xref: /qemu/hw/core/machine.c (revision c9ea365dce32bc114dacd7cfca7c478a82459b47)
1 /*
2  * QEMU Machine
3  *
4  * Copyright (C) 2014 Red Hat Inc
5  *
6  * Authors:
7  *   Marcel Apfelbaum <marcel.a@redhat.com>
8  *
9  * This work is licensed under the terms of the GNU GPL, version 2 or later.
10  * See the COPYING file in the top-level directory.
11  */
12 
13 #include "qemu/osdep.h"
14 #include "qemu/units.h"
15 #include "qemu/accel.h"
16 #include "sysemu/replay.h"
17 #include "hw/boards.h"
18 #include "hw/loader.h"
19 #include "qemu/error-report.h"
20 #include "qapi/error.h"
21 #include "qapi/qapi-visit-machine.h"
22 #include "qemu/madvise.h"
23 #include "qom/object_interfaces.h"
24 #include "sysemu/cpus.h"
25 #include "sysemu/sysemu.h"
26 #include "sysemu/reset.h"
27 #include "sysemu/runstate.h"
28 #include "sysemu/xen.h"
29 #include "sysemu/qtest.h"
30 #include "hw/pci/pci_bridge.h"
31 #include "hw/mem/nvdimm.h"
32 #include "migration/global_state.h"
33 #include "exec/confidential-guest-support.h"
34 #include "hw/virtio/virtio-pci.h"
35 #include "hw/virtio/virtio-net.h"
36 #include "hw/virtio/virtio-iommu.h"
37 #include "audio/audio.h"
38 
39 GlobalProperty hw_compat_9_2[] = {};
40 const size_t hw_compat_9_2_len = G_N_ELEMENTS(hw_compat_9_2);
41 
42 GlobalProperty hw_compat_9_1[] = {
43     { TYPE_PCI_DEVICE, "x-pcie-ext-tag", "false" },
44 };
45 const size_t hw_compat_9_1_len = G_N_ELEMENTS(hw_compat_9_1);
46 
47 GlobalProperty hw_compat_9_0[] = {
48     {"arm-cpu", "backcompat-cntfrq", "true" },
49     { "scsi-hd", "migrate-emulated-scsi-request", "false" },
50     { "scsi-cd", "migrate-emulated-scsi-request", "false" },
51     {"vfio-pci", "skip-vsc-check", "false" },
52     { "virtio-pci", "x-pcie-pm-no-soft-reset", "off" },
53     {"sd-card", "spec_version", "2" },
54 };
55 const size_t hw_compat_9_0_len = G_N_ELEMENTS(hw_compat_9_0);
56 
57 GlobalProperty hw_compat_8_2[] = {
58     { "migration", "zero-page-detection", "legacy"},
59     { TYPE_VIRTIO_IOMMU_PCI, "granule", "4k" },
60     { TYPE_VIRTIO_IOMMU_PCI, "aw-bits", "64" },
61     { "virtio-gpu-device", "x-scanout-vmstate-version", "1" },
62 };
63 const size_t hw_compat_8_2_len = G_N_ELEMENTS(hw_compat_8_2);
64 
65 GlobalProperty hw_compat_8_1[] = {
66     { TYPE_PCI_BRIDGE, "x-pci-express-writeable-slt-bug", "true" },
67     { "ramfb", "x-migrate", "off" },
68     { "vfio-pci-nohotplug", "x-ramfb-migrate", "off" },
69     { "igb", "x-pcie-flr-init", "off" },
70     { TYPE_VIRTIO_NET, "host_uso", "off"},
71     { TYPE_VIRTIO_NET, "guest_uso4", "off"},
72     { TYPE_VIRTIO_NET, "guest_uso6", "off"},
73 };
74 const size_t hw_compat_8_1_len = G_N_ELEMENTS(hw_compat_8_1);
75 
76 GlobalProperty hw_compat_8_0[] = {
77     { "migration", "multifd-flush-after-each-section", "on"},
78     { TYPE_PCI_DEVICE, "x-pcie-ari-nextfn-1", "on" },
79 };
80 const size_t hw_compat_8_0_len = G_N_ELEMENTS(hw_compat_8_0);
81 
82 GlobalProperty hw_compat_7_2[] = {
83     { "e1000e", "migrate-timadj", "off" },
84     { "virtio-mem", "x-early-migration", "false" },
85     { "migration", "x-preempt-pre-7-2", "true" },
86     { TYPE_PCI_DEVICE, "x-pcie-err-unc-mask", "off" },
87 };
88 const size_t hw_compat_7_2_len = G_N_ELEMENTS(hw_compat_7_2);
89 
90 GlobalProperty hw_compat_7_1[] = {
91     { "virtio-device", "queue_reset", "false" },
92     { "virtio-rng-pci", "vectors", "0" },
93     { "virtio-rng-pci-transitional", "vectors", "0" },
94     { "virtio-rng-pci-non-transitional", "vectors", "0" },
95 };
96 const size_t hw_compat_7_1_len = G_N_ELEMENTS(hw_compat_7_1);
97 
98 GlobalProperty hw_compat_7_0[] = {
99     { "arm-gicv3-common", "force-8-bit-prio", "on" },
100     { "nvme-ns", "eui64-default", "on"},
101 };
102 const size_t hw_compat_7_0_len = G_N_ELEMENTS(hw_compat_7_0);
103 
104 GlobalProperty hw_compat_6_2[] = {
105     { "PIIX4_PM", "x-not-migrate-acpi-index", "on"},
106 };
107 const size_t hw_compat_6_2_len = G_N_ELEMENTS(hw_compat_6_2);
108 
109 GlobalProperty hw_compat_6_1[] = {
110     { "vhost-user-vsock-device", "seqpacket", "off" },
111     { "nvme-ns", "shared", "off" },
112 };
113 const size_t hw_compat_6_1_len = G_N_ELEMENTS(hw_compat_6_1);
114 
115 GlobalProperty hw_compat_6_0[] = {
116     { "gpex-pcihost", "allow-unmapped-accesses", "false" },
117     { "i8042", "extended-state", "false"},
118     { "nvme-ns", "eui64-default", "off"},
119     { "e1000", "init-vet", "off" },
120     { "e1000e", "init-vet", "off" },
121     { "vhost-vsock-device", "seqpacket", "off" },
122 };
123 const size_t hw_compat_6_0_len = G_N_ELEMENTS(hw_compat_6_0);
124 
125 GlobalProperty hw_compat_5_2[] = {
126     { "ICH9-LPC", "smm-compat", "on"},
127     { "PIIX4_PM", "smm-compat", "on"},
128     { "virtio-blk-device", "report-discard-granularity", "off" },
129     { "virtio-net-pci-base", "vectors", "3"},
130     { "nvme", "msix-exclusive-bar", "on"},
131 };
132 const size_t hw_compat_5_2_len = G_N_ELEMENTS(hw_compat_5_2);
133 
134 GlobalProperty hw_compat_5_1[] = {
135     { "vhost-scsi", "num_queues", "1"},
136     { "vhost-user-blk", "num-queues", "1"},
137     { "vhost-user-scsi", "num_queues", "1"},
138     { "virtio-blk-device", "num-queues", "1"},
139     { "virtio-scsi-device", "num_queues", "1"},
140     { "nvme", "use-intel-id", "on"},
141     { "pvpanic", "events", "1"}, /* PVPANIC_PANICKED */
142     { "pl011", "migrate-clk", "off" },
143     { "virtio-pci", "x-ats-page-aligned", "off"},
144 };
145 const size_t hw_compat_5_1_len = G_N_ELEMENTS(hw_compat_5_1);
146 
147 GlobalProperty hw_compat_5_0[] = {
148     { "pci-host-bridge", "x-config-reg-migration-enabled", "off" },
149     { "virtio-balloon-device", "page-poison", "false" },
150     { "vmport", "x-read-set-eax", "off" },
151     { "vmport", "x-signal-unsupported-cmd", "off" },
152     { "vmport", "x-report-vmx-type", "off" },
153     { "vmport", "x-cmds-v2", "off" },
154     { "virtio-device", "x-disable-legacy-check", "true" },
155 };
156 const size_t hw_compat_5_0_len = G_N_ELEMENTS(hw_compat_5_0);
157 
158 GlobalProperty hw_compat_4_2[] = {
159     { "virtio-blk-device", "queue-size", "128"},
160     { "virtio-scsi-device", "virtqueue_size", "128"},
161     { "virtio-blk-device", "x-enable-wce-if-config-wce", "off" },
162     { "virtio-blk-device", "seg-max-adjust", "off"},
163     { "virtio-scsi-device", "seg_max_adjust", "off"},
164     { "vhost-blk-device", "seg_max_adjust", "off"},
165     { "usb-host", "suppress-remote-wake", "off" },
166     { "usb-redir", "suppress-remote-wake", "off" },
167     { "qxl", "revision", "4" },
168     { "qxl-vga", "revision", "4" },
169     { "fw_cfg", "acpi-mr-restore", "false" },
170     { "virtio-device", "use-disabled-flag", "false" },
171 };
172 const size_t hw_compat_4_2_len = G_N_ELEMENTS(hw_compat_4_2);
173 
174 GlobalProperty hw_compat_4_1[] = {
175     { "virtio-pci", "x-pcie-flr-init", "off" },
176 };
177 const size_t hw_compat_4_1_len = G_N_ELEMENTS(hw_compat_4_1);
178 
179 GlobalProperty hw_compat_4_0[] = {
180     { "VGA",            "edid", "false" },
181     { "secondary-vga",  "edid", "false" },
182     { "bochs-display",  "edid", "false" },
183     { "virtio-vga",     "edid", "false" },
184     { "virtio-gpu-device", "edid", "false" },
185     { "virtio-device", "use-started", "false" },
186     { "virtio-balloon-device", "qemu-4-0-config-size", "true" },
187     { "pl031", "migrate-tick-offset", "false" },
188 };
189 const size_t hw_compat_4_0_len = G_N_ELEMENTS(hw_compat_4_0);
190 
191 GlobalProperty hw_compat_3_1[] = {
192     { "pcie-root-port", "x-speed", "2_5" },
193     { "pcie-root-port", "x-width", "1" },
194     { "memory-backend-file", "x-use-canonical-path-for-ramblock-id", "true" },
195     { "memory-backend-memfd", "x-use-canonical-path-for-ramblock-id", "true" },
196     { "tpm-crb", "ppi", "false" },
197     { "tpm-tis", "ppi", "false" },
198     { "usb-kbd", "serial", "42" },
199     { "usb-mouse", "serial", "42" },
200     { "usb-tablet", "serial", "42" },
201     { "virtio-blk-device", "discard", "false" },
202     { "virtio-blk-device", "write-zeroes", "false" },
203     { "virtio-balloon-device", "qemu-4-0-config-size", "false" },
204     { "pcie-root-port-base", "disable-acs", "true" }, /* Added in 4.1 */
205 };
206 const size_t hw_compat_3_1_len = G_N_ELEMENTS(hw_compat_3_1);
207 
208 GlobalProperty hw_compat_3_0[] = {};
209 const size_t hw_compat_3_0_len = G_N_ELEMENTS(hw_compat_3_0);
210 
211 GlobalProperty hw_compat_2_12[] = {
212     { "hda-audio", "use-timer", "false" },
213     { "cirrus-vga", "global-vmstate", "true" },
214     { "VGA", "global-vmstate", "true" },
215     { "vmware-svga", "global-vmstate", "true" },
216     { "qxl-vga", "global-vmstate", "true" },
217 };
218 const size_t hw_compat_2_12_len = G_N_ELEMENTS(hw_compat_2_12);
219 
220 GlobalProperty hw_compat_2_11[] = {
221     { "hpet", "hpet-offset-saved", "false" },
222     { "virtio-blk-pci", "vectors", "2" },
223     { "vhost-user-blk-pci", "vectors", "2" },
224     { "e1000", "migrate_tso_props", "off" },
225 };
226 const size_t hw_compat_2_11_len = G_N_ELEMENTS(hw_compat_2_11);
227 
228 GlobalProperty hw_compat_2_10[] = {
229     { "virtio-mouse-device", "wheel-axis", "false" },
230     { "virtio-tablet-device", "wheel-axis", "false" },
231 };
232 const size_t hw_compat_2_10_len = G_N_ELEMENTS(hw_compat_2_10);
233 
234 GlobalProperty hw_compat_2_9[] = {
235     { "pci-bridge", "shpc", "off" },
236     { "intel-iommu", "pt", "off" },
237     { "virtio-net-device", "x-mtu-bypass-backend", "off" },
238     { "pcie-root-port", "x-migrate-msix", "false" },
239 };
240 const size_t hw_compat_2_9_len = G_N_ELEMENTS(hw_compat_2_9);
241 
242 GlobalProperty hw_compat_2_8[] = {
243     { "fw_cfg_mem", "x-file-slots", "0x10" },
244     { "fw_cfg_io", "x-file-slots", "0x10" },
245     { "pflash_cfi01", "old-multiple-chip-handling", "on" },
246     { "pci-bridge", "shpc", "on" },
247     { TYPE_PCI_DEVICE, "x-pcie-extcap-init", "off" },
248     { "virtio-pci", "x-pcie-deverr-init", "off" },
249     { "virtio-pci", "x-pcie-lnkctl-init", "off" },
250     { "virtio-pci", "x-pcie-pm-init", "off" },
251     { "cirrus-vga", "vgamem_mb", "8" },
252     { "isa-cirrus-vga", "vgamem_mb", "8" },
253 };
254 const size_t hw_compat_2_8_len = G_N_ELEMENTS(hw_compat_2_8);
255 
256 GlobalProperty hw_compat_2_7[] = {
257     { "virtio-pci", "page-per-vq", "on" },
258     { "virtio-serial-device", "emergency-write", "off" },
259     { "ioapic", "version", "0x11" },
260     { "intel-iommu", "x-buggy-eim", "true" },
261     { "virtio-pci", "x-ignore-backend-features", "on" },
262 };
263 const size_t hw_compat_2_7_len = G_N_ELEMENTS(hw_compat_2_7);
264 
265 GlobalProperty hw_compat_2_6[] = {
266     { "virtio-mmio", "format_transport_address", "off" },
267     /* Optional because not all virtio-pci devices support legacy mode */
268     { "virtio-pci", "disable-modern", "on",  .optional = true },
269     { "virtio-pci", "disable-legacy", "off", .optional = true },
270 };
271 const size_t hw_compat_2_6_len = G_N_ELEMENTS(hw_compat_2_6);
272 
273 GlobalProperty hw_compat_2_5[] = {
274     { "isa-fdc", "fallback", "144" },
275     { "pvscsi", "x-old-pci-configuration", "on" },
276     { "pvscsi", "x-disable-pcie", "on" },
277     { "vmxnet3", "x-old-msi-offsets", "on" },
278     { "vmxnet3", "x-disable-pcie", "on" },
279 };
280 const size_t hw_compat_2_5_len = G_N_ELEMENTS(hw_compat_2_5);
281 
282 GlobalProperty hw_compat_2_4[] = {
283     { "e1000", "extra_mac_registers", "off" },
284     { "virtio-pci", "x-disable-pcie", "on" },
285     { "virtio-pci", "migrate-extra", "off" },
286     { "fw_cfg_mem", "dma_enabled", "off" },
287     { "fw_cfg_io", "dma_enabled", "off" }
288 };
289 const size_t hw_compat_2_4_len = G_N_ELEMENTS(hw_compat_2_4);
290 
291 MachineState *current_machine;
292 
293 static char *machine_get_kernel(Object *obj, Error **errp)
294 {
295     MachineState *ms = MACHINE(obj);
296 
297     return g_strdup(ms->kernel_filename);
298 }
299 
300 static void machine_set_kernel(Object *obj, const char *value, Error **errp)
301 {
302     MachineState *ms = MACHINE(obj);
303 
304     g_free(ms->kernel_filename);
305     ms->kernel_filename = g_strdup(value);
306 }
307 
308 static char *machine_get_initrd(Object *obj, Error **errp)
309 {
310     MachineState *ms = MACHINE(obj);
311 
312     return g_strdup(ms->initrd_filename);
313 }
314 
315 static void machine_set_initrd(Object *obj, const char *value, Error **errp)
316 {
317     MachineState *ms = MACHINE(obj);
318 
319     g_free(ms->initrd_filename);
320     ms->initrd_filename = g_strdup(value);
321 }
322 
323 static char *machine_get_append(Object *obj, Error **errp)
324 {
325     MachineState *ms = MACHINE(obj);
326 
327     return g_strdup(ms->kernel_cmdline);
328 }
329 
330 static void machine_set_append(Object *obj, const char *value, Error **errp)
331 {
332     MachineState *ms = MACHINE(obj);
333 
334     g_free(ms->kernel_cmdline);
335     ms->kernel_cmdline = g_strdup(value);
336 }
337 
338 static char *machine_get_dtb(Object *obj, Error **errp)
339 {
340     MachineState *ms = MACHINE(obj);
341 
342     return g_strdup(ms->dtb);
343 }
344 
345 static void machine_set_dtb(Object *obj, const char *value, Error **errp)
346 {
347     MachineState *ms = MACHINE(obj);
348 
349     g_free(ms->dtb);
350     ms->dtb = g_strdup(value);
351 }
352 
353 static char *machine_get_dumpdtb(Object *obj, Error **errp)
354 {
355     MachineState *ms = MACHINE(obj);
356 
357     return g_strdup(ms->dumpdtb);
358 }
359 
360 static void machine_set_dumpdtb(Object *obj, const char *value, Error **errp)
361 {
362     MachineState *ms = MACHINE(obj);
363 
364     g_free(ms->dumpdtb);
365     ms->dumpdtb = g_strdup(value);
366 }
367 
368 static void machine_get_phandle_start(Object *obj, Visitor *v,
369                                       const char *name, void *opaque,
370                                       Error **errp)
371 {
372     MachineState *ms = MACHINE(obj);
373     int64_t value = ms->phandle_start;
374 
375     visit_type_int(v, name, &value, errp);
376 }
377 
378 static void machine_set_phandle_start(Object *obj, Visitor *v,
379                                       const char *name, void *opaque,
380                                       Error **errp)
381 {
382     MachineState *ms = MACHINE(obj);
383     int64_t value;
384 
385     if (!visit_type_int(v, name, &value, errp)) {
386         return;
387     }
388 
389     ms->phandle_start = value;
390 }
391 
392 static char *machine_get_dt_compatible(Object *obj, Error **errp)
393 {
394     MachineState *ms = MACHINE(obj);
395 
396     return g_strdup(ms->dt_compatible);
397 }
398 
399 static void machine_set_dt_compatible(Object *obj, const char *value, Error **errp)
400 {
401     MachineState *ms = MACHINE(obj);
402 
403     g_free(ms->dt_compatible);
404     ms->dt_compatible = g_strdup(value);
405 }
406 
407 static bool machine_get_dump_guest_core(Object *obj, Error **errp)
408 {
409     MachineState *ms = MACHINE(obj);
410 
411     return ms->dump_guest_core;
412 }
413 
414 static void machine_set_dump_guest_core(Object *obj, bool value, Error **errp)
415 {
416     MachineState *ms = MACHINE(obj);
417 
418     if (!value && QEMU_MADV_DONTDUMP == QEMU_MADV_INVALID) {
419         error_setg(errp, "Dumping guest memory cannot be disabled on this host");
420         return;
421     }
422     ms->dump_guest_core = value;
423 }
424 
425 static bool machine_get_mem_merge(Object *obj, Error **errp)
426 {
427     MachineState *ms = MACHINE(obj);
428 
429     return ms->mem_merge;
430 }
431 
432 static void machine_set_mem_merge(Object *obj, bool value, Error **errp)
433 {
434     MachineState *ms = MACHINE(obj);
435 
436     if (value && QEMU_MADV_MERGEABLE == QEMU_MADV_INVALID) {
437         error_setg(errp, "Memory merging is not supported on this host");
438         return;
439     }
440     ms->mem_merge = value;
441 }
442 
443 static bool machine_get_usb(Object *obj, Error **errp)
444 {
445     MachineState *ms = MACHINE(obj);
446 
447     return ms->usb;
448 }
449 
450 static void machine_set_usb(Object *obj, bool value, Error **errp)
451 {
452     MachineState *ms = MACHINE(obj);
453 
454     ms->usb = value;
455     ms->usb_disabled = !value;
456 }
457 
458 static bool machine_get_graphics(Object *obj, Error **errp)
459 {
460     MachineState *ms = MACHINE(obj);
461 
462     return ms->enable_graphics;
463 }
464 
465 static void machine_set_graphics(Object *obj, bool value, Error **errp)
466 {
467     MachineState *ms = MACHINE(obj);
468 
469     ms->enable_graphics = value;
470 }
471 
472 static char *machine_get_firmware(Object *obj, Error **errp)
473 {
474     MachineState *ms = MACHINE(obj);
475 
476     return g_strdup(ms->firmware);
477 }
478 
479 static void machine_set_firmware(Object *obj, const char *value, Error **errp)
480 {
481     MachineState *ms = MACHINE(obj);
482 
483     g_free(ms->firmware);
484     ms->firmware = g_strdup(value);
485 }
486 
487 static void machine_set_suppress_vmdesc(Object *obj, bool value, Error **errp)
488 {
489     MachineState *ms = MACHINE(obj);
490 
491     ms->suppress_vmdesc = value;
492 }
493 
494 static bool machine_get_suppress_vmdesc(Object *obj, Error **errp)
495 {
496     MachineState *ms = MACHINE(obj);
497 
498     return ms->suppress_vmdesc;
499 }
500 
501 static char *machine_get_memory_encryption(Object *obj, Error **errp)
502 {
503     MachineState *ms = MACHINE(obj);
504 
505     if (ms->cgs) {
506         return g_strdup(object_get_canonical_path_component(OBJECT(ms->cgs)));
507     }
508 
509     return NULL;
510 }
511 
512 static void machine_set_memory_encryption(Object *obj, const char *value,
513                                         Error **errp)
514 {
515     Object *cgs =
516         object_resolve_path_component(object_get_objects_root(), value);
517 
518     if (!cgs) {
519         error_setg(errp, "No such memory encryption object '%s'", value);
520         return;
521     }
522 
523     object_property_set_link(obj, "confidential-guest-support", cgs, errp);
524 }
525 
526 static void machine_check_confidential_guest_support(const Object *obj,
527                                                      const char *name,
528                                                      Object *new_target,
529                                                      Error **errp)
530 {
531     /*
532      * So far the only constraint is that the target has the
533      * TYPE_CONFIDENTIAL_GUEST_SUPPORT interface, and that's checked
534      * by the QOM core
535      */
536 }
537 
538 static bool machine_get_nvdimm(Object *obj, Error **errp)
539 {
540     MachineState *ms = MACHINE(obj);
541 
542     return ms->nvdimms_state->is_enabled;
543 }
544 
545 static void machine_set_nvdimm(Object *obj, bool value, Error **errp)
546 {
547     MachineState *ms = MACHINE(obj);
548 
549     ms->nvdimms_state->is_enabled = value;
550 }
551 
552 static bool machine_get_hmat(Object *obj, Error **errp)
553 {
554     MachineState *ms = MACHINE(obj);
555 
556     return ms->numa_state->hmat_enabled;
557 }
558 
559 static void machine_set_hmat(Object *obj, bool value, Error **errp)
560 {
561     MachineState *ms = MACHINE(obj);
562 
563     ms->numa_state->hmat_enabled = value;
564 }
565 
566 static void machine_get_mem(Object *obj, Visitor *v, const char *name,
567                             void *opaque, Error **errp)
568 {
569     MachineState *ms = MACHINE(obj);
570     MemorySizeConfiguration mem = {
571         .has_size = true,
572         .size = ms->ram_size,
573         .has_max_size = !!ms->ram_slots,
574         .max_size = ms->maxram_size,
575         .has_slots = !!ms->ram_slots,
576         .slots = ms->ram_slots,
577     };
578     MemorySizeConfiguration *p_mem = &mem;
579 
580     visit_type_MemorySizeConfiguration(v, name, &p_mem, &error_abort);
581 }
582 
583 static void machine_set_mem(Object *obj, Visitor *v, const char *name,
584                             void *opaque, Error **errp)
585 {
586     ERRP_GUARD();
587     MachineState *ms = MACHINE(obj);
588     MachineClass *mc = MACHINE_GET_CLASS(obj);
589     MemorySizeConfiguration *mem;
590 
591     if (!visit_type_MemorySizeConfiguration(v, name, &mem, errp)) {
592         return;
593     }
594 
595     if (!mem->has_size) {
596         mem->has_size = true;
597         mem->size = mc->default_ram_size;
598     }
599     mem->size = QEMU_ALIGN_UP(mem->size, 8192);
600     if (mc->fixup_ram_size) {
601         mem->size = mc->fixup_ram_size(mem->size);
602     }
603     if ((ram_addr_t)mem->size != mem->size) {
604         error_setg(errp, "ram size %llu exceeds permitted maximum %llu",
605                    (unsigned long long)mem->size,
606                    (unsigned long long)RAM_ADDR_MAX);
607         goto out_free;
608     }
609 
610     if (mem->has_max_size) {
611         if ((ram_addr_t)mem->max_size != mem->max_size) {
612             error_setg(errp, "ram size %llu exceeds permitted maximum %llu",
613                        (unsigned long long)mem->max_size,
614                        (unsigned long long)RAM_ADDR_MAX);
615             goto out_free;
616         }
617         if (mem->max_size < mem->size) {
618             error_setg(errp, "invalid value of maxmem: "
619                        "maximum memory size (0x%" PRIx64 ") must be at least "
620                        "the initial memory size (0x%" PRIx64 ")",
621                        mem->max_size, mem->size);
622             goto out_free;
623         }
624         if (mem->has_slots && mem->slots && mem->max_size == mem->size) {
625             error_setg(errp, "invalid value of maxmem: "
626                        "memory slots were specified but maximum memory size "
627                        "(0x%" PRIx64 ") is equal to the initial memory size "
628                        "(0x%" PRIx64 ")", mem->max_size, mem->size);
629             goto out_free;
630         }
631         ms->maxram_size = mem->max_size;
632     } else {
633         if (mem->has_slots) {
634             error_setg(errp, "slots specified but no max-size");
635             goto out_free;
636         }
637         ms->maxram_size = mem->size;
638     }
639     ms->ram_size = mem->size;
640     ms->ram_slots = mem->has_slots ? mem->slots : 0;
641 out_free:
642     qapi_free_MemorySizeConfiguration(mem);
643 }
644 
645 static char *machine_get_nvdimm_persistence(Object *obj, Error **errp)
646 {
647     MachineState *ms = MACHINE(obj);
648 
649     return g_strdup(ms->nvdimms_state->persistence_string);
650 }
651 
652 static void machine_set_nvdimm_persistence(Object *obj, const char *value,
653                                            Error **errp)
654 {
655     MachineState *ms = MACHINE(obj);
656     NVDIMMState *nvdimms_state = ms->nvdimms_state;
657 
658     if (strcmp(value, "cpu") == 0) {
659         nvdimms_state->persistence = 3;
660     } else if (strcmp(value, "mem-ctrl") == 0) {
661         nvdimms_state->persistence = 2;
662     } else {
663         error_setg(errp, "-machine nvdimm-persistence=%s: unsupported option",
664                    value);
665         return;
666     }
667 
668     g_free(nvdimms_state->persistence_string);
669     nvdimms_state->persistence_string = g_strdup(value);
670 }
671 
672 void machine_class_allow_dynamic_sysbus_dev(MachineClass *mc, const char *type)
673 {
674     QAPI_LIST_PREPEND(mc->allowed_dynamic_sysbus_devices, g_strdup(type));
675 }
676 
677 bool device_is_dynamic_sysbus(MachineClass *mc, DeviceState *dev)
678 {
679     Object *obj = OBJECT(dev);
680 
681     if (!object_dynamic_cast(obj, TYPE_SYS_BUS_DEVICE)) {
682         return false;
683     }
684 
685     return device_type_is_dynamic_sysbus(mc, object_get_typename(obj));
686 }
687 
688 bool device_type_is_dynamic_sysbus(MachineClass *mc, const char *type)
689 {
690     bool allowed = false;
691     strList *wl;
692     ObjectClass *klass = object_class_by_name(type);
693 
694     for (wl = mc->allowed_dynamic_sysbus_devices;
695          !allowed && wl;
696          wl = wl->next) {
697         allowed |= !!object_class_dynamic_cast(klass, wl->value);
698     }
699 
700     return allowed;
701 }
702 
703 static char *machine_get_audiodev(Object *obj, Error **errp)
704 {
705     MachineState *ms = MACHINE(obj);
706 
707     return g_strdup(ms->audiodev);
708 }
709 
710 static void machine_set_audiodev(Object *obj, const char *value,
711                                  Error **errp)
712 {
713     MachineState *ms = MACHINE(obj);
714 
715     if (!audio_state_by_name(value, errp)) {
716         return;
717     }
718 
719     g_free(ms->audiodev);
720     ms->audiodev = g_strdup(value);
721 }
722 
723 HotpluggableCPUList *machine_query_hotpluggable_cpus(MachineState *machine)
724 {
725     int i;
726     HotpluggableCPUList *head = NULL;
727     MachineClass *mc = MACHINE_GET_CLASS(machine);
728 
729     /* force board to initialize possible_cpus if it hasn't been done yet */
730     mc->possible_cpu_arch_ids(machine);
731 
732     for (i = 0; i < machine->possible_cpus->len; i++) {
733         CPUState *cpu;
734         HotpluggableCPU *cpu_item = g_new0(typeof(*cpu_item), 1);
735 
736         cpu_item->type = g_strdup(machine->possible_cpus->cpus[i].type);
737         cpu_item->vcpus_count = machine->possible_cpus->cpus[i].vcpus_count;
738         cpu_item->props = g_memdup(&machine->possible_cpus->cpus[i].props,
739                                    sizeof(*cpu_item->props));
740 
741         cpu = machine->possible_cpus->cpus[i].cpu;
742         if (cpu) {
743             cpu_item->qom_path = object_get_canonical_path(OBJECT(cpu));
744         }
745         QAPI_LIST_PREPEND(head, cpu_item);
746     }
747     return head;
748 }
749 
750 /**
751  * machine_set_cpu_numa_node:
752  * @machine: machine object to modify
753  * @props: specifies which cpu objects to assign to
754  *         numa node specified by @props.node_id
755  * @errp: if an error occurs, a pointer to an area to store the error
756  *
757  * Associate NUMA node specified by @props.node_id with cpu slots that
758  * match socket/core/thread-ids specified by @props. It's recommended to use
759  * query-hotpluggable-cpus.props values to specify affected cpu slots,
760  * which would lead to exact 1:1 mapping of cpu slots to NUMA node.
761  *
762  * However for CLI convenience it's possible to pass in subset of properties,
763  * which would affect all cpu slots that match it.
764  * Ex for pc machine:
765  *    -smp 4,cores=2,sockets=2 -numa node,nodeid=0 -numa node,nodeid=1 \
766  *    -numa cpu,node-id=0,socket_id=0 \
767  *    -numa cpu,node-id=1,socket_id=1
768  * will assign all child cores of socket 0 to node 0 and
769  * of socket 1 to node 1.
770  *
771  * On attempt of reassigning (already assigned) cpu slot to another NUMA node,
772  * return error.
773  * Empty subset is disallowed and function will return with error in this case.
774  */
775 void machine_set_cpu_numa_node(MachineState *machine,
776                                const CpuInstanceProperties *props, Error **errp)
777 {
778     MachineClass *mc = MACHINE_GET_CLASS(machine);
779     NodeInfo *numa_info = machine->numa_state->nodes;
780     bool match = false;
781     int i;
782 
783     if (!mc->possible_cpu_arch_ids) {
784         error_setg(errp, "mapping of CPUs to NUMA node is not supported");
785         return;
786     }
787 
788     /* disabling node mapping is not supported, forbid it */
789     assert(props->has_node_id);
790 
791     /* force board to initialize possible_cpus if it hasn't been done yet */
792     mc->possible_cpu_arch_ids(machine);
793 
794     for (i = 0; i < machine->possible_cpus->len; i++) {
795         CPUArchId *slot = &machine->possible_cpus->cpus[i];
796 
797         /* reject unsupported by board properties */
798         if (props->has_thread_id && !slot->props.has_thread_id) {
799             error_setg(errp, "thread-id is not supported");
800             return;
801         }
802 
803         if (props->has_core_id && !slot->props.has_core_id) {
804             error_setg(errp, "core-id is not supported");
805             return;
806         }
807 
808         if (props->has_module_id && !slot->props.has_module_id) {
809             error_setg(errp, "module-id is not supported");
810             return;
811         }
812 
813         if (props->has_cluster_id && !slot->props.has_cluster_id) {
814             error_setg(errp, "cluster-id is not supported");
815             return;
816         }
817 
818         if (props->has_socket_id && !slot->props.has_socket_id) {
819             error_setg(errp, "socket-id is not supported");
820             return;
821         }
822 
823         if (props->has_die_id && !slot->props.has_die_id) {
824             error_setg(errp, "die-id is not supported");
825             return;
826         }
827 
828         /* skip slots with explicit mismatch */
829         if (props->has_thread_id && props->thread_id != slot->props.thread_id) {
830                 continue;
831         }
832 
833         if (props->has_core_id && props->core_id != slot->props.core_id) {
834                 continue;
835         }
836 
837         if (props->has_module_id &&
838             props->module_id != slot->props.module_id) {
839                 continue;
840         }
841 
842         if (props->has_cluster_id &&
843             props->cluster_id != slot->props.cluster_id) {
844                 continue;
845         }
846 
847         if (props->has_die_id && props->die_id != slot->props.die_id) {
848                 continue;
849         }
850 
851         if (props->has_socket_id && props->socket_id != slot->props.socket_id) {
852                 continue;
853         }
854 
855         /* reject assignment if slot is already assigned, for compatibility
856          * of legacy cpu_index mapping with SPAPR core based mapping do not
857          * error out if cpu thread and matched core have the same node-id */
858         if (slot->props.has_node_id &&
859             slot->props.node_id != props->node_id) {
860             error_setg(errp, "CPU is already assigned to node-id: %" PRId64,
861                        slot->props.node_id);
862             return;
863         }
864 
865         /* assign slot to node as it's matched '-numa cpu' key */
866         match = true;
867         slot->props.node_id = props->node_id;
868         slot->props.has_node_id = props->has_node_id;
869 
870         if (machine->numa_state->hmat_enabled) {
871             if ((numa_info[props->node_id].initiator < MAX_NODES) &&
872                 (props->node_id != numa_info[props->node_id].initiator)) {
873                 error_setg(errp, "The initiator of CPU NUMA node %" PRId64
874                            " should be itself (got %" PRIu16 ")",
875                            props->node_id, numa_info[props->node_id].initiator);
876                 return;
877             }
878             numa_info[props->node_id].has_cpu = true;
879             numa_info[props->node_id].initiator = props->node_id;
880         }
881     }
882 
883     if (!match) {
884         error_setg(errp, "no match found");
885     }
886 }
887 
888 static void machine_get_smp(Object *obj, Visitor *v, const char *name,
889                             void *opaque, Error **errp)
890 {
891     MachineState *ms = MACHINE(obj);
892     SMPConfiguration *config = &(SMPConfiguration){
893         .has_cpus = true, .cpus = ms->smp.cpus,
894         .has_drawers = true, .drawers = ms->smp.drawers,
895         .has_books = true, .books = ms->smp.books,
896         .has_sockets = true, .sockets = ms->smp.sockets,
897         .has_dies = true, .dies = ms->smp.dies,
898         .has_clusters = true, .clusters = ms->smp.clusters,
899         .has_modules = true, .modules = ms->smp.modules,
900         .has_cores = true, .cores = ms->smp.cores,
901         .has_threads = true, .threads = ms->smp.threads,
902         .has_maxcpus = true, .maxcpus = ms->smp.max_cpus,
903     };
904 
905     if (!visit_type_SMPConfiguration(v, name, &config, &error_abort)) {
906         return;
907     }
908 }
909 
910 static void machine_set_smp(Object *obj, Visitor *v, const char *name,
911                             void *opaque, Error **errp)
912 {
913     MachineState *ms = MACHINE(obj);
914     g_autoptr(SMPConfiguration) config = NULL;
915 
916     if (!visit_type_SMPConfiguration(v, name, &config, errp)) {
917         return;
918     }
919 
920     machine_parse_smp_config(ms, config, errp);
921 }
922 
923 static void machine_get_smp_cache(Object *obj, Visitor *v, const char *name,
924                                   void *opaque, Error **errp)
925 {
926     MachineState *ms = MACHINE(obj);
927     SmpCache *cache = &ms->smp_cache;
928     SmpCachePropertiesList *head = NULL;
929     SmpCachePropertiesList **tail = &head;
930 
931     for (int i = 0; i < CACHE_LEVEL_AND_TYPE__MAX; i++) {
932         SmpCacheProperties *node = g_new(SmpCacheProperties, 1);
933 
934         node->cache = cache->props[i].cache;
935         node->topology = cache->props[i].topology;
936         QAPI_LIST_APPEND(tail, node);
937     }
938 
939     visit_type_SmpCachePropertiesList(v, name, &head, errp);
940     qapi_free_SmpCachePropertiesList(head);
941 }
942 
943 static void machine_set_smp_cache(Object *obj, Visitor *v, const char *name,
944                                   void *opaque, Error **errp)
945 {
946     MachineState *ms = MACHINE(obj);
947     SmpCachePropertiesList *caches;
948 
949     if (!visit_type_SmpCachePropertiesList(v, name, &caches, errp)) {
950         return;
951     }
952 
953     machine_parse_smp_cache(ms, caches, errp);
954     qapi_free_SmpCachePropertiesList(caches);
955 }
956 
957 static void machine_get_boot(Object *obj, Visitor *v, const char *name,
958                             void *opaque, Error **errp)
959 {
960     MachineState *ms = MACHINE(obj);
961     BootConfiguration *config = &ms->boot_config;
962     visit_type_BootConfiguration(v, name, &config, &error_abort);
963 }
964 
965 static void machine_free_boot_config(MachineState *ms)
966 {
967     g_free(ms->boot_config.order);
968     g_free(ms->boot_config.once);
969     g_free(ms->boot_config.splash);
970 }
971 
972 static void machine_copy_boot_config(MachineState *ms, BootConfiguration *config)
973 {
974     MachineClass *machine_class = MACHINE_GET_CLASS(ms);
975 
976     machine_free_boot_config(ms);
977     ms->boot_config = *config;
978     if (!config->order) {
979         ms->boot_config.order = g_strdup(machine_class->default_boot_order);
980     }
981 }
982 
983 static void machine_set_boot(Object *obj, Visitor *v, const char *name,
984                             void *opaque, Error **errp)
985 {
986     ERRP_GUARD();
987     MachineState *ms = MACHINE(obj);
988     BootConfiguration *config = NULL;
989 
990     if (!visit_type_BootConfiguration(v, name, &config, errp)) {
991         return;
992     }
993     if (config->order) {
994         validate_bootdevices(config->order, errp);
995         if (*errp) {
996             goto out_free;
997         }
998     }
999     if (config->once) {
1000         validate_bootdevices(config->once, errp);
1001         if (*errp) {
1002             goto out_free;
1003         }
1004     }
1005 
1006     machine_copy_boot_config(ms, config);
1007     /* Strings live in ms->boot_config.  */
1008     free(config);
1009     return;
1010 
1011 out_free:
1012     qapi_free_BootConfiguration(config);
1013 }
1014 
1015 void machine_add_audiodev_property(MachineClass *mc)
1016 {
1017     ObjectClass *oc = OBJECT_CLASS(mc);
1018 
1019     object_class_property_add_str(oc, "audiodev",
1020                                   machine_get_audiodev,
1021                                   machine_set_audiodev);
1022     object_class_property_set_description(oc, "audiodev",
1023                                           "Audiodev to use for default machine devices");
1024 }
1025 
1026 static bool create_default_memdev(MachineState *ms, const char *path,
1027                                   Error **errp)
1028 {
1029     Object *obj;
1030     MachineClass *mc = MACHINE_GET_CLASS(ms);
1031     bool r = false;
1032 
1033     obj = object_new(path ? TYPE_MEMORY_BACKEND_FILE : TYPE_MEMORY_BACKEND_RAM);
1034     if (path) {
1035         if (!object_property_set_str(obj, "mem-path", path, errp)) {
1036             goto out;
1037         }
1038     }
1039     if (!object_property_set_int(obj, "size", ms->ram_size, errp)) {
1040         goto out;
1041     }
1042     object_property_add_child(object_get_objects_root(), mc->default_ram_id,
1043                               obj);
1044     /* Ensure backend's memory region name is equal to mc->default_ram_id */
1045     if (!object_property_set_bool(obj, "x-use-canonical-path-for-ramblock-id",
1046                              false, errp)) {
1047         goto out;
1048     }
1049     if (!user_creatable_complete(USER_CREATABLE(obj), errp)) {
1050         goto out;
1051     }
1052     r = object_property_set_link(OBJECT(ms), "memory-backend", obj, errp);
1053 
1054 out:
1055     object_unref(obj);
1056     return r;
1057 }
1058 
1059 static void machine_class_init(ObjectClass *oc, void *data)
1060 {
1061     MachineClass *mc = MACHINE_CLASS(oc);
1062 
1063     /* Default 128 MB as guest ram size */
1064     mc->default_ram_size = 128 * MiB;
1065     mc->rom_file_has_mr = true;
1066     /*
1067      * SMBIOS 3.1.0 7.18.5 Memory Device — Extended Size
1068      * use max possible value that could be encoded into
1069      * 'Extended Size' field (2047Tb).
1070      */
1071     mc->smbios_memory_device_size = 2047 * TiB;
1072 
1073     /* numa node memory size aligned on 8MB by default.
1074      * On Linux, each node's border has to be 8MB aligned
1075      */
1076     mc->numa_mem_align_shift = 23;
1077 
1078     mc->create_default_memdev = create_default_memdev;
1079 
1080     object_class_property_add_str(oc, "kernel",
1081         machine_get_kernel, machine_set_kernel);
1082     object_class_property_set_description(oc, "kernel",
1083         "Linux kernel image file");
1084 
1085     object_class_property_add_str(oc, "initrd",
1086         machine_get_initrd, machine_set_initrd);
1087     object_class_property_set_description(oc, "initrd",
1088         "Linux initial ramdisk file");
1089 
1090     object_class_property_add_str(oc, "append",
1091         machine_get_append, machine_set_append);
1092     object_class_property_set_description(oc, "append",
1093         "Linux kernel command line");
1094 
1095     object_class_property_add_str(oc, "dtb",
1096         machine_get_dtb, machine_set_dtb);
1097     object_class_property_set_description(oc, "dtb",
1098         "Linux kernel device tree file");
1099 
1100     object_class_property_add_str(oc, "dumpdtb",
1101         machine_get_dumpdtb, machine_set_dumpdtb);
1102     object_class_property_set_description(oc, "dumpdtb",
1103         "Dump current dtb to a file and quit");
1104 
1105     object_class_property_add(oc, "boot", "BootConfiguration",
1106         machine_get_boot, machine_set_boot,
1107         NULL, NULL);
1108     object_class_property_set_description(oc, "boot",
1109         "Boot configuration");
1110 
1111     object_class_property_add(oc, "smp", "SMPConfiguration",
1112         machine_get_smp, machine_set_smp,
1113         NULL, NULL);
1114     object_class_property_set_description(oc, "smp",
1115         "CPU topology");
1116 
1117     object_class_property_add(oc, "smp-cache", "SmpCachePropertiesWrapper",
1118         machine_get_smp_cache, machine_set_smp_cache, NULL, NULL);
1119     object_class_property_set_description(oc, "smp-cache",
1120         "Cache properties list for SMP machine");
1121 
1122     object_class_property_add(oc, "phandle-start", "int",
1123         machine_get_phandle_start, machine_set_phandle_start,
1124         NULL, NULL);
1125     object_class_property_set_description(oc, "phandle-start",
1126         "The first phandle ID we may generate dynamically");
1127 
1128     object_class_property_add_str(oc, "dt-compatible",
1129         machine_get_dt_compatible, machine_set_dt_compatible);
1130     object_class_property_set_description(oc, "dt-compatible",
1131         "Overrides the \"compatible\" property of the dt root node");
1132 
1133     object_class_property_add_bool(oc, "dump-guest-core",
1134         machine_get_dump_guest_core, machine_set_dump_guest_core);
1135     object_class_property_set_description(oc, "dump-guest-core",
1136         "Include guest memory in a core dump");
1137 
1138     object_class_property_add_bool(oc, "mem-merge",
1139         machine_get_mem_merge, machine_set_mem_merge);
1140     object_class_property_set_description(oc, "mem-merge",
1141         "Enable/disable memory merge support");
1142 
1143     object_class_property_add_bool(oc, "usb",
1144         machine_get_usb, machine_set_usb);
1145     object_class_property_set_description(oc, "usb",
1146         "Set on/off to enable/disable usb");
1147 
1148     object_class_property_add_bool(oc, "graphics",
1149         machine_get_graphics, machine_set_graphics);
1150     object_class_property_set_description(oc, "graphics",
1151         "Set on/off to enable/disable graphics emulation");
1152 
1153     object_class_property_add_str(oc, "firmware",
1154         machine_get_firmware, machine_set_firmware);
1155     object_class_property_set_description(oc, "firmware",
1156         "Firmware image");
1157 
1158     object_class_property_add_bool(oc, "suppress-vmdesc",
1159         machine_get_suppress_vmdesc, machine_set_suppress_vmdesc);
1160     object_class_property_set_description(oc, "suppress-vmdesc",
1161         "Set on to disable self-describing migration");
1162 
1163     object_class_property_add_link(oc, "confidential-guest-support",
1164                                    TYPE_CONFIDENTIAL_GUEST_SUPPORT,
1165                                    offsetof(MachineState, cgs),
1166                                    machine_check_confidential_guest_support,
1167                                    OBJ_PROP_LINK_STRONG);
1168     object_class_property_set_description(oc, "confidential-guest-support",
1169                                           "Set confidential guest scheme to support");
1170 
1171     /* For compatibility */
1172     object_class_property_add_str(oc, "memory-encryption",
1173         machine_get_memory_encryption, machine_set_memory_encryption);
1174     object_class_property_set_description(oc, "memory-encryption",
1175         "Set memory encryption object to use");
1176 
1177     object_class_property_add_link(oc, "memory-backend", TYPE_MEMORY_BACKEND,
1178                                    offsetof(MachineState, memdev), object_property_allow_set_link,
1179                                    OBJ_PROP_LINK_STRONG);
1180     object_class_property_set_description(oc, "memory-backend",
1181                                           "Set RAM backend"
1182                                           "Valid value is ID of hostmem based backend");
1183 
1184     object_class_property_add(oc, "memory", "MemorySizeConfiguration",
1185         machine_get_mem, machine_set_mem,
1186         NULL, NULL);
1187     object_class_property_set_description(oc, "memory",
1188         "Memory size configuration");
1189 }
1190 
1191 static void machine_class_base_init(ObjectClass *oc, void *data)
1192 {
1193     MachineClass *mc = MACHINE_CLASS(oc);
1194     mc->max_cpus = mc->max_cpus ?: 1;
1195     mc->min_cpus = mc->min_cpus ?: 1;
1196     mc->default_cpus = mc->default_cpus ?: 1;
1197 
1198     if (!object_class_is_abstract(oc)) {
1199         const char *cname = object_class_get_name(oc);
1200         assert(g_str_has_suffix(cname, TYPE_MACHINE_SUFFIX));
1201         mc->name = g_strndup(cname,
1202                             strlen(cname) - strlen(TYPE_MACHINE_SUFFIX));
1203         mc->compat_props = g_ptr_array_new();
1204     }
1205 }
1206 
1207 static void machine_initfn(Object *obj)
1208 {
1209     MachineState *ms = MACHINE(obj);
1210     MachineClass *mc = MACHINE_GET_CLASS(obj);
1211 
1212     container_get(obj, "/peripheral");
1213     container_get(obj, "/peripheral-anon");
1214 
1215     ms->dump_guest_core = true;
1216     ms->mem_merge = (QEMU_MADV_MERGEABLE != QEMU_MADV_INVALID);
1217     ms->enable_graphics = true;
1218     ms->kernel_cmdline = g_strdup("");
1219     ms->ram_size = mc->default_ram_size;
1220     ms->maxram_size = mc->default_ram_size;
1221 
1222     if (mc->nvdimm_supported) {
1223         ms->nvdimms_state = g_new0(NVDIMMState, 1);
1224         object_property_add_bool(obj, "nvdimm",
1225                                  machine_get_nvdimm, machine_set_nvdimm);
1226         object_property_set_description(obj, "nvdimm",
1227                                         "Set on/off to enable/disable "
1228                                         "NVDIMM instantiation");
1229 
1230         object_property_add_str(obj, "nvdimm-persistence",
1231                                 machine_get_nvdimm_persistence,
1232                                 machine_set_nvdimm_persistence);
1233         object_property_set_description(obj, "nvdimm-persistence",
1234                                         "Set NVDIMM persistence"
1235                                         "Valid values are cpu, mem-ctrl");
1236     }
1237 
1238     if (mc->cpu_index_to_instance_props && mc->get_default_cpu_node_id) {
1239         ms->numa_state = g_new0(NumaState, 1);
1240         object_property_add_bool(obj, "hmat",
1241                                  machine_get_hmat, machine_set_hmat);
1242         object_property_set_description(obj, "hmat",
1243                                         "Set on/off to enable/disable "
1244                                         "ACPI Heterogeneous Memory Attribute "
1245                                         "Table (HMAT)");
1246     }
1247 
1248     /* default to mc->default_cpus */
1249     ms->smp.cpus = mc->default_cpus;
1250     ms->smp.max_cpus = mc->default_cpus;
1251     ms->smp.drawers = 1;
1252     ms->smp.books = 1;
1253     ms->smp.sockets = 1;
1254     ms->smp.dies = 1;
1255     ms->smp.clusters = 1;
1256     ms->smp.modules = 1;
1257     ms->smp.cores = 1;
1258     ms->smp.threads = 1;
1259 
1260     for (int i = 0; i < CACHE_LEVEL_AND_TYPE__MAX; i++) {
1261         ms->smp_cache.props[i].cache = (CacheLevelAndType)i;
1262         ms->smp_cache.props[i].topology = CPU_TOPOLOGY_LEVEL_DEFAULT;
1263     }
1264 
1265     machine_copy_boot_config(ms, &(BootConfiguration){ 0 });
1266 }
1267 
1268 static void machine_finalize(Object *obj)
1269 {
1270     MachineState *ms = MACHINE(obj);
1271 
1272     machine_free_boot_config(ms);
1273     g_free(ms->kernel_filename);
1274     g_free(ms->initrd_filename);
1275     g_free(ms->kernel_cmdline);
1276     g_free(ms->dtb);
1277     g_free(ms->dumpdtb);
1278     g_free(ms->dt_compatible);
1279     g_free(ms->firmware);
1280     g_free(ms->device_memory);
1281     g_free(ms->nvdimms_state);
1282     g_free(ms->numa_state);
1283     g_free(ms->audiodev);
1284 }
1285 
1286 bool machine_usb(MachineState *machine)
1287 {
1288     return machine->usb;
1289 }
1290 
1291 int machine_phandle_start(MachineState *machine)
1292 {
1293     return machine->phandle_start;
1294 }
1295 
1296 bool machine_dump_guest_core(MachineState *machine)
1297 {
1298     return machine->dump_guest_core;
1299 }
1300 
1301 bool machine_mem_merge(MachineState *machine)
1302 {
1303     return machine->mem_merge;
1304 }
1305 
1306 bool machine_require_guest_memfd(MachineState *machine)
1307 {
1308     return machine->cgs && machine->cgs->require_guest_memfd;
1309 }
1310 
1311 static char *cpu_slot_to_string(const CPUArchId *cpu)
1312 {
1313     GString *s = g_string_new(NULL);
1314     if (cpu->props.has_socket_id) {
1315         g_string_append_printf(s, "socket-id: %"PRId64, cpu->props.socket_id);
1316     }
1317     if (cpu->props.has_die_id) {
1318         if (s->len) {
1319             g_string_append_printf(s, ", ");
1320         }
1321         g_string_append_printf(s, "die-id: %"PRId64, cpu->props.die_id);
1322     }
1323     if (cpu->props.has_cluster_id) {
1324         if (s->len) {
1325             g_string_append_printf(s, ", ");
1326         }
1327         g_string_append_printf(s, "cluster-id: %"PRId64, cpu->props.cluster_id);
1328     }
1329     if (cpu->props.has_module_id) {
1330         if (s->len) {
1331             g_string_append_printf(s, ", ");
1332         }
1333         g_string_append_printf(s, "module-id: %"PRId64, cpu->props.module_id);
1334     }
1335     if (cpu->props.has_core_id) {
1336         if (s->len) {
1337             g_string_append_printf(s, ", ");
1338         }
1339         g_string_append_printf(s, "core-id: %"PRId64, cpu->props.core_id);
1340     }
1341     if (cpu->props.has_thread_id) {
1342         if (s->len) {
1343             g_string_append_printf(s, ", ");
1344         }
1345         g_string_append_printf(s, "thread-id: %"PRId64, cpu->props.thread_id);
1346     }
1347     return g_string_free(s, false);
1348 }
1349 
1350 static void numa_validate_initiator(NumaState *numa_state)
1351 {
1352     int i;
1353     NodeInfo *numa_info = numa_state->nodes;
1354 
1355     for (i = 0; i < numa_state->num_nodes; i++) {
1356         if (numa_info[i].initiator == MAX_NODES) {
1357             continue;
1358         }
1359 
1360         if (!numa_info[numa_info[i].initiator].present) {
1361             error_report("NUMA node %" PRIu16 " is missing, use "
1362                          "'-numa node' option to declare it first",
1363                          numa_info[i].initiator);
1364             exit(1);
1365         }
1366 
1367         if (!numa_info[numa_info[i].initiator].has_cpu) {
1368             error_report("The initiator of NUMA node %d is invalid", i);
1369             exit(1);
1370         }
1371     }
1372 }
1373 
1374 static void machine_numa_finish_cpu_init(MachineState *machine)
1375 {
1376     int i;
1377     bool default_mapping;
1378     GString *s = g_string_new(NULL);
1379     MachineClass *mc = MACHINE_GET_CLASS(machine);
1380     const CPUArchIdList *possible_cpus = mc->possible_cpu_arch_ids(machine);
1381 
1382     assert(machine->numa_state->num_nodes);
1383     for (i = 0; i < possible_cpus->len; i++) {
1384         if (possible_cpus->cpus[i].props.has_node_id) {
1385             break;
1386         }
1387     }
1388     default_mapping = (i == possible_cpus->len);
1389 
1390     for (i = 0; i < possible_cpus->len; i++) {
1391         const CPUArchId *cpu_slot = &possible_cpus->cpus[i];
1392 
1393         if (!cpu_slot->props.has_node_id) {
1394             /* fetch default mapping from board and enable it */
1395             CpuInstanceProperties props = cpu_slot->props;
1396 
1397             props.node_id = mc->get_default_cpu_node_id(machine, i);
1398             if (!default_mapping) {
1399                 /* record slots with not set mapping,
1400                  * TODO: make it hard error in future */
1401                 char *cpu_str = cpu_slot_to_string(cpu_slot);
1402                 g_string_append_printf(s, "%sCPU %d [%s]",
1403                                        s->len ? ", " : "", i, cpu_str);
1404                 g_free(cpu_str);
1405 
1406                 /* non mapped cpus used to fallback to node 0 */
1407                 props.node_id = 0;
1408             }
1409 
1410             props.has_node_id = true;
1411             machine_set_cpu_numa_node(machine, &props, &error_fatal);
1412         }
1413     }
1414 
1415     if (machine->numa_state->hmat_enabled) {
1416         numa_validate_initiator(machine->numa_state);
1417     }
1418 
1419     if (s->len && !qtest_enabled()) {
1420         warn_report("CPU(s) not present in any NUMA nodes: %s",
1421                     s->str);
1422         warn_report("All CPU(s) up to maxcpus should be described "
1423                     "in NUMA config, ability to start up with partial NUMA "
1424                     "mappings is obsoleted and will be removed in future");
1425     }
1426     g_string_free(s, true);
1427 }
1428 
1429 static void validate_cpu_cluster_to_numa_boundary(MachineState *ms)
1430 {
1431     MachineClass *mc = MACHINE_GET_CLASS(ms);
1432     NumaState *state = ms->numa_state;
1433     const CPUArchIdList *possible_cpus = mc->possible_cpu_arch_ids(ms);
1434     const CPUArchId *cpus = possible_cpus->cpus;
1435     int i, j;
1436 
1437     if (qtest_enabled() || state->num_nodes <= 1 || possible_cpus->len <= 1) {
1438         return;
1439     }
1440 
1441     /*
1442      * The Linux scheduling domain can't be parsed when the multiple CPUs
1443      * in one cluster have been associated with different NUMA nodes. However,
1444      * it's fine to associate one NUMA node with CPUs in different clusters.
1445      */
1446     for (i = 0; i < possible_cpus->len; i++) {
1447         for (j = i + 1; j < possible_cpus->len; j++) {
1448             if (cpus[i].props.has_socket_id &&
1449                 cpus[i].props.has_cluster_id &&
1450                 cpus[i].props.has_node_id &&
1451                 cpus[j].props.has_socket_id &&
1452                 cpus[j].props.has_cluster_id &&
1453                 cpus[j].props.has_node_id &&
1454                 cpus[i].props.socket_id == cpus[j].props.socket_id &&
1455                 cpus[i].props.cluster_id == cpus[j].props.cluster_id &&
1456                 cpus[i].props.node_id != cpus[j].props.node_id) {
1457                 warn_report("CPU-%d and CPU-%d in socket-%" PRId64 "-cluster-%" PRId64
1458                              " have been associated with node-%" PRId64 " and node-%" PRId64
1459                              " respectively. It can cause OSes like Linux to"
1460                              " misbehave", i, j, cpus[i].props.socket_id,
1461                              cpus[i].props.cluster_id, cpus[i].props.node_id,
1462                              cpus[j].props.node_id);
1463             }
1464         }
1465     }
1466 }
1467 
1468 MemoryRegion *machine_consume_memdev(MachineState *machine,
1469                                      HostMemoryBackend *backend)
1470 {
1471     MemoryRegion *ret = host_memory_backend_get_memory(backend);
1472 
1473     if (host_memory_backend_is_mapped(backend)) {
1474         error_report("memory backend %s can't be used multiple times.",
1475                      object_get_canonical_path_component(OBJECT(backend)));
1476         exit(EXIT_FAILURE);
1477     }
1478     host_memory_backend_set_mapped(backend, true);
1479     vmstate_register_ram_global(ret);
1480     return ret;
1481 }
1482 
1483 const char *machine_class_default_cpu_type(MachineClass *mc)
1484 {
1485     if (mc->valid_cpu_types && !mc->valid_cpu_types[1]) {
1486         /* Only a single CPU type allowed: use it as default. */
1487         return mc->valid_cpu_types[0];
1488     }
1489     return mc->default_cpu_type;
1490 }
1491 
1492 static bool is_cpu_type_supported(const MachineState *machine, Error **errp)
1493 {
1494     MachineClass *mc = MACHINE_GET_CLASS(machine);
1495     ObjectClass *oc = object_class_by_name(machine->cpu_type);
1496     CPUClass *cc;
1497     int i;
1498 
1499     /*
1500      * Check if the user specified CPU type is supported when the valid
1501      * CPU types have been determined. Note that the user specified CPU
1502      * type is provided through '-cpu' option.
1503      */
1504     if (mc->valid_cpu_types) {
1505         assert(mc->valid_cpu_types[0] != NULL);
1506         for (i = 0; mc->valid_cpu_types[i]; i++) {
1507             if (object_class_dynamic_cast(oc, mc->valid_cpu_types[i])) {
1508                 break;
1509             }
1510         }
1511 
1512         /* The user specified CPU type isn't valid */
1513         if (!mc->valid_cpu_types[i]) {
1514             g_autofree char *requested = cpu_model_from_type(machine->cpu_type);
1515             error_setg(errp, "Invalid CPU model: %s", requested);
1516             if (!mc->valid_cpu_types[1]) {
1517                 g_autofree char *model = cpu_model_from_type(
1518                                                  mc->valid_cpu_types[0]);
1519                 error_append_hint(errp, "The only valid type is: %s\n", model);
1520             } else {
1521                 error_append_hint(errp, "The valid models are: ");
1522                 for (i = 0; mc->valid_cpu_types[i]; i++) {
1523                     g_autofree char *model = cpu_model_from_type(
1524                                                  mc->valid_cpu_types[i]);
1525                     error_append_hint(errp, "%s%s",
1526                                       model,
1527                                       mc->valid_cpu_types[i + 1] ? ", " : "");
1528                 }
1529                 error_append_hint(errp, "\n");
1530             }
1531 
1532             return false;
1533         }
1534     }
1535 
1536     /* Check if CPU type is deprecated and warn if so */
1537     cc = CPU_CLASS(oc);
1538     assert(cc != NULL);
1539     if (cc->deprecation_note) {
1540         warn_report("CPU model %s is deprecated -- %s",
1541                     machine->cpu_type, cc->deprecation_note);
1542     }
1543 
1544     return true;
1545 }
1546 
1547 void machine_run_board_init(MachineState *machine, const char *mem_path, Error **errp)
1548 {
1549     ERRP_GUARD();
1550     MachineClass *machine_class = MACHINE_GET_CLASS(machine);
1551 
1552     /* This checkpoint is required by replay to separate prior clock
1553        reading from the other reads, because timer polling functions query
1554        clock values from the log. */
1555     replay_checkpoint(CHECKPOINT_INIT);
1556 
1557     if (!xen_enabled()) {
1558         /* On 32-bit hosts, QEMU is limited by virtual address space */
1559         if (machine->ram_size > (2047 << 20) && HOST_LONG_BITS == 32) {
1560             error_setg(errp, "at most 2047 MB RAM can be simulated");
1561             return;
1562         }
1563     }
1564 
1565     if (machine->memdev) {
1566         ram_addr_t backend_size = object_property_get_uint(OBJECT(machine->memdev),
1567                                                            "size",  &error_abort);
1568         if (backend_size != machine->ram_size) {
1569             error_setg(errp, "Machine memory size does not match the size of the memory backend");
1570             return;
1571         }
1572     } else if (machine_class->default_ram_id && machine->ram_size &&
1573                numa_uses_legacy_mem()) {
1574         if (object_property_find(object_get_objects_root(),
1575                                  machine_class->default_ram_id)) {
1576             error_setg(errp, "object's id '%s' is reserved for the default"
1577                 " RAM backend, it can't be used for any other purposes",
1578                 machine_class->default_ram_id);
1579             error_append_hint(errp,
1580                 "Change the object's 'id' to something else or disable"
1581                 " automatic creation of the default RAM backend by setting"
1582                 " 'memory-backend=%s' with '-machine'.\n",
1583                 machine_class->default_ram_id);
1584             return;
1585         }
1586 
1587         if (!machine_class->create_default_memdev(current_machine, mem_path,
1588                                                   errp)) {
1589             return;
1590         }
1591     }
1592 
1593     if (machine->numa_state) {
1594         numa_complete_configuration(machine);
1595         if (machine->numa_state->num_nodes) {
1596             machine_numa_finish_cpu_init(machine);
1597             if (machine_class->cpu_cluster_has_numa_boundary) {
1598                 validate_cpu_cluster_to_numa_boundary(machine);
1599             }
1600         }
1601     }
1602 
1603     if (!machine->ram && machine->memdev) {
1604         machine->ram = machine_consume_memdev(machine, machine->memdev);
1605     }
1606 
1607     /* Check if the CPU type is supported */
1608     if (machine->cpu_type && !is_cpu_type_supported(machine, errp)) {
1609         return;
1610     }
1611 
1612     if (machine->cgs) {
1613         /*
1614          * With confidential guests, the host can't see the real
1615          * contents of RAM, so there's no point in it trying to merge
1616          * areas.
1617          */
1618         machine_set_mem_merge(OBJECT(machine), false, &error_abort);
1619 
1620         /*
1621          * Virtio devices can't count on directly accessing guest
1622          * memory, so they need iommu_platform=on to use normal DMA
1623          * mechanisms.  That requires also disabling legacy virtio
1624          * support for those virtio pci devices which allow it.
1625          */
1626         object_register_sugar_prop(TYPE_VIRTIO_PCI, "disable-legacy",
1627                                    "on", true);
1628         object_register_sugar_prop(TYPE_VIRTIO_DEVICE, "iommu_platform",
1629                                    "on", false);
1630     }
1631 
1632     accel_init_interfaces(ACCEL_GET_CLASS(machine->accelerator));
1633     machine_class->init(machine);
1634     phase_advance(PHASE_MACHINE_INITIALIZED);
1635 }
1636 
1637 static NotifierList machine_init_done_notifiers =
1638     NOTIFIER_LIST_INITIALIZER(machine_init_done_notifiers);
1639 
1640 void qemu_add_machine_init_done_notifier(Notifier *notify)
1641 {
1642     notifier_list_add(&machine_init_done_notifiers, notify);
1643     if (phase_check(PHASE_MACHINE_READY)) {
1644         notify->notify(notify, NULL);
1645     }
1646 }
1647 
1648 void qemu_remove_machine_init_done_notifier(Notifier *notify)
1649 {
1650     notifier_remove(notify);
1651 }
1652 
1653 void qdev_machine_creation_done(void)
1654 {
1655     cpu_synchronize_all_post_init();
1656 
1657     if (current_machine->boot_config.once) {
1658         qemu_boot_set(current_machine->boot_config.once, &error_fatal);
1659         qemu_register_reset(restore_boot_order, g_strdup(current_machine->boot_config.order));
1660     }
1661 
1662     /*
1663      * ok, initial machine setup is done, starting from now we can
1664      * only create hotpluggable devices
1665      */
1666     phase_advance(PHASE_MACHINE_READY);
1667     qdev_assert_realized_properly();
1668 
1669     /* TODO: once all bus devices are qdevified, this should be done
1670      * when bus is created by qdev.c */
1671     /*
1672      * This is where we arrange for the sysbus to be reset when the
1673      * whole simulation is reset. In turn, resetting the sysbus will cause
1674      * all devices hanging off it (and all their child buses, recursively)
1675      * to be reset. Note that this will *not* reset any Device objects
1676      * which are not attached to some part of the qbus tree!
1677      */
1678     qemu_register_resettable(OBJECT(sysbus_get_default()));
1679 
1680     notifier_list_notify(&machine_init_done_notifiers, NULL);
1681 
1682     if (rom_check_and_register_reset() != 0) {
1683         exit(1);
1684     }
1685 
1686     replay_start();
1687 
1688     /* This checkpoint is required by replay to separate prior clock
1689        reading from the other reads, because timer polling functions query
1690        clock values from the log. */
1691     replay_checkpoint(CHECKPOINT_RESET);
1692     qemu_system_reset(SHUTDOWN_CAUSE_NONE);
1693     register_global_state();
1694 }
1695 
1696 static const TypeInfo machine_info = {
1697     .name = TYPE_MACHINE,
1698     .parent = TYPE_OBJECT,
1699     .abstract = true,
1700     .class_size = sizeof(MachineClass),
1701     .class_init    = machine_class_init,
1702     .class_base_init = machine_class_base_init,
1703     .instance_size = sizeof(MachineState),
1704     .instance_init = machine_initfn,
1705     .instance_finalize = machine_finalize,
1706 };
1707 
1708 static void machine_register_types(void)
1709 {
1710     type_register_static(&machine_info);
1711 }
1712 
1713 type_init(machine_register_types)
1714