xref: /qemu/hw/core/machine.c (revision 79854b95)
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/option.h"
15 #include "qapi/qmp/qerror.h"
16 #include "sysemu/replay.h"
17 #include "qemu/units.h"
18 #include "hw/boards.h"
19 #include "hw/loader.h"
20 #include "qapi/error.h"
21 #include "qapi/qapi-visit-common.h"
22 #include "qapi/qapi-visit-machine.h"
23 #include "qapi/visitor.h"
24 #include "hw/sysbus.h"
25 #include "sysemu/cpus.h"
26 #include "sysemu/sysemu.h"
27 #include "sysemu/reset.h"
28 #include "sysemu/runstate.h"
29 #include "sysemu/numa.h"
30 #include "qemu/error-report.h"
31 #include "sysemu/qtest.h"
32 #include "hw/pci/pci.h"
33 #include "hw/mem/nvdimm.h"
34 #include "migration/global_state.h"
35 #include "migration/vmstate.h"
36 #include "exec/confidential-guest-support.h"
37 #include "hw/virtio/virtio.h"
38 #include "hw/virtio/virtio-pci.h"
39 
40 GlobalProperty hw_compat_6_2[] = {};
41 const size_t hw_compat_6_2_len = G_N_ELEMENTS(hw_compat_6_2);
42 
43 GlobalProperty hw_compat_6_1[] = {
44     { "vhost-user-vsock-device", "seqpacket", "off" },
45     { "nvme-ns", "shared", "off" },
46 };
47 const size_t hw_compat_6_1_len = G_N_ELEMENTS(hw_compat_6_1);
48 
49 GlobalProperty hw_compat_6_0[] = {
50     { "gpex-pcihost", "allow-unmapped-accesses", "false" },
51     { "i8042", "extended-state", "false"},
52     { "nvme-ns", "eui64-default", "off"},
53     { "e1000", "init-vet", "off" },
54     { "e1000e", "init-vet", "off" },
55     { "vhost-vsock-device", "seqpacket", "off" },
56 };
57 const size_t hw_compat_6_0_len = G_N_ELEMENTS(hw_compat_6_0);
58 
59 GlobalProperty hw_compat_5_2[] = {
60     { "ICH9-LPC", "smm-compat", "on"},
61     { "PIIX4_PM", "smm-compat", "on"},
62     { "virtio-blk-device", "report-discard-granularity", "off" },
63     { "virtio-net-pci-base", "vectors", "3"},
64 };
65 const size_t hw_compat_5_2_len = G_N_ELEMENTS(hw_compat_5_2);
66 
67 GlobalProperty hw_compat_5_1[] = {
68     { "vhost-scsi", "num_queues", "1"},
69     { "vhost-user-blk", "num-queues", "1"},
70     { "vhost-user-scsi", "num_queues", "1"},
71     { "virtio-blk-device", "num-queues", "1"},
72     { "virtio-scsi-device", "num_queues", "1"},
73     { "nvme", "use-intel-id", "on"},
74     { "pvpanic", "events", "1"}, /* PVPANIC_PANICKED */
75     { "pl011", "migrate-clk", "off" },
76     { "virtio-pci", "x-ats-page-aligned", "off"},
77 };
78 const size_t hw_compat_5_1_len = G_N_ELEMENTS(hw_compat_5_1);
79 
80 GlobalProperty hw_compat_5_0[] = {
81     { "pci-host-bridge", "x-config-reg-migration-enabled", "off" },
82     { "virtio-balloon-device", "page-poison", "false" },
83     { "vmport", "x-read-set-eax", "off" },
84     { "vmport", "x-signal-unsupported-cmd", "off" },
85     { "vmport", "x-report-vmx-type", "off" },
86     { "vmport", "x-cmds-v2", "off" },
87     { "virtio-device", "x-disable-legacy-check", "true" },
88 };
89 const size_t hw_compat_5_0_len = G_N_ELEMENTS(hw_compat_5_0);
90 
91 GlobalProperty hw_compat_4_2[] = {
92     { "virtio-blk-device", "queue-size", "128"},
93     { "virtio-scsi-device", "virtqueue_size", "128"},
94     { "virtio-blk-device", "x-enable-wce-if-config-wce", "off" },
95     { "virtio-blk-device", "seg-max-adjust", "off"},
96     { "virtio-scsi-device", "seg_max_adjust", "off"},
97     { "vhost-blk-device", "seg_max_adjust", "off"},
98     { "usb-host", "suppress-remote-wake", "off" },
99     { "usb-redir", "suppress-remote-wake", "off" },
100     { "qxl", "revision", "4" },
101     { "qxl-vga", "revision", "4" },
102     { "fw_cfg", "acpi-mr-restore", "false" },
103     { "virtio-device", "use-disabled-flag", "false" },
104 };
105 const size_t hw_compat_4_2_len = G_N_ELEMENTS(hw_compat_4_2);
106 
107 GlobalProperty hw_compat_4_1[] = {
108     { "virtio-pci", "x-pcie-flr-init", "off" },
109 };
110 const size_t hw_compat_4_1_len = G_N_ELEMENTS(hw_compat_4_1);
111 
112 GlobalProperty hw_compat_4_0[] = {
113     { "VGA",            "edid", "false" },
114     { "secondary-vga",  "edid", "false" },
115     { "bochs-display",  "edid", "false" },
116     { "virtio-vga",     "edid", "false" },
117     { "virtio-gpu-device", "edid", "false" },
118     { "virtio-device", "use-started", "false" },
119     { "virtio-balloon-device", "qemu-4-0-config-size", "true" },
120     { "pl031", "migrate-tick-offset", "false" },
121 };
122 const size_t hw_compat_4_0_len = G_N_ELEMENTS(hw_compat_4_0);
123 
124 GlobalProperty hw_compat_3_1[] = {
125     { "pcie-root-port", "x-speed", "2_5" },
126     { "pcie-root-port", "x-width", "1" },
127     { "memory-backend-file", "x-use-canonical-path-for-ramblock-id", "true" },
128     { "memory-backend-memfd", "x-use-canonical-path-for-ramblock-id", "true" },
129     { "tpm-crb", "ppi", "false" },
130     { "tpm-tis", "ppi", "false" },
131     { "usb-kbd", "serial", "42" },
132     { "usb-mouse", "serial", "42" },
133     { "usb-tablet", "serial", "42" },
134     { "virtio-blk-device", "discard", "false" },
135     { "virtio-blk-device", "write-zeroes", "false" },
136     { "virtio-balloon-device", "qemu-4-0-config-size", "false" },
137     { "pcie-root-port-base", "disable-acs", "true" }, /* Added in 4.1 */
138 };
139 const size_t hw_compat_3_1_len = G_N_ELEMENTS(hw_compat_3_1);
140 
141 GlobalProperty hw_compat_3_0[] = {};
142 const size_t hw_compat_3_0_len = G_N_ELEMENTS(hw_compat_3_0);
143 
144 GlobalProperty hw_compat_2_12[] = {
145     { "migration", "decompress-error-check", "off" },
146     { "hda-audio", "use-timer", "false" },
147     { "cirrus-vga", "global-vmstate", "true" },
148     { "VGA", "global-vmstate", "true" },
149     { "vmware-svga", "global-vmstate", "true" },
150     { "qxl-vga", "global-vmstate", "true" },
151 };
152 const size_t hw_compat_2_12_len = G_N_ELEMENTS(hw_compat_2_12);
153 
154 GlobalProperty hw_compat_2_11[] = {
155     { "hpet", "hpet-offset-saved", "false" },
156     { "virtio-blk-pci", "vectors", "2" },
157     { "vhost-user-blk-pci", "vectors", "2" },
158     { "e1000", "migrate_tso_props", "off" },
159 };
160 const size_t hw_compat_2_11_len = G_N_ELEMENTS(hw_compat_2_11);
161 
162 GlobalProperty hw_compat_2_10[] = {
163     { "virtio-mouse-device", "wheel-axis", "false" },
164     { "virtio-tablet-device", "wheel-axis", "false" },
165 };
166 const size_t hw_compat_2_10_len = G_N_ELEMENTS(hw_compat_2_10);
167 
168 GlobalProperty hw_compat_2_9[] = {
169     { "pci-bridge", "shpc", "off" },
170     { "intel-iommu", "pt", "off" },
171     { "virtio-net-device", "x-mtu-bypass-backend", "off" },
172     { "pcie-root-port", "x-migrate-msix", "false" },
173 };
174 const size_t hw_compat_2_9_len = G_N_ELEMENTS(hw_compat_2_9);
175 
176 GlobalProperty hw_compat_2_8[] = {
177     { "fw_cfg_mem", "x-file-slots", "0x10" },
178     { "fw_cfg_io", "x-file-slots", "0x10" },
179     { "pflash_cfi01", "old-multiple-chip-handling", "on" },
180     { "pci-bridge", "shpc", "on" },
181     { TYPE_PCI_DEVICE, "x-pcie-extcap-init", "off" },
182     { "virtio-pci", "x-pcie-deverr-init", "off" },
183     { "virtio-pci", "x-pcie-lnkctl-init", "off" },
184     { "virtio-pci", "x-pcie-pm-init", "off" },
185     { "cirrus-vga", "vgamem_mb", "8" },
186     { "isa-cirrus-vga", "vgamem_mb", "8" },
187 };
188 const size_t hw_compat_2_8_len = G_N_ELEMENTS(hw_compat_2_8);
189 
190 GlobalProperty hw_compat_2_7[] = {
191     { "virtio-pci", "page-per-vq", "on" },
192     { "virtio-serial-device", "emergency-write", "off" },
193     { "ioapic", "version", "0x11" },
194     { "intel-iommu", "x-buggy-eim", "true" },
195     { "virtio-pci", "x-ignore-backend-features", "on" },
196 };
197 const size_t hw_compat_2_7_len = G_N_ELEMENTS(hw_compat_2_7);
198 
199 GlobalProperty hw_compat_2_6[] = {
200     { "virtio-mmio", "format_transport_address", "off" },
201     /* Optional because not all virtio-pci devices support legacy mode */
202     { "virtio-pci", "disable-modern", "on",  .optional = true },
203     { "virtio-pci", "disable-legacy", "off", .optional = true },
204 };
205 const size_t hw_compat_2_6_len = G_N_ELEMENTS(hw_compat_2_6);
206 
207 GlobalProperty hw_compat_2_5[] = {
208     { "isa-fdc", "fallback", "144" },
209     { "pvscsi", "x-old-pci-configuration", "on" },
210     { "pvscsi", "x-disable-pcie", "on" },
211     { "vmxnet3", "x-old-msi-offsets", "on" },
212     { "vmxnet3", "x-disable-pcie", "on" },
213 };
214 const size_t hw_compat_2_5_len = G_N_ELEMENTS(hw_compat_2_5);
215 
216 GlobalProperty hw_compat_2_4[] = {
217     /* Optional because the 'scsi' property is Linux-only */
218     { "virtio-blk-device", "scsi", "true", .optional = true },
219     { "e1000", "extra_mac_registers", "off" },
220     { "virtio-pci", "x-disable-pcie", "on" },
221     { "virtio-pci", "migrate-extra", "off" },
222     { "fw_cfg_mem", "dma_enabled", "off" },
223     { "fw_cfg_io", "dma_enabled", "off" }
224 };
225 const size_t hw_compat_2_4_len = G_N_ELEMENTS(hw_compat_2_4);
226 
227 GlobalProperty hw_compat_2_3[] = {
228     { "virtio-blk-pci", "any_layout", "off" },
229     { "virtio-balloon-pci", "any_layout", "off" },
230     { "virtio-serial-pci", "any_layout", "off" },
231     { "virtio-9p-pci", "any_layout", "off" },
232     { "virtio-rng-pci", "any_layout", "off" },
233     { TYPE_PCI_DEVICE, "x-pcie-lnksta-dllla", "off" },
234     { "migration", "send-configuration", "off" },
235     { "migration", "send-section-footer", "off" },
236     { "migration", "store-global-state", "off" },
237 };
238 const size_t hw_compat_2_3_len = G_N_ELEMENTS(hw_compat_2_3);
239 
240 GlobalProperty hw_compat_2_2[] = {};
241 const size_t hw_compat_2_2_len = G_N_ELEMENTS(hw_compat_2_2);
242 
243 GlobalProperty hw_compat_2_1[] = {
244     { "intel-hda", "old_msi_addr", "on" },
245     { "VGA", "qemu-extended-regs", "off" },
246     { "secondary-vga", "qemu-extended-regs", "off" },
247     { "virtio-scsi-pci", "any_layout", "off" },
248     { "usb-mouse", "usb_version", "1" },
249     { "usb-kbd", "usb_version", "1" },
250     { "virtio-pci", "virtio-pci-bus-master-bug-migration", "on" },
251 };
252 const size_t hw_compat_2_1_len = G_N_ELEMENTS(hw_compat_2_1);
253 
254 MachineState *current_machine;
255 
256 static char *machine_get_kernel(Object *obj, Error **errp)
257 {
258     MachineState *ms = MACHINE(obj);
259 
260     return g_strdup(ms->kernel_filename);
261 }
262 
263 static void machine_set_kernel(Object *obj, const char *value, Error **errp)
264 {
265     MachineState *ms = MACHINE(obj);
266 
267     g_free(ms->kernel_filename);
268     ms->kernel_filename = g_strdup(value);
269 }
270 
271 static char *machine_get_initrd(Object *obj, Error **errp)
272 {
273     MachineState *ms = MACHINE(obj);
274 
275     return g_strdup(ms->initrd_filename);
276 }
277 
278 static void machine_set_initrd(Object *obj, const char *value, Error **errp)
279 {
280     MachineState *ms = MACHINE(obj);
281 
282     g_free(ms->initrd_filename);
283     ms->initrd_filename = g_strdup(value);
284 }
285 
286 static char *machine_get_append(Object *obj, Error **errp)
287 {
288     MachineState *ms = MACHINE(obj);
289 
290     return g_strdup(ms->kernel_cmdline);
291 }
292 
293 static void machine_set_append(Object *obj, const char *value, Error **errp)
294 {
295     MachineState *ms = MACHINE(obj);
296 
297     g_free(ms->kernel_cmdline);
298     ms->kernel_cmdline = g_strdup(value);
299 }
300 
301 static char *machine_get_dtb(Object *obj, Error **errp)
302 {
303     MachineState *ms = MACHINE(obj);
304 
305     return g_strdup(ms->dtb);
306 }
307 
308 static void machine_set_dtb(Object *obj, const char *value, Error **errp)
309 {
310     MachineState *ms = MACHINE(obj);
311 
312     g_free(ms->dtb);
313     ms->dtb = g_strdup(value);
314 }
315 
316 static char *machine_get_dumpdtb(Object *obj, Error **errp)
317 {
318     MachineState *ms = MACHINE(obj);
319 
320     return g_strdup(ms->dumpdtb);
321 }
322 
323 static void machine_set_dumpdtb(Object *obj, const char *value, Error **errp)
324 {
325     MachineState *ms = MACHINE(obj);
326 
327     g_free(ms->dumpdtb);
328     ms->dumpdtb = g_strdup(value);
329 }
330 
331 static void machine_get_phandle_start(Object *obj, Visitor *v,
332                                       const char *name, void *opaque,
333                                       Error **errp)
334 {
335     MachineState *ms = MACHINE(obj);
336     int64_t value = ms->phandle_start;
337 
338     visit_type_int(v, name, &value, errp);
339 }
340 
341 static void machine_set_phandle_start(Object *obj, Visitor *v,
342                                       const char *name, void *opaque,
343                                       Error **errp)
344 {
345     MachineState *ms = MACHINE(obj);
346     int64_t value;
347 
348     if (!visit_type_int(v, name, &value, errp)) {
349         return;
350     }
351 
352     ms->phandle_start = value;
353 }
354 
355 static char *machine_get_dt_compatible(Object *obj, Error **errp)
356 {
357     MachineState *ms = MACHINE(obj);
358 
359     return g_strdup(ms->dt_compatible);
360 }
361 
362 static void machine_set_dt_compatible(Object *obj, const char *value, Error **errp)
363 {
364     MachineState *ms = MACHINE(obj);
365 
366     g_free(ms->dt_compatible);
367     ms->dt_compatible = g_strdup(value);
368 }
369 
370 static bool machine_get_dump_guest_core(Object *obj, Error **errp)
371 {
372     MachineState *ms = MACHINE(obj);
373 
374     return ms->dump_guest_core;
375 }
376 
377 static void machine_set_dump_guest_core(Object *obj, bool value, Error **errp)
378 {
379     MachineState *ms = MACHINE(obj);
380 
381     ms->dump_guest_core = value;
382 }
383 
384 static bool machine_get_mem_merge(Object *obj, Error **errp)
385 {
386     MachineState *ms = MACHINE(obj);
387 
388     return ms->mem_merge;
389 }
390 
391 static void machine_set_mem_merge(Object *obj, bool value, Error **errp)
392 {
393     MachineState *ms = MACHINE(obj);
394 
395     ms->mem_merge = value;
396 }
397 
398 static bool machine_get_usb(Object *obj, Error **errp)
399 {
400     MachineState *ms = MACHINE(obj);
401 
402     return ms->usb;
403 }
404 
405 static void machine_set_usb(Object *obj, bool value, Error **errp)
406 {
407     MachineState *ms = MACHINE(obj);
408 
409     ms->usb = value;
410     ms->usb_disabled = !value;
411 }
412 
413 static bool machine_get_graphics(Object *obj, Error **errp)
414 {
415     MachineState *ms = MACHINE(obj);
416 
417     return ms->enable_graphics;
418 }
419 
420 static void machine_set_graphics(Object *obj, bool value, Error **errp)
421 {
422     MachineState *ms = MACHINE(obj);
423 
424     ms->enable_graphics = value;
425 }
426 
427 static char *machine_get_firmware(Object *obj, Error **errp)
428 {
429     MachineState *ms = MACHINE(obj);
430 
431     return g_strdup(ms->firmware);
432 }
433 
434 static void machine_set_firmware(Object *obj, const char *value, Error **errp)
435 {
436     MachineState *ms = MACHINE(obj);
437 
438     g_free(ms->firmware);
439     ms->firmware = g_strdup(value);
440 }
441 
442 static void machine_set_suppress_vmdesc(Object *obj, bool value, Error **errp)
443 {
444     MachineState *ms = MACHINE(obj);
445 
446     ms->suppress_vmdesc = value;
447 }
448 
449 static bool machine_get_suppress_vmdesc(Object *obj, Error **errp)
450 {
451     MachineState *ms = MACHINE(obj);
452 
453     return ms->suppress_vmdesc;
454 }
455 
456 static char *machine_get_memory_encryption(Object *obj, Error **errp)
457 {
458     MachineState *ms = MACHINE(obj);
459 
460     if (ms->cgs) {
461         return g_strdup(object_get_canonical_path_component(OBJECT(ms->cgs)));
462     }
463 
464     return NULL;
465 }
466 
467 static void machine_set_memory_encryption(Object *obj, const char *value,
468                                         Error **errp)
469 {
470     Object *cgs =
471         object_resolve_path_component(object_get_objects_root(), value);
472 
473     if (!cgs) {
474         error_setg(errp, "No such memory encryption object '%s'", value);
475         return;
476     }
477 
478     object_property_set_link(obj, "confidential-guest-support", cgs, errp);
479 }
480 
481 static void machine_check_confidential_guest_support(const Object *obj,
482                                                      const char *name,
483                                                      Object *new_target,
484                                                      Error **errp)
485 {
486     /*
487      * So far the only constraint is that the target has the
488      * TYPE_CONFIDENTIAL_GUEST_SUPPORT interface, and that's checked
489      * by the QOM core
490      */
491 }
492 
493 static bool machine_get_nvdimm(Object *obj, Error **errp)
494 {
495     MachineState *ms = MACHINE(obj);
496 
497     return ms->nvdimms_state->is_enabled;
498 }
499 
500 static void machine_set_nvdimm(Object *obj, bool value, Error **errp)
501 {
502     MachineState *ms = MACHINE(obj);
503 
504     ms->nvdimms_state->is_enabled = value;
505 }
506 
507 static bool machine_get_hmat(Object *obj, Error **errp)
508 {
509     MachineState *ms = MACHINE(obj);
510 
511     return ms->numa_state->hmat_enabled;
512 }
513 
514 static void machine_set_hmat(Object *obj, bool value, Error **errp)
515 {
516     MachineState *ms = MACHINE(obj);
517 
518     ms->numa_state->hmat_enabled = value;
519 }
520 
521 static char *machine_get_nvdimm_persistence(Object *obj, Error **errp)
522 {
523     MachineState *ms = MACHINE(obj);
524 
525     return g_strdup(ms->nvdimms_state->persistence_string);
526 }
527 
528 static void machine_set_nvdimm_persistence(Object *obj, const char *value,
529                                            Error **errp)
530 {
531     MachineState *ms = MACHINE(obj);
532     NVDIMMState *nvdimms_state = ms->nvdimms_state;
533 
534     if (strcmp(value, "cpu") == 0) {
535         nvdimms_state->persistence = 3;
536     } else if (strcmp(value, "mem-ctrl") == 0) {
537         nvdimms_state->persistence = 2;
538     } else {
539         error_setg(errp, "-machine nvdimm-persistence=%s: unsupported option",
540                    value);
541         return;
542     }
543 
544     g_free(nvdimms_state->persistence_string);
545     nvdimms_state->persistence_string = g_strdup(value);
546 }
547 
548 void machine_class_allow_dynamic_sysbus_dev(MachineClass *mc, const char *type)
549 {
550     QAPI_LIST_PREPEND(mc->allowed_dynamic_sysbus_devices, g_strdup(type));
551 }
552 
553 bool device_is_dynamic_sysbus(MachineClass *mc, DeviceState *dev)
554 {
555     Object *obj = OBJECT(dev);
556 
557     if (!object_dynamic_cast(obj, TYPE_SYS_BUS_DEVICE)) {
558         return false;
559     }
560 
561     return device_type_is_dynamic_sysbus(mc, object_get_typename(obj));
562 }
563 
564 bool device_type_is_dynamic_sysbus(MachineClass *mc, const char *type)
565 {
566     bool allowed = false;
567     strList *wl;
568     ObjectClass *klass = object_class_by_name(type);
569 
570     for (wl = mc->allowed_dynamic_sysbus_devices;
571          !allowed && wl;
572          wl = wl->next) {
573         allowed |= !!object_class_dynamic_cast(klass, wl->value);
574     }
575 
576     return allowed;
577 }
578 
579 static char *machine_get_memdev(Object *obj, Error **errp)
580 {
581     MachineState *ms = MACHINE(obj);
582 
583     return g_strdup(ms->ram_memdev_id);
584 }
585 
586 static void machine_set_memdev(Object *obj, const char *value, Error **errp)
587 {
588     MachineState *ms = MACHINE(obj);
589 
590     g_free(ms->ram_memdev_id);
591     ms->ram_memdev_id = g_strdup(value);
592 }
593 
594 HotpluggableCPUList *machine_query_hotpluggable_cpus(MachineState *machine)
595 {
596     int i;
597     HotpluggableCPUList *head = NULL;
598     MachineClass *mc = MACHINE_GET_CLASS(machine);
599 
600     /* force board to initialize possible_cpus if it hasn't been done yet */
601     mc->possible_cpu_arch_ids(machine);
602 
603     for (i = 0; i < machine->possible_cpus->len; i++) {
604         Object *cpu;
605         HotpluggableCPU *cpu_item = g_new0(typeof(*cpu_item), 1);
606 
607         cpu_item->type = g_strdup(machine->possible_cpus->cpus[i].type);
608         cpu_item->vcpus_count = machine->possible_cpus->cpus[i].vcpus_count;
609         cpu_item->props = g_memdup(&machine->possible_cpus->cpus[i].props,
610                                    sizeof(*cpu_item->props));
611 
612         cpu = machine->possible_cpus->cpus[i].cpu;
613         if (cpu) {
614             cpu_item->has_qom_path = true;
615             cpu_item->qom_path = object_get_canonical_path(cpu);
616         }
617         QAPI_LIST_PREPEND(head, cpu_item);
618     }
619     return head;
620 }
621 
622 /**
623  * machine_set_cpu_numa_node:
624  * @machine: machine object to modify
625  * @props: specifies which cpu objects to assign to
626  *         numa node specified by @props.node_id
627  * @errp: if an error occurs, a pointer to an area to store the error
628  *
629  * Associate NUMA node specified by @props.node_id with cpu slots that
630  * match socket/core/thread-ids specified by @props. It's recommended to use
631  * query-hotpluggable-cpus.props values to specify affected cpu slots,
632  * which would lead to exact 1:1 mapping of cpu slots to NUMA node.
633  *
634  * However for CLI convenience it's possible to pass in subset of properties,
635  * which would affect all cpu slots that match it.
636  * Ex for pc machine:
637  *    -smp 4,cores=2,sockets=2 -numa node,nodeid=0 -numa node,nodeid=1 \
638  *    -numa cpu,node-id=0,socket_id=0 \
639  *    -numa cpu,node-id=1,socket_id=1
640  * will assign all child cores of socket 0 to node 0 and
641  * of socket 1 to node 1.
642  *
643  * On attempt of reassigning (already assigned) cpu slot to another NUMA node,
644  * return error.
645  * Empty subset is disallowed and function will return with error in this case.
646  */
647 void machine_set_cpu_numa_node(MachineState *machine,
648                                const CpuInstanceProperties *props, Error **errp)
649 {
650     MachineClass *mc = MACHINE_GET_CLASS(machine);
651     NodeInfo *numa_info = machine->numa_state->nodes;
652     bool match = false;
653     int i;
654 
655     if (!mc->possible_cpu_arch_ids) {
656         error_setg(errp, "mapping of CPUs to NUMA node is not supported");
657         return;
658     }
659 
660     /* disabling node mapping is not supported, forbid it */
661     assert(props->has_node_id);
662 
663     /* force board to initialize possible_cpus if it hasn't been done yet */
664     mc->possible_cpu_arch_ids(machine);
665 
666     for (i = 0; i < machine->possible_cpus->len; i++) {
667         CPUArchId *slot = &machine->possible_cpus->cpus[i];
668 
669         /* reject unsupported by board properties */
670         if (props->has_thread_id && !slot->props.has_thread_id) {
671             error_setg(errp, "thread-id is not supported");
672             return;
673         }
674 
675         if (props->has_core_id && !slot->props.has_core_id) {
676             error_setg(errp, "core-id is not supported");
677             return;
678         }
679 
680         if (props->has_socket_id && !slot->props.has_socket_id) {
681             error_setg(errp, "socket-id is not supported");
682             return;
683         }
684 
685         if (props->has_die_id && !slot->props.has_die_id) {
686             error_setg(errp, "die-id is not supported");
687             return;
688         }
689 
690         /* skip slots with explicit mismatch */
691         if (props->has_thread_id && props->thread_id != slot->props.thread_id) {
692                 continue;
693         }
694 
695         if (props->has_core_id && props->core_id != slot->props.core_id) {
696                 continue;
697         }
698 
699         if (props->has_die_id && props->die_id != slot->props.die_id) {
700                 continue;
701         }
702 
703         if (props->has_socket_id && props->socket_id != slot->props.socket_id) {
704                 continue;
705         }
706 
707         /* reject assignment if slot is already assigned, for compatibility
708          * of legacy cpu_index mapping with SPAPR core based mapping do not
709          * error out if cpu thread and matched core have the same node-id */
710         if (slot->props.has_node_id &&
711             slot->props.node_id != props->node_id) {
712             error_setg(errp, "CPU is already assigned to node-id: %" PRId64,
713                        slot->props.node_id);
714             return;
715         }
716 
717         /* assign slot to node as it's matched '-numa cpu' key */
718         match = true;
719         slot->props.node_id = props->node_id;
720         slot->props.has_node_id = props->has_node_id;
721 
722         if (machine->numa_state->hmat_enabled) {
723             if ((numa_info[props->node_id].initiator < MAX_NODES) &&
724                 (props->node_id != numa_info[props->node_id].initiator)) {
725                 error_setg(errp, "The initiator of CPU NUMA node %" PRId64
726                            " should be itself (got %" PRIu16 ")",
727                            props->node_id, numa_info[props->node_id].initiator);
728                 return;
729             }
730             numa_info[props->node_id].has_cpu = true;
731             numa_info[props->node_id].initiator = props->node_id;
732         }
733     }
734 
735     if (!match) {
736         error_setg(errp, "no match found");
737     }
738 }
739 
740 static void machine_get_smp(Object *obj, Visitor *v, const char *name,
741                             void *opaque, Error **errp)
742 {
743     MachineState *ms = MACHINE(obj);
744     SMPConfiguration *config = &(SMPConfiguration){
745         .has_cpus = true, .cpus = ms->smp.cpus,
746         .has_sockets = true, .sockets = ms->smp.sockets,
747         .has_dies = true, .dies = ms->smp.dies,
748         .has_clusters = true, .clusters = ms->smp.clusters,
749         .has_cores = true, .cores = ms->smp.cores,
750         .has_threads = true, .threads = ms->smp.threads,
751         .has_maxcpus = true, .maxcpus = ms->smp.max_cpus,
752     };
753 
754     if (!visit_type_SMPConfiguration(v, name, &config, &error_abort)) {
755         return;
756     }
757 }
758 
759 static void machine_set_smp(Object *obj, Visitor *v, const char *name,
760                             void *opaque, Error **errp)
761 {
762     MachineState *ms = MACHINE(obj);
763     g_autoptr(SMPConfiguration) config = NULL;
764 
765     if (!visit_type_SMPConfiguration(v, name, &config, errp)) {
766         return;
767     }
768 
769     machine_parse_smp_config(ms, config, errp);
770 }
771 
772 static void machine_class_init(ObjectClass *oc, void *data)
773 {
774     MachineClass *mc = MACHINE_CLASS(oc);
775 
776     /* Default 128 MB as guest ram size */
777     mc->default_ram_size = 128 * MiB;
778     mc->rom_file_has_mr = true;
779 
780     /* numa node memory size aligned on 8MB by default.
781      * On Linux, each node's border has to be 8MB aligned
782      */
783     mc->numa_mem_align_shift = 23;
784 
785     object_class_property_add_str(oc, "kernel",
786         machine_get_kernel, machine_set_kernel);
787     object_class_property_set_description(oc, "kernel",
788         "Linux kernel image file");
789 
790     object_class_property_add_str(oc, "initrd",
791         machine_get_initrd, machine_set_initrd);
792     object_class_property_set_description(oc, "initrd",
793         "Linux initial ramdisk file");
794 
795     object_class_property_add_str(oc, "append",
796         machine_get_append, machine_set_append);
797     object_class_property_set_description(oc, "append",
798         "Linux kernel command line");
799 
800     object_class_property_add_str(oc, "dtb",
801         machine_get_dtb, machine_set_dtb);
802     object_class_property_set_description(oc, "dtb",
803         "Linux kernel device tree file");
804 
805     object_class_property_add_str(oc, "dumpdtb",
806         machine_get_dumpdtb, machine_set_dumpdtb);
807     object_class_property_set_description(oc, "dumpdtb",
808         "Dump current dtb to a file and quit");
809 
810     object_class_property_add(oc, "smp", "SMPConfiguration",
811         machine_get_smp, machine_set_smp,
812         NULL, NULL);
813     object_class_property_set_description(oc, "smp",
814         "CPU topology");
815 
816     object_class_property_add(oc, "phandle-start", "int",
817         machine_get_phandle_start, machine_set_phandle_start,
818         NULL, NULL);
819     object_class_property_set_description(oc, "phandle-start",
820         "The first phandle ID we may generate dynamically");
821 
822     object_class_property_add_str(oc, "dt-compatible",
823         machine_get_dt_compatible, machine_set_dt_compatible);
824     object_class_property_set_description(oc, "dt-compatible",
825         "Overrides the \"compatible\" property of the dt root node");
826 
827     object_class_property_add_bool(oc, "dump-guest-core",
828         machine_get_dump_guest_core, machine_set_dump_guest_core);
829     object_class_property_set_description(oc, "dump-guest-core",
830         "Include guest memory in a core dump");
831 
832     object_class_property_add_bool(oc, "mem-merge",
833         machine_get_mem_merge, machine_set_mem_merge);
834     object_class_property_set_description(oc, "mem-merge",
835         "Enable/disable memory merge support");
836 
837     object_class_property_add_bool(oc, "usb",
838         machine_get_usb, machine_set_usb);
839     object_class_property_set_description(oc, "usb",
840         "Set on/off to enable/disable usb");
841 
842     object_class_property_add_bool(oc, "graphics",
843         machine_get_graphics, machine_set_graphics);
844     object_class_property_set_description(oc, "graphics",
845         "Set on/off to enable/disable graphics emulation");
846 
847     object_class_property_add_str(oc, "firmware",
848         machine_get_firmware, machine_set_firmware);
849     object_class_property_set_description(oc, "firmware",
850         "Firmware image");
851 
852     object_class_property_add_bool(oc, "suppress-vmdesc",
853         machine_get_suppress_vmdesc, machine_set_suppress_vmdesc);
854     object_class_property_set_description(oc, "suppress-vmdesc",
855         "Set on to disable self-describing migration");
856 
857     object_class_property_add_link(oc, "confidential-guest-support",
858                                    TYPE_CONFIDENTIAL_GUEST_SUPPORT,
859                                    offsetof(MachineState, cgs),
860                                    machine_check_confidential_guest_support,
861                                    OBJ_PROP_LINK_STRONG);
862     object_class_property_set_description(oc, "confidential-guest-support",
863                                           "Set confidential guest scheme to support");
864 
865     /* For compatibility */
866     object_class_property_add_str(oc, "memory-encryption",
867         machine_get_memory_encryption, machine_set_memory_encryption);
868     object_class_property_set_description(oc, "memory-encryption",
869         "Set memory encryption object to use");
870 
871     object_class_property_add_str(oc, "memory-backend",
872                                   machine_get_memdev, machine_set_memdev);
873     object_class_property_set_description(oc, "memory-backend",
874                                           "Set RAM backend"
875                                           "Valid value is ID of hostmem based backend");
876 }
877 
878 static void machine_class_base_init(ObjectClass *oc, void *data)
879 {
880     MachineClass *mc = MACHINE_CLASS(oc);
881     mc->max_cpus = mc->max_cpus ?: 1;
882     mc->min_cpus = mc->min_cpus ?: 1;
883     mc->default_cpus = mc->default_cpus ?: 1;
884 
885     if (!object_class_is_abstract(oc)) {
886         const char *cname = object_class_get_name(oc);
887         assert(g_str_has_suffix(cname, TYPE_MACHINE_SUFFIX));
888         mc->name = g_strndup(cname,
889                             strlen(cname) - strlen(TYPE_MACHINE_SUFFIX));
890         mc->compat_props = g_ptr_array_new();
891     }
892 }
893 
894 static void machine_initfn(Object *obj)
895 {
896     MachineState *ms = MACHINE(obj);
897     MachineClass *mc = MACHINE_GET_CLASS(obj);
898 
899     container_get(obj, "/peripheral");
900     container_get(obj, "/peripheral-anon");
901 
902     ms->dump_guest_core = true;
903     ms->mem_merge = true;
904     ms->enable_graphics = true;
905     ms->kernel_cmdline = g_strdup("");
906 
907     if (mc->nvdimm_supported) {
908         Object *obj = OBJECT(ms);
909 
910         ms->nvdimms_state = g_new0(NVDIMMState, 1);
911         object_property_add_bool(obj, "nvdimm",
912                                  machine_get_nvdimm, machine_set_nvdimm);
913         object_property_set_description(obj, "nvdimm",
914                                         "Set on/off to enable/disable "
915                                         "NVDIMM instantiation");
916 
917         object_property_add_str(obj, "nvdimm-persistence",
918                                 machine_get_nvdimm_persistence,
919                                 machine_set_nvdimm_persistence);
920         object_property_set_description(obj, "nvdimm-persistence",
921                                         "Set NVDIMM persistence"
922                                         "Valid values are cpu, mem-ctrl");
923     }
924 
925     if (mc->cpu_index_to_instance_props && mc->get_default_cpu_node_id) {
926         ms->numa_state = g_new0(NumaState, 1);
927         object_property_add_bool(obj, "hmat",
928                                  machine_get_hmat, machine_set_hmat);
929         object_property_set_description(obj, "hmat",
930                                         "Set on/off to enable/disable "
931                                         "ACPI Heterogeneous Memory Attribute "
932                                         "Table (HMAT)");
933     }
934 
935     /* default to mc->default_cpus */
936     ms->smp.cpus = mc->default_cpus;
937     ms->smp.max_cpus = mc->default_cpus;
938     ms->smp.sockets = 1;
939     ms->smp.dies = 1;
940     ms->smp.clusters = 1;
941     ms->smp.cores = 1;
942     ms->smp.threads = 1;
943 }
944 
945 static void machine_finalize(Object *obj)
946 {
947     MachineState *ms = MACHINE(obj);
948 
949     g_free(ms->kernel_filename);
950     g_free(ms->initrd_filename);
951     g_free(ms->kernel_cmdline);
952     g_free(ms->dtb);
953     g_free(ms->dumpdtb);
954     g_free(ms->dt_compatible);
955     g_free(ms->firmware);
956     g_free(ms->device_memory);
957     g_free(ms->nvdimms_state);
958     g_free(ms->numa_state);
959 }
960 
961 bool machine_usb(MachineState *machine)
962 {
963     return machine->usb;
964 }
965 
966 int machine_phandle_start(MachineState *machine)
967 {
968     return machine->phandle_start;
969 }
970 
971 bool machine_dump_guest_core(MachineState *machine)
972 {
973     return machine->dump_guest_core;
974 }
975 
976 bool machine_mem_merge(MachineState *machine)
977 {
978     return machine->mem_merge;
979 }
980 
981 static char *cpu_slot_to_string(const CPUArchId *cpu)
982 {
983     GString *s = g_string_new(NULL);
984     if (cpu->props.has_socket_id) {
985         g_string_append_printf(s, "socket-id: %"PRId64, cpu->props.socket_id);
986     }
987     if (cpu->props.has_die_id) {
988         if (s->len) {
989             g_string_append_printf(s, ", ");
990         }
991         g_string_append_printf(s, "die-id: %"PRId64, cpu->props.die_id);
992     }
993     if (cpu->props.has_core_id) {
994         if (s->len) {
995             g_string_append_printf(s, ", ");
996         }
997         g_string_append_printf(s, "core-id: %"PRId64, cpu->props.core_id);
998     }
999     if (cpu->props.has_thread_id) {
1000         if (s->len) {
1001             g_string_append_printf(s, ", ");
1002         }
1003         g_string_append_printf(s, "thread-id: %"PRId64, cpu->props.thread_id);
1004     }
1005     return g_string_free(s, false);
1006 }
1007 
1008 static void numa_validate_initiator(NumaState *numa_state)
1009 {
1010     int i;
1011     NodeInfo *numa_info = numa_state->nodes;
1012 
1013     for (i = 0; i < numa_state->num_nodes; i++) {
1014         if (numa_info[i].initiator == MAX_NODES) {
1015             error_report("The initiator of NUMA node %d is missing, use "
1016                          "'-numa node,initiator' option to declare it", i);
1017             exit(1);
1018         }
1019 
1020         if (!numa_info[numa_info[i].initiator].present) {
1021             error_report("NUMA node %" PRIu16 " is missing, use "
1022                          "'-numa node' option to declare it first",
1023                          numa_info[i].initiator);
1024             exit(1);
1025         }
1026 
1027         if (!numa_info[numa_info[i].initiator].has_cpu) {
1028             error_report("The initiator of NUMA node %d is invalid", i);
1029             exit(1);
1030         }
1031     }
1032 }
1033 
1034 static void machine_numa_finish_cpu_init(MachineState *machine)
1035 {
1036     int i;
1037     bool default_mapping;
1038     GString *s = g_string_new(NULL);
1039     MachineClass *mc = MACHINE_GET_CLASS(machine);
1040     const CPUArchIdList *possible_cpus = mc->possible_cpu_arch_ids(machine);
1041 
1042     assert(machine->numa_state->num_nodes);
1043     for (i = 0; i < possible_cpus->len; i++) {
1044         if (possible_cpus->cpus[i].props.has_node_id) {
1045             break;
1046         }
1047     }
1048     default_mapping = (i == possible_cpus->len);
1049 
1050     for (i = 0; i < possible_cpus->len; i++) {
1051         const CPUArchId *cpu_slot = &possible_cpus->cpus[i];
1052 
1053         if (!cpu_slot->props.has_node_id) {
1054             /* fetch default mapping from board and enable it */
1055             CpuInstanceProperties props = cpu_slot->props;
1056 
1057             props.node_id = mc->get_default_cpu_node_id(machine, i);
1058             if (!default_mapping) {
1059                 /* record slots with not set mapping,
1060                  * TODO: make it hard error in future */
1061                 char *cpu_str = cpu_slot_to_string(cpu_slot);
1062                 g_string_append_printf(s, "%sCPU %d [%s]",
1063                                        s->len ? ", " : "", i, cpu_str);
1064                 g_free(cpu_str);
1065 
1066                 /* non mapped cpus used to fallback to node 0 */
1067                 props.node_id = 0;
1068             }
1069 
1070             props.has_node_id = true;
1071             machine_set_cpu_numa_node(machine, &props, &error_fatal);
1072         }
1073     }
1074 
1075     if (machine->numa_state->hmat_enabled) {
1076         numa_validate_initiator(machine->numa_state);
1077     }
1078 
1079     if (s->len && !qtest_enabled()) {
1080         warn_report("CPU(s) not present in any NUMA nodes: %s",
1081                     s->str);
1082         warn_report("All CPU(s) up to maxcpus should be described "
1083                     "in NUMA config, ability to start up with partial NUMA "
1084                     "mappings is obsoleted and will be removed in future");
1085     }
1086     g_string_free(s, true);
1087 }
1088 
1089 MemoryRegion *machine_consume_memdev(MachineState *machine,
1090                                      HostMemoryBackend *backend)
1091 {
1092     MemoryRegion *ret = host_memory_backend_get_memory(backend);
1093 
1094     if (host_memory_backend_is_mapped(backend)) {
1095         error_report("memory backend %s can't be used multiple times.",
1096                      object_get_canonical_path_component(OBJECT(backend)));
1097         exit(EXIT_FAILURE);
1098     }
1099     host_memory_backend_set_mapped(backend, true);
1100     vmstate_register_ram_global(ret);
1101     return ret;
1102 }
1103 
1104 void machine_run_board_init(MachineState *machine)
1105 {
1106     MachineClass *machine_class = MACHINE_GET_CLASS(machine);
1107     ObjectClass *oc = object_class_by_name(machine->cpu_type);
1108     CPUClass *cc;
1109 
1110     /* This checkpoint is required by replay to separate prior clock
1111        reading from the other reads, because timer polling functions query
1112        clock values from the log. */
1113     replay_checkpoint(CHECKPOINT_INIT);
1114 
1115     if (machine->ram_memdev_id) {
1116         Object *o;
1117         o = object_resolve_path_type(machine->ram_memdev_id,
1118                                      TYPE_MEMORY_BACKEND, NULL);
1119         machine->ram = machine_consume_memdev(machine, MEMORY_BACKEND(o));
1120     }
1121 
1122     if (machine->numa_state) {
1123         numa_complete_configuration(machine);
1124         if (machine->numa_state->num_nodes) {
1125             machine_numa_finish_cpu_init(machine);
1126         }
1127     }
1128 
1129     /* If the machine supports the valid_cpu_types check and the user
1130      * specified a CPU with -cpu check here that the user CPU is supported.
1131      */
1132     if (machine_class->valid_cpu_types && machine->cpu_type) {
1133         int i;
1134 
1135         for (i = 0; machine_class->valid_cpu_types[i]; i++) {
1136             if (object_class_dynamic_cast(oc,
1137                                           machine_class->valid_cpu_types[i])) {
1138                 /* The user specificed CPU is in the valid field, we are
1139                  * good to go.
1140                  */
1141                 break;
1142             }
1143         }
1144 
1145         if (!machine_class->valid_cpu_types[i]) {
1146             /* The user specified CPU is not valid */
1147             error_report("Invalid CPU type: %s", machine->cpu_type);
1148             error_printf("The valid types are: %s",
1149                          machine_class->valid_cpu_types[0]);
1150             for (i = 1; machine_class->valid_cpu_types[i]; i++) {
1151                 error_printf(", %s", machine_class->valid_cpu_types[i]);
1152             }
1153             error_printf("\n");
1154 
1155             exit(1);
1156         }
1157     }
1158 
1159     /* Check if CPU type is deprecated and warn if so */
1160     cc = CPU_CLASS(oc);
1161     if (cc && cc->deprecation_note) {
1162         warn_report("CPU model %s is deprecated -- %s", machine->cpu_type,
1163                     cc->deprecation_note);
1164     }
1165 
1166     if (machine->cgs) {
1167         /*
1168          * With confidential guests, the host can't see the real
1169          * contents of RAM, so there's no point in it trying to merge
1170          * areas.
1171          */
1172         machine_set_mem_merge(OBJECT(machine), false, &error_abort);
1173 
1174         /*
1175          * Virtio devices can't count on directly accessing guest
1176          * memory, so they need iommu_platform=on to use normal DMA
1177          * mechanisms.  That requires also disabling legacy virtio
1178          * support for those virtio pci devices which allow it.
1179          */
1180         object_register_sugar_prop(TYPE_VIRTIO_PCI, "disable-legacy",
1181                                    "on", true);
1182         object_register_sugar_prop(TYPE_VIRTIO_DEVICE, "iommu_platform",
1183                                    "on", false);
1184     }
1185 
1186     accel_init_interfaces(ACCEL_GET_CLASS(machine->accelerator));
1187     machine_class->init(machine);
1188     phase_advance(PHASE_MACHINE_INITIALIZED);
1189 }
1190 
1191 static NotifierList machine_init_done_notifiers =
1192     NOTIFIER_LIST_INITIALIZER(machine_init_done_notifiers);
1193 
1194 void qemu_add_machine_init_done_notifier(Notifier *notify)
1195 {
1196     notifier_list_add(&machine_init_done_notifiers, notify);
1197     if (phase_check(PHASE_MACHINE_READY)) {
1198         notify->notify(notify, NULL);
1199     }
1200 }
1201 
1202 void qemu_remove_machine_init_done_notifier(Notifier *notify)
1203 {
1204     notifier_remove(notify);
1205 }
1206 
1207 void qdev_machine_creation_done(void)
1208 {
1209     cpu_synchronize_all_post_init();
1210 
1211     if (current_machine->boot_once) {
1212         qemu_boot_set(current_machine->boot_once, &error_fatal);
1213         qemu_register_reset(restore_boot_order, g_strdup(current_machine->boot_order));
1214     }
1215 
1216     /*
1217      * ok, initial machine setup is done, starting from now we can
1218      * only create hotpluggable devices
1219      */
1220     phase_advance(PHASE_MACHINE_READY);
1221     qdev_assert_realized_properly();
1222 
1223     /* TODO: once all bus devices are qdevified, this should be done
1224      * when bus is created by qdev.c */
1225     /*
1226      * TODO: If we had a main 'reset container' that the whole system
1227      * lived in, we could reset that using the multi-phase reset
1228      * APIs. For the moment, we just reset the sysbus, which will cause
1229      * all devices hanging off it (and all their child buses, recursively)
1230      * to be reset. Note that this will *not* reset any Device objects
1231      * which are not attached to some part of the qbus tree!
1232      */
1233     qemu_register_reset(resettable_cold_reset_fn, sysbus_get_default());
1234 
1235     notifier_list_notify(&machine_init_done_notifiers, NULL);
1236 
1237     if (rom_check_and_register_reset() != 0) {
1238         exit(1);
1239     }
1240 
1241     replay_start();
1242 
1243     /* This checkpoint is required by replay to separate prior clock
1244        reading from the other reads, because timer polling functions query
1245        clock values from the log. */
1246     replay_checkpoint(CHECKPOINT_RESET);
1247     qemu_system_reset(SHUTDOWN_CAUSE_NONE);
1248     register_global_state();
1249 }
1250 
1251 static const TypeInfo machine_info = {
1252     .name = TYPE_MACHINE,
1253     .parent = TYPE_OBJECT,
1254     .abstract = true,
1255     .class_size = sizeof(MachineClass),
1256     .class_init    = machine_class_init,
1257     .class_base_init = machine_class_base_init,
1258     .instance_size = sizeof(MachineState),
1259     .instance_init = machine_initfn,
1260     .instance_finalize = machine_finalize,
1261 };
1262 
1263 static void machine_register_types(void)
1264 {
1265     type_register_static(&machine_info);
1266 }
1267 
1268 type_init(machine_register_types)
1269