpu) { 725 cpu_item->qom_path = object_get_canonical_path(cpu); 726 } 727 QAPI_LIST_PREPEND(head, cpu_item); 728 } 729 return head; 730 } 731 732 /** 733 * machine_set_cpu_numa_node: 734 * @machine: machine object to modify 735 * @props: specifies which cpu objects to assign to 736 * numa node specified by @props.node_id 737 * @errp: if an error occurs, a pointer to an area to store the error 738 * 739 * Associate NUMA node specified by @props.node_id with cpu slots that 740 * match socket/core/thread-ids specified by @props. It's recommended to use 741 * query-hotpluggable-cpus.props values to specify affected cpu slots, 742 * which would lead to exact 1:1 mapping of cpu slots to NUMA node. 743 * 744 * However for CLI convenience it's possible to pass in subset of properties, 745 * which would affect all cpu slots that match it. 746 * Ex for pc machine: 747 * -smp 4,cores=2,sockets=2 -numa node,nodeid=0 -numa node,nodeid=1 \ 748 * -numa cpu,node-id=0,socket_id=0 \ 749 * -numa cpu,node-id=1,socket_id=1 750 * will assign all child cores of socket 0 to node 0 and 751 * of socket 1 to node 1. 752 * 753 * On attempt of reassigning (already assigned) cpu slot to another NUMA node, 754 * return error. 755 * Empty subset is disallowed and function will return with error in this case. 756 */ 757 void machine_set_cpu_numa_node(MachineState *machine, 758 const CpuInstanceProperties *props, Error **errp) 759 { 760 MachineClass *mc = MACHINE_GET_CLASS(machine); 761 NodeInfo *numa_info = machine->numa_state->nodes; 762 bool match = false; 763 int i; 764 765 if (!mc->possible_cpu_arch_ids) { 766 error_setg(errp, "mapping of CPUs to NUMA node is not supported"); 767 return; 768 } 769 770 /* disabling node mapping is not supported, forbid it */ 771 assert(props->has_node_id); 772 773 /* force board to initialize possible_cpus if it hasn't been done yet */ 774 mc->possible_cpu_arch_ids(machine); 775 776 for (i = 0; i < machine->possible_cpus->len; i++) { 777 CPUArchId *slot = &machine->possible_cpus->cpus[i]; 778 779 /* reject unsupported by board properties */ 780 if (props->has_thread_id && !slot->props.has_thread_id) { 781 error_setg(errp, "thread-id is not supported"); 782 return; 783 } 784 785 if (props->has_core_id && !slot->props.has_core_id) { 786 error_setg(errp, "core-id is not supported"); 787 return; 788 } 789 790 if (props->has_cluster_id && !slot->props.has_cluster_id) { 791 error_setg(errp, "cluster-id is not supported"); 792 return; 793 } 794 795 if (props->has_socket_id && !slot->props.has_socket_id) { 796 error_setg(errp, "socket-id is not supported"); 797 return; 798 } 799 800 if (props->has_die_id && !slot->props.has_die_id) { 801 error_setg(errp, "die-id is not supported"); 802 return; 803 } 804 805 /* skip slots with explicit mismatch */ 806 if (props->has_thread_id && props->thread_id != slot->props.thread_id) { 807 continue; 808 } 809 810 if (props->has_core_id && props->core_id != slot->props.core_id) { 811 continue; 812 } 813 814 if (props->has_cluster_id && 815 props->cluster_id != slot->props.cluster_id) { 816 continue; 817 } 818 819 if (props->has_die_id && props->die_id != slot->props.die_id) { 820 continue; 821 } 822 823 if (props->has_socket_id && props->socket_id != slot->props.socket_id) { 824 continue; 825 } 826 827 /* reject assignment if slot is already assigned, for compatibility 828 * of legacy cpu_index mapping with SPAPR core based mapping do not 829 * error out if cpu thread and matched core have the same node-id */ 830 if (slot->props.has_node_id && 831 slot->props.node_id != props->node_id) { 832 error_setg(errp, "CPU is already assigned to node-id: %" PRId64, 833 slot->props.node_id); 834 return; 835 } 836 837 /* assign slot to node as it's matched '-numa cpu' key */ 838 match = true; 839 slot->props.node_id = props->node_id; 840 slot->props.has_node_id = props->has_node_id; 841 842 if (machine->numa_state->hmat_enabled) { 843 if ((numa_info[props->node_id].initiator < MAX_NODES) && 844 (props->node_id != numa_info[props->node_id].initiator)) { 845 error_setg(errp, "The initiator of CPU NUMA node %" PRId64 846 " should be itself (got %" PRIu16 ")", 847 props->node_id, numa_info[props->node_id].initiator); 848 return; 849 } 850 numa_info[props->node_id].has_cpu = true; 851 numa_info[props->node_id].initiator = props->node_id; 852 } 853 } 854 855 if (!match) { 856 error_setg(errp, "no match found"); 857 } 858 } 859 860 static void machine_get_smp(Object *obj, Visitor *v, const char *name, 861 void *opaque, Error **errp) 862 { 863 MachineState *ms = MACHINE(obj); 864 SMPConfiguration *config = &(SMPConfiguration){ 865 .has_cpus = true, .cpus = ms->smp.cpus, 866 .has_sockets = true, .sockets = ms->smp.sockets, 867 .has_dies = true, .dies = ms->smp.dies, 868 .has_clusters = true, .clusters = ms->smp.clusters, 869 .has_cores = true, .cores = ms->smp.cores, 870 .has_threads = true, .threads = ms->smp.threads, 871 .has_maxcpus = true, .maxcpus = ms->smp.max_cpus, 872 }; 873 874 if (!visit_type_SMPConfiguration(v, name, &config, &error_abort)) { 875 return; 876 } 877 } 878 879 static void machine_set_smp(Object *obj, Visitor *v, const char *name, 880 void *opaque, Error **errp) 881 { 882 MachineState *ms = MACHINE(obj); 883 g_autoptr(SMPConfiguration) config = NULL; 884 885 if (!visit_type_SMPConfiguration(v, name, &config, errp)) { 886 return; 887 } 888 889 machine_parse_smp_config(ms, config, errp); 890 } 891 892 static void machine_get_boot(Object *obj, Visitor *v, const char *name, 893 void *opaque, Error **errp) 894 { 895 MachineState *ms = MACHINE(obj); 896 BootConfiguration *config = &ms->boot_config; 897 visit_type_BootConfiguration(v, name, &config, &error_abort); 898 } 899 900 static void machine_free_boot_config(MachineState *ms) 901 { 902 g_free(ms->boot_config.order); 903 g_free(ms->boot_config.once); 904 g_free(ms->boot_config.splash); 905 } 906 907 static void machine_copy_boot_config(MachineState *ms, BootConfiguration *config) 908 { 909 MachineClass *machine_class = MACHINE_GET_CLASS(ms); 910 911 machine_free_boot_config(ms); 912 ms->boot_config = *config; 913 if (!config->order) { 914 ms->boot_config.order = g_strdup(machine_class->default_boot_order); 915 } 916 } 917 918 static void machine_set_boot(Object *obj, Visitor *v, const char *name, 919 void *opaque, Error **errp) 920 { 921 ERRP_GUARD(); 922 MachineState *ms = MACHINE(obj); 923 BootConfiguration *config = NULL; 924 925 if (!visit_type_BootConfiguration(v, name, &config, errp)) { 926 return; 927 } 928 if (config->order) { 929 validate_bootdevices(config->order, errp); 930 if (*errp) { 931 goto out_free; 932 } 933 } 934 if (config->once) { 935 validate_bootdevices(config->once, errp); 936 if (*errp) { 937 goto out_free; 938 } 939 } 940 941 machine_copy_boot_config(ms, config); 942 /* Strings live in ms->boot_config. */ 943 free(config); 944 return; 945 946 out_free: 947 qapi_free_BootConfiguration(config); 948 } 949 950 void machine_add_audiodev_property(MachineClass *mc) 951 { 952 ObjectClass *oc = OBJECT_CLASS(mc); 953 954 object_class_property_add_str(oc, "audiodev", 955 machine_get_audiodev, 956 machine_set_audiodev); 957 object_class_property_set_description(oc, "audiodev", 958 "Audiodev to use for default machine devices"); 959 } 960 961 static void machine_class_init(ObjectClass *oc, void *data) 962 { 963 MachineClass *mc = MACHINE_CLASS(oc); 964 965 /* Default 128 MB as guest ram size */ 966 mc->default_ram_size = 128 * MiB; 967 mc->rom_file_has_mr = true; 968 969 /* numa node memory size aligned on 8MB by default. 970 * On Linux, each node's border has to be 8MB aligned 971 */ 972 mc->numa_mem_align_shift = 23; 973 974 object_class_property_add_str(oc, "kernel", 975 machine_get_kernel, machine_set_kernel); 976 object_class_property_set_description(oc, "kernel", 977 "Linux kernel image file"); 978 979 object_class_property_add_str(oc, "initrd", 980 machine_get_initrd, machine_set_initrd); 981 object_class_property_set_description(oc, "initrd", 982 "Linux initial ramdisk file"); 983 984 object_class_property_add_str(oc, "append", 985 machine_get_append, machine_set_append); 986 object_class_property_set_description(oc, "append", 987 "Linux kernel command line"); 988 989 object_class_property_add_str(oc, "dtb", 990 machine_get_dtb, machine_set_dtb); 991 object_class_property_set_description(oc, "dtb", 992 "Linux kernel device tree file"); 993 994 object_class_property_add_str(oc, "dumpdtb", 995 machine_get_dumpdtb, machine_set_dumpdtb); 996 object_class_property_set_description(oc, "dumpdtb", 997 "Dump current dtb to a file and quit"); 998 999 object_class_property_add(oc, "boot", "BootConfiguration", 1000 machine_get_boot, machine_set_boot, 1001 NULL, NULL); 1002 object_class_property_set_description(oc, "boot", 1003 "Boot configuration"); 1004 1005 object_class_property_add(oc, "smp", "SMPConfiguration", 1006 machine_get_smp, machine_set_smp, 1007 NULL, NULL); 1008 object_class_property_set_description(oc, "smp", 1009 "CPU topology"); 1010 1011 object_class_property_add(oc, "phandle-start", "int", 1012 machine_get_phandle_start, machine_set_phandle_start, 1013 NULL, NULL); 1014 object_class_property_set_description(oc, "phandle-start", 1015 "The first phandle ID we may generate dynamically"); 1016 1017 object_class_property_add_str(oc, "dt-compatible", 1018 machine_get_dt_compatible, machine_set_dt_compatible); 1019 object_class_property_set_description(oc, "dt-compatible", 1020 "Overrides the \"compatible\" property of the dt root node"); 1021 1022 object_class_property_add_bool(oc, "dump-guest-core", 1023 machine_get_dump_guest_core, machine_set_dump_guest_core); 1024 object_class_property_set_description(oc, "dump-guest-core", 1025 "Include guest memory in a core dump"); 1026 1027 object_class_property_add_bool(oc, "mem-merge", 1028 machine_get_mem_merge, machine_set_mem_merge); 1029 object_class_property_set_description(oc, "mem-merge", 1030 "Enable/disable memory merge support"); 1031 1032 object_class_property_add_bool(oc, "usb", 1033 machine_get_usb, machine_set_usb); 1034 object_class_property_set_description(oc, "usb", 1035 "Set on/off to enable/disable usb"); 1036 1037 object_class_property_add_bool(oc, "graphics", 1038 machine_get_graphics, machine_set_graphics); 1039 object_class_property_set_description(oc, "graphics", 1040 "Set on/off to enable/disable graphics emulation"); 1041 1042 object_class_property_add_str(oc, "firmware", 1043 machine_get_firmware, machine_set_firmware); 1044 object_class_property_set_description(oc, "firmware", 1045 "Firmware image"); 1046 1047 object_class_property_add_bool(oc, "suppress-vmdesc", 1048 machine_get_suppress_vmdesc, machine_set_suppress_vmdesc); 1049 object_class_property_set_description(oc, "suppress-vmdesc", 1050 "Set on to disable self-describing migration"); 1051 1052 object_class_property_add_link(oc, "confidential-guest-support", 1053 TYPE_CONFIDENTIAL_GUEST_SUPPORT, 1054 offsetof(MachineState, cgs), 1055 machine_check_confidential_guest_support, 1056 OBJ_PROP_LINK_STRONG); 1057 object_class_property_set_description(oc, "confidential-guest-support", 1058 "Set confidential guest scheme to support"); 1059 1060 /* For compatibility */ 1061 object_class_property_add_str(oc, "memory-encryption", 1062 machine_get_memory_encryption, machine_set_memory_encryption); 1063 object_class_property_set_description(oc, "memory-encryption", 1064 "Set memory encryption object to use"); 1065 1066 object_class_property_add_link(oc, "memory-backend", TYPE_MEMORY_BACKEND, 1067 offsetof(MachineState, memdev), object_property_allow_set_link, 1068 OBJ_PROP_LINK_STRONG); 1069 object_class_property_set_description(oc, "memory-backend", 1070 "Set RAM backend" 1071 "Valid value is ID of hostmem based backend"); 1072 1073 object_class_property_add(oc, "memory", "MemorySizeConfiguration", 1074 machine_get_mem, machine_set_mem, 1075 NULL, NULL); 1076 object_class_property_set_description(oc, "memory", 1077 "Memory size configuration"); 1078 } 1079 1080 static void machine_class_base_init(ObjectClass *oc, void *data) 1081 { 1082 MachineClass *mc = MACHINE_CLASS(oc); 1083 mc->max_cpus = mc->max_cpus ?: 1; 1084 mc->min_cpus = mc->min_cpus ?: 1; 1085 mc->default_cpus = mc->default_cpus ?: 1; 1086 1087 if (!object_class_is_abstract(oc)) { 1088 const char *cname = object_class_get_name(oc); 1089 assert(g_str_has_suffix(cname, TYPE_MACHINE_SUFFIX)); 1090 mc->name = g_strndup(cname, 1091 strlen(cname) - strlen(TYPE_MACHINE_SUFFIX)); 1092 mc->compat_props = g_ptr_array_new(); 1093 } 1094 } 1095 1096 static void machine_initfn(Object *obj) 1097 { 1098 MachineState *ms = MACHINE(obj); 1099 MachineClass *mc = MACHINE_GET_CLASS(obj); 1100 1101 container_get(obj, "/peripheral"); 1102 container_get(obj, "/peripheral-anon"); 1103 1104 ms->dump_guest_core = true; 1105 ms->mem_merge = true; 1106 ms->enable_graphics = true; 1107 ms->kernel_cmdline = g_strdup(""); 1108 ms->ram_size = mc->default_ram_size; 1109 ms->maxram_size = mc->default_ram_size; 1110 1111 if (mc->nvdimm_supported) { 1112 ms->nvdimms_state = g_new0(NVDIMMState, 1); 1113 object_property_add_bool(obj, "nvdimm", 1114 machine_get_nvdimm, machine_set_nvdimm); 1115 object_property_set_description(obj, "nvdimm", 1116 "Set on/off to enable/disable " 1117 "NVDIMM instantiation"); 1118 1119 object_property_add_str(obj, "nvdimm-persistence", 1120 machine_get_nvdimm_persistence, 1121 machine_set_nvdimm_persistence); 1122 object_property_set_description(obj, "nvdimm-persistence", 1123 "Set NVDIMM persistence" 1124 "Valid values are cpu, mem-ctrl"); 1125 } 1126 1127 if (mc->cpu_index_to_instance_props && mc->get_default_cpu_node_id) { 1128 ms->numa_state = g_new0(NumaState, 1); 1129 object_property_add_bool(obj, "hmat", 1130 machine_get_hmat, machine_set_hmat); 1131 object_property_set_description(obj, "hmat", 1132 "Set on/off to enable/disable " 1133 "ACPI Heterogeneous Memory Attribute " 1134 "Table (HMAT)"); 1135 } 1136 1137 /* default to mc->default_cpus */ 1138 ms->smp.cpus = mc->default_cpus; 1139 ms->smp.max_cpus = mc->default_cpus; 1140 ms->smp.sockets = 1; 1141 ms->smp.dies = 1; 1142 ms->smp.clusters = 1; 1143 ms->smp.cores = 1; 1144 ms->smp.threads = 1; 1145 1146 machine_copy_boot_config(ms, &(BootConfiguration){ 0 }); 1147 } 1148 1149 static void machine_finalize(Object *obj) 1150 { 1151 MachineState *ms = MACHINE(obj); 1152 1153 machine_free_boot_config(ms); 1154 g_free(ms->kernel_filename); 1155 g_free(ms->initrd_filename); 1156 g_free(ms->kernel_cmdline); 1157 g_free(ms->dtb); 1158 g_free(ms->dumpdtb); 1159 g_free(ms->dt_compatible); 1160 g_free(ms->firmware); 1161 g_free(ms->device_memory); 1162 g_free(ms->nvdimms_state); 1163 g_free(ms->numa_state); 1164 g_free(ms->audiodev); 1165 } 1166 1167 bool machine_usb(MachineState *machine) 1168 { 1169 return machine->usb; 1170 } 1171 1172 int machine_phandle_start(MachineState *machine) 1173 { 1174 return machine->phandle_start; 1175 } 1176 1177 bool machine_dump_guest_core(MachineState *machine) 1178 { 1179 return machine->dump_guest_core; 1180 } 1181 1182 bool machine_mem_merge(MachineState *machine) 1183 { 1184 return machine->mem_merge; 1185 } 1186 1187 static char *cpu_slot_to_string(const CPUArchId *cpu) 1188 { 1189 GString *s = g_string_new(NULL); 1190 if (cpu->props.has_socket_id) { 1191 g_string_append_printf(s, "socket-id: %"PRId64, cpu->props.socket_id); 1192 } 1193 if (cpu->props.has_die_id) { 1194 if (s->len) { 1195 g_string_append_printf(s, ", "); 1196 } 1197 g_string_append_printf(s, "die-id: %"PRId64, cpu->props.die_id); 1198 } 1199 if (cpu->props.has_cluster_id) { 1200 if (s->len) { 1201 g_string_append_printf(s, ", "); 1202 } 1203 g_string_append_printf(s, "cluster-id: %"PRId64, cpu->props.cluster_id); 1204 } 1205 if (cpu->props.has_core_id) { 1206 if (s->len) { 1207 g_string_append_printf(s, ", "); 1208 } 1209 g_string_append_printf(s, "core-id: %"PRId64, cpu->props.core_id); 1210 } 1211 if (cpu->props.has_thread_id) { 1212 if (s->len) { 1213 g_string_append_printf(s, ", "); 1214 } 1215 g_string_append_printf(s, "thread-id: %"PRId64, cpu->props.thread_id); 1216 } 1217 return g_string_free(s, false); 1218 } 1219 1220 static void numa_validate_initiator(NumaState *numa_state) 1221 { 1222 int i; 1223 NodeInfo *numa_info = numa_state->nodes; 1224 1225 for (i = 0; i < numa_state->num_nodes; i++) { 1226 if (numa_info[i].initiator == MAX_NODES) { 1227 continue; 1228 } 1229 1230 if (!numa_info[numa_info[i].initiator].present) { 1231 error_report("NUMA node %" PRIu16 " is missing, use " 1232 "'-numa node' option to declare it first", 1233 numa_info[i].initiator); 1234 exit(1); 1235 } 1236 1237 if (!numa_info[numa_info[i].initiator].has_cpu) { 1238 error_report("The initiator of NUMA node %d is invalid", i); 1239 exit(1); 1240 } 1241 } 1242 } 1243 1244 static void machine_numa_finish_cpu_init(MachineState *machine) 1245 { 1246 int i; 1247 bool default_mapping; 1248 GString *s = g_string_new(NULL); 1249 MachineClass *mc = MACHINE_GET_CLASS(machine); 1250 const CPUArchIdList *possible_cpus = mc->possible_cpu_arch_ids(machine); 1251 1252 assert(machine->numa_state->num_nodes); 1253 for (i = 0; i < possible_cpus->len; i++) { 1254 if (possible_cpus->cpus[i].props.has_node_id) { 1255 break; 1256 } 1257 } 1258 default_mapping = (i == possible_cpus->len); 1259 1260 for (i = 0; i < possible_cpus->len; i++) { 1261 const CPUArchId *cpu_slot = &possible_cpus->cpus[i]; 1262 1263 if (!cpu_slot->props.has_node_id) { 1264 /* fetch default mapping from board and enable it */ 1265 CpuInstanceProperties props = cpu_slot->props; 1266 1267 props.node_id = mc->get_default_cpu_node_id(machine, i); 1268 if (!default_mapping) { 1269 /* record slots with not set mapping, 1270 * TODO: make it hard error in future */ 1271 char *cpu_str = cpu_slot_to_string(cpu_slot); 1272 g_string_append_printf(s, "%sCPU %d [%s]", 1273 s->len ? ", " : "", i, cpu_str); 1274 g_free(cpu_str); 1275 1276 /* non mapped cpus used to fallback to node 0 */ 1277 props.node_id = 0; 1278 } 1279 1280 props.has_node_id = true; 1281 machine_set_cpu_numa_node(machine, &props, &error_fatal); 1282 } 1283 } 1284 1285 if (machine->numa_state->hmat_enabled) { 1286 numa_validate_initiator(machine->numa_state); 1287 } 1288 1289 if (s->len && !qtest_enabled()) { 1290 warn_report("CPU(s) not present in any NUMA nodes: %s", 1291 s->str); 1292 warn_report("All CPU(s) up to maxcpus should be described " 1293 "in NUMA config, ability to start up with partial NUMA " 1294 "mappings is obsoleted and will be removed in future"); 1295 } 1296 g_string_free(s, true); 1297 } 1298 1299 static void validate_cpu_cluster_to_numa_boundary(MachineState *ms) 1300 { 1301 MachineClass *mc = MACHINE_GET_CLASS(ms); 1302 NumaState *state = ms->numa_state; 1303 const CPUArchIdList *possible_cpus = mc->possible_cpu_arch_ids(ms); 1304 const CPUArchId *cpus = possible_cpus->cpus; 1305 int i, j; 1306 1307 if (state->num_nodes <= 1 || possible_cpus->len <= 1) { 1308 return; 1309 } 1310 1311 /* 1312 * The Linux scheduling domain can't be parsed when the multiple CPUs 1313 * in one cluster have been associated with different NUMA nodes. However, 1314 * it's fine to associate one NUMA node with CPUs in different clusters. 1315 */ 1316 for (i = 0; i < possible_cpus->len; i++) { 1317 for (j = i + 1; j < possible_cpus->len; j++) { 1318 if (cpus[i].props.has_socket_id && 1319 cpus[i].props.has_cluster_id && 1320 cpus[i].props.has_node_id && 1321 cpus[j].props.has_socket_id && 1322 cpus[j].props.has_cluster_id && 1323 cpus[j].props.has_node_id && 1324 cpus[i].props.socket_id == cpus[j].props.socket_id && 1325 cpus[i].props.cluster_id == cpus[j].props.cluster_id && 1326 cpus[i].props.node_id != cpus[j].props.node_id) { 1327 warn_report("CPU-%d and CPU-%d in socket-%" PRId64 "-cluster-%" PRId64 1328 " have been associated with node-%" PRId64 " and node-%" PRId64 1329 " respectively. It can cause OSes like Linux to" 1330 " misbehave", i, j, cpus[i].props.socket_id, 1331 cpus[i].props.cluster_id, cpus[i].props.node_id, 1332 cpus[j].props.node_id); 1333 } 1334 } 1335 } 1336 } 1337 1338 MemoryRegion *machine_consume_memdev(MachineState *machine, 1339 HostMemoryBackend *backend) 1340 { 1341 MemoryRegion *ret = host_memory_backend_get_memory(backend); 1342 1343 if (host_memory_backend_is_mapped(backend)) { 1344 error_report("memory backend %s can't be used multiple times.", 1345 object_get_canonical_path_component(OBJECT(backend))); 1346 exit(EXIT_FAILURE); 1347 } 1348 host_memory_backend_set_mapped(backend, true); 1349 vmstate_register_ram_global(ret); 1350 return ret; 1351 } 1352 1353 static bool create_default_memdev(MachineState *ms, const char *path, Error **errp) 1354 { 1355 Object *obj; 1356 MachineClass *mc = MACHINE_GET_CLASS(ms); 1357 bool r = false; 1358 1359 obj = object_new(path ? TYPE_MEMORY_BACKEND_FILE : TYPE_MEMORY_BACKEND_RAM); 1360 if (path) { 1361 if (!object_property_set_str(obj, "mem-path", path, errp)) { 1362 goto out; 1363 } 1364 } 1365 if (!object_property_set_int(obj, "size", ms->ram_size, errp)) { 1366 goto out; 1367 } 1368 object_property_add_child(object_get_objects_root(), mc->default_ram_id, 1369 obj); 1370 /* Ensure backend's memory region name is equal to mc->default_ram_id */ 1371 if (!object_property_set_bool(obj, "x-use-canonical-path-for-ramblock-id", 1372 false, errp)) { 1373 goto out; 1374 } 1375 if (!user_creatable_complete(USER_CREATABLE(obj), errp)) { 1376 goto out; 1377 } 1378 r = object_property_set_link(OBJECT(ms), "memory-backend", obj, errp); 1379 1380 out: 1381 object_unref(obj); 1382 return r; 1383 } 1384 1385 1386 void machine_run_board_init(MachineState *machine, const char *mem_path, Error **errp) 1387 { 1388 ERRP_GUARD(); 1389 MachineClass *machine_class = MACHINE_GET_CLASS(machine); 1390 ObjectClass *oc = object_class_by_name(machine->cpu_type); 1391 CPUClass *cc; 1392 1393 /* This checkpoint is required by replay to separate prior clock 1394 reading from the other reads, because timer polling functions query 1395 clock values from the log. */ 1396 replay_checkpoint(CHECKPOINT_INIT); 1397 1398 if (!xen_enabled()) { 1399 /* On 32-bit hosts, QEMU is limited by virtual address space */ 1400 if (machine->ram_size > (2047 << 20) && HOST_LONG_BITS == 32) { 1401 error_setg(errp, "at most 2047 MB RAM can be simulated"); 1402 return; 1403 } 1404 } 1405 1406 if (machine->memdev) { 1407 ram_addr_t backend_size = object_property_get_uint(OBJECT(machine->memdev), 1408 "size", &error_abort); 1409 if (backend_size != machine->ram_size) { 1410 error_setg(errp, "Machine memory size does not match the size of the memory backend"); 1411 return; 1412 } 1413 } else if (machine_class->default_ram_id && machine->ram_size && 1414 numa_uses_legacy_mem()) { 1415 if (object_property_find(object_get_objects_root(), 1416 machine_class->default_ram_id)) { 1417 error_setg(errp, "object's id '%s' is reserved for the default" 1418 " RAM backend, it can't be used for any other purposes", 1419 machine_class->default_ram_id); 1420 error_append_hint(errp, 1421 "Change the object's 'id' to something else or disable" 1422 " automatic creation of the default RAM backend by setting" 1423 " 'memory-backend=%s' with '-machine'.\n", 1424 machine_class->default_ram_id); 1425 return; 1426 } 1427 if (!create_default_memdev(current_machine, mem_path, errp)) { 1428 return; 1429 } 1430 } 1431 1432 if (machine->numa_state) { 1433 numa_complete_configuration(machine); 1434 if (machine->numa_state->num_nodes) { 1435 machine_numa_finish_cpu_init(machine); 1436 if (machine_class->cpu_cluster_has_numa_boundary) { 1437 validate_cpu_cluster_to_numa_boundary(machine); 1438 } 1439 } 1440 } 1441 1442 if (!machine->ram && machine->memdev) { 1443 machine->ram = machine_consume_memdev(machine, machine->memdev); 1444 } 1445 1446 /* If the machine supports the valid_cpu_types check and the user 1447 * specified a CPU with -cpu check here that the user CPU is supported. 1448 */ 1449 if (machine_class->valid_cpu_types && machine->cpu_type) { 1450 int i; 1451 1452 for (i = 0; machine_class->valid_cpu_types[i]; i++) { 1453 if (object_class_dynamic_cast(oc, 1454 machine_class->valid_cpu_types[i])) { 1455 /* The user specified CPU is in the valid field, we are 1456 * good to go. 1457 */ 1458 break; 1459 } 1460 } 1461 1462 if (!machine_class->valid_cpu_types[i]) { 1463 /* The user specified CPU is not valid */ 1464 error_report("Invalid CPU type: %s", machine->cpu_type); 1465 error_printf("The valid types are: %s", 1466 machine_class->valid_cpu_types[0]); 1467 for (i = 1; machine_class->valid_cpu_types[i]; i++) { 1468 error_printf(", %s", machine_class->valid_cpu_types[i]); 1469 } 1470 error_printf("\n"); 1471 1472 exit(1); 1473 } 1474 } 1475 1476 /* Check if CPU type is deprecated and warn if so */ 1477 cc = CPU_CLASS(oc); 1478 if (cc && cc->deprecation_note) { 1479 warn_report("CPU model %s is deprecated -- %s", machine->cpu_type, 1480 cc->deprecation_note); 1481 } 1482 1483 if (machine->cgs) { 1484 /* 1485 * With confidential guests, the host can't see the real 1486 * contents of RAM, so there's no point in it trying to merge 1487 * areas. 1488 */ 1489 machine_set_mem_merge(OBJECT(machine), false, &error_abort); 1490 1491 /* 1492 * Virtio devices can't count on directly accessing guest 1493 * memory, so they need iommu_platform=on to use normal DMA 1494 * mechanisms. That requires also disabling legacy virtio 1495 * support for those virtio pci devices which allow it. 1496 */ 1497 object_register_sugar_prop(TYPE_VIRTIO_PCI, "disable-legacy", 1498 "on", true); 1499 object_register_sugar_prop(TYPE_VIRTIO_DEVICE, "iommu_platform", 1500 "on", false); 1501 } 1502 1503 accel_init_interfaces(ACCEL_GET_CLASS(machine->accelerator)); 1504 machine_class->init(machine); 1505 phase_advance(PHASE_MACHINE_INITIALIZED); 1506 } 1507 1508 static NotifierList machine_init_done_notifiers = 1509 NOTIFIER_LIST_INITIALIZER(machine_init_done_notifiers); 1510 1511 void qemu_add_machine_init_done_notifier(Notifier *notify) 1512 { 1513 notifier_list_add(&machine_init_done_notifiers, notify); 1514 if (phase_check(PHASE_MACHINE_READY)) { 1515 notify->notify(notify, NULL); 1516 } 1517 } 1518 1519 void qemu_remove_machine_init_done_notifier(Notifier *notify) 1520 { 1521 notifier_remove(notify); 1522 } 1523 1524 void qdev_machine_creation_done(void) 1525 { 1526 cpu_synchronize_all_post_init(); 1527 1528 if (current_machine->boot_config.once) { 1529 qemu_boot_set(current_machine->boot_config.once, &error_fatal); 1530 qemu_register_reset(restore_boot_order, g_strdup(current_machine->boot_config.order)); 1531 } 1532 1533 /* 1534 * ok, initial machine setup is done, starting from now we can 1535 * only create hotpluggable devices 1536 */ 1537 phase_advance(PHASE_MACHINE_READY); 1538 qdev_assert_realized_properly(); 1539 1540 /* TODO: once all bus devices are qdevified, this should be done 1541 * when bus is created by qdev.c */ 1542 /* 1543 * TODO: If we had a main 'reset container' that the whole system 1544 * lived in, we could reset that using the multi-phase reset 1545 * APIs. For the moment, we just reset the sysbus, which will cause 1546 * all devices hanging off it (and all their child buses, recursively) 1547 * to be reset. Note that this will *not* reset any Device objects 1548 * which are not attached to some part of the qbus tree! 1549 */ 1550 qemu_register_reset(resettable_cold_reset_fn, sysbus_get_default()); 1551 1552 notifier_list_notify(&machine_init_done_notifiers, NULL); 1553 1554 if (rom_check_and_register_reset() != 0) { 1555 exit(1); 1556 } 1557 1558 replay_start(); 1559 1560 /* This checkpoint is required by replay to separate prior clock 1561 reading from the other reads, because timer polling functions query 1562 clock values from the log. */ 1563 replay_checkpoint(CHECKPOINT_RESET); 1564 qemu_system_reset(SHUTDOWN_CAUSE_NONE); 1565 register_global_state(); 1566 } 1567 1568 static const TypeInfo machine_info = { 1569 .name = TYPE_MACHINE, 1570 .parent = TYPE_OBJECT, 1571 .abstract = true, 1572 .class_size = sizeof(MachineClass), 1573 .class_init = machine_class_init, 1574 .class_base_init = machine_class_base_init, 1575 .instance_size = sizeof(MachineState), 1576 .instance_init = machine_initfn, 1577 .instance_finalize = machine_finalize, 1578 }; 1579 1580 static void machine_register_types(void) 1581 { 1582 type_register_static(&machine_info); 1583 } 1584 1585 type_init(machine_register_types) 1586