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/units.h" 15 #include "qemu/accel.h" 16 #include "system/replay.h" 17 #include "hw/boards.h" 18 #include "hw/loader.h" 19 #include "qemu/error-report.h" 20 #include "qapi/error.h" 21 #include "qapi/qapi-visit-machine.h" 22 #include "qemu/madvise.h" 23 #include "qom/object_interfaces.h" 24 #include "system/cpus.h" 25 #include "system/system.h" 26 #include "system/reset.h" 27 #include "system/runstate.h" 28 #include "system/xen.h" 29 #include "system/qtest.h" 30 #include "hw/pci/pci_bridge.h" 31 #include "hw/mem/nvdimm.h" 32 #include "migration/global_state.h" 33 #include "system/confidential-guest-support.h" 34 #include "hw/virtio/virtio-pci.h" 35 #include "hw/virtio/virtio-net.h" 36 #include "hw/virtio/virtio-iommu.h" 37 #include "audio/audio.h" 38 39 GlobalProperty hw_compat_9_2[] = { 40 {"arm-cpu", "backcompat-pauth-default-use-qarma5", "true"}, 41 { "virtio-balloon-pci", "vectors", "0" }, 42 { "virtio-balloon-pci-transitional", "vectors", "0" }, 43 { "virtio-balloon-pci-non-transitional", "vectors", "0" }, 44 { "virtio-mem-pci", "vectors", "0" }, 45 }; 46 const size_t hw_compat_9_2_len = G_N_ELEMENTS(hw_compat_9_2); 47 48 GlobalProperty hw_compat_9_1[] = { 49 { TYPE_PCI_DEVICE, "x-pcie-ext-tag", "false" }, 50 }; 51 const size_t hw_compat_9_1_len = G_N_ELEMENTS(hw_compat_9_1); 52 53 GlobalProperty hw_compat_9_0[] = { 54 {"arm-cpu", "backcompat-cntfrq", "true" }, 55 { "scsi-hd", "migrate-emulated-scsi-request", "false" }, 56 { "scsi-cd", "migrate-emulated-scsi-request", "false" }, 57 {"vfio-pci", "skip-vsc-check", "false" }, 58 { "virtio-pci", "x-pcie-pm-no-soft-reset", "off" }, 59 {"sd-card", "spec_version", "2" }, 60 }; 61 const size_t hw_compat_9_0_len = G_N_ELEMENTS(hw_compat_9_0); 62 63 GlobalProperty hw_compat_8_2[] = { 64 { "migration", "zero-page-detection", "legacy"}, 65 { TYPE_VIRTIO_IOMMU_PCI, "granule", "4k" }, 66 { TYPE_VIRTIO_IOMMU_PCI, "aw-bits", "64" }, 67 { "virtio-gpu-device", "x-scanout-vmstate-version", "1" }, 68 }; 69 const size_t hw_compat_8_2_len = G_N_ELEMENTS(hw_compat_8_2); 70 71 GlobalProperty hw_compat_8_1[] = { 72 { TYPE_PCI_BRIDGE, "x-pci-express-writeable-slt-bug", "true" }, 73 { "ramfb", "x-migrate", "off" }, 74 { "vfio-pci-nohotplug", "x-ramfb-migrate", "off" }, 75 { "igb", "x-pcie-flr-init", "off" }, 76 { TYPE_VIRTIO_NET, "host_uso", "off"}, 77 { TYPE_VIRTIO_NET, "guest_uso4", "off"}, 78 { TYPE_VIRTIO_NET, "guest_uso6", "off"}, 79 }; 80 const size_t hw_compat_8_1_len = G_N_ELEMENTS(hw_compat_8_1); 81 82 GlobalProperty hw_compat_8_0[] = { 83 { "migration", "multifd-flush-after-each-section", "on"}, 84 { TYPE_PCI_DEVICE, "x-pcie-ari-nextfn-1", "on" }, 85 }; 86 const size_t hw_compat_8_0_len = G_N_ELEMENTS(hw_compat_8_0); 87 88 GlobalProperty hw_compat_7_2[] = { 89 { "e1000e", "migrate-timadj", "off" }, 90 { "virtio-mem", "x-early-migration", "false" }, 91 { "migration", "x-preempt-pre-7-2", "true" }, 92 { TYPE_PCI_DEVICE, "x-pcie-err-unc-mask", "off" }, 93 }; 94 const size_t hw_compat_7_2_len = G_N_ELEMENTS(hw_compat_7_2); 95 96 GlobalProperty hw_compat_7_1[] = { 97 { "virtio-device", "queue_reset", "false" }, 98 { "virtio-rng-pci", "vectors", "0" }, 99 { "virtio-rng-pci-transitional", "vectors", "0" }, 100 { "virtio-rng-pci-non-transitional", "vectors", "0" }, 101 }; 102 const size_t hw_compat_7_1_len = G_N_ELEMENTS(hw_compat_7_1); 103 104 GlobalProperty hw_compat_7_0[] = { 105 { "arm-gicv3-common", "force-8-bit-prio", "on" }, 106 { "nvme-ns", "eui64-default", "on"}, 107 }; 108 const size_t hw_compat_7_0_len = G_N_ELEMENTS(hw_compat_7_0); 109 110 GlobalProperty hw_compat_6_2[] = { 111 { "PIIX4_PM", "x-not-migrate-acpi-index", "on"}, 112 }; 113 const size_t hw_compat_6_2_len = G_N_ELEMENTS(hw_compat_6_2); 114 115 GlobalProperty hw_compat_6_1[] = { 116 { "vhost-user-vsock-device", "seqpacket", "off" }, 117 { "nvme-ns", "shared", "off" }, 118 }; 119 const size_t hw_compat_6_1_len = G_N_ELEMENTS(hw_compat_6_1); 120 121 GlobalProperty hw_compat_6_0[] = { 122 { "gpex-pcihost", "allow-unmapped-accesses", "false" }, 123 { "i8042", "extended-state", "false"}, 124 { "nvme-ns", "eui64-default", "off"}, 125 { "e1000", "init-vet", "off" }, 126 { "e1000e", "init-vet", "off" }, 127 { "vhost-vsock-device", "seqpacket", "off" }, 128 }; 129 const size_t hw_compat_6_0_len = G_N_ELEMENTS(hw_compat_6_0); 130 131 GlobalProperty hw_compat_5_2[] = { 132 { "ICH9-LPC", "smm-compat", "on"}, 133 { "PIIX4_PM", "smm-compat", "on"}, 134 { "virtio-blk-device", "report-discard-granularity", "off" }, 135 { "virtio-net-pci-base", "vectors", "3"}, 136 { "nvme", "msix-exclusive-bar", "on"}, 137 }; 138 const size_t hw_compat_5_2_len = G_N_ELEMENTS(hw_compat_5_2); 139 140 GlobalProperty hw_compat_5_1[] = { 141 { "vhost-scsi", "num_queues", "1"}, 142 { "vhost-user-blk", "num-queues", "1"}, 143 { "vhost-user-scsi", "num_queues", "1"}, 144 { "virtio-blk-device", "num-queues", "1"}, 145 { "virtio-scsi-device", "num_queues", "1"}, 146 { "nvme", "use-intel-id", "on"}, 147 { "pvpanic", "events", "1"}, /* PVPANIC_PANICKED */ 148 { "pl011", "migrate-clk", "off" }, 149 { "virtio-pci", "x-ats-page-aligned", "off"}, 150 }; 151 const size_t hw_compat_5_1_len = G_N_ELEMENTS(hw_compat_5_1); 152 153 GlobalProperty hw_compat_5_0[] = { 154 { "pci-host-bridge", "x-config-reg-migration-enabled", "off" }, 155 { "virtio-balloon-device", "page-poison", "false" }, 156 { "vmport", "x-read-set-eax", "off" }, 157 { "vmport", "x-signal-unsupported-cmd", "off" }, 158 { "vmport", "x-report-vmx-type", "off" }, 159 { "vmport", "x-cmds-v2", "off" }, 160 { "virtio-device", "x-disable-legacy-check", "true" }, 161 }; 162 const size_t hw_compat_5_0_len = G_N_ELEMENTS(hw_compat_5_0); 163 164 GlobalProperty hw_compat_4_2[] = { 165 { "virtio-blk-device", "queue-size", "128"}, 166 { "virtio-scsi-device", "virtqueue_size", "128"}, 167 { "virtio-blk-device", "x-enable-wce-if-config-wce", "off" }, 168 { "virtio-blk-device", "seg-max-adjust", "off"}, 169 { "virtio-scsi-device", "seg_max_adjust", "off"}, 170 { "vhost-blk-device", "seg_max_adjust", "off"}, 171 { "usb-host", "suppress-remote-wake", "off" }, 172 { "usb-redir", "suppress-remote-wake", "off" }, 173 { "qxl", "revision", "4" }, 174 { "qxl-vga", "revision", "4" }, 175 { "fw_cfg", "acpi-mr-restore", "false" }, 176 { "virtio-device", "use-disabled-flag", "false" }, 177 }; 178 const size_t hw_compat_4_2_len = G_N_ELEMENTS(hw_compat_4_2); 179 180 GlobalProperty hw_compat_4_1[] = { 181 { "virtio-pci", "x-pcie-flr-init", "off" }, 182 }; 183 const size_t hw_compat_4_1_len = G_N_ELEMENTS(hw_compat_4_1); 184 185 GlobalProperty hw_compat_4_0[] = { 186 { "VGA", "edid", "false" }, 187 { "secondary-vga", "edid", "false" }, 188 { "bochs-display", "edid", "false" }, 189 { "virtio-vga", "edid", "false" }, 190 { "virtio-gpu-device", "edid", "false" }, 191 { "virtio-device", "use-started", "false" }, 192 { "virtio-balloon-device", "qemu-4-0-config-size", "true" }, 193 { "pl031", "migrate-tick-offset", "false" }, 194 }; 195 const size_t hw_compat_4_0_len = G_N_ELEMENTS(hw_compat_4_0); 196 197 GlobalProperty hw_compat_3_1[] = { 198 { "pcie-root-port", "x-speed", "2_5" }, 199 { "pcie-root-port", "x-width", "1" }, 200 { "memory-backend-file", "x-use-canonical-path-for-ramblock-id", "true" }, 201 { "memory-backend-memfd", "x-use-canonical-path-for-ramblock-id", "true" }, 202 { "tpm-crb", "ppi", "false" }, 203 { "tpm-tis", "ppi", "false" }, 204 { "usb-kbd", "serial", "42" }, 205 { "usb-mouse", "serial", "42" }, 206 { "usb-tablet", "serial", "42" }, 207 { "virtio-blk-device", "discard", "false" }, 208 { "virtio-blk-device", "write-zeroes", "false" }, 209 { "virtio-balloon-device", "qemu-4-0-config-size", "false" }, 210 { "pcie-root-port-base", "disable-acs", "true" }, /* Added in 4.1 */ 211 }; 212 const size_t hw_compat_3_1_len = G_N_ELEMENTS(hw_compat_3_1); 213 214 GlobalProperty hw_compat_3_0[] = {}; 215 const size_t hw_compat_3_0_len = G_N_ELEMENTS(hw_compat_3_0); 216 217 GlobalProperty hw_compat_2_12[] = { 218 { "hda-audio", "use-timer", "false" }, 219 { "cirrus-vga", "global-vmstate", "true" }, 220 { "VGA", "global-vmstate", "true" }, 221 { "vmware-svga", "global-vmstate", "true" }, 222 { "qxl-vga", "global-vmstate", "true" }, 223 }; 224 const size_t hw_compat_2_12_len = G_N_ELEMENTS(hw_compat_2_12); 225 226 GlobalProperty hw_compat_2_11[] = { 227 { "hpet", "hpet-offset-saved", "false" }, 228 { "virtio-blk-pci", "vectors", "2" }, 229 { "vhost-user-blk-pci", "vectors", "2" }, 230 { "e1000", "migrate_tso_props", "off" }, 231 }; 232 const size_t hw_compat_2_11_len = G_N_ELEMENTS(hw_compat_2_11); 233 234 GlobalProperty hw_compat_2_10[] = { 235 { "virtio-mouse-device", "wheel-axis", "false" }, 236 { "virtio-tablet-device", "wheel-axis", "false" }, 237 }; 238 const size_t hw_compat_2_10_len = G_N_ELEMENTS(hw_compat_2_10); 239 240 GlobalProperty hw_compat_2_9[] = { 241 { "pci-bridge", "shpc", "off" }, 242 { "intel-iommu", "pt", "off" }, 243 { "virtio-net-device", "x-mtu-bypass-backend", "off" }, 244 { "pcie-root-port", "x-migrate-msix", "false" }, 245 }; 246 const size_t hw_compat_2_9_len = G_N_ELEMENTS(hw_compat_2_9); 247 248 GlobalProperty hw_compat_2_8[] = { 249 { "fw_cfg_mem", "x-file-slots", "0x10" }, 250 { "fw_cfg_io", "x-file-slots", "0x10" }, 251 { "pflash_cfi01", "old-multiple-chip-handling", "on" }, 252 { "pci-bridge", "shpc", "on" }, 253 { TYPE_PCI_DEVICE, "x-pcie-extcap-init", "off" }, 254 { "virtio-pci", "x-pcie-deverr-init", "off" }, 255 { "virtio-pci", "x-pcie-lnkctl-init", "off" }, 256 { "virtio-pci", "x-pcie-pm-init", "off" }, 257 { "cirrus-vga", "vgamem_mb", "8" }, 258 { "isa-cirrus-vga", "vgamem_mb", "8" }, 259 }; 260 const size_t hw_compat_2_8_len = G_N_ELEMENTS(hw_compat_2_8); 261 262 GlobalProperty hw_compat_2_7[] = { 263 { "virtio-pci", "page-per-vq", "on" }, 264 { "virtio-serial-device", "emergency-write", "off" }, 265 { "ioapic", "version", "0x11" }, 266 { "intel-iommu", "x-buggy-eim", "true" }, 267 { "virtio-pci", "x-ignore-backend-features", "on" }, 268 }; 269 const size_t hw_compat_2_7_len = G_N_ELEMENTS(hw_compat_2_7); 270 271 GlobalProperty hw_compat_2_6[] = { 272 { "virtio-mmio", "format_transport_address", "off" }, 273 /* Optional because not all virtio-pci devices support legacy mode */ 274 { "virtio-pci", "disable-modern", "on", .optional = true }, 275 { "virtio-pci", "disable-legacy", "off", .optional = true }, 276 }; 277 const size_t hw_compat_2_6_len = G_N_ELEMENTS(hw_compat_2_6); 278 279 GlobalProperty hw_compat_2_5[] = { 280 { "isa-fdc", "fallback", "144" }, 281 { "pvscsi", "x-old-pci-configuration", "on" }, 282 { "pvscsi", "x-disable-pcie", "on" }, 283 { "vmxnet3", "x-old-msi-offsets", "on" }, 284 { "vmxnet3", "x-disable-pcie", "on" }, 285 }; 286 const size_t hw_compat_2_5_len = G_N_ELEMENTS(hw_compat_2_5); 287 288 GlobalProperty hw_compat_2_4[] = { 289 { "e1000", "extra_mac_registers", "off" }, 290 { "virtio-pci", "x-disable-pcie", "on" }, 291 { "virtio-pci", "migrate-extra", "off" }, 292 { "fw_cfg_mem", "dma_enabled", "off" }, 293 { "fw_cfg_io", "dma_enabled", "off" } 294 }; 295 const size_t hw_compat_2_4_len = G_N_ELEMENTS(hw_compat_2_4); 296 297 MachineState *current_machine; 298 299 static char *machine_get_kernel(Object *obj, Error **errp) 300 { 301 MachineState *ms = MACHINE(obj); 302 303 return g_strdup(ms->kernel_filename); 304 } 305 306 static void machine_set_kernel(Object *obj, const char *value, Error **errp) 307 { 308 MachineState *ms = MACHINE(obj); 309 310 g_free(ms->kernel_filename); 311 ms->kernel_filename = g_strdup(value); 312 } 313 314 static char *machine_get_shim(Object *obj, Error **errp) 315 { 316 MachineState *ms = MACHINE(obj); 317 318 return g_strdup(ms->shim_filename); 319 } 320 321 static void machine_set_shim(Object *obj, const char *value, Error **errp) 322 { 323 MachineState *ms = MACHINE(obj); 324 325 g_free(ms->shim_filename); 326 ms->shim_filename = g_strdup(value); 327 } 328 329 static char *machine_get_initrd(Object *obj, Error **errp) 330 { 331 MachineState *ms = MACHINE(obj); 332 333 return g_strdup(ms->initrd_filename); 334 } 335 336 static void machine_set_initrd(Object *obj, const char *value, Error **errp) 337 { 338 MachineState *ms = MACHINE(obj); 339 340 g_free(ms->initrd_filename); 341 ms->initrd_filename = g_strdup(value); 342 } 343 344 static char *machine_get_append(Object *obj, Error **errp) 345 { 346 MachineState *ms = MACHINE(obj); 347 348 return g_strdup(ms->kernel_cmdline); 349 } 350 351 static void machine_set_append(Object *obj, const char *value, Error **errp) 352 { 353 MachineState *ms = MACHINE(obj); 354 355 g_free(ms->kernel_cmdline); 356 ms->kernel_cmdline = g_strdup(value); 357 } 358 359 static char *machine_get_dtb(Object *obj, Error **errp) 360 { 361 MachineState *ms = MACHINE(obj); 362 363 return g_strdup(ms->dtb); 364 } 365 366 static void machine_set_dtb(Object *obj, const char *value, Error **errp) 367 { 368 MachineState *ms = MACHINE(obj); 369 370 g_free(ms->dtb); 371 ms->dtb = g_strdup(value); 372 } 373 374 static char *machine_get_dumpdtb(Object *obj, Error **errp) 375 { 376 MachineState *ms = MACHINE(obj); 377 378 return g_strdup(ms->dumpdtb); 379 } 380 381 static void machine_set_dumpdtb(Object *obj, const char *value, Error **errp) 382 { 383 MachineState *ms = MACHINE(obj); 384 385 g_free(ms->dumpdtb); 386 ms->dumpdtb = g_strdup(value); 387 } 388 389 static void machine_get_phandle_start(Object *obj, Visitor *v, 390 const char *name, void *opaque, 391 Error **errp) 392 { 393 MachineState *ms = MACHINE(obj); 394 int64_t value = ms->phandle_start; 395 396 visit_type_int(v, name, &value, errp); 397 } 398 399 static void machine_set_phandle_start(Object *obj, Visitor *v, 400 const char *name, void *opaque, 401 Error **errp) 402 { 403 MachineState *ms = MACHINE(obj); 404 int64_t value; 405 406 if (!visit_type_int(v, name, &value, errp)) { 407 return; 408 } 409 410 ms->phandle_start = value; 411 } 412 413 static char *machine_get_dt_compatible(Object *obj, Error **errp) 414 { 415 MachineState *ms = MACHINE(obj); 416 417 return g_strdup(ms->dt_compatible); 418 } 419 420 static void machine_set_dt_compatible(Object *obj, const char *value, Error **errp) 421 { 422 MachineState *ms = MACHINE(obj); 423 424 g_free(ms->dt_compatible); 425 ms->dt_compatible = g_strdup(value); 426 } 427 428 static bool machine_get_dump_guest_core(Object *obj, Error **errp) 429 { 430 MachineState *ms = MACHINE(obj); 431 432 return ms->dump_guest_core; 433 } 434 435 static void machine_set_dump_guest_core(Object *obj, bool value, Error **errp) 436 { 437 MachineState *ms = MACHINE(obj); 438 439 if (!value && QEMU_MADV_DONTDUMP == QEMU_MADV_INVALID) { 440 error_setg(errp, "Dumping guest memory cannot be disabled on this host"); 441 return; 442 } 443 ms->dump_guest_core = value; 444 } 445 446 static bool machine_get_mem_merge(Object *obj, Error **errp) 447 { 448 MachineState *ms = MACHINE(obj); 449 450 return ms->mem_merge; 451 } 452 453 static void machine_set_mem_merge(Object *obj, bool value, Error **errp) 454 { 455 MachineState *ms = MACHINE(obj); 456 457 if (value && QEMU_MADV_MERGEABLE == QEMU_MADV_INVALID) { 458 error_setg(errp, "Memory merging is not supported on this host"); 459 return; 460 } 461 ms->mem_merge = value; 462 } 463 464 static bool machine_get_usb(Object *obj, Error **errp) 465 { 466 MachineState *ms = MACHINE(obj); 467 468 return ms->usb; 469 } 470 471 static void machine_set_usb(Object *obj, bool value, Error **errp) 472 { 473 MachineState *ms = MACHINE(obj); 474 475 ms->usb = value; 476 ms->usb_disabled = !value; 477 } 478 479 static bool machine_get_graphics(Object *obj, Error **errp) 480 { 481 MachineState *ms = MACHINE(obj); 482 483 return ms->enable_graphics; 484 } 485 486 static void machine_set_graphics(Object *obj, bool value, Error **errp) 487 { 488 MachineState *ms = MACHINE(obj); 489 490 ms->enable_graphics = value; 491 } 492 493 static char *machine_get_firmware(Object *obj, Error **errp) 494 { 495 MachineState *ms = MACHINE(obj); 496 497 return g_strdup(ms->firmware); 498 } 499 500 static void machine_set_firmware(Object *obj, const char *value, Error **errp) 501 { 502 MachineState *ms = MACHINE(obj); 503 504 g_free(ms->firmware); 505 ms->firmware = g_strdup(value); 506 } 507 508 static void machine_set_suppress_vmdesc(Object *obj, bool value, Error **errp) 509 { 510 MachineState *ms = MACHINE(obj); 511 512 ms->suppress_vmdesc = value; 513 } 514 515 static bool machine_get_suppress_vmdesc(Object *obj, Error **errp) 516 { 517 MachineState *ms = MACHINE(obj); 518 519 return ms->suppress_vmdesc; 520 } 521 522 static char *machine_get_memory_encryption(Object *obj, Error **errp) 523 { 524 MachineState *ms = MACHINE(obj); 525 526 if (ms->cgs) { 527 return g_strdup(object_get_canonical_path_component(OBJECT(ms->cgs))); 528 } 529 530 return NULL; 531 } 532 533 static void machine_set_memory_encryption(Object *obj, const char *value, 534 Error **errp) 535 { 536 Object *cgs = 537 object_resolve_path_component(object_get_objects_root(), value); 538 539 if (!cgs) { 540 error_setg(errp, "No such memory encryption object '%s'", value); 541 return; 542 } 543 544 object_property_set_link(obj, "confidential-guest-support", cgs, errp); 545 } 546 547 static void machine_check_confidential_guest_support(const Object *obj, 548 const char *name, 549 Object *new_target, 550 Error **errp) 551 { 552 /* 553 * So far the only constraint is that the target has the 554 * TYPE_CONFIDENTIAL_GUEST_SUPPORT interface, and that's checked 555 * by the QOM core 556 */ 557 } 558 559 static bool machine_get_nvdimm(Object *obj, Error **errp) 560 { 561 MachineState *ms = MACHINE(obj); 562 563 return ms->nvdimms_state->is_enabled; 564 } 565 566 static void machine_set_nvdimm(Object *obj, bool value, Error **errp) 567 { 568 MachineState *ms = MACHINE(obj); 569 570 ms->nvdimms_state->is_enabled = value; 571 } 572 573 static bool machine_get_hmat(Object *obj, Error **errp) 574 { 575 MachineState *ms = MACHINE(obj); 576 577 return ms->numa_state->hmat_enabled; 578 } 579 580 static void machine_set_hmat(Object *obj, bool value, Error **errp) 581 { 582 MachineState *ms = MACHINE(obj); 583 584 ms->numa_state->hmat_enabled = value; 585 } 586 587 static void machine_get_mem(Object *obj, Visitor *v, const char *name, 588 void *opaque, Error **errp) 589 { 590 MachineState *ms = MACHINE(obj); 591 MemorySizeConfiguration mem = { 592 .has_size = true, 593 .size = ms->ram_size, 594 .has_max_size = !!ms->ram_slots, 595 .max_size = ms->maxram_size, 596 .has_slots = !!ms->ram_slots, 597 .slots = ms->ram_slots, 598 }; 599 MemorySizeConfiguration *p_mem = &mem; 600 601 visit_type_MemorySizeConfiguration(v, name, &p_mem, &error_abort); 602 } 603 604 static void machine_set_mem(Object *obj, Visitor *v, const char *name, 605 void *opaque, Error **errp) 606 { 607 ERRP_GUARD(); 608 MachineState *ms = MACHINE(obj); 609 MachineClass *mc = MACHINE_GET_CLASS(obj); 610 MemorySizeConfiguration *mem; 611 612 if (!visit_type_MemorySizeConfiguration(v, name, &mem, errp)) { 613 return; 614 } 615 616 if (!mem->has_size) { 617 mem->has_size = true; 618 mem->size = mc->default_ram_size; 619 } 620 mem->size = QEMU_ALIGN_UP(mem->size, 8192); 621 if (mc->fixup_ram_size) { 622 mem->size = mc->fixup_ram_size(mem->size); 623 } 624 if ((ram_addr_t)mem->size != mem->size) { 625 error_setg(errp, "ram size %llu exceeds permitted maximum %llu", 626 (unsigned long long)mem->size, 627 (unsigned long long)RAM_ADDR_MAX); 628 goto out_free; 629 } 630 631 if (mem->has_max_size) { 632 if ((ram_addr_t)mem->max_size != mem->max_size) { 633 error_setg(errp, "ram size %llu exceeds permitted maximum %llu", 634 (unsigned long long)mem->max_size, 635 (unsigned long long)RAM_ADDR_MAX); 636 goto out_free; 637 } 638 if (mem->max_size < mem->size) { 639 error_setg(errp, "invalid value of maxmem: " 640 "maximum memory size (0x%" PRIx64 ") must be at least " 641 "the initial memory size (0x%" PRIx64 ")", 642 mem->max_size, mem->size); 643 goto out_free; 644 } 645 if (mem->has_slots && mem->slots && mem->max_size == mem->size) { 646 error_setg(errp, "invalid value of maxmem: " 647 "memory slots were specified but maximum memory size " 648 "(0x%" PRIx64 ") is equal to the initial memory size " 649 "(0x%" PRIx64 ")", mem->max_size, mem->size); 650 goto out_free; 651 } 652 ms->maxram_size = mem->max_size; 653 } else { 654 if (mem->has_slots) { 655 error_setg(errp, "slots specified but no max-size"); 656 goto out_free; 657 } 658 ms->maxram_size = mem->size; 659 } 660 ms->ram_size = mem->size; 661 ms->ram_slots = mem->has_slots ? mem->slots : 0; 662 out_free: 663 qapi_free_MemorySizeConfiguration(mem); 664 } 665 666 static char *machine_get_nvdimm_persistence(Object *obj, Error **errp) 667 { 668 MachineState *ms = MACHINE(obj); 669 670 return g_strdup(ms->nvdimms_state->persistence_string); 671 } 672 673 static void machine_set_nvdimm_persistence(Object *obj, const char *value, 674 Error **errp) 675 { 676 MachineState *ms = MACHINE(obj); 677 NVDIMMState *nvdimms_state = ms->nvdimms_state; 678 679 if (strcmp(value, "cpu") == 0) { 680 nvdimms_state->persistence = 3; 681 } else if (strcmp(value, "mem-ctrl") == 0) { 682 nvdimms_state->persistence = 2; 683 } else { 684 error_setg(errp, "-machine nvdimm-persistence=%s: unsupported option", 685 value); 686 return; 687 } 688 689 g_free(nvdimms_state->persistence_string); 690 nvdimms_state->persistence_string = g_strdup(value); 691 } 692 693 void machine_class_allow_dynamic_sysbus_dev(MachineClass *mc, const char *type) 694 { 695 QAPI_LIST_PREPEND(mc->allowed_dynamic_sysbus_devices, g_strdup(type)); 696 } 697 698 bool device_is_dynamic_sysbus(MachineClass *mc, DeviceState *dev) 699 { 700 Object *obj = OBJECT(dev); 701 702 if (!object_dynamic_cast(obj, TYPE_SYS_BUS_DEVICE)) { 703 return false; 704 } 705 706 return device_type_is_dynamic_sysbus(mc, object_get_typename(obj)); 707 } 708 709 bool device_type_is_dynamic_sysbus(MachineClass *mc, const char *type) 710 { 711 bool allowed = false; 712 strList *wl; 713 ObjectClass *klass = object_class_by_name(type); 714 715 for (wl = mc->allowed_dynamic_sysbus_devices; 716 !allowed && wl; 717 wl = wl->next) { 718 allowed |= !!object_class_dynamic_cast(klass, wl->value); 719 } 720 721 return allowed; 722 } 723 724 static char *machine_get_audiodev(Object *obj, Error **errp) 725 { 726 MachineState *ms = MACHINE(obj); 727 728 return g_strdup(ms->audiodev); 729 } 730 731 static void machine_set_audiodev(Object *obj, const char *value, 732 Error **errp) 733 { 734 MachineState *ms = MACHINE(obj); 735 736 if (!audio_state_by_name(value, errp)) { 737 return; 738 } 739 740 g_free(ms->audiodev); 741 ms->audiodev = g_strdup(value); 742 } 743 744 HotpluggableCPUList *machine_query_hotpluggable_cpus(MachineState *machine) 745 { 746 int i; 747 HotpluggableCPUList *head = NULL; 748 MachineClass *mc = MACHINE_GET_CLASS(machine); 749 750 /* force board to initialize possible_cpus if it hasn't been done yet */ 751 mc->possible_cpu_arch_ids(machine); 752 753 for (i = 0; i < machine->possible_cpus->len; i++) { 754 CPUState *cpu; 755 HotpluggableCPU *cpu_item = g_new0(typeof(*cpu_item), 1); 756 757 cpu_item->type = g_strdup(machine->possible_cpus->cpus[i].type); 758 cpu_item->vcpus_count = machine->possible_cpus->cpus[i].vcpus_count; 759 cpu_item->props = g_memdup(&machine->possible_cpus->cpus[i].props, 760 sizeof(*cpu_item->props)); 761 762 cpu = machine->possible_cpus->cpus[i].cpu; 763 if (cpu) { 764 cpu_item->qom_path = object_get_canonical_path(OBJECT(cpu)); 765 } 766 QAPI_LIST_PREPEND(head, cpu_item); 767 } 768 return head; 769 } 770 771 /** 772 * machine_set_cpu_numa_node: 773 * @machine: machine object to modify 774 * @props: specifies which cpu objects to assign to 775 * numa node specified by @props.node_id 776 * @errp: if an error occurs, a pointer to an area to store the error 777 * 778 * Associate NUMA node specified by @props.node_id with cpu slots that 779 * match socket/core/thread-ids specified by @props. It's recommended to use 780 * query-hotpluggable-cpus.props values to specify affected cpu slots, 781 * which would lead to exact 1:1 mapping of cpu slots to NUMA node. 782 * 783 * However for CLI convenience it's possible to pass in subset of properties, 784 * which would affect all cpu slots that match it. 785 * Ex for pc machine: 786 * -smp 4,cores=2,sockets=2 -numa node,nodeid=0 -numa node,nodeid=1 \ 787 * -numa cpu,node-id=0,socket_id=0 \ 788 * -numa cpu,node-id=1,socket_id=1 789 * will assign all child cores of socket 0 to node 0 and 790 * of socket 1 to node 1. 791 * 792 * On attempt of reassigning (already assigned) cpu slot to another NUMA node, 793 * return error. 794 * Empty subset is disallowed and function will return with error in this case. 795 */ 796 void machine_set_cpu_numa_node(MachineState *machine, 797 const CpuInstanceProperties *props, Error **errp) 798 { 799 MachineClass *mc = MACHINE_GET_CLASS(machine); 800 NodeInfo *numa_info = machine->numa_state->nodes; 801 bool match = false; 802 int i; 803 804 if (!mc->possible_cpu_arch_ids) { 805 error_setg(errp, "mapping of CPUs to NUMA node is not supported"); 806 return; 807 } 808 809 /* disabling node mapping is not supported, forbid it */ 810 assert(props->has_node_id); 811 812 /* force board to initialize possible_cpus if it hasn't been done yet */ 813 mc->possible_cpu_arch_ids(machine); 814 815 for (i = 0; i < machine->possible_cpus->len; i++) { 816 CPUArchId *slot = &machine->possible_cpus->cpus[i]; 817 818 /* reject unsupported by board properties */ 819 if (props->has_thread_id && !slot->props.has_thread_id) { 820 error_setg(errp, "thread-id is not supported"); 821 return; 822 } 823 824 if (props->has_core_id && !slot->props.has_core_id) { 825 error_setg(errp, "core-id is not supported"); 826 return; 827 } 828 829 if (props->has_module_id && !slot->props.has_module_id) { 830 error_setg(errp, "module-id is not supported"); 831 return; 832 } 833 834 if (props->has_cluster_id && !slot->props.has_cluster_id) { 835 error_setg(errp, "cluster-id is not supported"); 836 return; 837 } 838 839 if (props->has_socket_id && !slot->props.has_socket_id) { 840 error_setg(errp, "socket-id is not supported"); 841 return; 842 } 843 844 if (props->has_die_id && !slot->props.has_die_id) { 845 error_setg(errp, "die-id is not supported"); 846 return; 847 } 848 849 /* skip slots with explicit mismatch */ 850 if (props->has_thread_id && props->thread_id != slot->props.thread_id) { 851 continue; 852 } 853 854 if (props->has_core_id && props->core_id != slot->props.core_id) { 855 continue; 856 } 857 858 if (props->has_module_id && 859 props->module_id != slot->props.module_id) { 860 continue; 861 } 862 863 if (props->has_cluster_id && 864 props->cluster_id != slot->props.cluster_id) { 865 continue; 866 } 867 868 if (props->has_die_id && props->die_id != slot->props.die_id) { 869 continue; 870 } 871 872 if (props->has_socket_id && props->socket_id != slot->props.socket_id) { 873 continue; 874 } 875 876 /* reject assignment if slot is already assigned, for compatibility 877 * of legacy cpu_index mapping with SPAPR core based mapping do not 878 * error out if cpu thread and matched core have the same node-id */ 879 if (slot->props.has_node_id && 880 slot->props.node_id != props->node_id) { 881 error_setg(errp, "CPU is already assigned to node-id: %" PRId64, 882 slot->props.node_id); 883 return; 884 } 885 886 /* assign slot to node as it's matched '-numa cpu' key */ 887 match = true; 888 slot->props.node_id = props->node_id; 889 slot->props.has_node_id = props->has_node_id; 890 891 if (machine->numa_state->hmat_enabled) { 892 if ((numa_info[props->node_id].initiator < MAX_NODES) && 893 (props->node_id != numa_info[props->node_id].initiator)) { 894 error_setg(errp, "The initiator of CPU NUMA node %" PRId64 895 " should be itself (got %" PRIu16 ")", 896 props->node_id, numa_info[props->node_id].initiator); 897 return; 898 } 899 numa_info[props->node_id].has_cpu = true; 900 numa_info[props->node_id].initiator = props->node_id; 901 } 902 } 903 904 if (!match) { 905 error_setg(errp, "no match found"); 906 } 907 } 908 909 static void machine_get_smp(Object *obj, Visitor *v, const char *name, 910 void *opaque, Error **errp) 911 { 912 MachineState *ms = MACHINE(obj); 913 SMPConfiguration *config = &(SMPConfiguration){ 914 .has_cpus = true, .cpus = ms->smp.cpus, 915 .has_drawers = true, .drawers = ms->smp.drawers, 916 .has_books = true, .books = ms->smp.books, 917 .has_sockets = true, .sockets = ms->smp.sockets, 918 .has_dies = true, .dies = ms->smp.dies, 919 .has_clusters = true, .clusters = ms->smp.clusters, 920 .has_modules = true, .modules = ms->smp.modules, 921 .has_cores = true, .cores = ms->smp.cores, 922 .has_threads = true, .threads = ms->smp.threads, 923 .has_maxcpus = true, .maxcpus = ms->smp.max_cpus, 924 }; 925 926 if (!visit_type_SMPConfiguration(v, name, &config, &error_abort)) { 927 return; 928 } 929 } 930 931 static void machine_set_smp(Object *obj, Visitor *v, const char *name, 932 void *opaque, Error **errp) 933 { 934 MachineState *ms = MACHINE(obj); 935 g_autoptr(SMPConfiguration) config = NULL; 936 937 if (!visit_type_SMPConfiguration(v, name, &config, errp)) { 938 return; 939 } 940 941 machine_parse_smp_config(ms, config, errp); 942 } 943 944 static void machine_get_smp_cache(Object *obj, Visitor *v, const char *name, 945 void *opaque, Error **errp) 946 { 947 MachineState *ms = MACHINE(obj); 948 SmpCache *cache = &ms->smp_cache; 949 SmpCachePropertiesList *head = NULL; 950 SmpCachePropertiesList **tail = &head; 951 952 for (int i = 0; i < CACHE_LEVEL_AND_TYPE__MAX; i++) { 953 SmpCacheProperties *node = g_new(SmpCacheProperties, 1); 954 955 node->cache = cache->props[i].cache; 956 node->topology = cache->props[i].topology; 957 QAPI_LIST_APPEND(tail, node); 958 } 959 960 visit_type_SmpCachePropertiesList(v, name, &head, errp); 961 qapi_free_SmpCachePropertiesList(head); 962 } 963 964 static void machine_set_smp_cache(Object *obj, Visitor *v, const char *name, 965 void *opaque, Error **errp) 966 { 967 MachineState *ms = MACHINE(obj); 968 SmpCachePropertiesList *caches; 969 970 if (!visit_type_SmpCachePropertiesList(v, name, &caches, errp)) { 971 return; 972 } 973 974 machine_parse_smp_cache(ms, caches, errp); 975 qapi_free_SmpCachePropertiesList(caches); 976 } 977 978 static void machine_get_boot(Object *obj, Visitor *v, const char *name, 979 void *opaque, Error **errp) 980 { 981 MachineState *ms = MACHINE(obj); 982 BootConfiguration *config = &ms->boot_config; 983 visit_type_BootConfiguration(v, name, &config, &error_abort); 984 } 985 986 static void machine_free_boot_config(MachineState *ms) 987 { 988 g_free(ms->boot_config.order); 989 g_free(ms->boot_config.once); 990 g_free(ms->boot_config.splash); 991 } 992 993 static void machine_copy_boot_config(MachineState *ms, BootConfiguration *config) 994 { 995 MachineClass *machine_class = MACHINE_GET_CLASS(ms); 996 997 machine_free_boot_config(ms); 998 ms->boot_config = *config; 999 if (!config->order) { 1000 ms->boot_config.order = g_strdup(machine_class->default_boot_order); 1001 } 1002 } 1003 1004 static void machine_set_boot(Object *obj, Visitor *v, const char *name, 1005 void *opaque, Error **errp) 1006 { 1007 ERRP_GUARD(); 1008 MachineState *ms = MACHINE(obj); 1009 BootConfiguration *config = NULL; 1010 1011 if (!visit_type_BootConfiguration(v, name, &config, errp)) { 1012 return; 1013 } 1014 if (config->order) { 1015 validate_bootdevices(config->order, errp); 1016 if (*errp) { 1017 goto out_free; 1018 } 1019 } 1020 if (config->once) { 1021 validate_bootdevices(config->once, errp); 1022 if (*errp) { 1023 goto out_free; 1024 } 1025 } 1026 1027 machine_copy_boot_config(ms, config); 1028 /* Strings live in ms->boot_config. */ 1029 free(config); 1030 return; 1031 1032 out_free: 1033 qapi_free_BootConfiguration(config); 1034 } 1035 1036 void machine_add_audiodev_property(MachineClass *mc) 1037 { 1038 ObjectClass *oc = OBJECT_CLASS(mc); 1039 1040 object_class_property_add_str(oc, "audiodev", 1041 machine_get_audiodev, 1042 machine_set_audiodev); 1043 object_class_property_set_description(oc, "audiodev", 1044 "Audiodev to use for default machine devices"); 1045 } 1046 1047 static bool create_default_memdev(MachineState *ms, const char *path, 1048 Error **errp) 1049 { 1050 Object *obj; 1051 MachineClass *mc = MACHINE_GET_CLASS(ms); 1052 bool r = false; 1053 1054 obj = object_new(path ? TYPE_MEMORY_BACKEND_FILE : TYPE_MEMORY_BACKEND_RAM); 1055 if (path) { 1056 if (!object_property_set_str(obj, "mem-path", path, errp)) { 1057 goto out; 1058 } 1059 } 1060 if (!object_property_set_int(obj, "size", ms->ram_size, errp)) { 1061 goto out; 1062 } 1063 object_property_add_child(object_get_objects_root(), mc->default_ram_id, 1064 obj); 1065 /* Ensure backend's memory region name is equal to mc->default_ram_id */ 1066 if (!object_property_set_bool(obj, "x-use-canonical-path-for-ramblock-id", 1067 false, errp)) { 1068 goto out; 1069 } 1070 if (!user_creatable_complete(USER_CREATABLE(obj), errp)) { 1071 goto out; 1072 } 1073 r = object_property_set_link(OBJECT(ms), "memory-backend", obj, errp); 1074 1075 out: 1076 object_unref(obj); 1077 return r; 1078 } 1079 1080 static void machine_class_init(ObjectClass *oc, void *data) 1081 { 1082 MachineClass *mc = MACHINE_CLASS(oc); 1083 1084 /* Default 128 MB as guest ram size */ 1085 mc->default_ram_size = 128 * MiB; 1086 mc->rom_file_has_mr = true; 1087 /* 1088 * SMBIOS 3.1.0 7.18.5 Memory Device — Extended Size 1089 * use max possible value that could be encoded into 1090 * 'Extended Size' field (2047Tb). 1091 */ 1092 mc->smbios_memory_device_size = 2047 * TiB; 1093 1094 /* numa node memory size aligned on 8MB by default. 1095 * On Linux, each node's border has to be 8MB aligned 1096 */ 1097 mc->numa_mem_align_shift = 23; 1098 1099 mc->create_default_memdev = create_default_memdev; 1100 1101 object_class_property_add_str(oc, "kernel", 1102 machine_get_kernel, machine_set_kernel); 1103 object_class_property_set_description(oc, "kernel", 1104 "Linux kernel image file"); 1105 1106 object_class_property_add_str(oc, "shim", 1107 machine_get_shim, machine_set_shim); 1108 object_class_property_set_description(oc, "shim", 1109 "shim.efi file"); 1110 1111 object_class_property_add_str(oc, "initrd", 1112 machine_get_initrd, machine_set_initrd); 1113 object_class_property_set_description(oc, "initrd", 1114 "Linux initial ramdisk file"); 1115 1116 object_class_property_add_str(oc, "append", 1117 machine_get_append, machine_set_append); 1118 object_class_property_set_description(oc, "append", 1119 "Linux kernel command line"); 1120 1121 object_class_property_add_str(oc, "dtb", 1122 machine_get_dtb, machine_set_dtb); 1123 object_class_property_set_description(oc, "dtb", 1124 "Linux kernel device tree file"); 1125 1126 object_class_property_add_str(oc, "dumpdtb", 1127 machine_get_dumpdtb, machine_set_dumpdtb); 1128 object_class_property_set_description(oc, "dumpdtb", 1129 "Dump current dtb to a file and quit"); 1130 1131 object_class_property_add(oc, "boot", "BootConfiguration", 1132 machine_get_boot, machine_set_boot, 1133 NULL, NULL); 1134 object_class_property_set_description(oc, "boot", 1135 "Boot configuration"); 1136 1137 object_class_property_add(oc, "smp", "SMPConfiguration", 1138 machine_get_smp, machine_set_smp, 1139 NULL, NULL); 1140 object_class_property_set_description(oc, "smp", 1141 "CPU topology"); 1142 1143 object_class_property_add(oc, "smp-cache", "SmpCachePropertiesWrapper", 1144 machine_get_smp_cache, machine_set_smp_cache, NULL, NULL); 1145 object_class_property_set_description(oc, "smp-cache", 1146 "Cache properties list for SMP machine"); 1147 1148 object_class_property_add(oc, "phandle-start", "int", 1149 machine_get_phandle_start, machine_set_phandle_start, 1150 NULL, NULL); 1151 object_class_property_set_description(oc, "phandle-start", 1152 "The first phandle ID we may generate dynamically"); 1153 1154 object_class_property_add_str(oc, "dt-compatible", 1155 machine_get_dt_compatible, machine_set_dt_compatible); 1156 object_class_property_set_description(oc, "dt-compatible", 1157 "Overrides the \"compatible\" property of the dt root node"); 1158 1159 object_class_property_add_bool(oc, "dump-guest-core", 1160 machine_get_dump_guest_core, machine_set_dump_guest_core); 1161 object_class_property_set_description(oc, "dump-guest-core", 1162 "Include guest memory in a core dump"); 1163 1164 object_class_property_add_bool(oc, "mem-merge", 1165 machine_get_mem_merge, machine_set_mem_merge); 1166 object_class_property_set_description(oc, "mem-merge", 1167 "Enable/disable memory merge support"); 1168 1169 object_class_property_add_bool(oc, "usb", 1170 machine_get_usb, machine_set_usb); 1171 object_class_property_set_description(oc, "usb", 1172 "Set on/off to enable/disable usb"); 1173 1174 object_class_property_add_bool(oc, "graphics", 1175 machine_get_graphics, machine_set_graphics); 1176 object_class_property_set_description(oc, "graphics", 1177 "Set on/off to enable/disable graphics emulation"); 1178 1179 object_class_property_add_str(oc, "firmware", 1180 machine_get_firmware, machine_set_firmware); 1181 object_class_property_set_description(oc, "firmware", 1182 "Firmware image"); 1183 1184 object_class_property_add_bool(oc, "suppress-vmdesc", 1185 machine_get_suppress_vmdesc, machine_set_suppress_vmdesc); 1186 object_class_property_set_description(oc, "suppress-vmdesc", 1187 "Set on to disable self-describing migration"); 1188 1189 object_class_property_add_link(oc, "confidential-guest-support", 1190 TYPE_CONFIDENTIAL_GUEST_SUPPORT, 1191 offsetof(MachineState, cgs), 1192 machine_check_confidential_guest_support, 1193 OBJ_PROP_LINK_STRONG); 1194 object_class_property_set_description(oc, "confidential-guest-support", 1195 "Set confidential guest scheme to support"); 1196 1197 /* For compatibility */ 1198 object_class_property_add_str(oc, "memory-encryption", 1199 machine_get_memory_encryption, machine_set_memory_encryption); 1200 object_class_property_set_description(oc, "memory-encryption", 1201 "Set memory encryption object to use"); 1202 1203 object_class_property_add_link(oc, "memory-backend", TYPE_MEMORY_BACKEND, 1204 offsetof(MachineState, memdev), object_property_allow_set_link, 1205 OBJ_PROP_LINK_STRONG); 1206 object_class_property_set_description(oc, "memory-backend", 1207 "Set RAM backend" 1208 "Valid value is ID of hostmem based backend"); 1209 1210 object_class_property_add(oc, "memory", "MemorySizeConfiguration", 1211 machine_get_mem, machine_set_mem, 1212 NULL, NULL); 1213 object_class_property_set_description(oc, "memory", 1214 "Memory size configuration"); 1215 } 1216 1217 static void machine_class_base_init(ObjectClass *oc, void *data) 1218 { 1219 MachineClass *mc = MACHINE_CLASS(oc); 1220 mc->max_cpus = mc->max_cpus ?: 1; 1221 mc->min_cpus = mc->min_cpus ?: 1; 1222 mc->default_cpus = mc->default_cpus ?: 1; 1223 1224 if (!object_class_is_abstract(oc)) { 1225 const char *cname = object_class_get_name(oc); 1226 assert(g_str_has_suffix(cname, TYPE_MACHINE_SUFFIX)); 1227 mc->name = g_strndup(cname, 1228 strlen(cname) - strlen(TYPE_MACHINE_SUFFIX)); 1229 mc->compat_props = g_ptr_array_new(); 1230 } 1231 } 1232 1233 static void machine_initfn(Object *obj) 1234 { 1235 MachineState *ms = MACHINE(obj); 1236 MachineClass *mc = MACHINE_GET_CLASS(obj); 1237 1238 ms->dump_guest_core = true; 1239 ms->mem_merge = (QEMU_MADV_MERGEABLE != QEMU_MADV_INVALID); 1240 ms->enable_graphics = true; 1241 ms->kernel_cmdline = g_strdup(""); 1242 ms->ram_size = mc->default_ram_size; 1243 ms->maxram_size = mc->default_ram_size; 1244 1245 if (mc->nvdimm_supported) { 1246 ms->nvdimms_state = g_new0(NVDIMMState, 1); 1247 object_property_add_bool(obj, "nvdimm", 1248 machine_get_nvdimm, machine_set_nvdimm); 1249 object_property_set_description(obj, "nvdimm", 1250 "Set on/off to enable/disable " 1251 "NVDIMM instantiation"); 1252 1253 object_property_add_str(obj, "nvdimm-persistence", 1254 machine_get_nvdimm_persistence, 1255 machine_set_nvdimm_persistence); 1256 object_property_set_description(obj, "nvdimm-persistence", 1257 "Set NVDIMM persistence" 1258 "Valid values are cpu, mem-ctrl"); 1259 } 1260 1261 if (mc->cpu_index_to_instance_props && mc->get_default_cpu_node_id) { 1262 ms->numa_state = g_new0(NumaState, 1); 1263 object_property_add_bool(obj, "hmat", 1264 machine_get_hmat, machine_set_hmat); 1265 object_property_set_description(obj, "hmat", 1266 "Set on/off to enable/disable " 1267 "ACPI Heterogeneous Memory Attribute " 1268 "Table (HMAT)"); 1269 } 1270 1271 /* default to mc->default_cpus */ 1272 ms->smp.cpus = mc->default_cpus; 1273 ms->smp.max_cpus = mc->default_cpus; 1274 ms->smp.drawers = 1; 1275 ms->smp.books = 1; 1276 ms->smp.sockets = 1; 1277 ms->smp.dies = 1; 1278 ms->smp.clusters = 1; 1279 ms->smp.modules = 1; 1280 ms->smp.cores = 1; 1281 ms->smp.threads = 1; 1282 1283 for (int i = 0; i < CACHE_LEVEL_AND_TYPE__MAX; i++) { 1284 ms->smp_cache.props[i].cache = (CacheLevelAndType)i; 1285 ms->smp_cache.props[i].topology = CPU_TOPOLOGY_LEVEL_DEFAULT; 1286 } 1287 1288 machine_copy_boot_config(ms, &(BootConfiguration){ 0 }); 1289 } 1290 1291 static void machine_finalize(Object *obj) 1292 { 1293 MachineState *ms = MACHINE(obj); 1294 1295 machine_free_boot_config(ms); 1296 g_free(ms->kernel_filename); 1297 g_free(ms->initrd_filename); 1298 g_free(ms->kernel_cmdline); 1299 g_free(ms->dtb); 1300 g_free(ms->dumpdtb); 1301 g_free(ms->dt_compatible); 1302 g_free(ms->firmware); 1303 g_free(ms->device_memory); 1304 g_free(ms->nvdimms_state); 1305 g_free(ms->numa_state); 1306 g_free(ms->audiodev); 1307 } 1308 1309 bool machine_usb(MachineState *machine) 1310 { 1311 return machine->usb; 1312 } 1313 1314 int machine_phandle_start(MachineState *machine) 1315 { 1316 return machine->phandle_start; 1317 } 1318 1319 bool machine_dump_guest_core(MachineState *machine) 1320 { 1321 return machine->dump_guest_core; 1322 } 1323 1324 bool machine_mem_merge(MachineState *machine) 1325 { 1326 return machine->mem_merge; 1327 } 1328 1329 bool machine_require_guest_memfd(MachineState *machine) 1330 { 1331 return machine->cgs && machine->cgs->require_guest_memfd; 1332 } 1333 1334 static char *cpu_slot_to_string(const CPUArchId *cpu) 1335 { 1336 GString *s = g_string_new(NULL); 1337 if (cpu->props.has_socket_id) { 1338 g_string_append_printf(s, "socket-id: %"PRId64, cpu->props.socket_id); 1339 } 1340 if (cpu->props.has_die_id) { 1341 if (s->len) { 1342 g_string_append_printf(s, ", "); 1343 } 1344 g_string_append_printf(s, "die-id: %"PRId64, cpu->props.die_id); 1345 } 1346 if (cpu->props.has_cluster_id) { 1347 if (s->len) { 1348 g_string_append_printf(s, ", "); 1349 } 1350 g_string_append_printf(s, "cluster-id: %"PRId64, cpu->props.cluster_id); 1351 } 1352 if (cpu->props.has_module_id) { 1353 if (s->len) { 1354 g_string_append_printf(s, ", "); 1355 } 1356 g_string_append_printf(s, "module-id: %"PRId64, cpu->props.module_id); 1357 } 1358 if (cpu->props.has_core_id) { 1359 if (s->len) { 1360 g_string_append_printf(s, ", "); 1361 } 1362 g_string_append_printf(s, "core-id: %"PRId64, cpu->props.core_id); 1363 } 1364 if (cpu->props.has_thread_id) { 1365 if (s->len) { 1366 g_string_append_printf(s, ", "); 1367 } 1368 g_string_append_printf(s, "thread-id: %"PRId64, cpu->props.thread_id); 1369 } 1370 return g_string_free(s, false); 1371 } 1372 1373 static void numa_validate_initiator(NumaState *numa_state) 1374 { 1375 int i; 1376 NodeInfo *numa_info = numa_state->nodes; 1377 1378 for (i = 0; i < numa_state->num_nodes; i++) { 1379 if (numa_info[i].initiator == MAX_NODES) { 1380 continue; 1381 } 1382 1383 if (!numa_info[numa_info[i].initiator].present) { 1384 error_report("NUMA node %" PRIu16 " is missing, use " 1385 "'-numa node' option to declare it first", 1386 numa_info[i].initiator); 1387 exit(1); 1388 } 1389 1390 if (!numa_info[numa_info[i].initiator].has_cpu) { 1391 error_report("The initiator of NUMA node %d is invalid", i); 1392 exit(1); 1393 } 1394 } 1395 } 1396 1397 static void machine_numa_finish_cpu_init(MachineState *machine) 1398 { 1399 int i; 1400 bool default_mapping; 1401 GString *s = g_string_new(NULL); 1402 MachineClass *mc = MACHINE_GET_CLASS(machine); 1403 const CPUArchIdList *possible_cpus = mc->possible_cpu_arch_ids(machine); 1404 1405 assert(machine->numa_state->num_nodes); 1406 for (i = 0; i < possible_cpus->len; i++) { 1407 if (possible_cpus->cpus[i].props.has_node_id) { 1408 break; 1409 } 1410 } 1411 default_mapping = (i == possible_cpus->len); 1412 1413 for (i = 0; i < possible_cpus->len; i++) { 1414 const CPUArchId *cpu_slot = &possible_cpus->cpus[i]; 1415 1416 if (!cpu_slot->props.has_node_id) { 1417 /* fetch default mapping from board and enable it */ 1418 CpuInstanceProperties props = cpu_slot->props; 1419 1420 props.node_id = mc->get_default_cpu_node_id(machine, i); 1421 if (!default_mapping) { 1422 /* record slots with not set mapping, 1423 * TODO: make it hard error in future */ 1424 char *cpu_str = cpu_slot_to_string(cpu_slot); 1425 g_string_append_printf(s, "%sCPU %d [%s]", 1426 s->len ? ", " : "", i, cpu_str); 1427 g_free(cpu_str); 1428 1429 /* non mapped cpus used to fallback to node 0 */ 1430 props.node_id = 0; 1431 } 1432 1433 props.has_node_id = true; 1434 machine_set_cpu_numa_node(machine, &props, &error_fatal); 1435 } 1436 } 1437 1438 if (machine->numa_state->hmat_enabled) { 1439 numa_validate_initiator(machine->numa_state); 1440 } 1441 1442 if (s->len && !qtest_enabled()) { 1443 warn_report("CPU(s) not present in any NUMA nodes: %s", 1444 s->str); 1445 warn_report("All CPU(s) up to maxcpus should be described " 1446 "in NUMA config, ability to start up with partial NUMA " 1447 "mappings is obsoleted and will be removed in future"); 1448 } 1449 g_string_free(s, true); 1450 } 1451 1452 static void validate_cpu_cluster_to_numa_boundary(MachineState *ms) 1453 { 1454 MachineClass *mc = MACHINE_GET_CLASS(ms); 1455 NumaState *state = ms->numa_state; 1456 const CPUArchIdList *possible_cpus = mc->possible_cpu_arch_ids(ms); 1457 const CPUArchId *cpus = possible_cpus->cpus; 1458 int i, j; 1459 1460 if (qtest_enabled() || state->num_nodes <= 1 || possible_cpus->len <= 1) { 1461 return; 1462 } 1463 1464 /* 1465 * The Linux scheduling domain can't be parsed when the multiple CPUs 1466 * in one cluster have been associated with different NUMA nodes. However, 1467 * it's fine to associate one NUMA node with CPUs in different clusters. 1468 */ 1469 for (i = 0; i < possible_cpus->len; i++) { 1470 for (j = i + 1; j < possible_cpus->len; j++) { 1471 if (cpus[i].props.has_socket_id && 1472 cpus[i].props.has_cluster_id && 1473 cpus[i].props.has_node_id && 1474 cpus[j].props.has_socket_id && 1475 cpus[j].props.has_cluster_id && 1476 cpus[j].props.has_node_id && 1477 cpus[i].props.socket_id == cpus[j].props.socket_id && 1478 cpus[i].props.cluster_id == cpus[j].props.cluster_id && 1479 cpus[i].props.node_id != cpus[j].props.node_id) { 1480 warn_report("CPU-%d and CPU-%d in socket-%" PRId64 "-cluster-%" PRId64 1481 " have been associated with node-%" PRId64 " and node-%" PRId64 1482 " respectively. It can cause OSes like Linux to" 1483 " misbehave", i, j, cpus[i].props.socket_id, 1484 cpus[i].props.cluster_id, cpus[i].props.node_id, 1485 cpus[j].props.node_id); 1486 } 1487 } 1488 } 1489 } 1490 1491 MemoryRegion *machine_consume_memdev(MachineState *machine, 1492 HostMemoryBackend *backend) 1493 { 1494 MemoryRegion *ret = host_memory_backend_get_memory(backend); 1495 1496 if (host_memory_backend_is_mapped(backend)) { 1497 error_report("memory backend %s can't be used multiple times.", 1498 object_get_canonical_path_component(OBJECT(backend))); 1499 exit(EXIT_FAILURE); 1500 } 1501 host_memory_backend_set_mapped(backend, true); 1502 vmstate_register_ram_global(ret); 1503 return ret; 1504 } 1505 1506 const char *machine_class_default_cpu_type(MachineClass *mc) 1507 { 1508 if (mc->valid_cpu_types && !mc->valid_cpu_types[1]) { 1509 /* Only a single CPU type allowed: use it as default. */ 1510 return mc->valid_cpu_types[0]; 1511 } 1512 return mc->default_cpu_type; 1513 } 1514 1515 static bool is_cpu_type_supported(const MachineState *machine, Error **errp) 1516 { 1517 MachineClass *mc = MACHINE_GET_CLASS(machine); 1518 ObjectClass *oc = object_class_by_name(machine->cpu_type); 1519 CPUClass *cc; 1520 int i; 1521 1522 /* 1523 * Check if the user specified CPU type is supported when the valid 1524 * CPU types have been determined. Note that the user specified CPU 1525 * type is provided through '-cpu' option. 1526 */ 1527 if (mc->valid_cpu_types) { 1528 assert(mc->valid_cpu_types[0] != NULL); 1529 for (i = 0; mc->valid_cpu_types[i]; i++) { 1530 if (object_class_dynamic_cast(oc, mc->valid_cpu_types[i])) { 1531 break; 1532 } 1533 } 1534 1535 /* The user specified CPU type isn't valid */ 1536 if (!mc->valid_cpu_types[i]) { 1537 g_autofree char *requested = cpu_model_from_type(machine->cpu_type); 1538 error_setg(errp, "Invalid CPU model: %s", requested); 1539 if (!mc->valid_cpu_types[1]) { 1540 g_autofree char *model = cpu_model_from_type( 1541 mc->valid_cpu_types[0]); 1542 error_append_hint(errp, "The only valid type is: %s\n", model); 1543 } else { 1544 error_append_hint(errp, "The valid models are: "); 1545 for (i = 0; mc->valid_cpu_types[i]; i++) { 1546 g_autofree char *model = cpu_model_from_type( 1547 mc->valid_cpu_types[i]); 1548 error_append_hint(errp, "%s%s", 1549 model, 1550 mc->valid_cpu_types[i + 1] ? ", " : ""); 1551 } 1552 error_append_hint(errp, "\n"); 1553 } 1554 1555 return false; 1556 } 1557 } 1558 1559 /* Check if CPU type is deprecated and warn if so */ 1560 cc = CPU_CLASS(oc); 1561 assert(cc != NULL); 1562 if (cc->deprecation_note) { 1563 warn_report("CPU model %s is deprecated -- %s", 1564 machine->cpu_type, cc->deprecation_note); 1565 } 1566 1567 return true; 1568 } 1569 1570 void machine_run_board_init(MachineState *machine, const char *mem_path, Error **errp) 1571 { 1572 ERRP_GUARD(); 1573 MachineClass *machine_class = MACHINE_GET_CLASS(machine); 1574 1575 /* This checkpoint is required by replay to separate prior clock 1576 reading from the other reads, because timer polling functions query 1577 clock values from the log. */ 1578 replay_checkpoint(CHECKPOINT_INIT); 1579 1580 if (!xen_enabled()) { 1581 /* On 32-bit hosts, QEMU is limited by virtual address space */ 1582 if (machine->ram_size > (2047 << 20) && HOST_LONG_BITS == 32) { 1583 error_setg(errp, "at most 2047 MB RAM can be simulated"); 1584 return; 1585 } 1586 } 1587 1588 if (machine->memdev) { 1589 ram_addr_t backend_size = object_property_get_uint(OBJECT(machine->memdev), 1590 "size", &error_abort); 1591 if (backend_size != machine->ram_size) { 1592 error_setg(errp, "Machine memory size does not match the size of the memory backend"); 1593 return; 1594 } 1595 } else if (machine_class->default_ram_id && machine->ram_size && 1596 numa_uses_legacy_mem()) { 1597 if (object_property_find(object_get_objects_root(), 1598 machine_class->default_ram_id)) { 1599 error_setg(errp, "object's id '%s' is reserved for the default" 1600 " RAM backend, it can't be used for any other purposes", 1601 machine_class->default_ram_id); 1602 error_append_hint(errp, 1603 "Change the object's 'id' to something else or disable" 1604 " automatic creation of the default RAM backend by setting" 1605 " 'memory-backend=%s' with '-machine'.\n", 1606 machine_class->default_ram_id); 1607 return; 1608 } 1609 1610 if (!machine_class->create_default_memdev(current_machine, mem_path, 1611 errp)) { 1612 return; 1613 } 1614 } 1615 1616 if (machine->numa_state) { 1617 numa_complete_configuration(machine); 1618 if (machine->numa_state->num_nodes) { 1619 machine_numa_finish_cpu_init(machine); 1620 if (machine_class->cpu_cluster_has_numa_boundary) { 1621 validate_cpu_cluster_to_numa_boundary(machine); 1622 } 1623 } 1624 } 1625 1626 if (!machine->ram && machine->memdev) { 1627 machine->ram = machine_consume_memdev(machine, machine->memdev); 1628 } 1629 1630 /* Check if the CPU type is supported */ 1631 if (machine->cpu_type && !is_cpu_type_supported(machine, errp)) { 1632 return; 1633 } 1634 1635 if (machine->cgs) { 1636 /* 1637 * With confidential guests, the host can't see the real 1638 * contents of RAM, so there's no point in it trying to merge 1639 * areas. 1640 */ 1641 machine_set_mem_merge(OBJECT(machine), false, &error_abort); 1642 1643 /* 1644 * Virtio devices can't count on directly accessing guest 1645 * memory, so they need iommu_platform=on to use normal DMA 1646 * mechanisms. That requires also disabling legacy virtio 1647 * support for those virtio pci devices which allow it. 1648 */ 1649 object_register_sugar_prop(TYPE_VIRTIO_PCI, "disable-legacy", 1650 "on", true); 1651 object_register_sugar_prop(TYPE_VIRTIO_DEVICE, "iommu_platform", 1652 "on", false); 1653 } 1654 1655 accel_init_interfaces(ACCEL_GET_CLASS(machine->accelerator)); 1656 machine_class->init(machine); 1657 phase_advance(PHASE_MACHINE_INITIALIZED); 1658 } 1659 1660 static NotifierList machine_init_done_notifiers = 1661 NOTIFIER_LIST_INITIALIZER(machine_init_done_notifiers); 1662 1663 void qemu_add_machine_init_done_notifier(Notifier *notify) 1664 { 1665 notifier_list_add(&machine_init_done_notifiers, notify); 1666 if (phase_check(PHASE_MACHINE_READY)) { 1667 notify->notify(notify, NULL); 1668 } 1669 } 1670 1671 void qemu_remove_machine_init_done_notifier(Notifier *notify) 1672 { 1673 notifier_remove(notify); 1674 } 1675 1676 void qdev_machine_creation_done(void) 1677 { 1678 cpu_synchronize_all_post_init(); 1679 1680 if (current_machine->boot_config.once) { 1681 qemu_boot_set(current_machine->boot_config.once, &error_fatal); 1682 qemu_register_reset(restore_boot_order, g_strdup(current_machine->boot_config.order)); 1683 } 1684 1685 /* 1686 * ok, initial machine setup is done, starting from now we can 1687 * only create hotpluggable devices 1688 */ 1689 phase_advance(PHASE_MACHINE_READY); 1690 qdev_assert_realized_properly(); 1691 1692 /* TODO: once all bus devices are qdevified, this should be done 1693 * when bus is created by qdev.c */ 1694 /* 1695 * This is where we arrange for the sysbus to be reset when the 1696 * whole simulation is reset. In turn, resetting the sysbus will cause 1697 * all devices hanging off it (and all their child buses, recursively) 1698 * to be reset. Note that this will *not* reset any Device objects 1699 * which are not attached to some part of the qbus tree! 1700 */ 1701 qemu_register_resettable(OBJECT(sysbus_get_default())); 1702 1703 notifier_list_notify(&machine_init_done_notifiers, NULL); 1704 1705 if (rom_check_and_register_reset() != 0) { 1706 exit(1); 1707 } 1708 1709 replay_start(); 1710 1711 /* This checkpoint is required by replay to separate prior clock 1712 reading from the other reads, because timer polling functions query 1713 clock values from the log. */ 1714 replay_checkpoint(CHECKPOINT_RESET); 1715 qemu_system_reset(SHUTDOWN_CAUSE_NONE); 1716 register_global_state(); 1717 } 1718 1719 static const TypeInfo machine_info = { 1720 .name = TYPE_MACHINE, 1721 .parent = TYPE_OBJECT, 1722 .abstract = true, 1723 .class_size = sizeof(MachineClass), 1724 .class_init = machine_class_init, 1725 .class_base_init = machine_class_base_init, 1726 .instance_size = sizeof(MachineState), 1727 .instance_init = machine_initfn, 1728 .instance_finalize = machine_finalize, 1729 }; 1730 1731 static void machine_register_types(void) 1732 { 1733 type_register_static(&machine_info); 1734 } 1735 1736 type_init(machine_register_types) 1737