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