1 /* 2 * QEMU Machine 3 * 4 * Copyright (C) 2014 Red Hat Inc 5 * 6 * Authors: 7 * Marcel Apfelbaum <marcel.a@redhat.com> 8 * 9 * This work is licensed under the terms of the GNU GPL, version 2 or later. 10 * See the COPYING file in the top-level directory. 11 */ 12 13 #include "qemu/osdep.h" 14 #include "qemu/accel.h" 15 #include "sysemu/replay.h" 16 #include "hw/boards.h" 17 #include "hw/loader.h" 18 #include "qapi/error.h" 19 #include "qapi/qapi-visit-machine.h" 20 #include "qom/object_interfaces.h" 21 #include "sysemu/cpus.h" 22 #include "sysemu/sysemu.h" 23 #include "sysemu/reset.h" 24 #include "sysemu/runstate.h" 25 #include "sysemu/xen.h" 26 #include "sysemu/qtest.h" 27 #include "hw/pci/pci_bridge.h" 28 #include "hw/mem/nvdimm.h" 29 #include "migration/global_state.h" 30 #include "exec/confidential-guest-support.h" 31 #include "hw/virtio/virtio-pci.h" 32 #include "hw/virtio/virtio-net.h" 33 #include "hw/virtio/virtio-iommu.h" 34 #include "audio/audio.h" 35 36 GlobalProperty hw_compat_9_0[] = { 37 {"arm-cpu", "backcompat-cntfrq", "true" }, 38 }; 39 const size_t hw_compat_9_0_len = G_N_ELEMENTS(hw_compat_9_0); 40 41 GlobalProperty hw_compat_8_2[] = { 42 { "migration", "zero-page-detection", "legacy"}, 43 { TYPE_VIRTIO_IOMMU_PCI, "granule", "4k" }, 44 { TYPE_VIRTIO_IOMMU_PCI, "aw-bits", "64" }, 45 }; 46 const size_t hw_compat_8_2_len = G_N_ELEMENTS(hw_compat_8_2); 47 48 GlobalProperty hw_compat_8_1[] = { 49 { TYPE_PCI_BRIDGE, "x-pci-express-writeable-slt-bug", "true" }, 50 { "ramfb", "x-migrate", "off" }, 51 { "vfio-pci-nohotplug", "x-ramfb-migrate", "off" }, 52 { "igb", "x-pcie-flr-init", "off" }, 53 }; 54 const size_t hw_compat_8_1_len = G_N_ELEMENTS(hw_compat_8_1); 55 56 GlobalProperty hw_compat_8_0[] = { 57 { "migration", "multifd-flush-after-each-section", "on"}, 58 { TYPE_PCI_DEVICE, "x-pcie-ari-nextfn-1", "on" }, 59 { TYPE_VIRTIO_NET, "host_uso", "off"}, 60 { TYPE_VIRTIO_NET, "guest_uso4", "off"}, 61 { TYPE_VIRTIO_NET, "guest_uso6", "off"}, 62 }; 63 const size_t hw_compat_8_0_len = G_N_ELEMENTS(hw_compat_8_0); 64 65 GlobalProperty hw_compat_7_2[] = { 66 { "e1000e", "migrate-timadj", "off" }, 67 { "virtio-mem", "x-early-migration", "false" }, 68 { "migration", "x-preempt-pre-7-2", "true" }, 69 { TYPE_PCI_DEVICE, "x-pcie-err-unc-mask", "off" }, 70 }; 71 const size_t hw_compat_7_2_len = G_N_ELEMENTS(hw_compat_7_2); 72 73 GlobalProperty hw_compat_7_1[] = { 74 { "virtio-device", "queue_reset", "false" }, 75 { "virtio-rng-pci", "vectors", "0" }, 76 { "virtio-rng-pci-transitional", "vectors", "0" }, 77 { "virtio-rng-pci-non-transitional", "vectors", "0" }, 78 }; 79 const size_t hw_compat_7_1_len = G_N_ELEMENTS(hw_compat_7_1); 80 81 GlobalProperty hw_compat_7_0[] = { 82 { "arm-gicv3-common", "force-8-bit-prio", "on" }, 83 { "nvme-ns", "eui64-default", "on"}, 84 }; 85 const size_t hw_compat_7_0_len = G_N_ELEMENTS(hw_compat_7_0); 86 87 GlobalProperty hw_compat_6_2[] = { 88 { "PIIX4_PM", "x-not-migrate-acpi-index", "on"}, 89 }; 90 const size_t hw_compat_6_2_len = G_N_ELEMENTS(hw_compat_6_2); 91 92 GlobalProperty hw_compat_6_1[] = { 93 { "vhost-user-vsock-device", "seqpacket", "off" }, 94 { "nvme-ns", "shared", "off" }, 95 }; 96 const size_t hw_compat_6_1_len = G_N_ELEMENTS(hw_compat_6_1); 97 98 GlobalProperty hw_compat_6_0[] = { 99 { "gpex-pcihost", "allow-unmapped-accesses", "false" }, 100 { "i8042", "extended-state", "false"}, 101 { "nvme-ns", "eui64-default", "off"}, 102 { "e1000", "init-vet", "off" }, 103 { "e1000e", "init-vet", "off" }, 104 { "vhost-vsock-device", "seqpacket", "off" }, 105 }; 106 const size_t hw_compat_6_0_len = G_N_ELEMENTS(hw_compat_6_0); 107 108 GlobalProperty hw_compat_5_2[] = { 109 { "ICH9-LPC", "smm-compat", "on"}, 110 { "PIIX4_PM", "smm-compat", "on"}, 111 { "virtio-blk-device", "report-discard-granularity", "off" }, 112 { "virtio-net-pci-base", "vectors", "3"}, 113 { "nvme", "msix-exclusive-bar", "on"}, 114 }; 115 const size_t hw_compat_5_2_len = G_N_ELEMENTS(hw_compat_5_2); 116 117 GlobalProperty hw_compat_5_1[] = { 118 { "vhost-scsi", "num_queues", "1"}, 119 { "vhost-user-blk", "num-queues", "1"}, 120 { "vhost-user-scsi", "num_queues", "1"}, 121 { "virtio-blk-device", "num-queues", "1"}, 122 { "virtio-scsi-device", "num_queues", "1"}, 123 { "nvme", "use-intel-id", "on"}, 124 { "pvpanic", "events", "1"}, /* PVPANIC_PANICKED */ 125 { "pl011", "migrate-clk", "off" }, 126 { "virtio-pci", "x-ats-page-aligned", "off"}, 127 }; 128 const size_t hw_compat_5_1_len = G_N_ELEMENTS(hw_compat_5_1); 129 130 GlobalProperty hw_compat_5_0[] = { 131 { "pci-host-bridge", "x-config-reg-migration-enabled", "off" }, 132 { "virtio-balloon-device", "page-poison", "false" }, 133 { "vmport", "x-read-set-eax", "off" }, 134 { "vmport", "x-signal-unsupported-cmd", "off" }, 135 { "vmport", "x-report-vmx-type", "off" }, 136 { "vmport", "x-cmds-v2", "off" }, 137 { "virtio-device", "x-disable-legacy-check", "true" }, 138 }; 139 const size_t hw_compat_5_0_len = G_N_ELEMENTS(hw_compat_5_0); 140 141 GlobalProperty hw_compat_4_2[] = { 142 { "virtio-blk-device", "queue-size", "128"}, 143 { "virtio-scsi-device", "virtqueue_size", "128"}, 144 { "virtio-blk-device", "x-enable-wce-if-config-wce", "off" }, 145 { "virtio-blk-device", "seg-max-adjust", "off"}, 146 { "virtio-scsi-device", "seg_max_adjust", "off"}, 147 { "vhost-blk-device", "seg_max_adjust", "off"}, 148 { "usb-host", "suppress-remote-wake", "off" }, 149 { "usb-redir", "suppress-remote-wake", "off" }, 150 { "qxl", "revision", "4" }, 151 { "qxl-vga", "revision", "4" }, 152 { "fw_cfg", "acpi-mr-restore", "false" }, 153 { "virtio-device", "use-disabled-flag", "false" }, 154 }; 155 const size_t hw_compat_4_2_len = G_N_ELEMENTS(hw_compat_4_2); 156 157 GlobalProperty hw_compat_4_1[] = { 158 { "virtio-pci", "x-pcie-flr-init", "off" }, 159 }; 160 const size_t hw_compat_4_1_len = G_N_ELEMENTS(hw_compat_4_1); 161 162 GlobalProperty hw_compat_4_0[] = { 163 { "VGA", "edid", "false" }, 164 { "secondary-vga", "edid", "false" }, 165 { "bochs-display", "edid", "false" }, 166 { "virtio-vga", "edid", "false" }, 167 { "virtio-gpu-device", "edid", "false" }, 168 { "virtio-device", "use-started", "false" }, 169 { "virtio-balloon-device", "qemu-4-0-config-size", "true" }, 170 { "pl031", "migrate-tick-offset", "false" }, 171 }; 172 const size_t hw_compat_4_0_len = G_N_ELEMENTS(hw_compat_4_0); 173 174 GlobalProperty hw_compat_3_1[] = { 175 { "pcie-root-port", "x-speed", "2_5" }, 176 { "pcie-root-port", "x-width", "1" }, 177 { "memory-backend-file", "x-use-canonical-path-for-ramblock-id", "true" }, 178 { "memory-backend-memfd", "x-use-canonical-path-for-ramblock-id", "true" }, 179 { "tpm-crb", "ppi", "false" }, 180 { "tpm-tis", "ppi", "false" }, 181 { "usb-kbd", "serial", "42" }, 182 { "usb-mouse", "serial", "42" }, 183 { "usb-tablet", "serial", "42" }, 184 { "virtio-blk-device", "discard", "false" }, 185 { "virtio-blk-device", "write-zeroes", "false" }, 186 { "virtio-balloon-device", "qemu-4-0-config-size", "false" }, 187 { "pcie-root-port-base", "disable-acs", "true" }, /* Added in 4.1 */ 188 }; 189 const size_t hw_compat_3_1_len = G_N_ELEMENTS(hw_compat_3_1); 190 191 GlobalProperty hw_compat_3_0[] = {}; 192 const size_t hw_compat_3_0_len = G_N_ELEMENTS(hw_compat_3_0); 193 194 GlobalProperty hw_compat_2_12[] = { 195 { "hda-audio", "use-timer", "false" }, 196 { "cirrus-vga", "global-vmstate", "true" }, 197 { "VGA", "global-vmstate", "true" }, 198 { "vmware-svga", "global-vmstate", "true" }, 199 { "qxl-vga", "global-vmstate", "true" }, 200 }; 201 const size_t hw_compat_2_12_len = G_N_ELEMENTS(hw_compat_2_12); 202 203 GlobalProperty hw_compat_2_11[] = { 204 { "hpet", "hpet-offset-saved", "false" }, 205 { "virtio-blk-pci", "vectors", "2" }, 206 { "vhost-user-blk-pci", "vectors", "2" }, 207 { "e1000", "migrate_tso_props", "off" }, 208 }; 209 const size_t hw_compat_2_11_len = G_N_ELEMENTS(hw_compat_2_11); 210 211 GlobalProperty hw_compat_2_10[] = { 212 { "virtio-mouse-device", "wheel-axis", "false" }, 213 { "virtio-tablet-device", "wheel-axis", "false" }, 214 }; 215 const size_t hw_compat_2_10_len = G_N_ELEMENTS(hw_compat_2_10); 216 217 GlobalProperty hw_compat_2_9[] = { 218 { "pci-bridge", "shpc", "off" }, 219 { "intel-iommu", "pt", "off" }, 220 { "virtio-net-device", "x-mtu-bypass-backend", "off" }, 221 { "pcie-root-port", "x-migrate-msix", "false" }, 222 }; 223 const size_t hw_compat_2_9_len = G_N_ELEMENTS(hw_compat_2_9); 224 225 GlobalProperty hw_compat_2_8[] = { 226 { "fw_cfg_mem", "x-file-slots", "0x10" }, 227 { "fw_cfg_io", "x-file-slots", "0x10" }, 228 { "pflash_cfi01", "old-multiple-chip-handling", "on" }, 229 { "pci-bridge", "shpc", "on" }, 230 { TYPE_PCI_DEVICE, "x-pcie-extcap-init", "off" }, 231 { "virtio-pci", "x-pcie-deverr-init", "off" }, 232 { "virtio-pci", "x-pcie-lnkctl-init", "off" }, 233 { "virtio-pci", "x-pcie-pm-init", "off" }, 234 { "cirrus-vga", "vgamem_mb", "8" }, 235 { "isa-cirrus-vga", "vgamem_mb", "8" }, 236 }; 237 const size_t hw_compat_2_8_len = G_N_ELEMENTS(hw_compat_2_8); 238 239 GlobalProperty hw_compat_2_7[] = { 240 { "virtio-pci", "page-per-vq", "on" }, 241 { "virtio-serial-device", "emergency-write", "off" }, 242 { "ioapic", "version", "0x11" }, 243 { "intel-iommu", "x-buggy-eim", "true" }, 244 { "virtio-pci", "x-ignore-backend-features", "on" }, 245 }; 246 const size_t hw_compat_2_7_len = G_N_ELEMENTS(hw_compat_2_7); 247 248 GlobalProperty hw_compat_2_6[] = { 249 { "virtio-mmio", "format_transport_address", "off" }, 250 /* Optional because not all virtio-pci devices support legacy mode */ 251 { "virtio-pci", "disable-modern", "on", .optional = true }, 252 { "virtio-pci", "disable-legacy", "off", .optional = true }, 253 }; 254 const size_t hw_compat_2_6_len = G_N_ELEMENTS(hw_compat_2_6); 255 256 GlobalProperty hw_compat_2_5[] = { 257 { "isa-fdc", "fallback", "144" }, 258 { "pvscsi", "x-old-pci-configuration", "on" }, 259 { "pvscsi", "x-disable-pcie", "on" }, 260 { "vmxnet3", "x-old-msi-offsets", "on" }, 261 { "vmxnet3", "x-disable-pcie", "on" }, 262 }; 263 const size_t hw_compat_2_5_len = G_N_ELEMENTS(hw_compat_2_5); 264 265 GlobalProperty hw_compat_2_4[] = { 266 /* Optional because the 'scsi' property is Linux-only */ 267 { "virtio-blk-device", "scsi", "true", .optional = true }, 268 { "e1000", "extra_mac_registers", "off" }, 269 { "virtio-pci", "x-disable-pcie", "on" }, 270 { "virtio-pci", "migrate-extra", "off" }, 271 { "fw_cfg_mem", "dma_enabled", "off" }, 272 { "fw_cfg_io", "dma_enabled", "off" } 273 }; 274 const size_t hw_compat_2_4_len = G_N_ELEMENTS(hw_compat_2_4); 275 276 GlobalProperty hw_compat_2_3[] = { 277 { "virtio-blk-pci", "any_layout", "off" }, 278 { "virtio-balloon-pci", "any_layout", "off" }, 279 { "virtio-serial-pci", "any_layout", "off" }, 280 { "virtio-9p-pci", "any_layout", "off" }, 281 { "virtio-rng-pci", "any_layout", "off" }, 282 { TYPE_PCI_DEVICE, "x-pcie-lnksta-dllla", "off" }, 283 { "migration", "send-configuration", "off" }, 284 { "migration", "send-section-footer", "off" }, 285 { "migration", "store-global-state", "off" }, 286 }; 287 const size_t hw_compat_2_3_len = G_N_ELEMENTS(hw_compat_2_3); 288 289 GlobalProperty hw_compat_2_2[] = {}; 290 const size_t hw_compat_2_2_len = G_N_ELEMENTS(hw_compat_2_2); 291 292 GlobalProperty hw_compat_2_1[] = { 293 { "intel-hda", "old_msi_addr", "on" }, 294 { "VGA", "qemu-extended-regs", "off" }, 295 { "secondary-vga", "qemu-extended-regs", "off" }, 296 { "virtio-scsi-pci", "any_layout", "off" }, 297 { "usb-mouse", "usb_version", "1" }, 298 { "usb-kbd", "usb_version", "1" }, 299 { "virtio-pci", "virtio-pci-bus-master-bug-migration", "on" }, 300 }; 301 const size_t hw_compat_2_1_len = G_N_ELEMENTS(hw_compat_2_1); 302 303 MachineState *current_machine; 304 305 static char *machine_get_kernel(Object *obj, Error **errp) 306 { 307 MachineState *ms = MACHINE(obj); 308 309 return g_strdup(ms->kernel_filename); 310 } 311 312 static void machine_set_kernel(Object *obj, const char *value, Error **errp) 313 { 314 MachineState *ms = MACHINE(obj); 315 316 g_free(ms->kernel_filename); 317 ms->kernel_filename = g_strdup(value); 318 } 319 320 static char *machine_get_initrd(Object *obj, Error **errp) 321 { 322 MachineState *ms = MACHINE(obj); 323 324 return g_strdup(ms->initrd_filename); 325 } 326 327 static void machine_set_initrd(Object *obj, const char *value, Error **errp) 328 { 329 MachineState *ms = MACHINE(obj); 330 331 g_free(ms->initrd_filename); 332 ms->initrd_filename = g_strdup(value); 333 } 334 335 static char *machine_get_append(Object *obj, Error **errp) 336 { 337 MachineState *ms = MACHINE(obj); 338 339 return g_strdup(ms->kernel_cmdline); 340 } 341 342 static void machine_set_append(Object *obj, const char *value, Error **errp) 343 { 344 MachineState *ms = MACHINE(obj); 345 346 g_free(ms->kernel_cmdline); 347 ms->kernel_cmdline = g_strdup(value); 348 } 349 350 static char *machine_get_dtb(Object *obj, Error **errp) 351 { 352 MachineState *ms = MACHINE(obj); 353 354 return g_strdup(ms->dtb); 355 } 356 357 static void machine_set_dtb(Object *obj, const char *value, Error **errp) 358 { 359 MachineState *ms = MACHINE(obj); 360 361 g_free(ms->dtb); 362 ms->dtb = g_strdup(value); 363 } 364 365 static char *machine_get_dumpdtb(Object *obj, Error **errp) 366 { 367 MachineState *ms = MACHINE(obj); 368 369 return g_strdup(ms->dumpdtb); 370 } 371 372 static void machine_set_dumpdtb(Object *obj, const char *value, Error **errp) 373 { 374 MachineState *ms = MACHINE(obj); 375 376 g_free(ms->dumpdtb); 377 ms->dumpdtb = g_strdup(value); 378 } 379 380 static void machine_get_phandle_start(Object *obj, Visitor *v, 381 const char *name, void *opaque, 382 Error **errp) 383 { 384 MachineState *ms = MACHINE(obj); 385 int64_t value = ms->phandle_start; 386 387 visit_type_int(v, name, &value, errp); 388 } 389 390 static void machine_set_phandle_start(Object *obj, Visitor *v, 391 const char *name, void *opaque, 392 Error **errp) 393 { 394 MachineState *ms = MACHINE(obj); 395 int64_t value; 396 397 if (!visit_type_int(v, name, &value, errp)) { 398 return; 399 } 400 401 ms->phandle_start = value; 402 } 403 404 static char *machine_get_dt_compatible(Object *obj, Error **errp) 405 { 406 MachineState *ms = MACHINE(obj); 407 408 return g_strdup(ms->dt_compatible); 409 } 410 411 static void machine_set_dt_compatible(Object *obj, const char *value, Error **errp) 412 { 413 MachineState *ms = MACHINE(obj); 414 415 g_free(ms->dt_compatible); 416 ms->dt_compatible = g_strdup(value); 417 } 418 419 static bool machine_get_dump_guest_core(Object *obj, Error **errp) 420 { 421 MachineState *ms = MACHINE(obj); 422 423 return ms->dump_guest_core; 424 } 425 426 static void machine_set_dump_guest_core(Object *obj, bool value, Error **errp) 427 { 428 MachineState *ms = MACHINE(obj); 429 430 ms->dump_guest_core = value; 431 } 432 433 static bool machine_get_mem_merge(Object *obj, Error **errp) 434 { 435 MachineState *ms = MACHINE(obj); 436 437 return ms->mem_merge; 438 } 439 440 static void machine_set_mem_merge(Object *obj, bool value, Error **errp) 441 { 442 MachineState *ms = MACHINE(obj); 443 444 ms->mem_merge = value; 445 } 446 447 static bool machine_get_usb(Object *obj, Error **errp) 448 { 449 MachineState *ms = MACHINE(obj); 450 451 return ms->usb; 452 } 453 454 static void machine_set_usb(Object *obj, bool value, Error **errp) 455 { 456 MachineState *ms = MACHINE(obj); 457 458 ms->usb = value; 459 ms->usb_disabled = !value; 460 } 461 462 static bool machine_get_graphics(Object *obj, Error **errp) 463 { 464 MachineState *ms = MACHINE(obj); 465 466 return ms->enable_graphics; 467 } 468 469 static void machine_set_graphics(Object *obj, bool value, Error **errp) 470 { 471 MachineState *ms = MACHINE(obj); 472 473 ms->enable_graphics = value; 474 } 475 476 static char *machine_get_firmware(Object *obj, Error **errp) 477 { 478 MachineState *ms = MACHINE(obj); 479 480 return g_strdup(ms->firmware); 481 } 482 483 static void machine_set_firmware(Object *obj, const char *value, Error **errp) 484 { 485 MachineState *ms = MACHINE(obj); 486 487 g_free(ms->firmware); 488 ms->firmware = g_strdup(value); 489 } 490 491 static void machine_set_suppress_vmdesc(Object *obj, bool value, Error **errp) 492 { 493 MachineState *ms = MACHINE(obj); 494 495 ms->suppress_vmdesc = value; 496 } 497 498 static bool machine_get_suppress_vmdesc(Object *obj, Error **errp) 499 { 500 MachineState *ms = MACHINE(obj); 501 502 return ms->suppress_vmdesc; 503 } 504 505 static char *machine_get_memory_encryption(Object *obj, Error **errp) 506 { 507 MachineState *ms = MACHINE(obj); 508 509 if (ms->cgs) { 510 return g_strdup(object_get_canonical_path_component(OBJECT(ms->cgs))); 511 } 512 513 return NULL; 514 } 515 516 static void machine_set_memory_encryption(Object *obj, const char *value, 517 Error **errp) 518 { 519 Object *cgs = 520 object_resolve_path_component(object_get_objects_root(), value); 521 522 if (!cgs) { 523 error_setg(errp, "No such memory encryption object '%s'", value); 524 return; 525 } 526 527 object_property_set_link(obj, "confidential-guest-support", cgs, errp); 528 } 529 530 static void machine_check_confidential_guest_support(const Object *obj, 531 const char *name, 532 Object *new_target, 533 Error **errp) 534 { 535 /* 536 * So far the only constraint is that the target has the 537 * TYPE_CONFIDENTIAL_GUEST_SUPPORT interface, and that's checked 538 * by the QOM core 539 */ 540 } 541 542 static bool machine_get_nvdimm(Object *obj, Error **errp) 543 { 544 MachineState *ms = MACHINE(obj); 545 546 return ms->nvdimms_state->is_enabled; 547 } 548 549 static void machine_set_nvdimm(Object *obj, bool value, Error **errp) 550 { 551 MachineState *ms = MACHINE(obj); 552 553 ms->nvdimms_state->is_enabled = value; 554 } 555 556 static bool machine_get_hmat(Object *obj, Error **errp) 557 { 558 MachineState *ms = MACHINE(obj); 559 560 return ms->numa_state->hmat_enabled; 561 } 562 563 static void machine_set_hmat(Object *obj, bool value, Error **errp) 564 { 565 MachineState *ms = MACHINE(obj); 566 567 ms->numa_state->hmat_enabled = value; 568 } 569 570 static void machine_get_mem(Object *obj, Visitor *v, const char *name, 571 void *opaque, Error **errp) 572 { 573 MachineState *ms = MACHINE(obj); 574 MemorySizeConfiguration mem = { 575 .has_size = true, 576 .size = ms->ram_size, 577 .has_max_size = !!ms->ram_slots, 578 .max_size = ms->maxram_size, 579 .has_slots = !!ms->ram_slots, 580 .slots = ms->ram_slots, 581 }; 582 MemorySizeConfiguration *p_mem = &mem; 583 584 visit_type_MemorySizeConfiguration(v, name, &p_mem, &error_abort); 585 } 586 587 static void machine_set_mem(Object *obj, Visitor *v, const char *name, 588 void *opaque, Error **errp) 589 { 590 ERRP_GUARD(); 591 MachineState *ms = MACHINE(obj); 592 MachineClass *mc = MACHINE_GET_CLASS(obj); 593 MemorySizeConfiguration *mem; 594 595 if (!visit_type_MemorySizeConfiguration(v, name, &mem, errp)) { 596 return; 597 } 598 599 if (!mem->has_size) { 600 mem->has_size = true; 601 mem->size = mc->default_ram_size; 602 } 603 mem->size = QEMU_ALIGN_UP(mem->size, 8192); 604 if (mc->fixup_ram_size) { 605 mem->size = mc->fixup_ram_size(mem->size); 606 } 607 if ((ram_addr_t)mem->size != mem->size) { 608 error_setg(errp, "ram size too large"); 609 goto out_free; 610 } 611 612 if (mem->has_max_size) { 613 if (mem->max_size < mem->size) { 614 error_setg(errp, "invalid value of maxmem: " 615 "maximum memory size (0x%" PRIx64 ") must be at least " 616 "the initial memory size (0x%" PRIx64 ")", 617 mem->max_size, mem->size); 618 goto out_free; 619 } 620 if (mem->has_slots && mem->slots && mem->max_size == mem->size) { 621 error_setg(errp, "invalid value of maxmem: " 622 "memory slots were specified but maximum memory size " 623 "(0x%" PRIx64 ") is equal to the initial memory size " 624 "(0x%" PRIx64 ")", mem->max_size, mem->size); 625 goto out_free; 626 } 627 ms->maxram_size = mem->max_size; 628 } else { 629 if (mem->has_slots) { 630 error_setg(errp, "slots specified but no max-size"); 631 goto out_free; 632 } 633 ms->maxram_size = mem->size; 634 } 635 ms->ram_size = mem->size; 636 ms->ram_slots = mem->has_slots ? mem->slots : 0; 637 out_free: 638 qapi_free_MemorySizeConfiguration(mem); 639 } 640 641 static char *machine_get_nvdimm_persistence(Object *obj, Error **errp) 642 { 643 MachineState *ms = MACHINE(obj); 644 645 return g_strdup(ms->nvdimms_state->persistence_string); 646 } 647 648 static void machine_set_nvdimm_persistence(Object *obj, const char *value, 649 Error **errp) 650 { 651 MachineState *ms = MACHINE(obj); 652 NVDIMMState *nvdimms_state = ms->nvdimms_state; 653 654 if (strcmp(value, "cpu") == 0) { 655 nvdimms_state->persistence = 3; 656 } else if (strcmp(value, "mem-ctrl") == 0) { 657 nvdimms_state->persistence = 2; 658 } else { 659 error_setg(errp, "-machine nvdimm-persistence=%s: unsupported option", 660 value); 661 return; 662 } 663 664 g_free(nvdimms_state->persistence_string); 665 nvdimms_state->persistence_string = g_strdup(value); 666 } 667 668 void machine_class_allow_dynamic_sysbus_dev(MachineClass *mc, const char *type) 669 { 670 QAPI_LIST_PREPEND(mc->allowed_dynamic_sysbus_devices, g_strdup(type)); 671 } 672 673 bool device_is_dynamic_sysbus(MachineClass *mc, DeviceState *dev) 674 { 675 Object *obj = OBJECT(dev); 676 677 if (!object_dynamic_cast(obj, TYPE_SYS_BUS_DEVICE)) { 678 return false; 679 } 680 681 return device_type_is_dynamic_sysbus(mc, object_get_typename(obj)); 682 } 683 684 bool device_type_is_dynamic_sysbus(MachineClass *mc, const char *type) 685 { 686 bool allowed = false; 687 strList *wl; 688 ObjectClass *klass = object_class_by_name(type); 689 690 for (wl = mc->allowed_dynamic_sysbus_devices; 691 !allowed && wl; 692 wl = wl->next) { 693 allowed |= !!object_class_dynamic_cast(klass, wl->value); 694 } 695 696 return allowed; 697 } 698 699 static char *machine_get_audiodev(Object *obj, Error **errp) 700 { 701 MachineState *ms = MACHINE(obj); 702 703 return g_strdup(ms->audiodev); 704 } 705 706 static void machine_set_audiodev(Object *obj, const char *value, 707 Error **errp) 708 { 709 MachineState *ms = MACHINE(obj); 710 711 if (!audio_state_by_name(value, errp)) { 712 return; 713 } 714 715 g_free(ms->audiodev); 716 ms->audiodev = g_strdup(value); 717 } 718 719 HotpluggableCPUList *machine_query_hotpluggable_cpus(MachineState *machine) 720 { 721 int i; 722 HotpluggableCPUList *head = NULL; 723 MachineClass *mc = MACHINE_GET_CLASS(machine); 724 725 /* force board to initialize possible_cpus if it hasn't been done yet */ 726 mc->possible_cpu_arch_ids(machine); 727 728 for (i = 0; i < machine->possible_cpus->len; i++) { 729 CPUState *cpu; 730 HotpluggableCPU *cpu_item = g_new0(typeof(*cpu_item), 1); 731 732 cpu_item->type = g_strdup(machine->possible_cpus->cpus[i].type); 733 cpu_item->vcpus_count = machine->possible_cpus->cpus[i].vcpus_count; 734 cpu_item->props = g_memdup(&machine->possible_cpus->cpus[i].props, 735 sizeof(*cpu_item->props)); 736 737 cpu = machine->possible_cpus->cpus[i].cpu; 738 if (cpu) { 739 cpu_item->qom_path = object_get_canonical_path(OBJECT(cpu)); 740 } 741 QAPI_LIST_PREPEND(head, cpu_item); 742 } 743 return head; 744 } 745 746 /** 747 * machine_set_cpu_numa_node: 748 * @machine: machine object to modify 749 * @props: specifies which cpu objects to assign to 750 * numa node specified by @props.node_id 751 * @errp: if an error occurs, a pointer to an area to store the error 752 * 753 * Associate NUMA node specified by @props.node_id with cpu slots that 754 * match socket/core/thread-ids specified by @props. It's recommended to use 755 * query-hotpluggable-cpus.props values to specify affected cpu slots, 756 * which would lead to exact 1:1 mapping of cpu slots to NUMA node. 757 * 758 * However for CLI convenience it's possible to pass in subset of properties, 759 * which would affect all cpu slots that match it. 760 * Ex for pc machine: 761 * -smp 4,cores=2,sockets=2 -numa node,nodeid=0 -numa node,nodeid=1 \ 762 * -numa cpu,node-id=0,socket_id=0 \ 763 * -numa cpu,node-id=1,socket_id=1 764 * will assign all child cores of socket 0 to node 0 and 765 * of socket 1 to node 1. 766 * 767 * On attempt of reassigning (already assigned) cpu slot to another NUMA node, 768 * return error. 769 * Empty subset is disallowed and function will return with error in this case. 770 */ 771 void machine_set_cpu_numa_node(MachineState *machine, 772 const CpuInstanceProperties *props, Error **errp) 773 { 774 MachineClass *mc = MACHINE_GET_CLASS(machine); 775 NodeInfo *numa_info = machine->numa_state->nodes; 776 bool match = false; 777 int i; 778 779 if (!mc->possible_cpu_arch_ids) { 780 error_setg(errp, "mapping of CPUs to NUMA node is not supported"); 781 return; 782 } 783 784 /* disabling node mapping is not supported, forbid it */ 785 assert(props->has_node_id); 786 787 /* force board to initialize possible_cpus if it hasn't been done yet */ 788 mc->possible_cpu_arch_ids(machine); 789 790 for (i = 0; i < machine->possible_cpus->len; i++) { 791 CPUArchId *slot = &machine->possible_cpus->cpus[i]; 792 793 /* reject unsupported by board properties */ 794 if (props->has_thread_id && !slot->props.has_thread_id) { 795 error_setg(errp, "thread-id is not supported"); 796 return; 797 } 798 799 if (props->has_core_id && !slot->props.has_core_id) { 800 error_setg(errp, "core-id is not supported"); 801 return; 802 } 803 804 if (props->has_module_id && !slot->props.has_module_id) { 805 error_setg(errp, "module-id is not supported"); 806 return; 807 } 808 809 if (props->has_cluster_id && !slot->props.has_cluster_id) { 810 error_setg(errp, "cluster-id is not supported"); 811 return; 812 } 813 814 if (props->has_socket_id && !slot->props.has_socket_id) { 815 error_setg(errp, "socket-id is not supported"); 816 return; 817 } 818 819 if (props->has_die_id && !slot->props.has_die_id) { 820 error_setg(errp, "die-id is not supported"); 821 return; 822 } 823 824 /* skip slots with explicit mismatch */ 825 if (props->has_thread_id && props->thread_id != slot->props.thread_id) { 826 continue; 827 } 828 829 if (props->has_core_id && props->core_id != slot->props.core_id) { 830 continue; 831 } 832 833 if (props->has_module_id && 834 props->module_id != slot->props.module_id) { 835 continue; 836 } 837 838 if (props->has_cluster_id && 839 props->cluster_id != slot->props.cluster_id) { 840 continue; 841 } 842 843 if (props->has_die_id && props->die_id != slot->props.die_id) { 844 continue; 845 } 846 847 if (props->has_socket_id && props->socket_id != slot->props.socket_id) { 848 continue; 849 } 850 851 /* reject assignment if slot is already assigned, for compatibility 852 * of legacy cpu_index mapping with SPAPR core based mapping do not 853 * error out if cpu thread and matched core have the same node-id */ 854 if (slot->props.has_node_id && 855 slot->props.node_id != props->node_id) { 856 error_setg(errp, "CPU is already assigned to node-id: %" PRId64, 857 slot->props.node_id); 858 return; 859 } 860 861 /* assign slot to node as it's matched '-numa cpu' key */ 862 match = true; 863 slot->props.node_id = props->node_id; 864 slot->props.has_node_id = props->has_node_id; 865 866 if (machine->numa_state->hmat_enabled) { 867 if ((numa_info[props->node_id].initiator < MAX_NODES) && 868 (props->node_id != numa_info[props->node_id].initiator)) { 869 error_setg(errp, "The initiator of CPU NUMA node %" PRId64 870 " should be itself (got %" PRIu16 ")", 871 props->node_id, numa_info[props->node_id].initiator); 872 return; 873 } 874 numa_info[props->node_id].has_cpu = true; 875 numa_info[props->node_id].initiator = props->node_id; 876 } 877 } 878 879 if (!match) { 880 error_setg(errp, "no match found"); 881 } 882 } 883 884 static void machine_get_smp(Object *obj, Visitor *v, const char *name, 885 void *opaque, Error **errp) 886 { 887 MachineState *ms = MACHINE(obj); 888 SMPConfiguration *config = &(SMPConfiguration){ 889 .has_cpus = true, .cpus = ms->smp.cpus, 890 .has_drawers = true, .drawers = ms->smp.drawers, 891 .has_books = true, .books = ms->smp.books, 892 .has_sockets = true, .sockets = ms->smp.sockets, 893 .has_dies = true, .dies = ms->smp.dies, 894 .has_clusters = true, .clusters = ms->smp.clusters, 895 .has_modules = true, .modules = ms->smp.modules, 896 .has_cores = true, .cores = ms->smp.cores, 897 .has_threads = true, .threads = ms->smp.threads, 898 .has_maxcpus = true, .maxcpus = ms->smp.max_cpus, 899 }; 900 901 if (!visit_type_SMPConfiguration(v, name, &config, &error_abort)) { 902 return; 903 } 904 } 905 906 static void machine_set_smp(Object *obj, Visitor *v, const char *name, 907 void *opaque, Error **errp) 908 { 909 MachineState *ms = MACHINE(obj); 910 g_autoptr(SMPConfiguration) config = NULL; 911 912 if (!visit_type_SMPConfiguration(v, name, &config, errp)) { 913 return; 914 } 915 916 machine_parse_smp_config(ms, config, errp); 917 } 918 919 static void machine_get_boot(Object *obj, Visitor *v, const char *name, 920 void *opaque, Error **errp) 921 { 922 MachineState *ms = MACHINE(obj); 923 BootConfiguration *config = &ms->boot_config; 924 visit_type_BootConfiguration(v, name, &config, &error_abort); 925 } 926 927 static void machine_free_boot_config(MachineState *ms) 928 { 929 g_free(ms->boot_config.order); 930 g_free(ms->boot_config.once); 931 g_free(ms->boot_config.splash); 932 } 933 934 static void machine_copy_boot_config(MachineState *ms, BootConfiguration *config) 935 { 936 MachineClass *machine_class = MACHINE_GET_CLASS(ms); 937 938 machine_free_boot_config(ms); 939 ms->boot_config = *config; 940 if (!config->order) { 941 ms->boot_config.order = g_strdup(machine_class->default_boot_order); 942 } 943 } 944 945 static void machine_set_boot(Object *obj, Visitor *v, const char *name, 946 void *opaque, Error **errp) 947 { 948 ERRP_GUARD(); 949 MachineState *ms = MACHINE(obj); 950 BootConfiguration *config = NULL; 951 952 if (!visit_type_BootConfiguration(v, name, &config, errp)) { 953 return; 954 } 955 if (config->order) { 956 validate_bootdevices(config->order, errp); 957 if (*errp) { 958 goto out_free; 959 } 960 } 961 if (config->once) { 962 validate_bootdevices(config->once, errp); 963 if (*errp) { 964 goto out_free; 965 } 966 } 967 968 machine_copy_boot_config(ms, config); 969 /* Strings live in ms->boot_config. */ 970 free(config); 971 return; 972 973 out_free: 974 qapi_free_BootConfiguration(config); 975 } 976 977 void machine_add_audiodev_property(MachineClass *mc) 978 { 979 ObjectClass *oc = OBJECT_CLASS(mc); 980 981 object_class_property_add_str(oc, "audiodev", 982 machine_get_audiodev, 983 machine_set_audiodev); 984 object_class_property_set_description(oc, "audiodev", 985 "Audiodev to use for default machine devices"); 986 } 987 988 static void machine_class_init(ObjectClass *oc, void *data) 989 { 990 MachineClass *mc = MACHINE_CLASS(oc); 991 992 /* Default 128 MB as guest ram size */ 993 mc->default_ram_size = 128 * MiB; 994 mc->rom_file_has_mr = true; 995 996 /* numa node memory size aligned on 8MB by default. 997 * On Linux, each node's border has to be 8MB aligned 998 */ 999 mc->numa_mem_align_shift = 23; 1000 1001 object_class_property_add_str(oc, "kernel", 1002 machine_get_kernel, machine_set_kernel); 1003 object_class_property_set_description(oc, "kernel", 1004 "Linux kernel image file"); 1005 1006 object_class_property_add_str(oc, "initrd", 1007 machine_get_initrd, machine_set_initrd); 1008 object_class_property_set_description(oc, "initrd", 1009 "Linux initial ramdisk file"); 1010 1011 object_class_property_add_str(oc, "append", 1012 machine_get_append, machine_set_append); 1013 object_class_property_set_description(oc, "append", 1014 "Linux kernel command line"); 1015 1016 object_class_property_add_str(oc, "dtb", 1017 machine_get_dtb, machine_set_dtb); 1018 object_class_property_set_description(oc, "dtb", 1019 "Linux kernel device tree file"); 1020 1021 object_class_property_add_str(oc, "dumpdtb", 1022 machine_get_dumpdtb, machine_set_dumpdtb); 1023 object_class_property_set_description(oc, "dumpdtb", 1024 "Dump current dtb to a file and quit"); 1025 1026 object_class_property_add(oc, "boot", "BootConfiguration", 1027 machine_get_boot, machine_set_boot, 1028 NULL, NULL); 1029 object_class_property_set_description(oc, "boot", 1030 "Boot configuration"); 1031 1032 object_class_property_add(oc, "smp", "SMPConfiguration", 1033 machine_get_smp, machine_set_smp, 1034 NULL, NULL); 1035 object_class_property_set_description(oc, "smp", 1036 "CPU topology"); 1037 1038 object_class_property_add(oc, "phandle-start", "int", 1039 machine_get_phandle_start, machine_set_phandle_start, 1040 NULL, NULL); 1041 object_class_property_set_description(oc, "phandle-start", 1042 "The first phandle ID we may generate dynamically"); 1043 1044 object_class_property_add_str(oc, "dt-compatible", 1045 machine_get_dt_compatible, machine_set_dt_compatible); 1046 object_class_property_set_description(oc, "dt-compatible", 1047 "Overrides the \"compatible\" property of the dt root node"); 1048 1049 object_class_property_add_bool(oc, "dump-guest-core", 1050 machine_get_dump_guest_core, machine_set_dump_guest_core); 1051 object_class_property_set_description(oc, "dump-guest-core", 1052 "Include guest memory in a core dump"); 1053 1054 object_class_property_add_bool(oc, "mem-merge", 1055 machine_get_mem_merge, machine_set_mem_merge); 1056 object_class_property_set_description(oc, "mem-merge", 1057 "Enable/disable memory merge support"); 1058 1059 object_class_property_add_bool(oc, "usb", 1060 machine_get_usb, machine_set_usb); 1061 object_class_property_set_description(oc, "usb", 1062 "Set on/off to enable/disable usb"); 1063 1064 object_class_property_add_bool(oc, "graphics", 1065 machine_get_graphics, machine_set_graphics); 1066 object_class_property_set_description(oc, "graphics", 1067 "Set on/off to enable/disable graphics emulation"); 1068 1069 object_class_property_add_str(oc, "firmware", 1070 machine_get_firmware, machine_set_firmware); 1071 object_class_property_set_description(oc, "firmware", 1072 "Firmware image"); 1073 1074 object_class_property_add_bool(oc, "suppress-vmdesc", 1075 machine_get_suppress_vmdesc, machine_set_suppress_vmdesc); 1076 object_class_property_set_description(oc, "suppress-vmdesc", 1077 "Set on to disable self-describing migration"); 1078 1079 object_class_property_add_link(oc, "confidential-guest-support", 1080 TYPE_CONFIDENTIAL_GUEST_SUPPORT, 1081 offsetof(MachineState, cgs), 1082 machine_check_confidential_guest_support, 1083 OBJ_PROP_LINK_STRONG); 1084 object_class_property_set_description(oc, "confidential-guest-support", 1085 "Set confidential guest scheme to support"); 1086 1087 /* For compatibility */ 1088 object_class_property_add_str(oc, "memory-encryption", 1089 machine_get_memory_encryption, machine_set_memory_encryption); 1090 object_class_property_set_description(oc, "memory-encryption", 1091 "Set memory encryption object to use"); 1092 1093 object_class_property_add_link(oc, "memory-backend", TYPE_MEMORY_BACKEND, 1094 offsetof(MachineState, memdev), object_property_allow_set_link, 1095 OBJ_PROP_LINK_STRONG); 1096 object_class_property_set_description(oc, "memory-backend", 1097 "Set RAM backend" 1098 "Valid value is ID of hostmem based backend"); 1099 1100 object_class_property_add(oc, "memory", "MemorySizeConfiguration", 1101 machine_get_mem, machine_set_mem, 1102 NULL, NULL); 1103 object_class_property_set_description(oc, "memory", 1104 "Memory size configuration"); 1105 } 1106 1107 static void machine_class_base_init(ObjectClass *oc, void *data) 1108 { 1109 MachineClass *mc = MACHINE_CLASS(oc); 1110 mc->max_cpus = mc->max_cpus ?: 1; 1111 mc->min_cpus = mc->min_cpus ?: 1; 1112 mc->default_cpus = mc->default_cpus ?: 1; 1113 1114 if (!object_class_is_abstract(oc)) { 1115 const char *cname = object_class_get_name(oc); 1116 assert(g_str_has_suffix(cname, TYPE_MACHINE_SUFFIX)); 1117 mc->name = g_strndup(cname, 1118 strlen(cname) - strlen(TYPE_MACHINE_SUFFIX)); 1119 mc->compat_props = g_ptr_array_new(); 1120 } 1121 } 1122 1123 static void machine_initfn(Object *obj) 1124 { 1125 MachineState *ms = MACHINE(obj); 1126 MachineClass *mc = MACHINE_GET_CLASS(obj); 1127 1128 container_get(obj, "/peripheral"); 1129 container_get(obj, "/peripheral-anon"); 1130 1131 ms->dump_guest_core = true; 1132 ms->mem_merge = true; 1133 ms->enable_graphics = true; 1134 ms->kernel_cmdline = g_strdup(""); 1135 ms->ram_size = mc->default_ram_size; 1136 ms->maxram_size = mc->default_ram_size; 1137 1138 if (mc->nvdimm_supported) { 1139 ms->nvdimms_state = g_new0(NVDIMMState, 1); 1140 object_property_add_bool(obj, "nvdimm", 1141 machine_get_nvdimm, machine_set_nvdimm); 1142 object_property_set_description(obj, "nvdimm", 1143 "Set on/off to enable/disable " 1144 "NVDIMM instantiation"); 1145 1146 object_property_add_str(obj, "nvdimm-persistence", 1147 machine_get_nvdimm_persistence, 1148 machine_set_nvdimm_persistence); 1149 object_property_set_description(obj, "nvdimm-persistence", 1150 "Set NVDIMM persistence" 1151 "Valid values are cpu, mem-ctrl"); 1152 } 1153 1154 if (mc->cpu_index_to_instance_props && mc->get_default_cpu_node_id) { 1155 ms->numa_state = g_new0(NumaState, 1); 1156 object_property_add_bool(obj, "hmat", 1157 machine_get_hmat, machine_set_hmat); 1158 object_property_set_description(obj, "hmat", 1159 "Set on/off to enable/disable " 1160 "ACPI Heterogeneous Memory Attribute " 1161 "Table (HMAT)"); 1162 } 1163 1164 /* default to mc->default_cpus */ 1165 ms->smp.cpus = mc->default_cpus; 1166 ms->smp.max_cpus = mc->default_cpus; 1167 ms->smp.drawers = 1; 1168 ms->smp.books = 1; 1169 ms->smp.sockets = 1; 1170 ms->smp.dies = 1; 1171 ms->smp.clusters = 1; 1172 ms->smp.modules = 1; 1173 ms->smp.cores = 1; 1174 ms->smp.threads = 1; 1175 1176 machine_copy_boot_config(ms, &(BootConfiguration){ 0 }); 1177 } 1178 1179 static void machine_finalize(Object *obj) 1180 { 1181 MachineState *ms = MACHINE(obj); 1182 1183 machine_free_boot_config(ms); 1184 g_free(ms->kernel_filename); 1185 g_free(ms->initrd_filename); 1186 g_free(ms->kernel_cmdline); 1187 g_free(ms->dtb); 1188 g_free(ms->dumpdtb); 1189 g_free(ms->dt_compatible); 1190 g_free(ms->firmware); 1191 g_free(ms->device_memory); 1192 g_free(ms->nvdimms_state); 1193 g_free(ms->numa_state); 1194 g_free(ms->audiodev); 1195 } 1196 1197 bool machine_usb(MachineState *machine) 1198 { 1199 return machine->usb; 1200 } 1201 1202 int machine_phandle_start(MachineState *machine) 1203 { 1204 return machine->phandle_start; 1205 } 1206 1207 bool machine_dump_guest_core(MachineState *machine) 1208 { 1209 return machine->dump_guest_core; 1210 } 1211 1212 bool machine_mem_merge(MachineState *machine) 1213 { 1214 return machine->mem_merge; 1215 } 1216 1217 bool machine_require_guest_memfd(MachineState *machine) 1218 { 1219 return machine->require_guest_memfd; 1220 } 1221 1222 static char *cpu_slot_to_string(const CPUArchId *cpu) 1223 { 1224 GString *s = g_string_new(NULL); 1225 if (cpu->props.has_socket_id) { 1226 g_string_append_printf(s, "socket-id: %"PRId64, cpu->props.socket_id); 1227 } 1228 if (cpu->props.has_die_id) { 1229 if (s->len) { 1230 g_string_append_printf(s, ", "); 1231 } 1232 g_string_append_printf(s, "die-id: %"PRId64, cpu->props.die_id); 1233 } 1234 if (cpu->props.has_cluster_id) { 1235 if (s->len) { 1236 g_string_append_printf(s, ", "); 1237 } 1238 g_string_append_printf(s, "cluster-id: %"PRId64, cpu->props.cluster_id); 1239 } 1240 if (cpu->props.has_module_id) { 1241 if (s->len) { 1242 g_string_append_printf(s, ", "); 1243 } 1244 g_string_append_printf(s, "module-id: %"PRId64, cpu->props.module_id); 1245 } 1246 if (cpu->props.has_core_id) { 1247 if (s->len) { 1248 g_string_append_printf(s, ", "); 1249 } 1250 g_string_append_printf(s, "core-id: %"PRId64, cpu->props.core_id); 1251 } 1252 if (cpu->props.has_thread_id) { 1253 if (s->len) { 1254 g_string_append_printf(s, ", "); 1255 } 1256 g_string_append_printf(s, "thread-id: %"PRId64, cpu->props.thread_id); 1257 } 1258 return g_string_free(s, false); 1259 } 1260 1261 static void numa_validate_initiator(NumaState *numa_state) 1262 { 1263 int i; 1264 NodeInfo *numa_info = numa_state->nodes; 1265 1266 for (i = 0; i < numa_state->num_nodes; i++) { 1267 if (numa_info[i].initiator == MAX_NODES) { 1268 continue; 1269 } 1270 1271 if (!numa_info[numa_info[i].initiator].present) { 1272 error_report("NUMA node %" PRIu16 " is missing, use " 1273 "'-numa node' option to declare it first", 1274 numa_info[i].initiator); 1275 exit(1); 1276 } 1277 1278 if (!numa_info[numa_info[i].initiator].has_cpu) { 1279 error_report("The initiator of NUMA node %d is invalid", i); 1280 exit(1); 1281 } 1282 } 1283 } 1284 1285 static void machine_numa_finish_cpu_init(MachineState *machine) 1286 { 1287 int i; 1288 bool default_mapping; 1289 GString *s = g_string_new(NULL); 1290 MachineClass *mc = MACHINE_GET_CLASS(machine); 1291 const CPUArchIdList *possible_cpus = mc->possible_cpu_arch_ids(machine); 1292 1293 assert(machine->numa_state->num_nodes); 1294 for (i = 0; i < possible_cpus->len; i++) { 1295 if (possible_cpus->cpus[i].props.has_node_id) { 1296 break; 1297 } 1298 } 1299 default_mapping = (i == possible_cpus->len); 1300 1301 for (i = 0; i < possible_cpus->len; i++) { 1302 const CPUArchId *cpu_slot = &possible_cpus->cpus[i]; 1303 1304 if (!cpu_slot->props.has_node_id) { 1305 /* fetch default mapping from board and enable it */ 1306 CpuInstanceProperties props = cpu_slot->props; 1307 1308 props.node_id = mc->get_default_cpu_node_id(machine, i); 1309 if (!default_mapping) { 1310 /* record slots with not set mapping, 1311 * TODO: make it hard error in future */ 1312 char *cpu_str = cpu_slot_to_string(cpu_slot); 1313 g_string_append_printf(s, "%sCPU %d [%s]", 1314 s->len ? ", " : "", i, cpu_str); 1315 g_free(cpu_str); 1316 1317 /* non mapped cpus used to fallback to node 0 */ 1318 props.node_id = 0; 1319 } 1320 1321 props.has_node_id = true; 1322 machine_set_cpu_numa_node(machine, &props, &error_fatal); 1323 } 1324 } 1325 1326 if (machine->numa_state->hmat_enabled) { 1327 numa_validate_initiator(machine->numa_state); 1328 } 1329 1330 if (s->len && !qtest_enabled()) { 1331 warn_report("CPU(s) not present in any NUMA nodes: %s", 1332 s->str); 1333 warn_report("All CPU(s) up to maxcpus should be described " 1334 "in NUMA config, ability to start up with partial NUMA " 1335 "mappings is obsoleted and will be removed in future"); 1336 } 1337 g_string_free(s, true); 1338 } 1339 1340 static void validate_cpu_cluster_to_numa_boundary(MachineState *ms) 1341 { 1342 MachineClass *mc = MACHINE_GET_CLASS(ms); 1343 NumaState *state = ms->numa_state; 1344 const CPUArchIdList *possible_cpus = mc->possible_cpu_arch_ids(ms); 1345 const CPUArchId *cpus = possible_cpus->cpus; 1346 int i, j; 1347 1348 if (qtest_enabled() || state->num_nodes <= 1 || possible_cpus->len <= 1) { 1349 return; 1350 } 1351 1352 /* 1353 * The Linux scheduling domain can't be parsed when the multiple CPUs 1354 * in one cluster have been associated with different NUMA nodes. However, 1355 * it's fine to associate one NUMA node with CPUs in different clusters. 1356 */ 1357 for (i = 0; i < possible_cpus->len; i++) { 1358 for (j = i + 1; j < possible_cpus->len; j++) { 1359 if (cpus[i].props.has_socket_id && 1360 cpus[i].props.has_cluster_id && 1361 cpus[i].props.has_node_id && 1362 cpus[j].props.has_socket_id && 1363 cpus[j].props.has_cluster_id && 1364 cpus[j].props.has_node_id && 1365 cpus[i].props.socket_id == cpus[j].props.socket_id && 1366 cpus[i].props.cluster_id == cpus[j].props.cluster_id && 1367 cpus[i].props.node_id != cpus[j].props.node_id) { 1368 warn_report("CPU-%d and CPU-%d in socket-%" PRId64 "-cluster-%" PRId64 1369 " have been associated with node-%" PRId64 " and node-%" PRId64 1370 " respectively. It can cause OSes like Linux to" 1371 " misbehave", i, j, cpus[i].props.socket_id, 1372 cpus[i].props.cluster_id, cpus[i].props.node_id, 1373 cpus[j].props.node_id); 1374 } 1375 } 1376 } 1377 } 1378 1379 MemoryRegion *machine_consume_memdev(MachineState *machine, 1380 HostMemoryBackend *backend) 1381 { 1382 MemoryRegion *ret = host_memory_backend_get_memory(backend); 1383 1384 if (host_memory_backend_is_mapped(backend)) { 1385 error_report("memory backend %s can't be used multiple times.", 1386 object_get_canonical_path_component(OBJECT(backend))); 1387 exit(EXIT_FAILURE); 1388 } 1389 host_memory_backend_set_mapped(backend, true); 1390 vmstate_register_ram_global(ret); 1391 return ret; 1392 } 1393 1394 static bool create_default_memdev(MachineState *ms, const char *path, Error **errp) 1395 { 1396 Object *obj; 1397 MachineClass *mc = MACHINE_GET_CLASS(ms); 1398 bool r = false; 1399 1400 obj = object_new(path ? TYPE_MEMORY_BACKEND_FILE : TYPE_MEMORY_BACKEND_RAM); 1401 if (path) { 1402 if (!object_property_set_str(obj, "mem-path", path, errp)) { 1403 goto out; 1404 } 1405 } 1406 if (!object_property_set_int(obj, "size", ms->ram_size, errp)) { 1407 goto out; 1408 } 1409 object_property_add_child(object_get_objects_root(), mc->default_ram_id, 1410 obj); 1411 /* Ensure backend's memory region name is equal to mc->default_ram_id */ 1412 if (!object_property_set_bool(obj, "x-use-canonical-path-for-ramblock-id", 1413 false, errp)) { 1414 goto out; 1415 } 1416 if (!user_creatable_complete(USER_CREATABLE(obj), errp)) { 1417 goto out; 1418 } 1419 r = object_property_set_link(OBJECT(ms), "memory-backend", obj, errp); 1420 1421 out: 1422 object_unref(obj); 1423 return r; 1424 } 1425 1426 const char *machine_class_default_cpu_type(MachineClass *mc) 1427 { 1428 if (mc->valid_cpu_types && !mc->valid_cpu_types[1]) { 1429 /* Only a single CPU type allowed: use it as default. */ 1430 return mc->valid_cpu_types[0]; 1431 } 1432 return mc->default_cpu_type; 1433 } 1434 1435 static bool is_cpu_type_supported(const MachineState *machine, Error **errp) 1436 { 1437 MachineClass *mc = MACHINE_GET_CLASS(machine); 1438 ObjectClass *oc = object_class_by_name(machine->cpu_type); 1439 CPUClass *cc; 1440 int i; 1441 1442 /* 1443 * Check if the user specified CPU type is supported when the valid 1444 * CPU types have been determined. Note that the user specified CPU 1445 * type is provided through '-cpu' option. 1446 */ 1447 if (mc->valid_cpu_types) { 1448 assert(mc->valid_cpu_types[0] != NULL); 1449 for (i = 0; mc->valid_cpu_types[i]; i++) { 1450 if (object_class_dynamic_cast(oc, mc->valid_cpu_types[i])) { 1451 break; 1452 } 1453 } 1454 1455 /* The user specified CPU type isn't valid */ 1456 if (!mc->valid_cpu_types[i]) { 1457 g_autofree char *requested = cpu_model_from_type(machine->cpu_type); 1458 error_setg(errp, "Invalid CPU model: %s", requested); 1459 if (!mc->valid_cpu_types[1]) { 1460 g_autofree char *model = cpu_model_from_type( 1461 mc->valid_cpu_types[0]); 1462 error_append_hint(errp, "The only valid type is: %s\n", model); 1463 } else { 1464 error_append_hint(errp, "The valid models are: "); 1465 for (i = 0; mc->valid_cpu_types[i]; i++) { 1466 g_autofree char *model = cpu_model_from_type( 1467 mc->valid_cpu_types[i]); 1468 error_append_hint(errp, "%s%s", 1469 model, 1470 mc->valid_cpu_types[i + 1] ? ", " : ""); 1471 } 1472 error_append_hint(errp, "\n"); 1473 } 1474 1475 return false; 1476 } 1477 } 1478 1479 /* Check if CPU type is deprecated and warn if so */ 1480 cc = CPU_CLASS(oc); 1481 assert(cc != NULL); 1482 if (cc->deprecation_note) { 1483 warn_report("CPU model %s is deprecated -- %s", 1484 machine->cpu_type, cc->deprecation_note); 1485 } 1486 1487 return true; 1488 } 1489 1490 void machine_run_board_init(MachineState *machine, const char *mem_path, Error **errp) 1491 { 1492 ERRP_GUARD(); 1493 MachineClass *machine_class = MACHINE_GET_CLASS(machine); 1494 1495 /* This checkpoint is required by replay to separate prior clock 1496 reading from the other reads, because timer polling functions query 1497 clock values from the log. */ 1498 replay_checkpoint(CHECKPOINT_INIT); 1499 1500 if (!xen_enabled()) { 1501 /* On 32-bit hosts, QEMU is limited by virtual address space */ 1502 if (machine->ram_size > (2047 << 20) && HOST_LONG_BITS == 32) { 1503 error_setg(errp, "at most 2047 MB RAM can be simulated"); 1504 return; 1505 } 1506 } 1507 1508 if (machine->memdev) { 1509 ram_addr_t backend_size = object_property_get_uint(OBJECT(machine->memdev), 1510 "size", &error_abort); 1511 if (backend_size != machine->ram_size) { 1512 error_setg(errp, "Machine memory size does not match the size of the memory backend"); 1513 return; 1514 } 1515 } else if (machine_class->default_ram_id && machine->ram_size && 1516 numa_uses_legacy_mem()) { 1517 if (object_property_find(object_get_objects_root(), 1518 machine_class->default_ram_id)) { 1519 error_setg(errp, "object's id '%s' is reserved for the default" 1520 " RAM backend, it can't be used for any other purposes", 1521 machine_class->default_ram_id); 1522 error_append_hint(errp, 1523 "Change the object's 'id' to something else or disable" 1524 " automatic creation of the default RAM backend by setting" 1525 " 'memory-backend=%s' with '-machine'.\n", 1526 machine_class->default_ram_id); 1527 return; 1528 } 1529 if (!create_default_memdev(current_machine, mem_path, errp)) { 1530 return; 1531 } 1532 } 1533 1534 if (machine->numa_state) { 1535 numa_complete_configuration(machine); 1536 if (machine->numa_state->num_nodes) { 1537 machine_numa_finish_cpu_init(machine); 1538 if (machine_class->cpu_cluster_has_numa_boundary) { 1539 validate_cpu_cluster_to_numa_boundary(machine); 1540 } 1541 } 1542 } 1543 1544 if (!machine->ram && machine->memdev) { 1545 machine->ram = machine_consume_memdev(machine, machine->memdev); 1546 } 1547 1548 /* Check if the CPU type is supported */ 1549 if (machine->cpu_type && !is_cpu_type_supported(machine, errp)) { 1550 return; 1551 } 1552 1553 if (machine->cgs) { 1554 /* 1555 * With confidential guests, the host can't see the real 1556 * contents of RAM, so there's no point in it trying to merge 1557 * areas. 1558 */ 1559 machine_set_mem_merge(OBJECT(machine), false, &error_abort); 1560 1561 /* 1562 * Virtio devices can't count on directly accessing guest 1563 * memory, so they need iommu_platform=on to use normal DMA 1564 * mechanisms. That requires also disabling legacy virtio 1565 * support for those virtio pci devices which allow it. 1566 */ 1567 object_register_sugar_prop(TYPE_VIRTIO_PCI, "disable-legacy", 1568 "on", true); 1569 object_register_sugar_prop(TYPE_VIRTIO_DEVICE, "iommu_platform", 1570 "on", false); 1571 } 1572 1573 accel_init_interfaces(ACCEL_GET_CLASS(machine->accelerator)); 1574 machine_class->init(machine); 1575 phase_advance(PHASE_MACHINE_INITIALIZED); 1576 } 1577 1578 static NotifierList machine_init_done_notifiers = 1579 NOTIFIER_LIST_INITIALIZER(machine_init_done_notifiers); 1580 1581 void qemu_add_machine_init_done_notifier(Notifier *notify) 1582 { 1583 notifier_list_add(&machine_init_done_notifiers, notify); 1584 if (phase_check(PHASE_MACHINE_READY)) { 1585 notify->notify(notify, NULL); 1586 } 1587 } 1588 1589 void qemu_remove_machine_init_done_notifier(Notifier *notify) 1590 { 1591 notifier_remove(notify); 1592 } 1593 1594 void qdev_machine_creation_done(void) 1595 { 1596 cpu_synchronize_all_post_init(); 1597 1598 if (current_machine->boot_config.once) { 1599 qemu_boot_set(current_machine->boot_config.once, &error_fatal); 1600 qemu_register_reset(restore_boot_order, g_strdup(current_machine->boot_config.order)); 1601 } 1602 1603 /* 1604 * ok, initial machine setup is done, starting from now we can 1605 * only create hotpluggable devices 1606 */ 1607 phase_advance(PHASE_MACHINE_READY); 1608 qdev_assert_realized_properly(); 1609 1610 /* TODO: once all bus devices are qdevified, this should be done 1611 * when bus is created by qdev.c */ 1612 /* 1613 * This is where we arrange for the sysbus to be reset when the 1614 * whole simulation is reset. In turn, resetting the sysbus will cause 1615 * all devices hanging off it (and all their child buses, recursively) 1616 * to be reset. Note that this will *not* reset any Device objects 1617 * which are not attached to some part of the qbus tree! 1618 */ 1619 qemu_register_resettable(OBJECT(sysbus_get_default())); 1620 1621 notifier_list_notify(&machine_init_done_notifiers, NULL); 1622 1623 if (rom_check_and_register_reset() != 0) { 1624 exit(1); 1625 } 1626 1627 replay_start(); 1628 1629 /* This checkpoint is required by replay to separate prior clock 1630 reading from the other reads, because timer polling functions query 1631 clock values from the log. */ 1632 replay_checkpoint(CHECKPOINT_RESET); 1633 qemu_system_reset(SHUTDOWN_CAUSE_NONE); 1634 register_global_state(); 1635 } 1636 1637 static const TypeInfo machine_info = { 1638 .name = TYPE_MACHINE, 1639 .parent = TYPE_OBJECT, 1640 .abstract = true, 1641 .class_size = sizeof(MachineClass), 1642 .class_init = machine_class_init, 1643 .class_base_init = machine_class_base_init, 1644 .instance_size = sizeof(MachineState), 1645 .instance_init = machine_initfn, 1646 .instance_finalize = machine_finalize, 1647 }; 1648 1649 static void machine_register_types(void) 1650 { 1651 type_register_static(&machine_info); 1652 } 1653 1654 type_init(machine_register_types) 1655