xref: /openbmc/qemu/hw/core/machine.c (revision 6e510855)
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 "qemu/accel.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 "qom/object_interfaces.h"
25 #include "hw/sysbus.h"
26 #include "sysemu/cpus.h"
27 #include "sysemu/sysemu.h"
28 #include "sysemu/reset.h"
29 #include "sysemu/runstate.h"
30 #include "sysemu/numa.h"
31 #include "sysemu/xen.h"
32 #include "qemu/error-report.h"
33 #include "sysemu/qtest.h"
34 #include "hw/pci/pci.h"
35 #include "hw/mem/nvdimm.h"
36 #include "migration/global_state.h"
37 #include "migration/vmstate.h"
38 #include "exec/confidential-guest-support.h"
39 #include "hw/virtio/virtio.h"
40 #include "hw/virtio/virtio-pci.h"
41 
42 GlobalProperty hw_compat_8_0[] = {
43     { "migration", "multifd-flush-after-each-section", "on"},
44     { TYPE_PCI_DEVICE, "x-pcie-ari-nextfn-1", "on" },
45 };
46 const size_t hw_compat_8_0_len = G_N_ELEMENTS(hw_compat_8_0);
47 
48 GlobalProperty hw_compat_7_2[] = {
49     { "e1000e", "migrate-timadj", "off" },
50     { "virtio-mem", "x-early-migration", "false" },
51     { "migration", "x-preempt-pre-7-2", "true" },
52     { TYPE_PCI_DEVICE, "x-pcie-err-unc-mask", "off" },
53 };
54 const size_t hw_compat_7_2_len = G_N_ELEMENTS(hw_compat_7_2);
55 
56 GlobalProperty hw_compat_7_1[] = {
57     { "virtio-device", "queue_reset", "false" },
58     { "virtio-rng-pci", "vectors", "0" },
59     { "virtio-rng-pci-transitional", "vectors", "0" },
60     { "virtio-rng-pci-non-transitional", "vectors", "0" },
61 };
62 const size_t hw_compat_7_1_len = G_N_ELEMENTS(hw_compat_7_1);
63 
64 GlobalProperty hw_compat_7_0[] = {
65     { "arm-gicv3-common", "force-8-bit-prio", "on" },
66     { "nvme-ns", "eui64-default", "on"},
67 };
68 const size_t hw_compat_7_0_len = G_N_ELEMENTS(hw_compat_7_0);
69 
70 GlobalProperty hw_compat_6_2[] = {
71     { "PIIX4_PM", "x-not-migrate-acpi-index", "on"},
72 };
73 const size_t hw_compat_6_2_len = G_N_ELEMENTS(hw_compat_6_2);
74 
75 GlobalProperty hw_compat_6_1[] = {
76     { "vhost-user-vsock-device", "seqpacket", "off" },
77     { "nvme-ns", "shared", "off" },
78 };
79 const size_t hw_compat_6_1_len = G_N_ELEMENTS(hw_compat_6_1);
80 
81 GlobalProperty hw_compat_6_0[] = {
82     { "gpex-pcihost", "allow-unmapped-accesses", "false" },
83     { "i8042", "extended-state", "false"},
84     { "nvme-ns", "eui64-default", "off"},
85     { "e1000", "init-vet", "off" },
86     { "e1000e", "init-vet", "off" },
87     { "vhost-vsock-device", "seqpacket", "off" },
88 };
89 const size_t hw_compat_6_0_len = G_N_ELEMENTS(hw_compat_6_0);
90 
91 GlobalProperty hw_compat_5_2[] = {
92     { "ICH9-LPC", "smm-compat", "on"},
93     { "PIIX4_PM", "smm-compat", "on"},
94     { "virtio-blk-device", "report-discard-granularity", "off" },
95     { "virtio-net-pci-base", "vectors", "3"},
96 };
97 const size_t hw_compat_5_2_len = G_N_ELEMENTS(hw_compat_5_2);
98 
99 GlobalProperty hw_compat_5_1[] = {
100     { "vhost-scsi", "num_queues", "1"},
101     { "vhost-user-blk", "num-queues", "1"},
102     { "vhost-user-scsi", "num_queues", "1"},
103     { "virtio-blk-device", "num-queues", "1"},
104     { "virtio-scsi-device", "num_queues", "1"},
105     { "nvme", "use-intel-id", "on"},
106     { "pvpanic", "events", "1"}, /* PVPANIC_PANICKED */
107     { "pl011", "migrate-clk", "off" },
108     { "virtio-pci", "x-ats-page-aligned", "off"},
109 };
110 const size_t hw_compat_5_1_len = G_N_ELEMENTS(hw_compat_5_1);
111 
112 GlobalProperty hw_compat_5_0[] = {
113     { "pci-host-bridge", "x-config-reg-migration-enabled", "off" },
114     { "virtio-balloon-device", "page-poison", "false" },
115     { "vmport", "x-read-set-eax", "off" },
116     { "vmport", "x-signal-unsupported-cmd", "off" },
117     { "vmport", "x-report-vmx-type", "off" },
118     { "vmport", "x-cmds-v2", "off" },
119     { "virtio-device", "x-disable-legacy-check", "true" },
120 };
121 const size_t hw_compat_5_0_len = G_N_ELEMENTS(hw_compat_5_0);
122 
123 GlobalProperty hw_compat_4_2[] = {
124     { "virtio-blk-device", "queue-size", "128"},
125     { "virtio-scsi-device", "virtqueue_size", "128"},
126     { "virtio-blk-device", "x-enable-wce-if-config-wce", "off" },
127     { "virtio-blk-device", "seg-max-adjust", "off"},
128     { "virtio-scsi-device", "seg_max_adjust", "off"},
129     { "vhost-blk-device", "seg_max_adjust", "off"},
130     { "usb-host", "suppress-remote-wake", "off" },
131     { "usb-redir", "suppress-remote-wake", "off" },
132     { "qxl", "revision", "4" },
133     { "qxl-vga", "revision", "4" },
134     { "fw_cfg", "acpi-mr-restore", "false" },
135     { "virtio-device", "use-disabled-flag", "false" },
136 };
137 const size_t hw_compat_4_2_len = G_N_ELEMENTS(hw_compat_4_2);
138 
139 GlobalProperty hw_compat_4_1[] = {
140     { "virtio-pci", "x-pcie-flr-init", "off" },
141 };
142 const size_t hw_compat_4_1_len = G_N_ELEMENTS(hw_compat_4_1);
143 
144 GlobalProperty hw_compat_4_0[] = {
145     { "VGA",            "edid", "false" },
146     { "secondary-vga",  "edid", "false" },
147     { "bochs-display",  "edid", "false" },
148     { "virtio-vga",     "edid", "false" },
149     { "virtio-gpu-device", "edid", "false" },
150     { "virtio-device", "use-started", "false" },
151     { "virtio-balloon-device", "qemu-4-0-config-size", "true" },
152     { "pl031", "migrate-tick-offset", "false" },
153 };
154 const size_t hw_compat_4_0_len = G_N_ELEMENTS(hw_compat_4_0);
155 
156 GlobalProperty hw_compat_3_1[] = {
157     { "pcie-root-port", "x-speed", "2_5" },
158     { "pcie-root-port", "x-width", "1" },
159     { "memory-backend-file", "x-use-canonical-path-for-ramblock-id", "true" },
160     { "memory-backend-memfd", "x-use-canonical-path-for-ramblock-id", "true" },
161     { "tpm-crb", "ppi", "false" },
162     { "tpm-tis", "ppi", "false" },
163     { "usb-kbd", "serial", "42" },
164     { "usb-mouse", "serial", "42" },
165     { "usb-tablet", "serial", "42" },
166     { "virtio-blk-device", "discard", "false" },
167     { "virtio-blk-device", "write-zeroes", "false" },
168     { "virtio-balloon-device", "qemu-4-0-config-size", "false" },
169     { "pcie-root-port-base", "disable-acs", "true" }, /* Added in 4.1 */
170 };
171 const size_t hw_compat_3_1_len = G_N_ELEMENTS(hw_compat_3_1);
172 
173 GlobalProperty hw_compat_3_0[] = {};
174 const size_t hw_compat_3_0_len = G_N_ELEMENTS(hw_compat_3_0);
175 
176 GlobalProperty hw_compat_2_12[] = {
177     { "migration", "decompress-error-check", "off" },
178     { "hda-audio", "use-timer", "false" },
179     { "cirrus-vga", "global-vmstate", "true" },
180     { "VGA", "global-vmstate", "true" },
181     { "vmware-svga", "global-vmstate", "true" },
182     { "qxl-vga", "global-vmstate", "true" },
183 };
184 const size_t hw_compat_2_12_len = G_N_ELEMENTS(hw_compat_2_12);
185 
186 GlobalProperty hw_compat_2_11[] = {
187     { "hpet", "hpet-offset-saved", "false" },
188     { "virtio-blk-pci", "vectors", "2" },
189     { "vhost-user-blk-pci", "vectors", "2" },
190     { "e1000", "migrate_tso_props", "off" },
191 };
192 const size_t hw_compat_2_11_len = G_N_ELEMENTS(hw_compat_2_11);
193 
194 GlobalProperty hw_compat_2_10[] = {
195     { "virtio-mouse-device", "wheel-axis", "false" },
196     { "virtio-tablet-device", "wheel-axis", "false" },
197 };
198 const size_t hw_compat_2_10_len = G_N_ELEMENTS(hw_compat_2_10);
199 
200 GlobalProperty hw_compat_2_9[] = {
201     { "pci-bridge", "shpc", "off" },
202     { "intel-iommu", "pt", "off" },
203     { "virtio-net-device", "x-mtu-bypass-backend", "off" },
204     { "pcie-root-port", "x-migrate-msix", "false" },
205 };
206 const size_t hw_compat_2_9_len = G_N_ELEMENTS(hw_compat_2_9);
207 
208 GlobalProperty hw_compat_2_8[] = {
209     { "fw_cfg_mem", "x-file-slots", "0x10" },
210     { "fw_cfg_io", "x-file-slots", "0x10" },
211     { "pflash_cfi01", "old-multiple-chip-handling", "on" },
212     { "pci-bridge", "shpc", "on" },
213     { TYPE_PCI_DEVICE, "x-pcie-extcap-init", "off" },
214     { "virtio-pci", "x-pcie-deverr-init", "off" },
215     { "virtio-pci", "x-pcie-lnkctl-init", "off" },
216     { "virtio-pci", "x-pcie-pm-init", "off" },
217     { "cirrus-vga", "vgamem_mb", "8" },
218     { "isa-cirrus-vga", "vgamem_mb", "8" },
219 };
220 const size_t hw_compat_2_8_len = G_N_ELEMENTS(hw_compat_2_8);
221 
222 GlobalProperty hw_compat_2_7[] = {
223     { "virtio-pci", "page-per-vq", "on" },
224     { "virtio-serial-device", "emergency-write", "off" },
225     { "ioapic", "version", "0x11" },
226     { "intel-iommu", "x-buggy-eim", "true" },
227     { "virtio-pci", "x-ignore-backend-features", "on" },
228 };
229 const size_t hw_compat_2_7_len = G_N_ELEMENTS(hw_compat_2_7);
230 
231 GlobalProperty hw_compat_2_6[] = {
232     { "virtio-mmio", "format_transport_address", "off" },
233     /* Optional because not all virtio-pci devices support legacy mode */
234     { "virtio-pci", "disable-modern", "on",  .optional = true },
235     { "virtio-pci", "disable-legacy", "off", .optional = true },
236 };
237 const size_t hw_compat_2_6_len = G_N_ELEMENTS(hw_compat_2_6);
238 
239 GlobalProperty hw_compat_2_5[] = {
240     { "isa-fdc", "fallback", "144" },
241     { "pvscsi", "x-old-pci-configuration", "on" },
242     { "pvscsi", "x-disable-pcie", "on" },
243     { "vmxnet3", "x-old-msi-offsets", "on" },
244     { "vmxnet3", "x-disable-pcie", "on" },
245 };
246 const size_t hw_compat_2_5_len = G_N_ELEMENTS(hw_compat_2_5);
247 
248 GlobalProperty hw_compat_2_4[] = {
249     /* Optional because the 'scsi' property is Linux-only */
250     { "virtio-blk-device", "scsi", "true", .optional = true },
251     { "e1000", "extra_mac_registers", "off" },
252     { "virtio-pci", "x-disable-pcie", "on" },
253     { "virtio-pci", "migrate-extra", "off" },
254     { "fw_cfg_mem", "dma_enabled", "off" },
255     { "fw_cfg_io", "dma_enabled", "off" }
256 };
257 const size_t hw_compat_2_4_len = G_N_ELEMENTS(hw_compat_2_4);
258 
259 GlobalProperty hw_compat_2_3[] = {
260     { "virtio-blk-pci", "any_layout", "off" },
261     { "virtio-balloon-pci", "any_layout", "off" },
262     { "virtio-serial-pci", "any_layout", "off" },
263     { "virtio-9p-pci", "any_layout", "off" },
264     { "virtio-rng-pci", "any_layout", "off" },
265     { TYPE_PCI_DEVICE, "x-pcie-lnksta-dllla", "off" },
266     { "migration", "send-configuration", "off" },
267     { "migration", "send-section-footer", "off" },
268     { "migration", "store-global-state", "off" },
269 };
270 const size_t hw_compat_2_3_len = G_N_ELEMENTS(hw_compat_2_3);
271 
272 GlobalProperty hw_compat_2_2[] = {};
273 const size_t hw_compat_2_2_len = G_N_ELEMENTS(hw_compat_2_2);
274 
275 GlobalProperty hw_compat_2_1[] = {
276     { "intel-hda", "old_msi_addr", "on" },
277     { "VGA", "qemu-extended-regs", "off" },
278     { "secondary-vga", "qemu-extended-regs", "off" },
279     { "virtio-scsi-pci", "any_layout", "off" },
280     { "usb-mouse", "usb_version", "1" },
281     { "usb-kbd", "usb_version", "1" },
282     { "virtio-pci", "virtio-pci-bus-master-bug-migration", "on" },
283 };
284 const size_t hw_compat_2_1_len = G_N_ELEMENTS(hw_compat_2_1);
285 
286 MachineState *current_machine;
287 
288 static char *machine_get_kernel(Object *obj, Error **errp)
289 {
290     MachineState *ms = MACHINE(obj);
291 
292     return g_strdup(ms->kernel_filename);
293 }
294 
295 static void machine_set_kernel(Object *obj, const char *value, Error **errp)
296 {
297     MachineState *ms = MACHINE(obj);
298 
299     g_free(ms->kernel_filename);
300     ms->kernel_filename = g_strdup(value);
301 }
302 
303 static char *machine_get_initrd(Object *obj, Error **errp)
304 {
305     MachineState *ms = MACHINE(obj);
306 
307     return g_strdup(ms->initrd_filename);
308 }
309 
310 static void machine_set_initrd(Object *obj, const char *value, Error **errp)
311 {
312     MachineState *ms = MACHINE(obj);
313 
314     g_free(ms->initrd_filename);
315     ms->initrd_filename = g_strdup(value);
316 }
317 
318 static char *machine_get_append(Object *obj, Error **errp)
319 {
320     MachineState *ms = MACHINE(obj);
321 
322     return g_strdup(ms->kernel_cmdline);
323 }
324 
325 static void machine_set_append(Object *obj, const char *value, Error **errp)
326 {
327     MachineState *ms = MACHINE(obj);
328 
329     g_free(ms->kernel_cmdline);
330     ms->kernel_cmdline = g_strdup(value);
331 }
332 
333 static char *machine_get_dtb(Object *obj, Error **errp)
334 {
335     MachineState *ms = MACHINE(obj);
336 
337     return g_strdup(ms->dtb);
338 }
339 
340 static void machine_set_dtb(Object *obj, const char *value, Error **errp)
341 {
342     MachineState *ms = MACHINE(obj);
343 
344     g_free(ms->dtb);
345     ms->dtb = g_strdup(value);
346 }
347 
348 static char *machine_get_dumpdtb(Object *obj, Error **errp)
349 {
350     MachineState *ms = MACHINE(obj);
351 
352     return g_strdup(ms->dumpdtb);
353 }
354 
355 static void machine_set_dumpdtb(Object *obj, const char *value, Error **errp)
356 {
357     MachineState *ms = MACHINE(obj);
358 
359     g_free(ms->dumpdtb);
360     ms->dumpdtb = g_strdup(value);
361 }
362 
363 static void machine_get_phandle_start(Object *obj, Visitor *v,
364                                       const char *name, void *opaque,
365                                       Error **errp)
366 {
367     MachineState *ms = MACHINE(obj);
368     int64_t value = ms->phandle_start;
369 
370     visit_type_int(v, name, &value, errp);
371 }
372 
373 static void machine_set_phandle_start(Object *obj, Visitor *v,
374                                       const char *name, void *opaque,
375                                       Error **errp)
376 {
377     MachineState *ms = MACHINE(obj);
378     int64_t value;
379 
380     if (!visit_type_int(v, name, &value, errp)) {
381         return;
382     }
383 
384     ms->phandle_start = value;
385 }
386 
387 static char *machine_get_dt_compatible(Object *obj, Error **errp)
388 {
389     MachineState *ms = MACHINE(obj);
390 
391     return g_strdup(ms->dt_compatible);
392 }
393 
394 static void machine_set_dt_compatible(Object *obj, const char *value, Error **errp)
395 {
396     MachineState *ms = MACHINE(obj);
397 
398     g_free(ms->dt_compatible);
399     ms->dt_compatible = g_strdup(value);
400 }
401 
402 static bool machine_get_dump_guest_core(Object *obj, Error **errp)
403 {
404     MachineState *ms = MACHINE(obj);
405 
406     return ms->dump_guest_core;
407 }
408 
409 static void machine_set_dump_guest_core(Object *obj, bool value, Error **errp)
410 {
411     MachineState *ms = MACHINE(obj);
412 
413     ms->dump_guest_core = value;
414 }
415 
416 static bool machine_get_mem_merge(Object *obj, Error **errp)
417 {
418     MachineState *ms = MACHINE(obj);
419 
420     return ms->mem_merge;
421 }
422 
423 static void machine_set_mem_merge(Object *obj, bool value, Error **errp)
424 {
425     MachineState *ms = MACHINE(obj);
426 
427     ms->mem_merge = value;
428 }
429 
430 static bool machine_get_usb(Object *obj, Error **errp)
431 {
432     MachineState *ms = MACHINE(obj);
433 
434     return ms->usb;
435 }
436 
437 static void machine_set_usb(Object *obj, bool value, Error **errp)
438 {
439     MachineState *ms = MACHINE(obj);
440 
441     ms->usb = value;
442     ms->usb_disabled = !value;
443 }
444 
445 static bool machine_get_graphics(Object *obj, Error **errp)
446 {
447     MachineState *ms = MACHINE(obj);
448 
449     return ms->enable_graphics;
450 }
451 
452 static void machine_set_graphics(Object *obj, bool value, Error **errp)
453 {
454     MachineState *ms = MACHINE(obj);
455 
456     ms->enable_graphics = value;
457 }
458 
459 static char *machine_get_firmware(Object *obj, Error **errp)
460 {
461     MachineState *ms = MACHINE(obj);
462 
463     return g_strdup(ms->firmware);
464 }
465 
466 static void machine_set_firmware(Object *obj, const char *value, Error **errp)
467 {
468     MachineState *ms = MACHINE(obj);
469 
470     g_free(ms->firmware);
471     ms->firmware = g_strdup(value);
472 }
473 
474 static void machine_set_suppress_vmdesc(Object *obj, bool value, Error **errp)
475 {
476     MachineState *ms = MACHINE(obj);
477 
478     ms->suppress_vmdesc = value;
479 }
480 
481 static bool machine_get_suppress_vmdesc(Object *obj, Error **errp)
482 {
483     MachineState *ms = MACHINE(obj);
484 
485     return ms->suppress_vmdesc;
486 }
487 
488 static char *machine_get_memory_encryption(Object *obj, Error **errp)
489 {
490     MachineState *ms = MACHINE(obj);
491 
492     if (ms->cgs) {
493         return g_strdup(object_get_canonical_path_component(OBJECT(ms->cgs)));
494     }
495 
496     return NULL;
497 }
498 
499 static void machine_set_memory_encryption(Object *obj, const char *value,
500                                         Error **errp)
501 {
502     Object *cgs =
503         object_resolve_path_component(object_get_objects_root(), value);
504 
505     if (!cgs) {
506         error_setg(errp, "No such memory encryption object '%s'", value);
507         return;
508     }
509 
510     object_property_set_link(obj, "confidential-guest-support", cgs, errp);
511 }
512 
513 static void machine_check_confidential_guest_support(const Object *obj,
514                                                      const char *name,
515                                                      Object *new_target,
516                                                      Error **errp)
517 {
518     /*
519      * So far the only constraint is that the target has the
520      * TYPE_CONFIDENTIAL_GUEST_SUPPORT interface, and that's checked
521      * by the QOM core
522      */
523 }
524 
525 static bool machine_get_nvdimm(Object *obj, Error **errp)
526 {
527     MachineState *ms = MACHINE(obj);
528 
529     return ms->nvdimms_state->is_enabled;
530 }
531 
532 static void machine_set_nvdimm(Object *obj, bool value, Error **errp)
533 {
534     MachineState *ms = MACHINE(obj);
535 
536     ms->nvdimms_state->is_enabled = value;
537 }
538 
539 static bool machine_get_hmat(Object *obj, Error **errp)
540 {
541     MachineState *ms = MACHINE(obj);
542 
543     return ms->numa_state->hmat_enabled;
544 }
545 
546 static void machine_set_hmat(Object *obj, bool value, Error **errp)
547 {
548     MachineState *ms = MACHINE(obj);
549 
550     ms->numa_state->hmat_enabled = value;
551 }
552 
553 static void machine_get_mem(Object *obj, Visitor *v, const char *name,
554                             void *opaque, Error **errp)
555 {
556     MachineState *ms = MACHINE(obj);
557     MemorySizeConfiguration mem = {
558         .has_size = true,
559         .size = ms->ram_size,
560         .has_max_size = !!ms->ram_slots,
561         .max_size = ms->maxram_size,
562         .has_slots = !!ms->ram_slots,
563         .slots = ms->ram_slots,
564     };
565     MemorySizeConfiguration *p_mem = &mem;
566 
567     visit_type_MemorySizeConfiguration(v, name, &p_mem, &error_abort);
568 }
569 
570 static void machine_set_mem(Object *obj, Visitor *v, const char *name,
571                             void *opaque, Error **errp)
572 {
573     ERRP_GUARD();
574     MachineState *ms = MACHINE(obj);
575     MachineClass *mc = MACHINE_GET_CLASS(obj);
576     MemorySizeConfiguration *mem;
577 
578     if (!visit_type_MemorySizeConfiguration(v, name, &mem, errp)) {
579         return;
580     }
581 
582     if (!mem->has_size) {
583         mem->has_size = true;
584         mem->size = mc->default_ram_size;
585     }
586     mem->size = QEMU_ALIGN_UP(mem->size, 8192);
587     if (mc->fixup_ram_size) {
588         mem->size = mc->fixup_ram_size(mem->size);
589     }
590     if ((ram_addr_t)mem->size != mem->size) {
591         error_setg(errp, "ram size too large");
592         goto out_free;
593     }
594 
595     if (mem->has_max_size) {
596         if (mem->max_size < mem->size) {
597             error_setg(errp, "invalid value of maxmem: "
598                        "maximum memory size (0x%" PRIx64 ") must be at least "
599                        "the initial memory size (0x%" PRIx64 ")",
600                        mem->max_size, mem->size);
601             goto out_free;
602         }
603         if (mem->has_slots && mem->slots && mem->max_size == mem->size) {
604             error_setg(errp, "invalid value of maxmem: "
605                        "memory slots were specified but maximum memory size "
606                        "(0x%" PRIx64 ") is equal to the initial memory size "
607                        "(0x%" PRIx64 ")", mem->max_size, mem->size);
608             goto out_free;
609         }
610         ms->maxram_size = mem->max_size;
611     } else {
612         if (mem->has_slots) {
613             error_setg(errp, "slots specified but no max-size");
614             goto out_free;
615         }
616         ms->maxram_size = mem->size;
617     }
618     ms->ram_size = mem->size;
619     ms->ram_slots = mem->has_slots ? mem->slots : 0;
620 out_free:
621     qapi_free_MemorySizeConfiguration(mem);
622 }
623 
624 static char *machine_get_nvdimm_persistence(Object *obj, Error **errp)
625 {
626     MachineState *ms = MACHINE(obj);
627 
628     return g_strdup(ms->nvdimms_state->persistence_string);
629 }
630 
631 static void machine_set_nvdimm_persistence(Object *obj, const char *value,
632                                            Error **errp)
633 {
634     MachineState *ms = MACHINE(obj);
635     NVDIMMState *nvdimms_state = ms->nvdimms_state;
636 
637     if (strcmp(value, "cpu") == 0) {
638         nvdimms_state->persistence = 3;
639     } else if (strcmp(value, "mem-ctrl") == 0) {
640         nvdimms_state->persistence = 2;
641     } else {
642         error_setg(errp, "-machine nvdimm-persistence=%s: unsupported option",
643                    value);
644         return;
645     }
646 
647     g_free(nvdimms_state->persistence_string);
648     nvdimms_state->persistence_string = g_strdup(value);
649 }
650 
651 void machine_class_allow_dynamic_sysbus_dev(MachineClass *mc, const char *type)
652 {
653     QAPI_LIST_PREPEND(mc->allowed_dynamic_sysbus_devices, g_strdup(type));
654 }
655 
656 bool device_is_dynamic_sysbus(MachineClass *mc, DeviceState *dev)
657 {
658     Object *obj = OBJECT(dev);
659 
660     if (!object_dynamic_cast(obj, TYPE_SYS_BUS_DEVICE)) {
661         return false;
662     }
663 
664     return device_type_is_dynamic_sysbus(mc, object_get_typename(obj));
665 }
666 
667 bool device_type_is_dynamic_sysbus(MachineClass *mc, const char *type)
668 {
669     bool allowed = false;
670     strList *wl;
671     ObjectClass *klass = object_class_by_name(type);
672 
673     for (wl = mc->allowed_dynamic_sysbus_devices;
674          !allowed && wl;
675          wl = wl->next) {
676         allowed |= !!object_class_dynamic_cast(klass, wl->value);
677     }
678 
679     return allowed;
680 }
681 
682 HotpluggableCPUList *machine_query_hotpluggable_cpus(MachineState *machine)
683 {
684     int i;
685     HotpluggableCPUList *head = NULL;
686     MachineClass *mc = MACHINE_GET_CLASS(machine);
687 
688     /* force board to initialize possible_cpus if it hasn't been done yet */
689     mc->possible_cpu_arch_ids(machine);
690 
691     for (i = 0; i < machine->possible_cpus->len; i++) {
692         Object *cpu;
693         HotpluggableCPU *cpu_item = g_new0(typeof(*cpu_item), 1);
694 
695         cpu_item->type = g_strdup(machine->possible_cpus->cpus[i].type);
696         cpu_item->vcpus_count = machine->possible_cpus->cpus[i].vcpus_count;
697         cpu_item->props = g_memdup(&machine->possible_cpus->cpus[i].props,
698                                    sizeof(*cpu_item->props));
699 
700         cpu = machine->possible_cpus->cpus[i].cpu;
701         if (cpu) {
702             cpu_item->qom_path = object_get_canonical_path(cpu);
703         }
704         QAPI_LIST_PREPEND(head, cpu_item);
705     }
706     return head;
707 }
708 
709 /**
710  * machine_set_cpu_numa_node:
711  * @machine: machine object to modify
712  * @props: specifies which cpu objects to assign to
713  *         numa node specified by @props.node_id
714  * @errp: if an error occurs, a pointer to an area to store the error
715  *
716  * Associate NUMA node specified by @props.node_id with cpu slots that
717  * match socket/core/thread-ids specified by @props. It's recommended to use
718  * query-hotpluggable-cpus.props values to specify affected cpu slots,
719  * which would lead to exact 1:1 mapping of cpu slots to NUMA node.
720  *
721  * However for CLI convenience it's possible to pass in subset of properties,
722  * which would affect all cpu slots that match it.
723  * Ex for pc machine:
724  *    -smp 4,cores=2,sockets=2 -numa node,nodeid=0 -numa node,nodeid=1 \
725  *    -numa cpu,node-id=0,socket_id=0 \
726  *    -numa cpu,node-id=1,socket_id=1
727  * will assign all child cores of socket 0 to node 0 and
728  * of socket 1 to node 1.
729  *
730  * On attempt of reassigning (already assigned) cpu slot to another NUMA node,
731  * return error.
732  * Empty subset is disallowed and function will return with error in this case.
733  */
734 void machine_set_cpu_numa_node(MachineState *machine,
735                                const CpuInstanceProperties *props, Error **errp)
736 {
737     MachineClass *mc = MACHINE_GET_CLASS(machine);
738     NodeInfo *numa_info = machine->numa_state->nodes;
739     bool match = false;
740     int i;
741 
742     if (!mc->possible_cpu_arch_ids) {
743         error_setg(errp, "mapping of CPUs to NUMA node is not supported");
744         return;
745     }
746 
747     /* disabling node mapping is not supported, forbid it */
748     assert(props->has_node_id);
749 
750     /* force board to initialize possible_cpus if it hasn't been done yet */
751     mc->possible_cpu_arch_ids(machine);
752 
753     for (i = 0; i < machine->possible_cpus->len; i++) {
754         CPUArchId *slot = &machine->possible_cpus->cpus[i];
755 
756         /* reject unsupported by board properties */
757         if (props->has_thread_id && !slot->props.has_thread_id) {
758             error_setg(errp, "thread-id is not supported");
759             return;
760         }
761 
762         if (props->has_core_id && !slot->props.has_core_id) {
763             error_setg(errp, "core-id is not supported");
764             return;
765         }
766 
767         if (props->has_cluster_id && !slot->props.has_cluster_id) {
768             error_setg(errp, "cluster-id is not supported");
769             return;
770         }
771 
772         if (props->has_socket_id && !slot->props.has_socket_id) {
773             error_setg(errp, "socket-id is not supported");
774             return;
775         }
776 
777         if (props->has_die_id && !slot->props.has_die_id) {
778             error_setg(errp, "die-id is not supported");
779             return;
780         }
781 
782         /* skip slots with explicit mismatch */
783         if (props->has_thread_id && props->thread_id != slot->props.thread_id) {
784                 continue;
785         }
786 
787         if (props->has_core_id && props->core_id != slot->props.core_id) {
788                 continue;
789         }
790 
791         if (props->has_cluster_id &&
792             props->cluster_id != slot->props.cluster_id) {
793                 continue;
794         }
795 
796         if (props->has_die_id && props->die_id != slot->props.die_id) {
797                 continue;
798         }
799 
800         if (props->has_socket_id && props->socket_id != slot->props.socket_id) {
801                 continue;
802         }
803 
804         /* reject assignment if slot is already assigned, for compatibility
805          * of legacy cpu_index mapping with SPAPR core based mapping do not
806          * error out if cpu thread and matched core have the same node-id */
807         if (slot->props.has_node_id &&
808             slot->props.node_id != props->node_id) {
809             error_setg(errp, "CPU is already assigned to node-id: %" PRId64,
810                        slot->props.node_id);
811             return;
812         }
813 
814         /* assign slot to node as it's matched '-numa cpu' key */
815         match = true;
816         slot->props.node_id = props->node_id;
817         slot->props.has_node_id = props->has_node_id;
818 
819         if (machine->numa_state->hmat_enabled) {
820             if ((numa_info[props->node_id].initiator < MAX_NODES) &&
821                 (props->node_id != numa_info[props->node_id].initiator)) {
822                 error_setg(errp, "The initiator of CPU NUMA node %" PRId64
823                            " should be itself (got %" PRIu16 ")",
824                            props->node_id, numa_info[props->node_id].initiator);
825                 return;
826             }
827             numa_info[props->node_id].has_cpu = true;
828             numa_info[props->node_id].initiator = props->node_id;
829         }
830     }
831 
832     if (!match) {
833         error_setg(errp, "no match found");
834     }
835 }
836 
837 static void machine_get_smp(Object *obj, Visitor *v, const char *name,
838                             void *opaque, Error **errp)
839 {
840     MachineState *ms = MACHINE(obj);
841     SMPConfiguration *config = &(SMPConfiguration){
842         .has_cpus = true, .cpus = ms->smp.cpus,
843         .has_sockets = true, .sockets = ms->smp.sockets,
844         .has_dies = true, .dies = ms->smp.dies,
845         .has_clusters = true, .clusters = ms->smp.clusters,
846         .has_cores = true, .cores = ms->smp.cores,
847         .has_threads = true, .threads = ms->smp.threads,
848         .has_maxcpus = true, .maxcpus = ms->smp.max_cpus,
849     };
850 
851     if (!visit_type_SMPConfiguration(v, name, &config, &error_abort)) {
852         return;
853     }
854 }
855 
856 static void machine_set_smp(Object *obj, Visitor *v, const char *name,
857                             void *opaque, Error **errp)
858 {
859     MachineState *ms = MACHINE(obj);
860     g_autoptr(SMPConfiguration) config = NULL;
861 
862     if (!visit_type_SMPConfiguration(v, name, &config, errp)) {
863         return;
864     }
865 
866     machine_parse_smp_config(ms, config, errp);
867 }
868 
869 static void machine_get_boot(Object *obj, Visitor *v, const char *name,
870                             void *opaque, Error **errp)
871 {
872     MachineState *ms = MACHINE(obj);
873     BootConfiguration *config = &ms->boot_config;
874     visit_type_BootConfiguration(v, name, &config, &error_abort);
875 }
876 
877 static void machine_free_boot_config(MachineState *ms)
878 {
879     g_free(ms->boot_config.order);
880     g_free(ms->boot_config.once);
881     g_free(ms->boot_config.splash);
882 }
883 
884 static void machine_copy_boot_config(MachineState *ms, BootConfiguration *config)
885 {
886     MachineClass *machine_class = MACHINE_GET_CLASS(ms);
887 
888     machine_free_boot_config(ms);
889     ms->boot_config = *config;
890     if (!config->order) {
891         ms->boot_config.order = g_strdup(machine_class->default_boot_order);
892     }
893 }
894 
895 static void machine_set_boot(Object *obj, Visitor *v, const char *name,
896                             void *opaque, Error **errp)
897 {
898     ERRP_GUARD();
899     MachineState *ms = MACHINE(obj);
900     BootConfiguration *config = NULL;
901 
902     if (!visit_type_BootConfiguration(v, name, &config, errp)) {
903         return;
904     }
905     if (config->order) {
906         validate_bootdevices(config->order, errp);
907         if (*errp) {
908             goto out_free;
909         }
910     }
911     if (config->once) {
912         validate_bootdevices(config->once, errp);
913         if (*errp) {
914             goto out_free;
915         }
916     }
917 
918     machine_copy_boot_config(ms, config);
919     /* Strings live in ms->boot_config.  */
920     free(config);
921     return;
922 
923 out_free:
924     qapi_free_BootConfiguration(config);
925 }
926 
927 static void machine_class_init(ObjectClass *oc, void *data)
928 {
929     MachineClass *mc = MACHINE_CLASS(oc);
930 
931     /* Default 128 MB as guest ram size */
932     mc->default_ram_size = 128 * MiB;
933     mc->rom_file_has_mr = true;
934 
935     /* numa node memory size aligned on 8MB by default.
936      * On Linux, each node's border has to be 8MB aligned
937      */
938     mc->numa_mem_align_shift = 23;
939 
940     object_class_property_add_str(oc, "kernel",
941         machine_get_kernel, machine_set_kernel);
942     object_class_property_set_description(oc, "kernel",
943         "Linux kernel image file");
944 
945     object_class_property_add_str(oc, "initrd",
946         machine_get_initrd, machine_set_initrd);
947     object_class_property_set_description(oc, "initrd",
948         "Linux initial ramdisk file");
949 
950     object_class_property_add_str(oc, "append",
951         machine_get_append, machine_set_append);
952     object_class_property_set_description(oc, "append",
953         "Linux kernel command line");
954 
955     object_class_property_add_str(oc, "dtb",
956         machine_get_dtb, machine_set_dtb);
957     object_class_property_set_description(oc, "dtb",
958         "Linux kernel device tree file");
959 
960     object_class_property_add_str(oc, "dumpdtb",
961         machine_get_dumpdtb, machine_set_dumpdtb);
962     object_class_property_set_description(oc, "dumpdtb",
963         "Dump current dtb to a file and quit");
964 
965     object_class_property_add(oc, "boot", "BootConfiguration",
966         machine_get_boot, machine_set_boot,
967         NULL, NULL);
968     object_class_property_set_description(oc, "boot",
969         "Boot configuration");
970 
971     object_class_property_add(oc, "smp", "SMPConfiguration",
972         machine_get_smp, machine_set_smp,
973         NULL, NULL);
974     object_class_property_set_description(oc, "smp",
975         "CPU topology");
976 
977     object_class_property_add(oc, "phandle-start", "int",
978         machine_get_phandle_start, machine_set_phandle_start,
979         NULL, NULL);
980     object_class_property_set_description(oc, "phandle-start",
981         "The first phandle ID we may generate dynamically");
982 
983     object_class_property_add_str(oc, "dt-compatible",
984         machine_get_dt_compatible, machine_set_dt_compatible);
985     object_class_property_set_description(oc, "dt-compatible",
986         "Overrides the \"compatible\" property of the dt root node");
987 
988     object_class_property_add_bool(oc, "dump-guest-core",
989         machine_get_dump_guest_core, machine_set_dump_guest_core);
990     object_class_property_set_description(oc, "dump-guest-core",
991         "Include guest memory in a core dump");
992 
993     object_class_property_add_bool(oc, "mem-merge",
994         machine_get_mem_merge, machine_set_mem_merge);
995     object_class_property_set_description(oc, "mem-merge",
996         "Enable/disable memory merge support");
997 
998     object_class_property_add_bool(oc, "usb",
999         machine_get_usb, machine_set_usb);
1000     object_class_property_set_description(oc, "usb",
1001         "Set on/off to enable/disable usb");
1002 
1003     object_class_property_add_bool(oc, "graphics",
1004         machine_get_graphics, machine_set_graphics);
1005     object_class_property_set_description(oc, "graphics",
1006         "Set on/off to enable/disable graphics emulation");
1007 
1008     object_class_property_add_str(oc, "firmware",
1009         machine_get_firmware, machine_set_firmware);
1010     object_class_property_set_description(oc, "firmware",
1011         "Firmware image");
1012 
1013     object_class_property_add_bool(oc, "suppress-vmdesc",
1014         machine_get_suppress_vmdesc, machine_set_suppress_vmdesc);
1015     object_class_property_set_description(oc, "suppress-vmdesc",
1016         "Set on to disable self-describing migration");
1017 
1018     object_class_property_add_link(oc, "confidential-guest-support",
1019                                    TYPE_CONFIDENTIAL_GUEST_SUPPORT,
1020                                    offsetof(MachineState, cgs),
1021                                    machine_check_confidential_guest_support,
1022                                    OBJ_PROP_LINK_STRONG);
1023     object_class_property_set_description(oc, "confidential-guest-support",
1024                                           "Set confidential guest scheme to support");
1025 
1026     /* For compatibility */
1027     object_class_property_add_str(oc, "memory-encryption",
1028         machine_get_memory_encryption, machine_set_memory_encryption);
1029     object_class_property_set_description(oc, "memory-encryption",
1030         "Set memory encryption object to use");
1031 
1032     object_class_property_add_link(oc, "memory-backend", TYPE_MEMORY_BACKEND,
1033                                    offsetof(MachineState, memdev), object_property_allow_set_link,
1034                                    OBJ_PROP_LINK_STRONG);
1035     object_class_property_set_description(oc, "memory-backend",
1036                                           "Set RAM backend"
1037                                           "Valid value is ID of hostmem based backend");
1038 
1039     object_class_property_add(oc, "memory", "MemorySizeConfiguration",
1040         machine_get_mem, machine_set_mem,
1041         NULL, NULL);
1042     object_class_property_set_description(oc, "memory",
1043         "Memory size configuration");
1044 }
1045 
1046 static void machine_class_base_init(ObjectClass *oc, void *data)
1047 {
1048     MachineClass *mc = MACHINE_CLASS(oc);
1049     mc->max_cpus = mc->max_cpus ?: 1;
1050     mc->min_cpus = mc->min_cpus ?: 1;
1051     mc->default_cpus = mc->default_cpus ?: 1;
1052 
1053     if (!object_class_is_abstract(oc)) {
1054         const char *cname = object_class_get_name(oc);
1055         assert(g_str_has_suffix(cname, TYPE_MACHINE_SUFFIX));
1056         mc->name = g_strndup(cname,
1057                             strlen(cname) - strlen(TYPE_MACHINE_SUFFIX));
1058         mc->compat_props = g_ptr_array_new();
1059     }
1060 }
1061 
1062 static void machine_initfn(Object *obj)
1063 {
1064     MachineState *ms = MACHINE(obj);
1065     MachineClass *mc = MACHINE_GET_CLASS(obj);
1066 
1067     container_get(obj, "/peripheral");
1068     container_get(obj, "/peripheral-anon");
1069 
1070     ms->dump_guest_core = true;
1071     ms->mem_merge = true;
1072     ms->enable_graphics = true;
1073     ms->kernel_cmdline = g_strdup("");
1074     ms->ram_size = mc->default_ram_size;
1075     ms->maxram_size = mc->default_ram_size;
1076 
1077     if (mc->nvdimm_supported) {
1078         Object *obj = OBJECT(ms);
1079 
1080         ms->nvdimms_state = g_new0(NVDIMMState, 1);
1081         object_property_add_bool(obj, "nvdimm",
1082                                  machine_get_nvdimm, machine_set_nvdimm);
1083         object_property_set_description(obj, "nvdimm",
1084                                         "Set on/off to enable/disable "
1085                                         "NVDIMM instantiation");
1086 
1087         object_property_add_str(obj, "nvdimm-persistence",
1088                                 machine_get_nvdimm_persistence,
1089                                 machine_set_nvdimm_persistence);
1090         object_property_set_description(obj, "nvdimm-persistence",
1091                                         "Set NVDIMM persistence"
1092                                         "Valid values are cpu, mem-ctrl");
1093     }
1094 
1095     if (mc->cpu_index_to_instance_props && mc->get_default_cpu_node_id) {
1096         ms->numa_state = g_new0(NumaState, 1);
1097         object_property_add_bool(obj, "hmat",
1098                                  machine_get_hmat, machine_set_hmat);
1099         object_property_set_description(obj, "hmat",
1100                                         "Set on/off to enable/disable "
1101                                         "ACPI Heterogeneous Memory Attribute "
1102                                         "Table (HMAT)");
1103     }
1104 
1105     /* default to mc->default_cpus */
1106     ms->smp.cpus = mc->default_cpus;
1107     ms->smp.max_cpus = mc->default_cpus;
1108     ms->smp.sockets = 1;
1109     ms->smp.dies = 1;
1110     ms->smp.clusters = 1;
1111     ms->smp.cores = 1;
1112     ms->smp.threads = 1;
1113 
1114     machine_copy_boot_config(ms, &(BootConfiguration){ 0 });
1115 }
1116 
1117 static void machine_finalize(Object *obj)
1118 {
1119     MachineState *ms = MACHINE(obj);
1120 
1121     machine_free_boot_config(ms);
1122     g_free(ms->kernel_filename);
1123     g_free(ms->initrd_filename);
1124     g_free(ms->kernel_cmdline);
1125     g_free(ms->dtb);
1126     g_free(ms->dumpdtb);
1127     g_free(ms->dt_compatible);
1128     g_free(ms->firmware);
1129     g_free(ms->device_memory);
1130     g_free(ms->nvdimms_state);
1131     g_free(ms->numa_state);
1132 }
1133 
1134 bool machine_usb(MachineState *machine)
1135 {
1136     return machine->usb;
1137 }
1138 
1139 int machine_phandle_start(MachineState *machine)
1140 {
1141     return machine->phandle_start;
1142 }
1143 
1144 bool machine_dump_guest_core(MachineState *machine)
1145 {
1146     return machine->dump_guest_core;
1147 }
1148 
1149 bool machine_mem_merge(MachineState *machine)
1150 {
1151     return machine->mem_merge;
1152 }
1153 
1154 static char *cpu_slot_to_string(const CPUArchId *cpu)
1155 {
1156     GString *s = g_string_new(NULL);
1157     if (cpu->props.has_socket_id) {
1158         g_string_append_printf(s, "socket-id: %"PRId64, cpu->props.socket_id);
1159     }
1160     if (cpu->props.has_die_id) {
1161         if (s->len) {
1162             g_string_append_printf(s, ", ");
1163         }
1164         g_string_append_printf(s, "die-id: %"PRId64, cpu->props.die_id);
1165     }
1166     if (cpu->props.has_cluster_id) {
1167         if (s->len) {
1168             g_string_append_printf(s, ", ");
1169         }
1170         g_string_append_printf(s, "cluster-id: %"PRId64, cpu->props.cluster_id);
1171     }
1172     if (cpu->props.has_core_id) {
1173         if (s->len) {
1174             g_string_append_printf(s, ", ");
1175         }
1176         g_string_append_printf(s, "core-id: %"PRId64, cpu->props.core_id);
1177     }
1178     if (cpu->props.has_thread_id) {
1179         if (s->len) {
1180             g_string_append_printf(s, ", ");
1181         }
1182         g_string_append_printf(s, "thread-id: %"PRId64, cpu->props.thread_id);
1183     }
1184     return g_string_free(s, false);
1185 }
1186 
1187 static void numa_validate_initiator(NumaState *numa_state)
1188 {
1189     int i;
1190     NodeInfo *numa_info = numa_state->nodes;
1191 
1192     for (i = 0; i < numa_state->num_nodes; i++) {
1193         if (numa_info[i].initiator == MAX_NODES) {
1194             continue;
1195         }
1196 
1197         if (!numa_info[numa_info[i].initiator].present) {
1198             error_report("NUMA node %" PRIu16 " is missing, use "
1199                          "'-numa node' option to declare it first",
1200                          numa_info[i].initiator);
1201             exit(1);
1202         }
1203 
1204         if (!numa_info[numa_info[i].initiator].has_cpu) {
1205             error_report("The initiator of NUMA node %d is invalid", i);
1206             exit(1);
1207         }
1208     }
1209 }
1210 
1211 static void machine_numa_finish_cpu_init(MachineState *machine)
1212 {
1213     int i;
1214     bool default_mapping;
1215     GString *s = g_string_new(NULL);
1216     MachineClass *mc = MACHINE_GET_CLASS(machine);
1217     const CPUArchIdList *possible_cpus = mc->possible_cpu_arch_ids(machine);
1218 
1219     assert(machine->numa_state->num_nodes);
1220     for (i = 0; i < possible_cpus->len; i++) {
1221         if (possible_cpus->cpus[i].props.has_node_id) {
1222             break;
1223         }
1224     }
1225     default_mapping = (i == possible_cpus->len);
1226 
1227     for (i = 0; i < possible_cpus->len; i++) {
1228         const CPUArchId *cpu_slot = &possible_cpus->cpus[i];
1229 
1230         if (!cpu_slot->props.has_node_id) {
1231             /* fetch default mapping from board and enable it */
1232             CpuInstanceProperties props = cpu_slot->props;
1233 
1234             props.node_id = mc->get_default_cpu_node_id(machine, i);
1235             if (!default_mapping) {
1236                 /* record slots with not set mapping,
1237                  * TODO: make it hard error in future */
1238                 char *cpu_str = cpu_slot_to_string(cpu_slot);
1239                 g_string_append_printf(s, "%sCPU %d [%s]",
1240                                        s->len ? ", " : "", i, cpu_str);
1241                 g_free(cpu_str);
1242 
1243                 /* non mapped cpus used to fallback to node 0 */
1244                 props.node_id = 0;
1245             }
1246 
1247             props.has_node_id = true;
1248             machine_set_cpu_numa_node(machine, &props, &error_fatal);
1249         }
1250     }
1251 
1252     if (machine->numa_state->hmat_enabled) {
1253         numa_validate_initiator(machine->numa_state);
1254     }
1255 
1256     if (s->len && !qtest_enabled()) {
1257         warn_report("CPU(s) not present in any NUMA nodes: %s",
1258                     s->str);
1259         warn_report("All CPU(s) up to maxcpus should be described "
1260                     "in NUMA config, ability to start up with partial NUMA "
1261                     "mappings is obsoleted and will be removed in future");
1262     }
1263     g_string_free(s, true);
1264 }
1265 
1266 static void validate_cpu_cluster_to_numa_boundary(MachineState *ms)
1267 {
1268     MachineClass *mc = MACHINE_GET_CLASS(ms);
1269     NumaState *state = ms->numa_state;
1270     const CPUArchIdList *possible_cpus = mc->possible_cpu_arch_ids(ms);
1271     const CPUArchId *cpus = possible_cpus->cpus;
1272     int i, j;
1273 
1274     if (state->num_nodes <= 1 || possible_cpus->len <= 1) {
1275         return;
1276     }
1277 
1278     /*
1279      * The Linux scheduling domain can't be parsed when the multiple CPUs
1280      * in one cluster have been associated with different NUMA nodes. However,
1281      * it's fine to associate one NUMA node with CPUs in different clusters.
1282      */
1283     for (i = 0; i < possible_cpus->len; i++) {
1284         for (j = i + 1; j < possible_cpus->len; j++) {
1285             if (cpus[i].props.has_socket_id &&
1286                 cpus[i].props.has_cluster_id &&
1287                 cpus[i].props.has_node_id &&
1288                 cpus[j].props.has_socket_id &&
1289                 cpus[j].props.has_cluster_id &&
1290                 cpus[j].props.has_node_id &&
1291                 cpus[i].props.socket_id == cpus[j].props.socket_id &&
1292                 cpus[i].props.cluster_id == cpus[j].props.cluster_id &&
1293                 cpus[i].props.node_id != cpus[j].props.node_id) {
1294                 warn_report("CPU-%d and CPU-%d in socket-%" PRId64 "-cluster-%" PRId64
1295                              " have been associated with node-%" PRId64 " and node-%" PRId64
1296                              " respectively. It can cause OSes like Linux to"
1297                              " misbehave", i, j, cpus[i].props.socket_id,
1298                              cpus[i].props.cluster_id, cpus[i].props.node_id,
1299                              cpus[j].props.node_id);
1300             }
1301         }
1302     }
1303 }
1304 
1305 MemoryRegion *machine_consume_memdev(MachineState *machine,
1306                                      HostMemoryBackend *backend)
1307 {
1308     MemoryRegion *ret = host_memory_backend_get_memory(backend);
1309 
1310     if (host_memory_backend_is_mapped(backend)) {
1311         error_report("memory backend %s can't be used multiple times.",
1312                      object_get_canonical_path_component(OBJECT(backend)));
1313         exit(EXIT_FAILURE);
1314     }
1315     host_memory_backend_set_mapped(backend, true);
1316     vmstate_register_ram_global(ret);
1317     return ret;
1318 }
1319 
1320 static bool create_default_memdev(MachineState *ms, const char *path, Error **errp)
1321 {
1322     Object *obj;
1323     MachineClass *mc = MACHINE_GET_CLASS(ms);
1324     bool r = false;
1325 
1326     obj = object_new(path ? TYPE_MEMORY_BACKEND_FILE : TYPE_MEMORY_BACKEND_RAM);
1327     if (path) {
1328         if (!object_property_set_str(obj, "mem-path", path, errp)) {
1329             goto out;
1330         }
1331     }
1332     if (!object_property_set_int(obj, "size", ms->ram_size, errp)) {
1333         goto out;
1334     }
1335     object_property_add_child(object_get_objects_root(), mc->default_ram_id,
1336                               obj);
1337     /* Ensure backend's memory region name is equal to mc->default_ram_id */
1338     if (!object_property_set_bool(obj, "x-use-canonical-path-for-ramblock-id",
1339                              false, errp)) {
1340         goto out;
1341     }
1342     if (!user_creatable_complete(USER_CREATABLE(obj), errp)) {
1343         goto out;
1344     }
1345     r = object_property_set_link(OBJECT(ms), "memory-backend", obj, errp);
1346 
1347 out:
1348     object_unref(obj);
1349     return r;
1350 }
1351 
1352 
1353 void machine_run_board_init(MachineState *machine, const char *mem_path, Error **errp)
1354 {
1355     MachineClass *machine_class = MACHINE_GET_CLASS(machine);
1356     ObjectClass *oc = object_class_by_name(machine->cpu_type);
1357     CPUClass *cc;
1358 
1359     /* This checkpoint is required by replay to separate prior clock
1360        reading from the other reads, because timer polling functions query
1361        clock values from the log. */
1362     replay_checkpoint(CHECKPOINT_INIT);
1363 
1364     if (!xen_enabled()) {
1365         /* On 32-bit hosts, QEMU is limited by virtual address space */
1366         if (machine->ram_size > (2047 << 20) && HOST_LONG_BITS == 32) {
1367             error_setg(errp, "at most 2047 MB RAM can be simulated");
1368             return;
1369         }
1370     }
1371 
1372     if (machine->memdev) {
1373         ram_addr_t backend_size = object_property_get_uint(OBJECT(machine->memdev),
1374                                                            "size",  &error_abort);
1375         if (backend_size != machine->ram_size) {
1376             error_setg(errp, "Machine memory size does not match the size of the memory backend");
1377             return;
1378         }
1379     } else if (machine_class->default_ram_id && machine->ram_size &&
1380                numa_uses_legacy_mem()) {
1381         if (object_property_find(object_get_objects_root(),
1382                                  machine_class->default_ram_id)) {
1383             error_setg(errp, "object name '%s' is reserved for the default"
1384                 " RAM backend, it can't be used for any other purposes."
1385                 " Change the object's 'id' to something else",
1386                 machine_class->default_ram_id);
1387             return;
1388         }
1389         if (!create_default_memdev(current_machine, mem_path, errp)) {
1390             return;
1391         }
1392     }
1393 
1394     if (machine->numa_state) {
1395         numa_complete_configuration(machine);
1396         if (machine->numa_state->num_nodes) {
1397             machine_numa_finish_cpu_init(machine);
1398             if (machine_class->cpu_cluster_has_numa_boundary) {
1399                 validate_cpu_cluster_to_numa_boundary(machine);
1400             }
1401         }
1402     }
1403 
1404     if (!machine->ram && machine->memdev) {
1405         machine->ram = machine_consume_memdev(machine, machine->memdev);
1406     }
1407 
1408     /* If the machine supports the valid_cpu_types check and the user
1409      * specified a CPU with -cpu check here that the user CPU is supported.
1410      */
1411     if (machine_class->valid_cpu_types && machine->cpu_type) {
1412         int i;
1413 
1414         for (i = 0; machine_class->valid_cpu_types[i]; i++) {
1415             if (object_class_dynamic_cast(oc,
1416                                           machine_class->valid_cpu_types[i])) {
1417                 /* The user specificed CPU is in the valid field, we are
1418                  * good to go.
1419                  */
1420                 break;
1421             }
1422         }
1423 
1424         if (!machine_class->valid_cpu_types[i]) {
1425             /* The user specified CPU is not valid */
1426             error_report("Invalid CPU type: %s", machine->cpu_type);
1427             error_printf("The valid types are: %s",
1428                          machine_class->valid_cpu_types[0]);
1429             for (i = 1; machine_class->valid_cpu_types[i]; i++) {
1430                 error_printf(", %s", machine_class->valid_cpu_types[i]);
1431             }
1432             error_printf("\n");
1433 
1434             exit(1);
1435         }
1436     }
1437 
1438     /* Check if CPU type is deprecated and warn if so */
1439     cc = CPU_CLASS(oc);
1440     if (cc && cc->deprecation_note) {
1441         warn_report("CPU model %s is deprecated -- %s", machine->cpu_type,
1442                     cc->deprecation_note);
1443     }
1444 
1445     if (machine->cgs) {
1446         /*
1447          * With confidential guests, the host can't see the real
1448          * contents of RAM, so there's no point in it trying to merge
1449          * areas.
1450          */
1451         machine_set_mem_merge(OBJECT(machine), false, &error_abort);
1452 
1453         /*
1454          * Virtio devices can't count on directly accessing guest
1455          * memory, so they need iommu_platform=on to use normal DMA
1456          * mechanisms.  That requires also disabling legacy virtio
1457          * support for those virtio pci devices which allow it.
1458          */
1459         object_register_sugar_prop(TYPE_VIRTIO_PCI, "disable-legacy",
1460                                    "on", true);
1461         object_register_sugar_prop(TYPE_VIRTIO_DEVICE, "iommu_platform",
1462                                    "on", false);
1463     }
1464 
1465     accel_init_interfaces(ACCEL_GET_CLASS(machine->accelerator));
1466     machine_class->init(machine);
1467     phase_advance(PHASE_MACHINE_INITIALIZED);
1468 }
1469 
1470 static NotifierList machine_init_done_notifiers =
1471     NOTIFIER_LIST_INITIALIZER(machine_init_done_notifiers);
1472 
1473 void qemu_add_machine_init_done_notifier(Notifier *notify)
1474 {
1475     notifier_list_add(&machine_init_done_notifiers, notify);
1476     if (phase_check(PHASE_MACHINE_READY)) {
1477         notify->notify(notify, NULL);
1478     }
1479 }
1480 
1481 void qemu_remove_machine_init_done_notifier(Notifier *notify)
1482 {
1483     notifier_remove(notify);
1484 }
1485 
1486 void qdev_machine_creation_done(void)
1487 {
1488     cpu_synchronize_all_post_init();
1489 
1490     if (current_machine->boot_config.once) {
1491         qemu_boot_set(current_machine->boot_config.once, &error_fatal);
1492         qemu_register_reset(restore_boot_order, g_strdup(current_machine->boot_config.order));
1493     }
1494 
1495     /*
1496      * ok, initial machine setup is done, starting from now we can
1497      * only create hotpluggable devices
1498      */
1499     phase_advance(PHASE_MACHINE_READY);
1500     qdev_assert_realized_properly();
1501 
1502     /* TODO: once all bus devices are qdevified, this should be done
1503      * when bus is created by qdev.c */
1504     /*
1505      * TODO: If we had a main 'reset container' that the whole system
1506      * lived in, we could reset that using the multi-phase reset
1507      * APIs. For the moment, we just reset the sysbus, which will cause
1508      * all devices hanging off it (and all their child buses, recursively)
1509      * to be reset. Note that this will *not* reset any Device objects
1510      * which are not attached to some part of the qbus tree!
1511      */
1512     qemu_register_reset(resettable_cold_reset_fn, sysbus_get_default());
1513 
1514     notifier_list_notify(&machine_init_done_notifiers, NULL);
1515 
1516     if (rom_check_and_register_reset() != 0) {
1517         exit(1);
1518     }
1519 
1520     replay_start();
1521 
1522     /* This checkpoint is required by replay to separate prior clock
1523        reading from the other reads, because timer polling functions query
1524        clock values from the log. */
1525     replay_checkpoint(CHECKPOINT_RESET);
1526     qemu_system_reset(SHUTDOWN_CAUSE_NONE);
1527     register_global_state();
1528 }
1529 
1530 static const TypeInfo machine_info = {
1531     .name = TYPE_MACHINE,
1532     .parent = TYPE_OBJECT,
1533     .abstract = true,
1534     .class_size = sizeof(MachineClass),
1535     .class_init    = machine_class_init,
1536     .class_base_init = machine_class_base_init,
1537     .instance_size = sizeof(MachineState),
1538     .instance_init = machine_initfn,
1539     .instance_finalize = machine_finalize,
1540 };
1541 
1542 static void machine_register_types(void)
1543 {
1544     type_register_static(&machine_info);
1545 }
1546 
1547 type_init(machine_register_types)
1548