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 { "igb", "x-pcie-flr-init", "off" }, 40 }; 41 const size_t hw_compat_8_1_len = G_N_ELEMENTS(hw_compat_8_1); 42 43 GlobalProperty hw_compat_8_0[] = { 44 { "migration", "multifd-flush-after-each-section", "on"}, 45 { TYPE_PCI_DEVICE, "x-pcie-ari-nextfn-1", "on" }, 46 { TYPE_VIRTIO_NET, "host_uso", "off"}, 47 { TYPE_VIRTIO_NET, "guest_uso4", "off"}, 48 { TYPE_VIRTIO_NET, "guest_uso6", "off"}, 49 }; 50 const size_t hw_compat_8_0_len = G_N_ELEMENTS(hw_compat_8_0); 51 52 GlobalProperty hw_compat_7_2[] = { 53 { "e1000e", "migrate-timadj", "off" }, 54 { "virtio-mem", "x-early-migration", "false" }, 55 { "migration", "x-preempt-pre-7-2", "true" }, 56 { TYPE_PCI_DEVICE, "x-pcie-err-unc-mask", "off" }, 57 }; 58 const size_t hw_compat_7_2_len = G_N_ELEMENTS(hw_compat_7_2); 59 60 GlobalProperty hw_compat_7_1[] = { 61 { "virtio-device", "queue_reset", "false" }, 62 { "virtio-rng-pci", "vectors", "0" }, 63 { "virtio-rng-pci-transitional", "vectors", "0" }, 64 { "virtio-rng-pci-non-transitional", "vectors", "0" }, 65 }; 66 const size_t hw_compat_7_1_len = G_N_ELEMENTS(hw_compat_7_1); 67 68 GlobalProperty hw_compat_7_0[] = { 69 { "arm-gicv3-common", "force-8-bit-prio", "on" }, 70 { "nvme-ns", "eui64-default", "on"}, 71 }; 72 const size_t hw_compat_7_0_len = G_N_ELEMENTS(hw_compat_7_0); 73 74 GlobalProperty hw_compat_6_2[] = { 75 { "PIIX4_PM", "x-not-migrate-acpi-index", "on"}, 76 }; 77 const size_t hw_compat_6_2_len = G_N_ELEMENTS(hw_compat_6_2); 78 79 GlobalProperty hw_compat_6_1[] = { 80 { "vhost-user-vsock-device", "seqpacket", "off" }, 81 { "nvme-ns", "shared", "off" }, 82 }; 83 const size_t hw_compat_6_1_len = G_N_ELEMENTS(hw_compat_6_1); 84 85 GlobalProperty hw_compat_6_0[] = { 86 { "gpex-pcihost", "allow-unmapped-accesses", "false" }, 87 { "i8042", "extended-state", "false"}, 88 { "nvme-ns", "eui64-default", "off"}, 89 { "e1000", "init-vet", "off" }, 90 { "e1000e", "init-vet", "off" }, 91 { "vhost-vsock-device", "seqpacket", "off" }, 92 }; 93 const size_t hw_compat_6_0_len = G_N_ELEMENTS(hw_compat_6_0); 94 95 GlobalProperty hw_compat_5_2[] = { 96 { "ICH9-LPC", "smm-compat", "on"}, 97 { "PIIX4_PM", "smm-compat", "on"}, 98 { "virtio-blk-device", "report-discard-granularity", "off" }, 99 { "virtio-net-pci-base", "vectors", "3"}, 100 }; 101 const size_t hw_compat_5_2_len = G_N_ELEMENTS(hw_compat_5_2); 102 103 GlobalProperty hw_compat_5_1[] = { 104 { "vhost-scsi", "num_queues", "1"}, 105 { "vhost-user-blk", "num-queues", "1"}, 106 { "vhost-user-scsi", "num_queues", "1"}, 107 { "virtio-blk-device", "num-queues", "1"}, 108 { "virtio-scsi-device", "num_queues", "1"}, 109 { "nvme", "use-intel-id", "on"}, 110 { "pvpanic", "events", "1"}, /* PVPANIC_PANICKED */ 111 { "pl011", "migrate-clk", "off" }, 112 { "virtio-pci", "x-ats-page-aligned", "off"}, 113 }; 114 const size_t hw_compat_5_1_len = G_N_ELEMENTS(hw_compat_5_1); 115 116 GlobalProperty hw_compat_5_0[] = { 117 { "pci-host-bridge", "x-config-reg-migration-enabled", "off" }, 118 { "virtio-balloon-device", "page-poison", "false" }, 119 { "vmport", "x-read-set-eax", "off" }, 120 { "vmport", "x-signal-unsupported-cmd", "off" }, 121 { "vmport", "x-report-vmx-type", "off" }, 122 { "vmport", "x-cmds-v2", "off" }, 123 { "virtio-device", "x-disable-legacy-check", "true" }, 124 }; 125 const size_t hw_compat_5_0_len = G_N_ELEMENTS(hw_compat_5_0); 126 127 GlobalProperty hw_compat_4_2[] = { 128 { "virtio-blk-device", "queue-size", "128"}, 129 { "virtio-scsi-device", "virtqueue_size", "128"}, 130 { "virtio-blk-device", "x-enable-wce-if-config-wce", "off" }, 131 { "virtio-blk-device", "seg-max-adjust", "off"}, 132 { "virtio-scsi-device", "seg_max_adjust", "off"}, 133 { "vhost-blk-device", "seg_max_adjust", "off"}, 134 { "usb-host", "suppress-remote-wake", "off" }, 135 { "usb-redir", "suppress-remote-wake", "off" }, 136 { "qxl", "revision", "4" }, 137 { "qxl-vga", "revision", "4" }, 138 { "fw_cfg", "acpi-mr-restore", "false" }, 139 { "virtio-device", "use-disabled-flag", "false" }, 140 }; 141 const size_t hw_compat_4_2_len = G_N_ELEMENTS(hw_compat_4_2); 142 143 GlobalProperty hw_compat_4_1[] = { 144 { "virtio-pci", "x-pcie-flr-init", "off" }, 145 }; 146 const size_t hw_compat_4_1_len = G_N_ELEMENTS(hw_compat_4_1); 147 148 GlobalProperty hw_compat_4_0[] = { 149 { "VGA", "edid", "false" }, 150 { "secondary-vga", "edid", "false" }, 151 { "bochs-display", "edid", "false" }, 152 { "virtio-vga", "edid", "false" }, 153 { "virtio-gpu-device", "edid", "false" }, 154 { "virtio-device", "use-started", "false" }, 155 { "virtio-balloon-device", "qemu-4-0-config-size", "true" }, 156 { "pl031", "migrate-tick-offset", "false" }, 157 }; 158 const size_t hw_compat_4_0_len = G_N_ELEMENTS(hw_compat_4_0); 159 160 GlobalProperty hw_compat_3_1[] = { 161 { "pcie-root-port", "x-speed", "2_5" }, 162 { "pcie-root-port", "x-width", "1" }, 163 { "memory-backend-file", "x-use-canonical-path-for-ramblock-id", "true" }, 164 { "memory-backend-memfd", "x-use-canonical-path-for-ramblock-id", "true" }, 165 { "tpm-crb", "ppi", "false" }, 166 { "tpm-tis", "ppi", "false" }, 167 { "usb-kbd", "serial", "42" }, 168 { "usb-mouse", "serial", "42" }, 169 { "usb-tablet", "serial", "42" }, 170 { "virtio-blk-device", "discard", "false" }, 171 { "virtio-blk-device", "write-zeroes", "false" }, 172 { "virtio-balloon-device", "qemu-4-0-config-size", "false" }, 173 { "pcie-root-port-base", "disable-acs", "true" }, /* Added in 4.1 */ 174 }; 175 const size_t hw_compat_3_1_len = G_N_ELEMENTS(hw_compat_3_1); 176 177 GlobalProperty hw_compat_3_0[] = {}; 178 const size_t hw_compat_3_0_len = G_N_ELEMENTS(hw_compat_3_0); 179 180 GlobalProperty hw_compat_2_12[] = { 181 { "migration", "decompress-error-check", "off" }, 182 { "hda-audio", "use-timer", "false" }, 183 { "cirrus-vga", "global-vmstate", "true" }, 184 { "VGA", "global-vmstate", "true" }, 185 { "vmware-svga", "global-vmstate", "true" }, 186 { "qxl-vga", "global-vmstate", "true" }, 187 }; 188 const size_t hw_compat_2_12_len = G_N_ELEMENTS(hw_compat_2_12); 189 190 GlobalProperty hw_compat_2_11[] = { 191 { "hpet", "hpet-offset-saved", "false" }, 192 { "virtio-blk-pci", "vectors", "2" }, 193 { "vhost-user-blk-pci", "vectors", "2" }, 194 { "e1000", "migrate_tso_props", "off" }, 195 }; 196 const size_t hw_compat_2_11_len = G_N_ELEMENTS(hw_compat_2_11); 197 198 GlobalProperty hw_compat_2_10[] = { 199 { "virtio-mouse-device", "wheel-axis", "false" }, 200 { "virtio-tablet-device", "wheel-axis", "false" }, 201 }; 202 const size_t hw_compat_2_10_len = G_N_ELEMENTS(hw_compat_2_10); 203 204 GlobalProperty hw_compat_2_9[] = { 205 { "pci-bridge", "shpc", "off" }, 206 { "intel-iommu", "pt", "off" }, 207 { "virtio-net-device", "x-mtu-bypass-backend", "off" }, 208 { "pcie-root-port", "x-migrate-msix", "false" }, 209 }; 210 const size_t hw_compat_2_9_len = G_N_ELEMENTS(hw_compat_2_9); 211 212 GlobalProperty hw_compat_2_8[] = { 213 { "fw_cfg_mem", "x-file-slots", "0x10" }, 214 { "fw_cfg_io", "x-file-slots", "0x10" }, 215 { "pflash_cfi01", "old-multiple-chip-handling", "on" }, 216 { "pci-bridge", "shpc", "on" }, 217 { TYPE_PCI_DEVICE, "x-pcie-extcap-init", "off" }, 218 { "virtio-pci", "x-pcie-deverr-init", "off" }, 219 { "virtio-pci", "x-pcie-lnkctl-init", "off" }, 220 { "virtio-pci", "x-pcie-pm-init", "off" }, 221 { "cirrus-vga", "vgamem_mb", "8" }, 222 { "isa-cirrus-vga", "vgamem_mb", "8" }, 223 }; 224 const size_t hw_compat_2_8_len = G_N_ELEMENTS(hw_compat_2_8); 225 226 GlobalProperty hw_compat_2_7[] = { 227 { "virtio-pci", "page-per-vq", "on" }, 228 { "virtio-serial-device", "emergency-write", "off" }, 229 { "ioapic", "version", "0x11" }, 230 { "intel-iommu", "x-buggy-eim", "true" }, 231 { "virtio-pci", "x-ignore-backend-features", "on" }, 232 }; 233 const size_t hw_compat_2_7_len = G_N_ELEMENTS(hw_compat_2_7); 234 235 GlobalProperty hw_compat_2_6[] = { 236 { "virtio-mmio", "format_transport_address", "off" }, 237 /* Optional because not all virtio-pci devices support legacy mode */ 238 { "virtio-pci", "disable-modern", "on", .optional = true }, 239 { "virtio-pci", "disable-legacy", "off", .optional = true }, 240 }; 241 const size_t hw_compat_2_6_len = G_N_ELEMENTS(hw_compat_2_6); 242 243 GlobalProperty hw_compat_2_5[] = { 244 { "isa-fdc", "fallback", "144" }, 245 { "pvscsi", "x-old-pci-configuration", "on" }, 246 { "pvscsi", "x-disable-pcie", "on" }, 247 { "vmxnet3", "x-old-msi-offsets", "on" }, 248 { "vmxnet3", "x-disable-pcie", "on" }, 249 }; 250 const size_t hw_compat_2_5_len = G_N_ELEMENTS(hw_compat_2_5); 251 252 GlobalProperty hw_compat_2_4[] = { 253 /* Optional because the 'scsi' property is Linux-only */ 254 { "virtio-blk-device", "scsi", "true", .optional = true }, 255 { "e1000", "extra_mac_registers", "off" }, 256 { "virtio-pci", "x-disable-pcie", "on" }, 257 { "virtio-pci", "migrate-extra", "off" }, 258 { "fw_cfg_mem", "dma_enabled", "off" }, 259 { "fw_cfg_io", "dma_enabled", "off" } 260 }; 261 const size_t hw_compat_2_4_len = G_N_ELEMENTS(hw_compat_2_4); 262 263 GlobalProperty hw_compat_2_3[] = { 264 { "virtio-blk-pci", "any_layout", "off" }, 265 { "virtio-balloon-pci", "any_layout", "off" }, 266 { "virtio-serial-pci", "any_layout", "off" }, 267 { "virtio-9p-pci", "any_layout", "off" }, 268 { "virtio-rng-pci", "any_layout", "off" }, 269 { TYPE_PCI_DEVICE, "x-pcie-lnksta-dllla", "off" }, 270 { "migration", "send-configuration", "off" }, 271 { "migration", "send-section-footer", "off" }, 272 { "migration", "store-global-state", "off" }, 273 }; 274 const size_t hw_compat_2_3_len = G_N_ELEMENTS(hw_compat_2_3); 275 276 GlobalProperty hw_compat_2_2[] = {}; 277 const size_t hw_compat_2_2_len = G_N_ELEMENTS(hw_compat_2_2); 278 279 GlobalProperty hw_compat_2_1[] = { 280 { "intel-hda", "old_msi_addr", "on" }, 281 { "VGA", "qemu-extended-regs", "off" }, 282 { "secondary-vga", "qemu-extended-regs", "off" }, 283 { "virtio-scsi-pci", "any_layout", "off" }, 284 { "usb-mouse", "usb_version", "1" }, 285 { "usb-kbd", "usb_version", "1" }, 286 { "virtio-pci", "virtio-pci-bus-master-bug-migration", "on" }, 287 }; 288 const size_t hw_compat_2_1_len = G_N_ELEMENTS(hw_compat_2_1); 289 290 MachineState *current_machine; 291 292 static char *machine_get_kernel(Object *obj, Error **errp) 293 { 294 MachineState *ms = MACHINE(obj); 295 296 return g_strdup(ms->kernel_filename); 297 } 298 299 static void machine_set_kernel(Object *obj, const char *value, Error **errp) 300 { 301 MachineState *ms = MACHINE(obj); 302 303 g_free(ms->kernel_filename); 304 ms->kernel_filename = g_strdup(value); 305 } 306 307 static char *machine_get_initrd(Object *obj, Error **errp) 308 { 309 MachineState *ms = MACHINE(obj); 310 311 return g_strdup(ms->initrd_filename); 312 } 313 314 static void machine_set_initrd(Object *obj, const char *value, Error **errp) 315 { 316 MachineState *ms = MACHINE(obj); 317 318 g_free(ms->initrd_filename); 319 ms->initrd_filename = g_strdup(value); 320 } 321 322 static char *machine_get_append(Object *obj, Error **errp) 323 { 324 MachineState *ms = MACHINE(obj); 325 326 return g_strdup(ms->kernel_cmdline); 327 } 328 329 static void machine_set_append(Object *obj, const char *value, Error **errp) 330 { 331 MachineState *ms = MACHINE(obj); 332 333 g_free(ms->kernel_cmdline); 334 ms->kernel_cmdline = g_strdup(value); 335 } 336 337 static char *machine_get_dtb(Object *obj, Error **errp) 338 { 339 MachineState *ms = MACHINE(obj); 340 341 return g_strdup(ms->dtb); 342 } 343 344 static void machine_set_dtb(Object *obj, const char *value, Error **errp) 345 { 346 MachineState *ms = MACHINE(obj); 347 348 g_free(ms->dtb); 349 ms->dtb = g_strdup(value); 350 } 351 352 static char *machine_get_dumpdtb(Object *obj, Error **errp) 353 { 354 MachineState *ms = MACHINE(obj); 355 356 return g_strdup(ms->dumpdtb); 357 } 358 359 static void machine_set_dumpdtb(Object *obj, const char *value, Error **errp) 360 { 361 MachineState *ms = MACHINE(obj); 362 363 g_free(ms->dumpdtb); 364 ms->dumpdtb = g_strdup(value); 365 } 366 367 static void machine_get_phandle_start(Object *obj, Visitor *v, 368 const char *name, void *opaque, 369 Error **errp) 370 { 371 MachineState *ms = MACHINE(obj); 372 int64_t value = ms->phandle_start; 373 374 visit_type_int(v, name, &value, errp); 375 } 376 377 static void machine_set_phandle_start(Object *obj, Visitor *v, 378 const char *name, void *opaque, 379 Error **errp) 380 { 381 MachineState *ms = MACHINE(obj); 382 int64_t value; 383 384 if (!visit_type_int(v, name, &value, errp)) { 385 return; 386 } 387 388 ms->phandle_start = value; 389 } 390 391 static char *machine_get_dt_compatible(Object *obj, Error **errp) 392 { 393 MachineState *ms = MACHINE(obj); 394 395 return g_strdup(ms->dt_compatible); 396 } 397 398 static void machine_set_dt_compatible(Object *obj, const char *value, Error **errp) 399 { 400 MachineState *ms = MACHINE(obj); 401 402 g_free(ms->dt_compatible); 403 ms->dt_compatible = g_strdup(value); 404 } 405 406 static bool machine_get_dump_guest_core(Object *obj, Error **errp) 407 { 408 MachineState *ms = MACHINE(obj); 409 410 return ms->dump_guest_core; 411 } 412 413 static void machine_set_dump_guest_core(Object *obj, bool value, Error **errp) 414 { 415 MachineState *ms = MACHINE(obj); 416 417 ms->dump_guest_core = value; 418 } 419 420 static bool machine_get_mem_merge(Object *obj, Error **errp) 421 { 422 MachineState *ms = MACHINE(obj); 423 424 return ms->mem_merge; 425 } 426 427 static void machine_set_mem_merge(Object *obj, bool value, Error **errp) 428 { 429 MachineState *ms = MACHINE(obj); 430 431 ms->mem_merge = value; 432 } 433 434 static bool machine_get_usb(Object *obj, Error **errp) 435 { 436 MachineState *ms = MACHINE(obj); 437 438 return ms->usb; 439 } 440 441 static void machine_set_usb(Object *obj, bool value, Error **errp) 442 { 443 MachineState *ms = MACHINE(obj); 444 445 ms->usb = value; 446 ms->usb_disabled = !value; 447 } 448 449 static bool machine_get_graphics(Object *obj, Error **errp) 450 { 451 MachineState *ms = MACHINE(obj); 452 453 return ms->enable_graphics; 454 } 455 456 static void machine_set_graphics(Object *obj, bool value, Error **errp) 457 { 458 MachineState *ms = MACHINE(obj); 459 460 ms->enable_graphics = value; 461 } 462 463 static char *machine_get_firmware(Object *obj, Error **errp) 464 { 465 MachineState *ms = MACHINE(obj); 466 467 return g_strdup(ms->firmware); 468 } 469 470 static void machine_set_firmware(Object *obj, const char *value, Error **errp) 471 { 472 MachineState *ms = MACHINE(obj); 473 474 g_free(ms->firmware); 475 ms->firmware = g_strdup(value); 476 } 477 478 static void machine_set_suppress_vmdesc(Object *obj, bool value, Error **errp) 479 { 480 MachineState *ms = MACHINE(obj); 481 482 ms->suppress_vmdesc = value; 483 } 484 485 static bool machine_get_suppress_vmdesc(Object *obj, Error **errp) 486 { 487 MachineState *ms = MACHINE(obj); 488 489 return ms->suppress_vmdesc; 490 } 491 492 static char *machine_get_memory_encryption(Object *obj, Error **errp) 493 { 494 MachineState *ms = MACHINE(obj); 495 496 if (ms->cgs) { 497 return g_strdup(object_get_canonical_path_component(OBJECT(ms->cgs))); 498 } 499 500 return NULL; 501 } 502 503 static void machine_set_memory_encryption(Object *obj, const char *value, 504 Error **errp) 505 { 506 Object *cgs = 507 object_resolve_path_component(object_get_objects_root(), value); 508 509 if (!cgs) { 510 error_setg(errp, "No such memory encryption object '%s'", value); 511 return; 512 } 513 514 object_property_set_link(obj, "confidential-guest-support", cgs, errp); 515 } 516 517 static void machine_check_confidential_guest_support(const Object *obj, 518 const char *name, 519 Object *new_target, 520 Error **errp) 521 { 522 /* 523 * So far the only constraint is that the target has the 524 * TYPE_CONFIDENTIAL_GUEST_SUPPORT interface, and that's checked 525 * by the QOM core 526 */ 527 } 528 529 static bool machine_get_nvdimm(Object *obj, Error **errp) 530 { 531 MachineState *ms = MACHINE(obj); 532 533 return ms->nvdimms_state->is_enabled; 534 } 535 536 static void machine_set_nvdimm(Object *obj, bool value, Error **errp) 537 { 538 MachineState *ms = MACHINE(obj); 539 540 ms->nvdimms_state->is_enabled = value; 541 } 542 543 static bool machine_get_hmat(Object *obj, Error **errp) 544 { 545 MachineState *ms = MACHINE(obj); 546 547 return ms->numa_state->hmat_enabled; 548 } 549 550 static void machine_set_hmat(Object *obj, bool value, Error **errp) 551 { 552 MachineState *ms = MACHINE(obj); 553 554 ms->numa_state->hmat_enabled = value; 555 } 556 557 static void machine_get_mem(Object *obj, Visitor *v, const char *name, 558 void *opaque, Error **errp) 559 { 560 MachineState *ms = MACHINE(obj); 561 MemorySizeConfiguration mem = { 562 .has_size = true, 563 .size = ms->ram_size, 564 .has_max_size = !!ms->ram_slots, 565 .max_size = ms->maxram_size, 566 .has_slots = !!ms->ram_slots, 567 .slots = ms->ram_slots, 568 }; 569 MemorySizeConfiguration *p_mem = &mem; 570 571 visit_type_MemorySizeConfiguration(v, name, &p_mem, &error_abort); 572 } 573 574 static void machine_set_mem(Object *obj, Visitor *v, const char *name, 575 void *opaque, Error **errp) 576 { 577 ERRP_GUARD(); 578 MachineState *ms = MACHINE(obj); 579 MachineClass *mc = MACHINE_GET_CLASS(obj); 580 MemorySizeConfiguration *mem; 581 582 if (!visit_type_MemorySizeConfiguration(v, name, &mem, errp)) { 583 return; 584 } 585 586 if (!mem->has_size) { 587 mem->has_size = true; 588 mem->size = mc->default_ram_size; 589 } 590 mem->size = QEMU_ALIGN_UP(mem->size, 8192); 591 if (mc->fixup_ram_size) { 592 mem->size = mc->fixup_ram_size(mem->size); 593 } 594 if ((ram_addr_t)mem->size != mem->size) { 595 error_setg(errp, "ram size too large"); 596 goto out_free; 597 } 598 599 if (mem->has_max_size) { 600 if (mem->max_size < mem->size) { 601 error_setg(errp, "invalid value of maxmem: " 602 "maximum memory size (0x%" PRIx64 ") must be at least " 603 "the initial memory size (0x%" PRIx64 ")", 604 mem->max_size, mem->size); 605 goto out_free; 606 } 607 if (mem->has_slots && mem->slots && mem->max_size == mem->size) { 608 error_setg(errp, "invalid value of maxmem: " 609 "memory slots were specified but maximum memory size " 610 "(0x%" PRIx64 ") is equal to the initial memory size " 611 "(0x%" PRIx64 ")", mem->max_size, mem->size); 612 goto out_free; 613 } 614 ms->maxram_size = mem->max_size; 615 } else { 616 if (mem->has_slots) { 617 error_setg(errp, "slots specified but no max-size"); 618 goto out_free; 619 } 620 ms->maxram_size = mem->size; 621 } 622 ms->ram_size = mem->size; 623 ms->ram_slots = mem->has_slots ? mem->slots : 0; 624 out_free: 625 qapi_free_MemorySizeConfiguration(mem); 626 } 627 628 static char *machine_get_nvdimm_persistence(Object *obj, Error **errp) 629 { 630 MachineState *ms = MACHINE(obj); 631 632 return g_strdup(ms->nvdimms_state->persistence_string); 633 } 634 635 static void machine_set_nvdimm_persistence(Object *obj, const char *value, 636 Error **errp) 637 { 638 MachineState *ms = MACHINE(obj); 639 NVDIMMState *nvdimms_state = ms->nvdimms_state; 640 641 if (strcmp(value, "cpu") == 0) { 642 nvdimms_state->persistence = 3; 643 } else if (strcmp(value, "mem-ctrl") == 0) { 644 nvdimms_state->persistence = 2; 645 } else { 646 error_setg(errp, "-machine nvdimm-persistence=%s: unsupported option", 647 value); 648 return; 649 } 650 651 g_free(nvdimms_state->persistence_string); 652 nvdimms_state->persistence_string = g_strdup(value); 653 } 654 655 void machine_class_allow_dynamic_sysbus_dev(MachineClass *mc, const char *type) 656 { 657 QAPI_LIST_PREPEND(mc->allowed_dynamic_sysbus_devices, g_strdup(type)); 658 } 659 660 bool device_is_dynamic_sysbus(MachineClass *mc, DeviceState *dev) 661 { 662 Object *obj = OBJECT(dev); 663 664 if (!object_dynamic_cast(obj, TYPE_SYS_BUS_DEVICE)) { 665 return false; 666 } 667 668 return device_type_is_dynamic_sysbus(mc, object_get_typename(obj)); 669 } 670 671 bool device_type_is_dynamic_sysbus(MachineClass *mc, const char *type) 672 { 673 bool allowed = false; 674 strList *wl; 675 ObjectClass *klass = object_class_by_name(type); 676 677 for (wl = mc->allowed_dynamic_sysbus_devices; 678 !allowed && wl; 679 wl = wl->next) { 680 allowed |= !!object_class_dynamic_cast(klass, wl->value); 681 } 682 683 return allowed; 684 } 685 686 static char *machine_get_audiodev(Object *obj, Error **errp) 687 { 688 MachineState *ms = MACHINE(obj); 689 690 return g_strdup(ms->audiodev); 691 } 692 693 static void machine_set_audiodev(Object *obj, const char *value, 694 Error **errp) 695 { 696 MachineState *ms = MACHINE(obj); 697 698 if (!audio_state_by_name(value, errp)) { 699 return; 700 } 701 702 g_free(ms->audiodev); 703 ms->audiodev = g_strdup(value); 704 } 705 706 HotpluggableCPUList *machine_query_hotpluggable_cpus(MachineState *machine) 707 { 708 int i; 709 HotpluggableCPUList *head = NULL; 710 MachineClass *mc = MACHINE_GET_CLASS(machine); 711 712 /* force board to initialize possible_cpus if it hasn't been done yet */ 713 mc->possible_cpu_arch_ids(machine); 714 715 for (i = 0; i < machine->possible_cpus->len; i++) { 716 Object *cpu; 717 HotpluggableCPU *cpu_item = g_new0(typeof(*cpu_item), 1); 718 719 cpu_item->type = g_strdup(machine->possible_cpus->cpus[i].type); 720 cpu_item->vcpus_count = machine->possible_cpus->cpus[i].vcpus_count; 721 cpu_item->props = g_memdup(&machine->possible_cpus->cpus[i].props, 722 sizeof(*cpu_item->props)); 723 724 cpu = machine->possible_cpus->cpus[i].cpu; 725 if (cpu) { 726 cpu_item->qom_path = object_get_canonical_path(cpu); 727 } 728 QAPI_LIST_PREPEND(head, cpu_item); 729 } 730 return head; 731 } 732 733 /** 734 * machine_set_cpu_numa_node: 735 * @machine: machine object to modify 736 * @props: specifies which cpu objects to assign to 737 * numa node specified by @props.node_id 738 * @errp: if an error occurs, a pointer to an area to store the error 739 * 740 * Associate NUMA node specified by @props.node_id with cpu slots that 741 * match socket/core/thread-ids specified by @props. It's recommended to use 742 * query-hotpluggable-cpus.props values to specify affected cpu slots, 743 * which would lead to exact 1:1 mapping of cpu slots to NUMA node. 744 * 745 * However for CLI convenience it's possible to pass in subset of properties, 746 * which would affect all cpu slots that match it. 747 * Ex for pc machine: 748 * -smp 4,cores=2,sockets=2 -numa node,nodeid=0 -numa node,nodeid=1 \ 749 * -numa cpu,node-id=0,socket_id=0 \ 750 * -numa cpu,node-id=1,socket_id=1 751 * will assign all child cores of socket 0 to node 0 and 752 * of socket 1 to node 1. 753 * 754 * On attempt of reassigning (already assigned) cpu slot to another NUMA node, 755 * return error. 756 * Empty subset is disallowed and function will return with error in this case. 757 */ 758 void machine_set_cpu_numa_node(MachineState *machine, 759 const CpuInstanceProperties *props, Error **errp) 760 { 761 MachineClass *mc = MACHINE_GET_CLASS(machine); 762 NodeInfo *numa_info = machine->numa_state->nodes; 763 bool match = false; 764 int i; 765 766 if (!mc->possible_cpu_arch_ids) { 767 error_setg(errp, "mapping of CPUs to NUMA node is not supported"); 768 return; 769 } 770 771 /* disabling node mapping is not supported, forbid it */ 772 assert(props->has_node_id); 773 774 /* force board to initialize possible_cpus if it hasn't been done yet */ 775 mc->possible_cpu_arch_ids(machine); 776 777 for (i = 0; i < machine->possible_cpus->len; i++) { 778 CPUArchId *slot = &machine->possible_cpus->cpus[i]; 779 780 /* reject unsupported by board properties */ 781 if (props->has_thread_id && !slot->props.has_thread_id) { 782 error_setg(errp, "thread-id is not supported"); 783 return; 784 } 785 786 if (props->has_core_id && !slot->props.has_core_id) { 787 error_setg(errp, "core-id is not supported"); 788 return; 789 } 790 791 if (props->has_cluster_id && !slot->props.has_cluster_id) { 792 error_setg(errp, "cluster-id is not supported"); 793 return; 794 } 795 796 if (props->has_socket_id && !slot->props.has_socket_id) { 797 error_setg(errp, "socket-id is not supported"); 798 return; 799 } 800 801 if (props->has_die_id && !slot->props.has_die_id) { 802 error_setg(errp, "die-id is not supported"); 803 return; 804 } 805 806 /* skip slots with explicit mismatch */ 807 if (props->has_thread_id && props->thread_id != slot->props.thread_id) { 808 continue; 809 } 810 811 if (props->has_core_id && props->core_id != slot->props.core_id) { 812 continue; 813 } 814 815 if (props->has_cluster_id && 816 props->cluster_id != slot->props.cluster_id) { 817 continue; 818 } 819 820 if (props->has_die_id && props->die_id != slot->props.die_id) { 821 continue; 822 } 823 824 if (props->has_socket_id && props->socket_id != slot->props.socket_id) { 825 continue; 826 } 827 828 /* reject assignment if slot is already assigned, for compatibility 829 * of legacy cpu_index mapping with SPAPR core based mapping do not 830 * error out if cpu thread and matched core have the same node-id */ 831 if (slot->props.has_node_id && 832 slot->props.node_id != props->node_id) { 833 error_setg(errp, "CPU is already assigned to node-id: %" PRId64, 834 slot->props.node_id); 835 return; 836 } 837 838 /* assign slot to node as it's matched '-numa cpu' key */ 839 match = true; 840 slot->props.node_id = props->node_id; 841 slot->props.has_node_id = props->has_node_id; 842 843 if (machine->numa_state->hmat_enabled) { 844 if ((numa_info[props->node_id].initiator < MAX_NODES) && 845 (props->node_id != numa_info[props->node_id].initiator)) { 846 error_setg(errp, "The initiator of CPU NUMA node %" PRId64 847 " should be itself (got %" PRIu16 ")", 848 props->node_id, numa_info[props->node_id].initiator); 849 return; 850 } 851 numa_info[props->node_id].has_cpu = true; 852 numa_info[props->node_id].initiator = props->node_id; 853 } 854 } 855 856 if (!match) { 857 error_setg(errp, "no match found"); 858 } 859 } 860 861 static void machine_get_smp(Object *obj, Visitor *v, const char *name, 862 void *opaque, Error **errp) 863 { 864 MachineState *ms = MACHINE(obj); 865 SMPConfiguration *config = &(SMPConfiguration){ 866 .has_cpus = true, .cpus = ms->smp.cpus, 867 .has_drawers = true, .drawers = ms->smp.drawers, 868 .has_books = true, .books = ms->smp.books, 869 .has_sockets = true, .sockets = ms->smp.sockets, 870 .has_dies = true, .dies = ms->smp.dies, 871 .has_clusters = true, .clusters = ms->smp.clusters, 872 .has_cores = true, .cores = ms->smp.cores, 873 .has_threads = true, .threads = ms->smp.threads, 874 .has_maxcpus = true, .maxcpus = ms->smp.max_cpus, 875 }; 876 877 if (!visit_type_SMPConfiguration(v, name, &config, &error_abort)) { 878 return; 879 } 880 } 881 882 static void machine_set_smp(Object *obj, Visitor *v, const char *name, 883 void *opaque, Error **errp) 884 { 885 MachineState *ms = MACHINE(obj); 886 g_autoptr(SMPConfiguration) config = NULL; 887 888 if (!visit_type_SMPConfiguration(v, name, &config, errp)) { 889 return; 890 } 891 892 machine_parse_smp_config(ms, config, errp); 893 } 894 895 static void machine_get_boot(Object *obj, Visitor *v, const char *name, 896 void *opaque, Error **errp) 897 { 898 MachineState *ms = MACHINE(obj); 899 BootConfiguration *config = &ms->boot_config; 900 visit_type_BootConfiguration(v, name, &config, &error_abort); 901 } 902 903 static void machine_free_boot_config(MachineState *ms) 904 { 905 g_free(ms->boot_config.order); 906 g_free(ms->boot_config.once); 907 g_free(ms->boot_config.splash); 908 } 909 910 static void machine_copy_boot_config(MachineState *ms, BootConfiguration *config) 911 { 912 MachineClass *machine_class = MACHINE_GET_CLASS(ms); 913 914 machine_free_boot_config(ms); 915 ms->boot_config = *config; 916 if (!config->order) { 917 ms->boot_config.order = g_strdup(machine_class->default_boot_order); 918 } 919 } 920 921 static void machine_set_boot(Object *obj, Visitor *v, const char *name, 922 void *opaque, Error **errp) 923 { 924 ERRP_GUARD(); 925 MachineState *ms = MACHINE(obj); 926 BootConfiguration *config = NULL; 927 928 if (!visit_type_BootConfiguration(v, name, &config, errp)) { 929 return; 930 } 931 if (config->order) { 932 validate_bootdevices(config->order, errp); 933 if (*errp) { 934 goto out_free; 935 } 936 } 937 if (config->once) { 938 validate_bootdevices(config->once, errp); 939 if (*errp) { 940 goto out_free; 941 } 942 } 943 944 machine_copy_boot_config(ms, config); 945 /* Strings live in ms->boot_config. */ 946 free(config); 947 return; 948 949 out_free: 950 qapi_free_BootConfiguration(config); 951 } 952 953 void machine_add_audiodev_property(MachineClass *mc) 954 { 955 ObjectClass *oc = OBJECT_CLASS(mc); 956 957 object_class_property_add_str(oc, "audiodev", 958 machine_get_audiodev, 959 machine_set_audiodev); 960 object_class_property_set_description(oc, "audiodev", 961 "Audiodev to use for default machine devices"); 962 } 963 964 static void machine_class_init(ObjectClass *oc, void *data) 965 { 966 MachineClass *mc = MACHINE_CLASS(oc); 967 968 /* Default 128 MB as guest ram size */ 969 mc->default_ram_size = 128 * MiB; 970 mc->rom_file_has_mr = true; 971 972 /* numa node memory size aligned on 8MB by default. 973 * On Linux, each node's border has to be 8MB aligned 974 */ 975 mc->numa_mem_align_shift = 23; 976 977 object_class_property_add_str(oc, "kernel", 978 machine_get_kernel, machine_set_kernel); 979 object_class_property_set_description(oc, "kernel", 980 "Linux kernel image file"); 981 982 object_class_property_add_str(oc, "initrd", 983 machine_get_initrd, machine_set_initrd); 984 object_class_property_set_description(oc, "initrd", 985 "Linux initial ramdisk file"); 986 987 object_class_property_add_str(oc, "append", 988 machine_get_append, machine_set_append); 989 object_class_property_set_description(oc, "append", 990 "Linux kernel command line"); 991 992 object_class_property_add_str(oc, "dtb", 993 machine_get_dtb, machine_set_dtb); 994 object_class_property_set_description(oc, "dtb", 995 "Linux kernel device tree file"); 996 997 object_class_property_add_str(oc, "dumpdtb", 998 machine_get_dumpdtb, machine_set_dumpdtb); 999 object_class_property_set_description(oc, "dumpdtb", 1000 "Dump current dtb to a file and quit"); 1001 1002 object_class_property_add(oc, "boot", "BootConfiguration", 1003 machine_get_boot, machine_set_boot, 1004 NULL, NULL); 1005 object_class_property_set_description(oc, "boot", 1006 "Boot configuration"); 1007 1008 object_class_property_add(oc, "smp", "SMPConfiguration", 1009 machine_get_smp, machine_set_smp, 1010 NULL, NULL); 1011 object_class_property_set_description(oc, "smp", 1012 "CPU topology"); 1013 1014 object_class_property_add(oc, "phandle-start", "int", 1015 machine_get_phandle_start, machine_set_phandle_start, 1016 NULL, NULL); 1017 object_class_property_set_description(oc, "phandle-start", 1018 "The first phandle ID we may generate dynamically"); 1019 1020 object_class_property_add_str(oc, "dt-compatible", 1021 machine_get_dt_compatible, machine_set_dt_compatible); 1022 object_class_property_set_description(oc, "dt-compatible", 1023 "Overrides the \"compatible\" property of the dt root node"); 1024 1025 object_class_property_add_bool(oc, "dump-guest-core", 1026 machine_get_dump_guest_core, machine_set_dump_guest_core); 1027 object_class_property_set_description(oc, "dump-guest-core", 1028 "Include guest memory in a core dump"); 1029 1030 object_class_property_add_bool(oc, "mem-merge", 1031 machine_get_mem_merge, machine_set_mem_merge); 1032 object_class_property_set_description(oc, "mem-merge", 1033 "Enable/disable memory merge support"); 1034 1035 object_class_property_add_bool(oc, "usb", 1036 machine_get_usb, machine_set_usb); 1037 object_class_property_set_description(oc, "usb", 1038 "Set on/off to enable/disable usb"); 1039 1040 object_class_property_add_bool(oc, "graphics", 1041 machine_get_graphics, machine_set_graphics); 1042 object_class_property_set_description(oc, "graphics", 1043 "Set on/off to enable/disable graphics emulation"); 1044 1045 object_class_property_add_str(oc, "firmware", 1046 machine_get_firmware, machine_set_firmware); 1047 object_class_property_set_description(oc, "firmware", 1048 "Firmware image"); 1049 1050 object_class_property_add_bool(oc, "suppress-vmdesc", 1051 machine_get_suppress_vmdesc, machine_set_suppress_vmdesc); 1052 object_class_property_set_description(oc, "suppress-vmdesc", 1053 "Set on to disable self-describing migration"); 1054 1055 object_class_property_add_link(oc, "confidential-guest-support", 1056 TYPE_CONFIDENTIAL_GUEST_SUPPORT, 1057 offsetof(MachineState, cgs), 1058 machine_check_confidential_guest_support, 1059 OBJ_PROP_LINK_STRONG); 1060 object_class_property_set_description(oc, "confidential-guest-support", 1061 "Set confidential guest scheme to support"); 1062 1063 /* For compatibility */ 1064 object_class_property_add_str(oc, "memory-encryption", 1065 machine_get_memory_encryption, machine_set_memory_encryption); 1066 object_class_property_set_description(oc, "memory-encryption", 1067 "Set memory encryption object to use"); 1068 1069 object_class_property_add_link(oc, "memory-backend", TYPE_MEMORY_BACKEND, 1070 offsetof(MachineState, memdev), object_property_allow_set_link, 1071 OBJ_PROP_LINK_STRONG); 1072 object_class_property_set_description(oc, "memory-backend", 1073 "Set RAM backend" 1074 "Valid value is ID of hostmem based backend"); 1075 1076 object_class_property_add(oc, "memory", "MemorySizeConfiguration", 1077 machine_get_mem, machine_set_mem, 1078 NULL, NULL); 1079 object_class_property_set_description(oc, "memory", 1080 "Memory size configuration"); 1081 } 1082 1083 static void machine_class_base_init(ObjectClass *oc, void *data) 1084 { 1085 MachineClass *mc = MACHINE_CLASS(oc); 1086 mc->max_cpus = mc->max_cpus ?: 1; 1087 mc->min_cpus = mc->min_cpus ?: 1; 1088 mc->default_cpus = mc->default_cpus ?: 1; 1089 1090 if (!object_class_is_abstract(oc)) { 1091 const char *cname = object_class_get_name(oc); 1092 assert(g_str_has_suffix(cname, TYPE_MACHINE_SUFFIX)); 1093 mc->name = g_strndup(cname, 1094 strlen(cname) - strlen(TYPE_MACHINE_SUFFIX)); 1095 mc->compat_props = g_ptr_array_new(); 1096 } 1097 } 1098 1099 static void machine_initfn(Object *obj) 1100 { 1101 MachineState *ms = MACHINE(obj); 1102 MachineClass *mc = MACHINE_GET_CLASS(obj); 1103 1104 container_get(obj, "/peripheral"); 1105 container_get(obj, "/peripheral-anon"); 1106 1107 ms->dump_guest_core = true; 1108 ms->mem_merge = true; 1109 ms->enable_graphics = true; 1110 ms->kernel_cmdline = g_strdup(""); 1111 ms->ram_size = mc->default_ram_size; 1112 ms->maxram_size = mc->default_ram_size; 1113 1114 if (mc->nvdimm_supported) { 1115 ms->nvdimms_state = g_new0(NVDIMMState, 1); 1116 object_property_add_bool(obj, "nvdimm", 1117 machine_get_nvdimm, machine_set_nvdimm); 1118 object_property_set_description(obj, "nvdimm", 1119 "Set on/off to enable/disable " 1120 "NVDIMM instantiation"); 1121 1122 object_property_add_str(obj, "nvdimm-persistence", 1123 machine_get_nvdimm_persistence, 1124 machine_set_nvdimm_persistence); 1125 object_property_set_description(obj, "nvdimm-persistence", 1126 "Set NVDIMM persistence" 1127 "Valid values are cpu, mem-ctrl"); 1128 } 1129 1130 if (mc->cpu_index_to_instance_props && mc->get_default_cpu_node_id) { 1131 ms->numa_state = g_new0(NumaState, 1); 1132 object_property_add_bool(obj, "hmat", 1133 machine_get_hmat, machine_set_hmat); 1134 object_property_set_description(obj, "hmat", 1135 "Set on/off to enable/disable " 1136 "ACPI Heterogeneous Memory Attribute " 1137 "Table (HMAT)"); 1138 } 1139 1140 /* default to mc->default_cpus */ 1141 ms->smp.cpus = mc->default_cpus; 1142 ms->smp.max_cpus = mc->default_cpus; 1143 ms->smp.drawers = 1; 1144 ms->smp.books = 1; 1145 ms->smp.sockets = 1; 1146 ms->smp.dies = 1; 1147 ms->smp.clusters = 1; 1148 ms->smp.cores = 1; 1149 ms->smp.threads = 1; 1150 1151 machine_copy_boot_config(ms, &(BootConfiguration){ 0 }); 1152 } 1153 1154 static void machine_finalize(Object *obj) 1155 { 1156 MachineState *ms = MACHINE(obj); 1157 1158 machine_free_boot_config(ms); 1159 g_free(ms->kernel_filename); 1160 g_free(ms->initrd_filename); 1161 g_free(ms->kernel_cmdline); 1162 g_free(ms->dtb); 1163 g_free(ms->dumpdtb); 1164 g_free(ms->dt_compatible); 1165 g_free(ms->firmware); 1166 g_free(ms->device_memory); 1167 g_free(ms->nvdimms_state); 1168 g_free(ms->numa_state); 1169 g_free(ms->audiodev); 1170 } 1171 1172 bool machine_usb(MachineState *machine) 1173 { 1174 return machine->usb; 1175 } 1176 1177 int machine_phandle_start(MachineState *machine) 1178 { 1179 return machine->phandle_start; 1180 } 1181 1182 bool machine_dump_guest_core(MachineState *machine) 1183 { 1184 return machine->dump_guest_core; 1185 } 1186 1187 bool machine_mem_merge(MachineState *machine) 1188 { 1189 return machine->mem_merge; 1190 } 1191 1192 static char *cpu_slot_to_string(const CPUArchId *cpu) 1193 { 1194 GString *s = g_string_new(NULL); 1195 if (cpu->props.has_socket_id) { 1196 g_string_append_printf(s, "socket-id: %"PRId64, cpu->props.socket_id); 1197 } 1198 if (cpu->props.has_die_id) { 1199 if (s->len) { 1200 g_string_append_printf(s, ", "); 1201 } 1202 g_string_append_printf(s, "die-id: %"PRId64, cpu->props.die_id); 1203 } 1204 if (cpu->props.has_cluster_id) { 1205 if (s->len) { 1206 g_string_append_printf(s, ", "); 1207 } 1208 g_string_append_printf(s, "cluster-id: %"PRId64, cpu->props.cluster_id); 1209 } 1210 if (cpu->props.has_core_id) { 1211 if (s->len) { 1212 g_string_append_printf(s, ", "); 1213 } 1214 g_string_append_printf(s, "core-id: %"PRId64, cpu->props.core_id); 1215 } 1216 if (cpu->props.has_thread_id) { 1217 if (s->len) { 1218 g_string_append_printf(s, ", "); 1219 } 1220 g_string_append_printf(s, "thread-id: %"PRId64, cpu->props.thread_id); 1221 } 1222 return g_string_free(s, false); 1223 } 1224 1225 static void numa_validate_initiator(NumaState *numa_state) 1226 { 1227 int i; 1228 NodeInfo *numa_info = numa_state->nodes; 1229 1230 for (i = 0; i < numa_state->num_nodes; i++) { 1231 if (numa_info[i].initiator == MAX_NODES) { 1232 continue; 1233 } 1234 1235 if (!numa_info[numa_info[i].initiator].present) { 1236 error_report("NUMA node %" PRIu16 " is missing, use " 1237 "'-numa node' option to declare it first", 1238 numa_info[i].initiator); 1239 exit(1); 1240 } 1241 1242 if (!numa_info[numa_info[i].initiator].has_cpu) { 1243 error_report("The initiator of NUMA node %d is invalid", i); 1244 exit(1); 1245 } 1246 } 1247 } 1248 1249 static void machine_numa_finish_cpu_init(MachineState *machine) 1250 { 1251 int i; 1252 bool default_mapping; 1253 GString *s = g_string_new(NULL); 1254 MachineClass *mc = MACHINE_GET_CLASS(machine); 1255 const CPUArchIdList *possible_cpus = mc->possible_cpu_arch_ids(machine); 1256 1257 assert(machine->numa_state->num_nodes); 1258 for (i = 0; i < possible_cpus->len; i++) { 1259 if (possible_cpus->cpus[i].props.has_node_id) { 1260 break; 1261 } 1262 } 1263 default_mapping = (i == possible_cpus->len); 1264 1265 for (i = 0; i < possible_cpus->len; i++) { 1266 const CPUArchId *cpu_slot = &possible_cpus->cpus[i]; 1267 1268 if (!cpu_slot->props.has_node_id) { 1269 /* fetch default mapping from board and enable it */ 1270 CpuInstanceProperties props = cpu_slot->props; 1271 1272 props.node_id = mc->get_default_cpu_node_id(machine, i); 1273 if (!default_mapping) { 1274 /* record slots with not set mapping, 1275 * TODO: make it hard error in future */ 1276 char *cpu_str = cpu_slot_to_string(cpu_slot); 1277 g_string_append_printf(s, "%sCPU %d [%s]", 1278 s->len ? ", " : "", i, cpu_str); 1279 g_free(cpu_str); 1280 1281 /* non mapped cpus used to fallback to node 0 */ 1282 props.node_id = 0; 1283 } 1284 1285 props.has_node_id = true; 1286 machine_set_cpu_numa_node(machine, &props, &error_fatal); 1287 } 1288 } 1289 1290 if (machine->numa_state->hmat_enabled) { 1291 numa_validate_initiator(machine->numa_state); 1292 } 1293 1294 if (s->len && !qtest_enabled()) { 1295 warn_report("CPU(s) not present in any NUMA nodes: %s", 1296 s->str); 1297 warn_report("All CPU(s) up to maxcpus should be described " 1298 "in NUMA config, ability to start up with partial NUMA " 1299 "mappings is obsoleted and will be removed in future"); 1300 } 1301 g_string_free(s, true); 1302 } 1303 1304 static void validate_cpu_cluster_to_numa_boundary(MachineState *ms) 1305 { 1306 MachineClass *mc = MACHINE_GET_CLASS(ms); 1307 NumaState *state = ms->numa_state; 1308 const CPUArchIdList *possible_cpus = mc->possible_cpu_arch_ids(ms); 1309 const CPUArchId *cpus = possible_cpus->cpus; 1310 int i, j; 1311 1312 if (state->num_nodes <= 1 || possible_cpus->len <= 1) { 1313 return; 1314 } 1315 1316 /* 1317 * The Linux scheduling domain can't be parsed when the multiple CPUs 1318 * in one cluster have been associated with different NUMA nodes. However, 1319 * it's fine to associate one NUMA node with CPUs in different clusters. 1320 */ 1321 for (i = 0; i < possible_cpus->len; i++) { 1322 for (j = i + 1; j < possible_cpus->len; j++) { 1323 if (cpus[i].props.has_socket_id && 1324 cpus[i].props.has_cluster_id && 1325 cpus[i].props.has_node_id && 1326 cpus[j].props.has_socket_id && 1327 cpus[j].props.has_cluster_id && 1328 cpus[j].props.has_node_id && 1329 cpus[i].props.socket_id == cpus[j].props.socket_id && 1330 cpus[i].props.cluster_id == cpus[j].props.cluster_id && 1331 cpus[i].props.node_id != cpus[j].props.node_id) { 1332 warn_report("CPU-%d and CPU-%d in socket-%" PRId64 "-cluster-%" PRId64 1333 " have been associated with node-%" PRId64 " and node-%" PRId64 1334 " respectively. It can cause OSes like Linux to" 1335 " misbehave", i, j, cpus[i].props.socket_id, 1336 cpus[i].props.cluster_id, cpus[i].props.node_id, 1337 cpus[j].props.node_id); 1338 } 1339 } 1340 } 1341 } 1342 1343 MemoryRegion *machine_consume_memdev(MachineState *machine, 1344 HostMemoryBackend *backend) 1345 { 1346 MemoryRegion *ret = host_memory_backend_get_memory(backend); 1347 1348 if (host_memory_backend_is_mapped(backend)) { 1349 error_report("memory backend %s can't be used multiple times.", 1350 object_get_canonical_path_component(OBJECT(backend))); 1351 exit(EXIT_FAILURE); 1352 } 1353 host_memory_backend_set_mapped(backend, true); 1354 vmstate_register_ram_global(ret); 1355 return ret; 1356 } 1357 1358 static bool create_default_memdev(MachineState *ms, const char *path, Error **errp) 1359 { 1360 Object *obj; 1361 MachineClass *mc = MACHINE_GET_CLASS(ms); 1362 bool r = false; 1363 1364 obj = object_new(path ? TYPE_MEMORY_BACKEND_FILE : TYPE_MEMORY_BACKEND_RAM); 1365 if (path) { 1366 if (!object_property_set_str(obj, "mem-path", path, errp)) { 1367 goto out; 1368 } 1369 } 1370 if (!object_property_set_int(obj, "size", ms->ram_size, errp)) { 1371 goto out; 1372 } 1373 object_property_add_child(object_get_objects_root(), mc->default_ram_id, 1374 obj); 1375 /* Ensure backend's memory region name is equal to mc->default_ram_id */ 1376 if (!object_property_set_bool(obj, "x-use-canonical-path-for-ramblock-id", 1377 false, errp)) { 1378 goto out; 1379 } 1380 if (!user_creatable_complete(USER_CREATABLE(obj), errp)) { 1381 goto out; 1382 } 1383 r = object_property_set_link(OBJECT(ms), "memory-backend", obj, errp); 1384 1385 out: 1386 object_unref(obj); 1387 return r; 1388 } 1389 1390 1391 void machine_run_board_init(MachineState *machine, const char *mem_path, Error **errp) 1392 { 1393 ERRP_GUARD(); 1394 MachineClass *machine_class = MACHINE_GET_CLASS(machine); 1395 ObjectClass *oc = object_class_by_name(machine->cpu_type); 1396 CPUClass *cc; 1397 1398 /* This checkpoint is required by replay to separate prior clock 1399 reading from the other reads, because timer polling functions query 1400 clock values from the log. */ 1401 replay_checkpoint(CHECKPOINT_INIT); 1402 1403 if (!xen_enabled()) { 1404 /* On 32-bit hosts, QEMU is limited by virtual address space */ 1405 if (machine->ram_size > (2047 << 20) && HOST_LONG_BITS == 32) { 1406 error_setg(errp, "at most 2047 MB RAM can be simulated"); 1407 return; 1408 } 1409 } 1410 1411 if (machine->memdev) { 1412 ram_addr_t backend_size = object_property_get_uint(OBJECT(machine->memdev), 1413 "size", &error_abort); 1414 if (backend_size != machine->ram_size) { 1415 error_setg(errp, "Machine memory size does not match the size of the memory backend"); 1416 return; 1417 } 1418 } else if (machine_class->default_ram_id && machine->ram_size && 1419 numa_uses_legacy_mem()) { 1420 if (object_property_find(object_get_objects_root(), 1421 machine_class->default_ram_id)) { 1422 error_setg(errp, "object's id '%s' is reserved for the default" 1423 " RAM backend, it can't be used for any other purposes", 1424 machine_class->default_ram_id); 1425 error_append_hint(errp, 1426 "Change the object's 'id' to something else or disable" 1427 " automatic creation of the default RAM backend by setting" 1428 " 'memory-backend=%s' with '-machine'.\n", 1429 machine_class->default_ram_id); 1430 return; 1431 } 1432 if (!create_default_memdev(current_machine, mem_path, errp)) { 1433 return; 1434 } 1435 } 1436 1437 if (machine->numa_state) { 1438 numa_complete_configuration(machine); 1439 if (machine->numa_state->num_nodes) { 1440 machine_numa_finish_cpu_init(machine); 1441 if (machine_class->cpu_cluster_has_numa_boundary) { 1442 validate_cpu_cluster_to_numa_boundary(machine); 1443 } 1444 } 1445 } 1446 1447 if (!machine->ram && machine->memdev) { 1448 machine->ram = machine_consume_memdev(machine, machine->memdev); 1449 } 1450 1451 /* If the machine supports the valid_cpu_types check and the user 1452 * specified a CPU with -cpu check here that the user CPU is supported. 1453 */ 1454 if (machine_class->valid_cpu_types && machine->cpu_type) { 1455 int i; 1456 1457 for (i = 0; machine_class->valid_cpu_types[i]; i++) { 1458 if (object_class_dynamic_cast(oc, 1459 machine_class->valid_cpu_types[i])) { 1460 /* The user specified CPU is in the valid field, we are 1461 * good to go. 1462 */ 1463 break; 1464 } 1465 } 1466 1467 if (!machine_class->valid_cpu_types[i]) { 1468 /* The user specified CPU is not valid */ 1469 error_report("Invalid CPU type: %s", machine->cpu_type); 1470 error_printf("The valid types are: %s", 1471 machine_class->valid_cpu_types[0]); 1472 for (i = 1; machine_class->valid_cpu_types[i]; i++) { 1473 error_printf(", %s", machine_class->valid_cpu_types[i]); 1474 } 1475 error_printf("\n"); 1476 1477 exit(1); 1478 } 1479 } 1480 1481 /* Check if CPU type is deprecated and warn if so */ 1482 cc = CPU_CLASS(oc); 1483 if (cc && cc->deprecation_note) { 1484 warn_report("CPU model %s is deprecated -- %s", machine->cpu_type, 1485 cc->deprecation_note); 1486 } 1487 1488 if (machine->cgs) { 1489 /* 1490 * With confidential guests, the host can't see the real 1491 * contents of RAM, so there's no point in it trying to merge 1492 * areas. 1493 */ 1494 machine_set_mem_merge(OBJECT(machine), false, &error_abort); 1495 1496 /* 1497 * Virtio devices can't count on directly accessing guest 1498 * memory, so they need iommu_platform=on to use normal DMA 1499 * mechanisms. That requires also disabling legacy virtio 1500 * support for those virtio pci devices which allow it. 1501 */ 1502 object_register_sugar_prop(TYPE_VIRTIO_PCI, "disable-legacy", 1503 "on", true); 1504 object_register_sugar_prop(TYPE_VIRTIO_DEVICE, "iommu_platform", 1505 "on", false); 1506 } 1507 1508 accel_init_interfaces(ACCEL_GET_CLASS(machine->accelerator)); 1509 machine_class->init(machine); 1510 phase_advance(PHASE_MACHINE_INITIALIZED); 1511 } 1512 1513 static NotifierList machine_init_done_notifiers = 1514 NOTIFIER_LIST_INITIALIZER(machine_init_done_notifiers); 1515 1516 void qemu_add_machine_init_done_notifier(Notifier *notify) 1517 { 1518 notifier_list_add(&machine_init_done_notifiers, notify); 1519 if (phase_check(PHASE_MACHINE_READY)) { 1520 notify->notify(notify, NULL); 1521 } 1522 } 1523 1524 void qemu_remove_machine_init_done_notifier(Notifier *notify) 1525 { 1526 notifier_remove(notify); 1527 } 1528 1529 void qdev_machine_creation_done(void) 1530 { 1531 cpu_synchronize_all_post_init(); 1532 1533 if (current_machine->boot_config.once) { 1534 qemu_boot_set(current_machine->boot_config.once, &error_fatal); 1535 qemu_register_reset(restore_boot_order, g_strdup(current_machine->boot_config.order)); 1536 } 1537 1538 /* 1539 * ok, initial machine setup is done, starting from now we can 1540 * only create hotpluggable devices 1541 */ 1542 phase_advance(PHASE_MACHINE_READY); 1543 qdev_assert_realized_properly(); 1544 1545 /* TODO: once all bus devices are qdevified, this should be done 1546 * when bus is created by qdev.c */ 1547 /* 1548 * TODO: If we had a main 'reset container' that the whole system 1549 * lived in, we could reset that using the multi-phase reset 1550 * APIs. For the moment, we just reset the sysbus, which will cause 1551 * all devices hanging off it (and all their child buses, recursively) 1552 * to be reset. Note that this will *not* reset any Device objects 1553 * which are not attached to some part of the qbus tree! 1554 */ 1555 qemu_register_reset(resettable_cold_reset_fn, sysbus_get_default()); 1556 1557 notifier_list_notify(&machine_init_done_notifiers, NULL); 1558 1559 if (rom_check_and_register_reset() != 0) { 1560 exit(1); 1561 } 1562 1563 replay_start(); 1564 1565 /* This checkpoint is required by replay to separate prior clock 1566 reading from the other reads, because timer polling functions query 1567 clock values from the log. */ 1568 replay_checkpoint(CHECKPOINT_RESET); 1569 qemu_system_reset(SHUTDOWN_CAUSE_NONE); 1570 register_global_state(); 1571 } 1572 1573 static const TypeInfo machine_info = { 1574 .name = TYPE_MACHINE, 1575 .parent = TYPE_OBJECT, 1576 .abstract = true, 1577 .class_size = sizeof(MachineClass), 1578 .class_init = machine_class_init, 1579 .class_base_init = machine_class_base_init, 1580 .instance_size = sizeof(MachineState), 1581 .instance_init = machine_initfn, 1582 .instance_finalize = machine_finalize, 1583 }; 1584 1585 static void machine_register_types(void) 1586 { 1587 type_register_static(&machine_info); 1588 } 1589 1590 type_init(machine_register_types) 1591