xref: /openbmc/qemu/hw/core/machine.c (revision 14a16403)
1 /*
2  * QEMU Machine
3  *
4  * Copyright (C) 2014 Red Hat Inc
5  *
6  * Authors:
7  *   Marcel Apfelbaum <marcel.a@redhat.com>
8  *
9  * This work is licensed under the terms of the GNU GPL, version 2 or later.
10  * See the COPYING file in the top-level directory.
11  */
12 
13 #include "qemu/osdep.h"
14 #include "qemu/accel.h"
15 #include "sysemu/replay.h"
16 #include "hw/boards.h"
17 #include "hw/loader.h"
18 #include "qapi/error.h"
19 #include "qapi/qapi-visit-machine.h"
20 #include "qom/object_interfaces.h"
21 #include "sysemu/cpus.h"
22 #include "sysemu/sysemu.h"
23 #include "sysemu/reset.h"
24 #include "sysemu/runstate.h"
25 #include "sysemu/xen.h"
26 #include "sysemu/qtest.h"
27 #include "hw/pci/pci_bridge.h"
28 #include "hw/mem/nvdimm.h"
29 #include "migration/global_state.h"
30 #include "exec/confidential-guest-support.h"
31 #include "hw/virtio/virtio-pci.h"
32 #include "hw/virtio/virtio-net.h"
33 #include "hw/virtio/virtio-iommu.h"
34 #include "audio/audio.h"
35 
36 GlobalProperty hw_compat_9_0[] = {};
37 const size_t hw_compat_9_0_len = G_N_ELEMENTS(hw_compat_9_0);
38 
39 GlobalProperty hw_compat_8_2[] = {
40     { "migration", "zero-page-detection", "legacy"},
41     { TYPE_VIRTIO_IOMMU_PCI, "granule", "4k" },
42     { TYPE_VIRTIO_IOMMU_PCI, "aw-bits", "64" },
43 };
44 const size_t hw_compat_8_2_len = G_N_ELEMENTS(hw_compat_8_2);
45 
46 GlobalProperty hw_compat_8_1[] = {
47     { TYPE_PCI_BRIDGE, "x-pci-express-writeable-slt-bug", "true" },
48     { "ramfb", "x-migrate", "off" },
49     { "vfio-pci-nohotplug", "x-ramfb-migrate", "off" },
50     { "igb", "x-pcie-flr-init", "off" },
51 };
52 const size_t hw_compat_8_1_len = G_N_ELEMENTS(hw_compat_8_1);
53 
54 GlobalProperty hw_compat_8_0[] = {
55     { "migration", "multifd-flush-after-each-section", "on"},
56     { TYPE_PCI_DEVICE, "x-pcie-ari-nextfn-1", "on" },
57     { TYPE_VIRTIO_NET, "host_uso", "off"},
58     { TYPE_VIRTIO_NET, "guest_uso4", "off"},
59     { TYPE_VIRTIO_NET, "guest_uso6", "off"},
60 };
61 const size_t hw_compat_8_0_len = G_N_ELEMENTS(hw_compat_8_0);
62 
63 GlobalProperty hw_compat_7_2[] = {
64     { "e1000e", "migrate-timadj", "off" },
65     { "virtio-mem", "x-early-migration", "false" },
66     { "migration", "x-preempt-pre-7-2", "true" },
67     { TYPE_PCI_DEVICE, "x-pcie-err-unc-mask", "off" },
68 };
69 const size_t hw_compat_7_2_len = G_N_ELEMENTS(hw_compat_7_2);
70 
71 GlobalProperty hw_compat_7_1[] = {
72     { "virtio-device", "queue_reset", "false" },
73     { "virtio-rng-pci", "vectors", "0" },
74     { "virtio-rng-pci-transitional", "vectors", "0" },
75     { "virtio-rng-pci-non-transitional", "vectors", "0" },
76 };
77 const size_t hw_compat_7_1_len = G_N_ELEMENTS(hw_compat_7_1);
78 
79 GlobalProperty hw_compat_7_0[] = {
80     { "arm-gicv3-common", "force-8-bit-prio", "on" },
81     { "nvme-ns", "eui64-default", "on"},
82 };
83 const size_t hw_compat_7_0_len = G_N_ELEMENTS(hw_compat_7_0);
84 
85 GlobalProperty hw_compat_6_2[] = {
86     { "PIIX4_PM", "x-not-migrate-acpi-index", "on"},
87 };
88 const size_t hw_compat_6_2_len = G_N_ELEMENTS(hw_compat_6_2);
89 
90 GlobalProperty hw_compat_6_1[] = {
91     { "vhost-user-vsock-device", "seqpacket", "off" },
92     { "nvme-ns", "shared", "off" },
93 };
94 const size_t hw_compat_6_1_len = G_N_ELEMENTS(hw_compat_6_1);
95 
96 GlobalProperty hw_compat_6_0[] = {
97     { "gpex-pcihost", "allow-unmapped-accesses", "false" },
98     { "i8042", "extended-state", "false"},
99     { "nvme-ns", "eui64-default", "off"},
100     { "e1000", "init-vet", "off" },
101     { "e1000e", "init-vet", "off" },
102     { "vhost-vsock-device", "seqpacket", "off" },
103 };
104 const size_t hw_compat_6_0_len = G_N_ELEMENTS(hw_compat_6_0);
105 
106 GlobalProperty hw_compat_5_2[] = {
107     { "ICH9-LPC", "smm-compat", "on"},
108     { "PIIX4_PM", "smm-compat", "on"},
109     { "virtio-blk-device", "report-discard-granularity", "off" },
110     { "virtio-net-pci-base", "vectors", "3"},
111     { "nvme", "msix-exclusive-bar", "on"},
112 };
113 const size_t hw_compat_5_2_len = G_N_ELEMENTS(hw_compat_5_2);
114 
115 GlobalProperty hw_compat_5_1[] = {
116     { "vhost-scsi", "num_queues", "1"},
117     { "vhost-user-blk", "num-queues", "1"},
118     { "vhost-user-scsi", "num_queues", "1"},
119     { "virtio-blk-device", "num-queues", "1"},
120     { "virtio-scsi-device", "num_queues", "1"},
121     { "nvme", "use-intel-id", "on"},
122     { "pvpanic", "events", "1"}, /* PVPANIC_PANICKED */
123     { "pl011", "migrate-clk", "off" },
124     { "virtio-pci", "x-ats-page-aligned", "off"},
125 };
126 const size_t hw_compat_5_1_len = G_N_ELEMENTS(hw_compat_5_1);
127 
128 GlobalProperty hw_compat_5_0[] = {
129     { "pci-host-bridge", "x-config-reg-migration-enabled", "off" },
130     { "virtio-balloon-device", "page-poison", "false" },
131     { "vmport", "x-read-set-eax", "off" },
132     { "vmport", "x-signal-unsupported-cmd", "off" },
133     { "vmport", "x-report-vmx-type", "off" },
134     { "vmport", "x-cmds-v2", "off" },
135     { "virtio-device", "x-disable-legacy-check", "true" },
136 };
137 const size_t hw_compat_5_0_len = G_N_ELEMENTS(hw_compat_5_0);
138 
139 GlobalProperty hw_compat_4_2[] = {
140     { "virtio-blk-device", "queue-size", "128"},
141     { "virtio-scsi-device", "virtqueue_size", "128"},
142     { "virtio-blk-device", "x-enable-wce-if-config-wce", "off" },
143     { "virtio-blk-device", "seg-max-adjust", "off"},
144     { "virtio-scsi-device", "seg_max_adjust", "off"},
145     { "vhost-blk-device", "seg_max_adjust", "off"},
146     { "usb-host", "suppress-remote-wake", "off" },
147     { "usb-redir", "suppress-remote-wake", "off" },
148     { "qxl", "revision", "4" },
149     { "qxl-vga", "revision", "4" },
150     { "fw_cfg", "acpi-mr-restore", "false" },
151     { "virtio-device", "use-disabled-flag", "false" },
152 };
153 const size_t hw_compat_4_2_len = G_N_ELEMENTS(hw_compat_4_2);
154 
155 GlobalProperty hw_compat_4_1[] = {
156     { "virtio-pci", "x-pcie-flr-init", "off" },
157 };
158 const size_t hw_compat_4_1_len = G_N_ELEMENTS(hw_compat_4_1);
159 
160 GlobalProperty hw_compat_4_0[] = {
161     { "VGA",            "edid", "false" },
162     { "secondary-vga",  "edid", "false" },
163     { "bochs-display",  "edid", "false" },
164     { "virtio-vga",     "edid", "false" },
165     { "virtio-gpu-device", "edid", "false" },
166     { "virtio-device", "use-started", "false" },
167     { "virtio-balloon-device", "qemu-4-0-config-size", "true" },
168     { "pl031", "migrate-tick-offset", "false" },
169 };
170 const size_t hw_compat_4_0_len = G_N_ELEMENTS(hw_compat_4_0);
171 
172 GlobalProperty hw_compat_3_1[] = {
173     { "pcie-root-port", "x-speed", "2_5" },
174     { "pcie-root-port", "x-width", "1" },
175     { "memory-backend-file", "x-use-canonical-path-for-ramblock-id", "true" },
176     { "memory-backend-memfd", "x-use-canonical-path-for-ramblock-id", "true" },
177     { "tpm-crb", "ppi", "false" },
178     { "tpm-tis", "ppi", "false" },
179     { "usb-kbd", "serial", "42" },
180     { "usb-mouse", "serial", "42" },
181     { "usb-tablet", "serial", "42" },
182     { "virtio-blk-device", "discard", "false" },
183     { "virtio-blk-device", "write-zeroes", "false" },
184     { "virtio-balloon-device", "qemu-4-0-config-size", "false" },
185     { "pcie-root-port-base", "disable-acs", "true" }, /* Added in 4.1 */
186 };
187 const size_t hw_compat_3_1_len = G_N_ELEMENTS(hw_compat_3_1);
188 
189 GlobalProperty hw_compat_3_0[] = {};
190 const size_t hw_compat_3_0_len = G_N_ELEMENTS(hw_compat_3_0);
191 
192 GlobalProperty hw_compat_2_12[] = {
193     { "migration", "decompress-error-check", "off" },
194     { "hda-audio", "use-timer", "false" },
195     { "cirrus-vga", "global-vmstate", "true" },
196     { "VGA", "global-vmstate", "true" },
197     { "vmware-svga", "global-vmstate", "true" },
198     { "qxl-vga", "global-vmstate", "true" },
199 };
200 const size_t hw_compat_2_12_len = G_N_ELEMENTS(hw_compat_2_12);
201 
202 GlobalProperty hw_compat_2_11[] = {
203     { "hpet", "hpet-offset-saved", "false" },
204     { "virtio-blk-pci", "vectors", "2" },
205     { "vhost-user-blk-pci", "vectors", "2" },
206     { "e1000", "migrate_tso_props", "off" },
207 };
208 const size_t hw_compat_2_11_len = G_N_ELEMENTS(hw_compat_2_11);
209 
210 GlobalProperty hw_compat_2_10[] = {
211     { "virtio-mouse-device", "wheel-axis", "false" },
212     { "virtio-tablet-device", "wheel-axis", "false" },
213 };
214 const size_t hw_compat_2_10_len = G_N_ELEMENTS(hw_compat_2_10);
215 
216 GlobalProperty hw_compat_2_9[] = {
217     { "pci-bridge", "shpc", "off" },
218     { "intel-iommu", "pt", "off" },
219     { "virtio-net-device", "x-mtu-bypass-backend", "off" },
220     { "pcie-root-port", "x-migrate-msix", "false" },
221 };
222 const size_t hw_compat_2_9_len = G_N_ELEMENTS(hw_compat_2_9);
223 
224 GlobalProperty hw_compat_2_8[] = {
225     { "fw_cfg_mem", "x-file-slots", "0x10" },
226     { "fw_cfg_io", "x-file-slots", "0x10" },
227     { "pflash_cfi01", "old-multiple-chip-handling", "on" },
228     { "pci-bridge", "shpc", "on" },
229     { TYPE_PCI_DEVICE, "x-pcie-extcap-init", "off" },
230     { "virtio-pci", "x-pcie-deverr-init", "off" },
231     { "virtio-pci", "x-pcie-lnkctl-init", "off" },
232     { "virtio-pci", "x-pcie-pm-init", "off" },
233     { "cirrus-vga", "vgamem_mb", "8" },
234     { "isa-cirrus-vga", "vgamem_mb", "8" },
235 };
236 const size_t hw_compat_2_8_len = G_N_ELEMENTS(hw_compat_2_8);
237 
238 GlobalProperty hw_compat_2_7[] = {
239     { "virtio-pci", "page-per-vq", "on" },
240     { "virtio-serial-device", "emergency-write", "off" },
241     { "ioapic", "version", "0x11" },
242     { "intel-iommu", "x-buggy-eim", "true" },
243     { "virtio-pci", "x-ignore-backend-features", "on" },
244 };
245 const size_t hw_compat_2_7_len = G_N_ELEMENTS(hw_compat_2_7);
246 
247 GlobalProperty hw_compat_2_6[] = {
248     { "virtio-mmio", "format_transport_address", "off" },
249     /* Optional because not all virtio-pci devices support legacy mode */
250     { "virtio-pci", "disable-modern", "on",  .optional = true },
251     { "virtio-pci", "disable-legacy", "off", .optional = true },
252 };
253 const size_t hw_compat_2_6_len = G_N_ELEMENTS(hw_compat_2_6);
254 
255 GlobalProperty hw_compat_2_5[] = {
256     { "isa-fdc", "fallback", "144" },
257     { "pvscsi", "x-old-pci-configuration", "on" },
258     { "pvscsi", "x-disable-pcie", "on" },
259     { "vmxnet3", "x-old-msi-offsets", "on" },
260     { "vmxnet3", "x-disable-pcie", "on" },
261 };
262 const size_t hw_compat_2_5_len = G_N_ELEMENTS(hw_compat_2_5);
263 
264 GlobalProperty hw_compat_2_4[] = {
265     /* Optional because the 'scsi' property is Linux-only */
266     { "virtio-blk-device", "scsi", "true", .optional = true },
267     { "e1000", "extra_mac_registers", "off" },
268     { "virtio-pci", "x-disable-pcie", "on" },
269     { "virtio-pci", "migrate-extra", "off" },
270     { "fw_cfg_mem", "dma_enabled", "off" },
271     { "fw_cfg_io", "dma_enabled", "off" }
272 };
273 const size_t hw_compat_2_4_len = G_N_ELEMENTS(hw_compat_2_4);
274 
275 GlobalProperty hw_compat_2_3[] = {
276     { "virtio-blk-pci", "any_layout", "off" },
277     { "virtio-balloon-pci", "any_layout", "off" },
278     { "virtio-serial-pci", "any_layout", "off" },
279     { "virtio-9p-pci", "any_layout", "off" },
280     { "virtio-rng-pci", "any_layout", "off" },
281     { TYPE_PCI_DEVICE, "x-pcie-lnksta-dllla", "off" },
282     { "migration", "send-configuration", "off" },
283     { "migration", "send-section-footer", "off" },
284     { "migration", "store-global-state", "off" },
285 };
286 const size_t hw_compat_2_3_len = G_N_ELEMENTS(hw_compat_2_3);
287 
288 GlobalProperty hw_compat_2_2[] = {};
289 const size_t hw_compat_2_2_len = G_N_ELEMENTS(hw_compat_2_2);
290 
291 GlobalProperty hw_compat_2_1[] = {
292     { "intel-hda", "old_msi_addr", "on" },
293     { "VGA", "qemu-extended-regs", "off" },
294     { "secondary-vga", "qemu-extended-regs", "off" },
295     { "virtio-scsi-pci", "any_layout", "off" },
296     { "usb-mouse", "usb_version", "1" },
297     { "usb-kbd", "usb_version", "1" },
298     { "virtio-pci", "virtio-pci-bus-master-bug-migration", "on" },
299 };
300 const size_t hw_compat_2_1_len = G_N_ELEMENTS(hw_compat_2_1);
301 
302 MachineState *current_machine;
303 
304 static char *machine_get_kernel(Object *obj, Error **errp)
305 {
306     MachineState *ms = MACHINE(obj);
307 
308     return g_strdup(ms->kernel_filename);
309 }
310 
311 static void machine_set_kernel(Object *obj, const char *value, Error **errp)
312 {
313     MachineState *ms = MACHINE(obj);
314 
315     g_free(ms->kernel_filename);
316     ms->kernel_filename = g_strdup(value);
317 }
318 
319 static char *machine_get_initrd(Object *obj, Error **errp)
320 {
321     MachineState *ms = MACHINE(obj);
322 
323     return g_strdup(ms->initrd_filename);
324 }
325 
326 static void machine_set_initrd(Object *obj, const char *value, Error **errp)
327 {
328     MachineState *ms = MACHINE(obj);
329 
330     g_free(ms->initrd_filename);
331     ms->initrd_filename = g_strdup(value);
332 }
333 
334 static char *machine_get_append(Object *obj, Error **errp)
335 {
336     MachineState *ms = MACHINE(obj);
337 
338     return g_strdup(ms->kernel_cmdline);
339 }
340 
341 static void machine_set_append(Object *obj, const char *value, Error **errp)
342 {
343     MachineState *ms = MACHINE(obj);
344 
345     g_free(ms->kernel_cmdline);
346     ms->kernel_cmdline = g_strdup(value);
347 }
348 
349 static char *machine_get_dtb(Object *obj, Error **errp)
350 {
351     MachineState *ms = MACHINE(obj);
352 
353     return g_strdup(ms->dtb);
354 }
355 
356 static void machine_set_dtb(Object *obj, const char *value, Error **errp)
357 {
358     MachineState *ms = MACHINE(obj);
359 
360     g_free(ms->dtb);
361     ms->dtb = g_strdup(value);
362 }
363 
364 static char *machine_get_dumpdtb(Object *obj, Error **errp)
365 {
366     MachineState *ms = MACHINE(obj);
367 
368     return g_strdup(ms->dumpdtb);
369 }
370 
371 static void machine_set_dumpdtb(Object *obj, const char *value, Error **errp)
372 {
373     MachineState *ms = MACHINE(obj);
374 
375     g_free(ms->dumpdtb);
376     ms->dumpdtb = g_strdup(value);
377 }
378 
379 static void machine_get_phandle_start(Object *obj, Visitor *v,
380                                       const char *name, void *opaque,
381                                       Error **errp)
382 {
383     MachineState *ms = MACHINE(obj);
384     int64_t value = ms->phandle_start;
385 
386     visit_type_int(v, name, &value, errp);
387 }
388 
389 static void machine_set_phandle_start(Object *obj, Visitor *v,
390                                       const char *name, void *opaque,
391                                       Error **errp)
392 {
393     MachineState *ms = MACHINE(obj);
394     int64_t value;
395 
396     if (!visit_type_int(v, name, &value, errp)) {
397         return;
398     }
399 
400     ms->phandle_start = value;
401 }
402 
403 static char *machine_get_dt_compatible(Object *obj, Error **errp)
404 {
405     MachineState *ms = MACHINE(obj);
406 
407     return g_strdup(ms->dt_compatible);
408 }
409 
410 static void machine_set_dt_compatible(Object *obj, const char *value, Error **errp)
411 {
412     MachineState *ms = MACHINE(obj);
413 
414     g_free(ms->dt_compatible);
415     ms->dt_compatible = g_strdup(value);
416 }
417 
418 static bool machine_get_dump_guest_core(Object *obj, Error **errp)
419 {
420     MachineState *ms = MACHINE(obj);
421 
422     return ms->dump_guest_core;
423 }
424 
425 static void machine_set_dump_guest_core(Object *obj, bool value, Error **errp)
426 {
427     MachineState *ms = MACHINE(obj);
428 
429     ms->dump_guest_core = value;
430 }
431 
432 static bool machine_get_mem_merge(Object *obj, Error **errp)
433 {
434     MachineState *ms = MACHINE(obj);
435 
436     return ms->mem_merge;
437 }
438 
439 static void machine_set_mem_merge(Object *obj, bool value, Error **errp)
440 {
441     MachineState *ms = MACHINE(obj);
442 
443     ms->mem_merge = value;
444 }
445 
446 static bool machine_get_usb(Object *obj, Error **errp)
447 {
448     MachineState *ms = MACHINE(obj);
449 
450     return ms->usb;
451 }
452 
453 static void machine_set_usb(Object *obj, bool value, Error **errp)
454 {
455     MachineState *ms = MACHINE(obj);
456 
457     ms->usb = value;
458     ms->usb_disabled = !value;
459 }
460 
461 static bool machine_get_graphics(Object *obj, Error **errp)
462 {
463     MachineState *ms = MACHINE(obj);
464 
465     return ms->enable_graphics;
466 }
467 
468 static void machine_set_graphics(Object *obj, bool value, Error **errp)
469 {
470     MachineState *ms = MACHINE(obj);
471 
472     ms->enable_graphics = value;
473 }
474 
475 static char *machine_get_firmware(Object *obj, Error **errp)
476 {
477     MachineState *ms = MACHINE(obj);
478 
479     return g_strdup(ms->firmware);
480 }
481 
482 static void machine_set_firmware(Object *obj, const char *value, Error **errp)
483 {
484     MachineState *ms = MACHINE(obj);
485 
486     g_free(ms->firmware);
487     ms->firmware = g_strdup(value);
488 }
489 
490 static void machine_set_suppress_vmdesc(Object *obj, bool value, Error **errp)
491 {
492     MachineState *ms = MACHINE(obj);
493 
494     ms->suppress_vmdesc = value;
495 }
496 
497 static bool machine_get_suppress_vmdesc(Object *obj, Error **errp)
498 {
499     MachineState *ms = MACHINE(obj);
500 
501     return ms->suppress_vmdesc;
502 }
503 
504 static char *machine_get_memory_encryption(Object *obj, Error **errp)
505 {
506     MachineState *ms = MACHINE(obj);
507 
508     if (ms->cgs) {
509         return g_strdup(object_get_canonical_path_component(OBJECT(ms->cgs)));
510     }
511 
512     return NULL;
513 }
514 
515 static void machine_set_memory_encryption(Object *obj, const char *value,
516                                         Error **errp)
517 {
518     Object *cgs =
519         object_resolve_path_component(object_get_objects_root(), value);
520 
521     if (!cgs) {
522         error_setg(errp, "No such memory encryption object '%s'", value);
523         return;
524     }
525 
526     object_property_set_link(obj, "confidential-guest-support", cgs, errp);
527 }
528 
529 static void machine_check_confidential_guest_support(const Object *obj,
530                                                      const char *name,
531                                                      Object *new_target,
532                                                      Error **errp)
533 {
534     /*
535      * So far the only constraint is that the target has the
536      * TYPE_CONFIDENTIAL_GUEST_SUPPORT interface, and that's checked
537      * by the QOM core
538      */
539 }
540 
541 static bool machine_get_nvdimm(Object *obj, Error **errp)
542 {
543     MachineState *ms = MACHINE(obj);
544 
545     return ms->nvdimms_state->is_enabled;
546 }
547 
548 static void machine_set_nvdimm(Object *obj, bool value, Error **errp)
549 {
550     MachineState *ms = MACHINE(obj);
551 
552     ms->nvdimms_state->is_enabled = value;
553 }
554 
555 static bool machine_get_hmat(Object *obj, Error **errp)
556 {
557     MachineState *ms = MACHINE(obj);
558 
559     return ms->numa_state->hmat_enabled;
560 }
561 
562 static void machine_set_hmat(Object *obj, bool value, Error **errp)
563 {
564     MachineState *ms = MACHINE(obj);
565 
566     ms->numa_state->hmat_enabled = value;
567 }
568 
569 static void machine_get_mem(Object *obj, Visitor *v, const char *name,
570                             void *opaque, Error **errp)
571 {
572     MachineState *ms = MACHINE(obj);
573     MemorySizeConfiguration mem = {
574         .has_size = true,
575         .size = ms->ram_size,
576         .has_max_size = !!ms->ram_slots,
577         .max_size = ms->maxram_size,
578         .has_slots = !!ms->ram_slots,
579         .slots = ms->ram_slots,
580     };
581     MemorySizeConfiguration *p_mem = &mem;
582 
583     visit_type_MemorySizeConfiguration(v, name, &p_mem, &error_abort);
584 }
585 
586 static void machine_set_mem(Object *obj, Visitor *v, const char *name,
587                             void *opaque, Error **errp)
588 {
589     ERRP_GUARD();
590     MachineState *ms = MACHINE(obj);
591     MachineClass *mc = MACHINE_GET_CLASS(obj);
592     MemorySizeConfiguration *mem;
593 
594     if (!visit_type_MemorySizeConfiguration(v, name, &mem, errp)) {
595         return;
596     }
597 
598     if (!mem->has_size) {
599         mem->has_size = true;
600         mem->size = mc->default_ram_size;
601     }
602     mem->size = QEMU_ALIGN_UP(mem->size, 8192);
603     if (mc->fixup_ram_size) {
604         mem->size = mc->fixup_ram_size(mem->size);
605     }
606     if ((ram_addr_t)mem->size != mem->size) {
607         error_setg(errp, "ram size too large");
608         goto out_free;
609     }
610 
611     if (mem->has_max_size) {
612         if (mem->max_size < mem->size) {
613             error_setg(errp, "invalid value of maxmem: "
614                        "maximum memory size (0x%" PRIx64 ") must be at least "
615                        "the initial memory size (0x%" PRIx64 ")",
616                        mem->max_size, mem->size);
617             goto out_free;
618         }
619         if (mem->has_slots && mem->slots && mem->max_size == mem->size) {
620             error_setg(errp, "invalid value of maxmem: "
621                        "memory slots were specified but maximum memory size "
622                        "(0x%" PRIx64 ") is equal to the initial memory size "
623                        "(0x%" PRIx64 ")", mem->max_size, mem->size);
624             goto out_free;
625         }
626         ms->maxram_size = mem->max_size;
627     } else {
628         if (mem->has_slots) {
629             error_setg(errp, "slots specified but no max-size");
630             goto out_free;
631         }
632         ms->maxram_size = mem->size;
633     }
634     ms->ram_size = mem->size;
635     ms->ram_slots = mem->has_slots ? mem->slots : 0;
636 out_free:
637     qapi_free_MemorySizeConfiguration(mem);
638 }
639 
640 static char *machine_get_nvdimm_persistence(Object *obj, Error **errp)
641 {
642     MachineState *ms = MACHINE(obj);
643 
644     return g_strdup(ms->nvdimms_state->persistence_string);
645 }
646 
647 static void machine_set_nvdimm_persistence(Object *obj, const char *value,
648                                            Error **errp)
649 {
650     MachineState *ms = MACHINE(obj);
651     NVDIMMState *nvdimms_state = ms->nvdimms_state;
652 
653     if (strcmp(value, "cpu") == 0) {
654         nvdimms_state->persistence = 3;
655     } else if (strcmp(value, "mem-ctrl") == 0) {
656         nvdimms_state->persistence = 2;
657     } else {
658         error_setg(errp, "-machine nvdimm-persistence=%s: unsupported option",
659                    value);
660         return;
661     }
662 
663     g_free(nvdimms_state->persistence_string);
664     nvdimms_state->persistence_string = g_strdup(value);
665 }
666 
667 void machine_class_allow_dynamic_sysbus_dev(MachineClass *mc, const char *type)
668 {
669     QAPI_LIST_PREPEND(mc->allowed_dynamic_sysbus_devices, g_strdup(type));
670 }
671 
672 bool device_is_dynamic_sysbus(MachineClass *mc, DeviceState *dev)
673 {
674     Object *obj = OBJECT(dev);
675 
676     if (!object_dynamic_cast(obj, TYPE_SYS_BUS_DEVICE)) {
677         return false;
678     }
679 
680     return device_type_is_dynamic_sysbus(mc, object_get_typename(obj));
681 }
682 
683 bool device_type_is_dynamic_sysbus(MachineClass *mc, const char *type)
684 {
685     bool allowed = false;
686     strList *wl;
687     ObjectClass *klass = object_class_by_name(type);
688 
689     for (wl = mc->allowed_dynamic_sysbus_devices;
690          !allowed && wl;
691          wl = wl->next) {
692         allowed |= !!object_class_dynamic_cast(klass, wl->value);
693     }
694 
695     return allowed;
696 }
697 
698 static char *machine_get_audiodev(Object *obj, Error **errp)
699 {
700     MachineState *ms = MACHINE(obj);
701 
702     return g_strdup(ms->audiodev);
703 }
704 
705 static void machine_set_audiodev(Object *obj, const char *value,
706                                  Error **errp)
707 {
708     MachineState *ms = MACHINE(obj);
709 
710     if (!audio_state_by_name(value, errp)) {
711         return;
712     }
713 
714     g_free(ms->audiodev);
715     ms->audiodev = g_strdup(value);
716 }
717 
718 HotpluggableCPUList *machine_query_hotpluggable_cpus(MachineState *machine)
719 {
720     int i;
721     HotpluggableCPUList *head = NULL;
722     MachineClass *mc = MACHINE_GET_CLASS(machine);
723 
724     /* force board to initialize possible_cpus if it hasn't been done yet */
725     mc->possible_cpu_arch_ids(machine);
726 
727     for (i = 0; i < machine->possible_cpus->len; i++) {
728         CPUState *cpu;
729         HotpluggableCPU *cpu_item = g_new0(typeof(*cpu_item), 1);
730 
731         cpu_item->type = g_strdup(machine->possible_cpus->cpus[i].type);
732         cpu_item->vcpus_count = machine->possible_cpus->cpus[i].vcpus_count;
733         cpu_item->props = g_memdup(&machine->possible_cpus->cpus[i].props,
734                                    sizeof(*cpu_item->props));
735 
736         cpu = machine->possible_cpus->cpus[i].cpu;
737         if (cpu) {
738             cpu_item->qom_path = object_get_canonical_path(OBJECT(cpu));
739         }
740         QAPI_LIST_PREPEND(head, cpu_item);
741     }
742     return head;
743 }
744 
745 /**
746  * machine_set_cpu_numa_node:
747  * @machine: machine object to modify
748  * @props: specifies which cpu objects to assign to
749  *         numa node specified by @props.node_id
750  * @errp: if an error occurs, a pointer to an area to store the error
751  *
752  * Associate NUMA node specified by @props.node_id with cpu slots that
753  * match socket/core/thread-ids specified by @props. It's recommended to use
754  * query-hotpluggable-cpus.props values to specify affected cpu slots,
755  * which would lead to exact 1:1 mapping of cpu slots to NUMA node.
756  *
757  * However for CLI convenience it's possible to pass in subset of properties,
758  * which would affect all cpu slots that match it.
759  * Ex for pc machine:
760  *    -smp 4,cores=2,sockets=2 -numa node,nodeid=0 -numa node,nodeid=1 \
761  *    -numa cpu,node-id=0,socket_id=0 \
762  *    -numa cpu,node-id=1,socket_id=1
763  * will assign all child cores of socket 0 to node 0 and
764  * of socket 1 to node 1.
765  *
766  * On attempt of reassigning (already assigned) cpu slot to another NUMA node,
767  * return error.
768  * Empty subset is disallowed and function will return with error in this case.
769  */
770 void machine_set_cpu_numa_node(MachineState *machine,
771                                const CpuInstanceProperties *props, Error **errp)
772 {
773     MachineClass *mc = MACHINE_GET_CLASS(machine);
774     NodeInfo *numa_info = machine->numa_state->nodes;
775     bool match = false;
776     int i;
777 
778     if (!mc->possible_cpu_arch_ids) {
779         error_setg(errp, "mapping of CPUs to NUMA node is not supported");
780         return;
781     }
782 
783     /* disabling node mapping is not supported, forbid it */
784     assert(props->has_node_id);
785 
786     /* force board to initialize possible_cpus if it hasn't been done yet */
787     mc->possible_cpu_arch_ids(machine);
788 
789     for (i = 0; i < machine->possible_cpus->len; i++) {
790         CPUArchId *slot = &machine->possible_cpus->cpus[i];
791 
792         /* reject unsupported by board properties */
793         if (props->has_thread_id && !slot->props.has_thread_id) {
794             error_setg(errp, "thread-id is not supported");
795             return;
796         }
797 
798         if (props->has_core_id && !slot->props.has_core_id) {
799             error_setg(errp, "core-id is not supported");
800             return;
801         }
802 
803         if (props->has_cluster_id && !slot->props.has_cluster_id) {
804             error_setg(errp, "cluster-id is not supported");
805             return;
806         }
807 
808         if (props->has_socket_id && !slot->props.has_socket_id) {
809             error_setg(errp, "socket-id is not supported");
810             return;
811         }
812 
813         if (props->has_die_id && !slot->props.has_die_id) {
814             error_setg(errp, "die-id is not supported");
815             return;
816         }
817 
818         /* skip slots with explicit mismatch */
819         if (props->has_thread_id && props->thread_id != slot->props.thread_id) {
820                 continue;
821         }
822 
823         if (props->has_core_id && props->core_id != slot->props.core_id) {
824                 continue;
825         }
826 
827         if (props->has_cluster_id &&
828             props->cluster_id != slot->props.cluster_id) {
829                 continue;
830         }
831 
832         if (props->has_die_id && props->die_id != slot->props.die_id) {
833                 continue;
834         }
835 
836         if (props->has_socket_id && props->socket_id != slot->props.socket_id) {
837                 continue;
838         }
839 
840         /* reject assignment if slot is already assigned, for compatibility
841          * of legacy cpu_index mapping with SPAPR core based mapping do not
842          * error out if cpu thread and matched core have the same node-id */
843         if (slot->props.has_node_id &&
844             slot->props.node_id != props->node_id) {
845             error_setg(errp, "CPU is already assigned to node-id: %" PRId64,
846                        slot->props.node_id);
847             return;
848         }
849 
850         /* assign slot to node as it's matched '-numa cpu' key */
851         match = true;
852         slot->props.node_id = props->node_id;
853         slot->props.has_node_id = props->has_node_id;
854 
855         if (machine->numa_state->hmat_enabled) {
856             if ((numa_info[props->node_id].initiator < MAX_NODES) &&
857                 (props->node_id != numa_info[props->node_id].initiator)) {
858                 error_setg(errp, "The initiator of CPU NUMA node %" PRId64
859                            " should be itself (got %" PRIu16 ")",
860                            props->node_id, numa_info[props->node_id].initiator);
861                 return;
862             }
863             numa_info[props->node_id].has_cpu = true;
864             numa_info[props->node_id].initiator = props->node_id;
865         }
866     }
867 
868     if (!match) {
869         error_setg(errp, "no match found");
870     }
871 }
872 
873 static void machine_get_smp(Object *obj, Visitor *v, const char *name,
874                             void *opaque, Error **errp)
875 {
876     MachineState *ms = MACHINE(obj);
877     SMPConfiguration *config = &(SMPConfiguration){
878         .has_cpus = true, .cpus = ms->smp.cpus,
879         .has_drawers = true, .drawers = ms->smp.drawers,
880         .has_books = true, .books = ms->smp.books,
881         .has_sockets = true, .sockets = ms->smp.sockets,
882         .has_dies = true, .dies = ms->smp.dies,
883         .has_clusters = true, .clusters = ms->smp.clusters,
884         .has_cores = true, .cores = ms->smp.cores,
885         .has_threads = true, .threads = ms->smp.threads,
886         .has_maxcpus = true, .maxcpus = ms->smp.max_cpus,
887     };
888 
889     if (!visit_type_SMPConfiguration(v, name, &config, &error_abort)) {
890         return;
891     }
892 }
893 
894 static void machine_set_smp(Object *obj, Visitor *v, const char *name,
895                             void *opaque, Error **errp)
896 {
897     MachineState *ms = MACHINE(obj);
898     g_autoptr(SMPConfiguration) config = NULL;
899 
900     if (!visit_type_SMPConfiguration(v, name, &config, errp)) {
901         return;
902     }
903 
904     machine_parse_smp_config(ms, config, errp);
905 }
906 
907 static void machine_get_boot(Object *obj, Visitor *v, const char *name,
908                             void *opaque, Error **errp)
909 {
910     MachineState *ms = MACHINE(obj);
911     BootConfiguration *config = &ms->boot_config;
912     visit_type_BootConfiguration(v, name, &config, &error_abort);
913 }
914 
915 static void machine_free_boot_config(MachineState *ms)
916 {
917     g_free(ms->boot_config.order);
918     g_free(ms->boot_config.once);
919     g_free(ms->boot_config.splash);
920 }
921 
922 static void machine_copy_boot_config(MachineState *ms, BootConfiguration *config)
923 {
924     MachineClass *machine_class = MACHINE_GET_CLASS(ms);
925 
926     machine_free_boot_config(ms);
927     ms->boot_config = *config;
928     if (!config->order) {
929         ms->boot_config.order = g_strdup(machine_class->default_boot_order);
930     }
931 }
932 
933 static void machine_set_boot(Object *obj, Visitor *v, const char *name,
934                             void *opaque, Error **errp)
935 {
936     ERRP_GUARD();
937     MachineState *ms = MACHINE(obj);
938     BootConfiguration *config = NULL;
939 
940     if (!visit_type_BootConfiguration(v, name, &config, errp)) {
941         return;
942     }
943     if (config->order) {
944         validate_bootdevices(config->order, errp);
945         if (*errp) {
946             goto out_free;
947         }
948     }
949     if (config->once) {
950         validate_bootdevices(config->once, errp);
951         if (*errp) {
952             goto out_free;
953         }
954     }
955 
956     machine_copy_boot_config(ms, config);
957     /* Strings live in ms->boot_config.  */
958     free(config);
959     return;
960 
961 out_free:
962     qapi_free_BootConfiguration(config);
963 }
964 
965 void machine_add_audiodev_property(MachineClass *mc)
966 {
967     ObjectClass *oc = OBJECT_CLASS(mc);
968 
969     object_class_property_add_str(oc, "audiodev",
970                                   machine_get_audiodev,
971                                   machine_set_audiodev);
972     object_class_property_set_description(oc, "audiodev",
973                                           "Audiodev to use for default machine devices");
974 }
975 
976 static void machine_class_init(ObjectClass *oc, void *data)
977 {
978     MachineClass *mc = MACHINE_CLASS(oc);
979 
980     /* Default 128 MB as guest ram size */
981     mc->default_ram_size = 128 * MiB;
982     mc->rom_file_has_mr = true;
983 
984     /* numa node memory size aligned on 8MB by default.
985      * On Linux, each node's border has to be 8MB aligned
986      */
987     mc->numa_mem_align_shift = 23;
988 
989     object_class_property_add_str(oc, "kernel",
990         machine_get_kernel, machine_set_kernel);
991     object_class_property_set_description(oc, "kernel",
992         "Linux kernel image file");
993 
994     object_class_property_add_str(oc, "initrd",
995         machine_get_initrd, machine_set_initrd);
996     object_class_property_set_description(oc, "initrd",
997         "Linux initial ramdisk file");
998 
999     object_class_property_add_str(oc, "append",
1000         machine_get_append, machine_set_append);
1001     object_class_property_set_description(oc, "append",
1002         "Linux kernel command line");
1003 
1004     object_class_property_add_str(oc, "dtb",
1005         machine_get_dtb, machine_set_dtb);
1006     object_class_property_set_description(oc, "dtb",
1007         "Linux kernel device tree file");
1008 
1009     object_class_property_add_str(oc, "dumpdtb",
1010         machine_get_dumpdtb, machine_set_dumpdtb);
1011     object_class_property_set_description(oc, "dumpdtb",
1012         "Dump current dtb to a file and quit");
1013 
1014     object_class_property_add(oc, "boot", "BootConfiguration",
1015         machine_get_boot, machine_set_boot,
1016         NULL, NULL);
1017     object_class_property_set_description(oc, "boot",
1018         "Boot configuration");
1019 
1020     object_class_property_add(oc, "smp", "SMPConfiguration",
1021         machine_get_smp, machine_set_smp,
1022         NULL, NULL);
1023     object_class_property_set_description(oc, "smp",
1024         "CPU topology");
1025 
1026     object_class_property_add(oc, "phandle-start", "int",
1027         machine_get_phandle_start, machine_set_phandle_start,
1028         NULL, NULL);
1029     object_class_property_set_description(oc, "phandle-start",
1030         "The first phandle ID we may generate dynamically");
1031 
1032     object_class_property_add_str(oc, "dt-compatible",
1033         machine_get_dt_compatible, machine_set_dt_compatible);
1034     object_class_property_set_description(oc, "dt-compatible",
1035         "Overrides the \"compatible\" property of the dt root node");
1036 
1037     object_class_property_add_bool(oc, "dump-guest-core",
1038         machine_get_dump_guest_core, machine_set_dump_guest_core);
1039     object_class_property_set_description(oc, "dump-guest-core",
1040         "Include guest memory in a core dump");
1041 
1042     object_class_property_add_bool(oc, "mem-merge",
1043         machine_get_mem_merge, machine_set_mem_merge);
1044     object_class_property_set_description(oc, "mem-merge",
1045         "Enable/disable memory merge support");
1046 
1047     object_class_property_add_bool(oc, "usb",
1048         machine_get_usb, machine_set_usb);
1049     object_class_property_set_description(oc, "usb",
1050         "Set on/off to enable/disable usb");
1051 
1052     object_class_property_add_bool(oc, "graphics",
1053         machine_get_graphics, machine_set_graphics);
1054     object_class_property_set_description(oc, "graphics",
1055         "Set on/off to enable/disable graphics emulation");
1056 
1057     object_class_property_add_str(oc, "firmware",
1058         machine_get_firmware, machine_set_firmware);
1059     object_class_property_set_description(oc, "firmware",
1060         "Firmware image");
1061 
1062     object_class_property_add_bool(oc, "suppress-vmdesc",
1063         machine_get_suppress_vmdesc, machine_set_suppress_vmdesc);
1064     object_class_property_set_description(oc, "suppress-vmdesc",
1065         "Set on to disable self-describing migration");
1066 
1067     object_class_property_add_link(oc, "confidential-guest-support",
1068                                    TYPE_CONFIDENTIAL_GUEST_SUPPORT,
1069                                    offsetof(MachineState, cgs),
1070                                    machine_check_confidential_guest_support,
1071                                    OBJ_PROP_LINK_STRONG);
1072     object_class_property_set_description(oc, "confidential-guest-support",
1073                                           "Set confidential guest scheme to support");
1074 
1075     /* For compatibility */
1076     object_class_property_add_str(oc, "memory-encryption",
1077         machine_get_memory_encryption, machine_set_memory_encryption);
1078     object_class_property_set_description(oc, "memory-encryption",
1079         "Set memory encryption object to use");
1080 
1081     object_class_property_add_link(oc, "memory-backend", TYPE_MEMORY_BACKEND,
1082                                    offsetof(MachineState, memdev), object_property_allow_set_link,
1083                                    OBJ_PROP_LINK_STRONG);
1084     object_class_property_set_description(oc, "memory-backend",
1085                                           "Set RAM backend"
1086                                           "Valid value is ID of hostmem based backend");
1087 
1088     object_class_property_add(oc, "memory", "MemorySizeConfiguration",
1089         machine_get_mem, machine_set_mem,
1090         NULL, NULL);
1091     object_class_property_set_description(oc, "memory",
1092         "Memory size configuration");
1093 }
1094 
1095 static void machine_class_base_init(ObjectClass *oc, void *data)
1096 {
1097     MachineClass *mc = MACHINE_CLASS(oc);
1098     mc->max_cpus = mc->max_cpus ?: 1;
1099     mc->min_cpus = mc->min_cpus ?: 1;
1100     mc->default_cpus = mc->default_cpus ?: 1;
1101 
1102     if (!object_class_is_abstract(oc)) {
1103         const char *cname = object_class_get_name(oc);
1104         assert(g_str_has_suffix(cname, TYPE_MACHINE_SUFFIX));
1105         mc->name = g_strndup(cname,
1106                             strlen(cname) - strlen(TYPE_MACHINE_SUFFIX));
1107         mc->compat_props = g_ptr_array_new();
1108     }
1109 }
1110 
1111 static void machine_initfn(Object *obj)
1112 {
1113     MachineState *ms = MACHINE(obj);
1114     MachineClass *mc = MACHINE_GET_CLASS(obj);
1115 
1116     container_get(obj, "/peripheral");
1117     container_get(obj, "/peripheral-anon");
1118 
1119     ms->dump_guest_core = true;
1120     ms->mem_merge = true;
1121     ms->enable_graphics = true;
1122     ms->kernel_cmdline = g_strdup("");
1123     ms->ram_size = mc->default_ram_size;
1124     ms->maxram_size = mc->default_ram_size;
1125 
1126     if (mc->nvdimm_supported) {
1127         ms->nvdimms_state = g_new0(NVDIMMState, 1);
1128         object_property_add_bool(obj, "nvdimm",
1129                                  machine_get_nvdimm, machine_set_nvdimm);
1130         object_property_set_description(obj, "nvdimm",
1131                                         "Set on/off to enable/disable "
1132                                         "NVDIMM instantiation");
1133 
1134         object_property_add_str(obj, "nvdimm-persistence",
1135                                 machine_get_nvdimm_persistence,
1136                                 machine_set_nvdimm_persistence);
1137         object_property_set_description(obj, "nvdimm-persistence",
1138                                         "Set NVDIMM persistence"
1139                                         "Valid values are cpu, mem-ctrl");
1140     }
1141 
1142     if (mc->cpu_index_to_instance_props && mc->get_default_cpu_node_id) {
1143         ms->numa_state = g_new0(NumaState, 1);
1144         object_property_add_bool(obj, "hmat",
1145                                  machine_get_hmat, machine_set_hmat);
1146         object_property_set_description(obj, "hmat",
1147                                         "Set on/off to enable/disable "
1148                                         "ACPI Heterogeneous Memory Attribute "
1149                                         "Table (HMAT)");
1150     }
1151 
1152     /* default to mc->default_cpus */
1153     ms->smp.cpus = mc->default_cpus;
1154     ms->smp.max_cpus = mc->default_cpus;
1155     ms->smp.drawers = 1;
1156     ms->smp.books = 1;
1157     ms->smp.sockets = 1;
1158     ms->smp.dies = 1;
1159     ms->smp.clusters = 1;
1160     ms->smp.cores = 1;
1161     ms->smp.threads = 1;
1162 
1163     machine_copy_boot_config(ms, &(BootConfiguration){ 0 });
1164 }
1165 
1166 static void machine_finalize(Object *obj)
1167 {
1168     MachineState *ms = MACHINE(obj);
1169 
1170     machine_free_boot_config(ms);
1171     g_free(ms->kernel_filename);
1172     g_free(ms->initrd_filename);
1173     g_free(ms->kernel_cmdline);
1174     g_free(ms->dtb);
1175     g_free(ms->dumpdtb);
1176     g_free(ms->dt_compatible);
1177     g_free(ms->firmware);
1178     g_free(ms->device_memory);
1179     g_free(ms->nvdimms_state);
1180     g_free(ms->numa_state);
1181     g_free(ms->audiodev);
1182 }
1183 
1184 bool machine_usb(MachineState *machine)
1185 {
1186     return machine->usb;
1187 }
1188 
1189 int machine_phandle_start(MachineState *machine)
1190 {
1191     return machine->phandle_start;
1192 }
1193 
1194 bool machine_dump_guest_core(MachineState *machine)
1195 {
1196     return machine->dump_guest_core;
1197 }
1198 
1199 bool machine_mem_merge(MachineState *machine)
1200 {
1201     return machine->mem_merge;
1202 }
1203 
1204 bool machine_require_guest_memfd(MachineState *machine)
1205 {
1206     return machine->require_guest_memfd;
1207 }
1208 
1209 static char *cpu_slot_to_string(const CPUArchId *cpu)
1210 {
1211     GString *s = g_string_new(NULL);
1212     if (cpu->props.has_socket_id) {
1213         g_string_append_printf(s, "socket-id: %"PRId64, cpu->props.socket_id);
1214     }
1215     if (cpu->props.has_die_id) {
1216         if (s->len) {
1217             g_string_append_printf(s, ", ");
1218         }
1219         g_string_append_printf(s, "die-id: %"PRId64, cpu->props.die_id);
1220     }
1221     if (cpu->props.has_cluster_id) {
1222         if (s->len) {
1223             g_string_append_printf(s, ", ");
1224         }
1225         g_string_append_printf(s, "cluster-id: %"PRId64, cpu->props.cluster_id);
1226     }
1227     if (cpu->props.has_core_id) {
1228         if (s->len) {
1229             g_string_append_printf(s, ", ");
1230         }
1231         g_string_append_printf(s, "core-id: %"PRId64, cpu->props.core_id);
1232     }
1233     if (cpu->props.has_thread_id) {
1234         if (s->len) {
1235             g_string_append_printf(s, ", ");
1236         }
1237         g_string_append_printf(s, "thread-id: %"PRId64, cpu->props.thread_id);
1238     }
1239     return g_string_free(s, false);
1240 }
1241 
1242 static void numa_validate_initiator(NumaState *numa_state)
1243 {
1244     int i;
1245     NodeInfo *numa_info = numa_state->nodes;
1246 
1247     for (i = 0; i < numa_state->num_nodes; i++) {
1248         if (numa_info[i].initiator == MAX_NODES) {
1249             continue;
1250         }
1251 
1252         if (!numa_info[numa_info[i].initiator].present) {
1253             error_report("NUMA node %" PRIu16 " is missing, use "
1254                          "'-numa node' option to declare it first",
1255                          numa_info[i].initiator);
1256             exit(1);
1257         }
1258 
1259         if (!numa_info[numa_info[i].initiator].has_cpu) {
1260             error_report("The initiator of NUMA node %d is invalid", i);
1261             exit(1);
1262         }
1263     }
1264 }
1265 
1266 static void machine_numa_finish_cpu_init(MachineState *machine)
1267 {
1268     int i;
1269     bool default_mapping;
1270     GString *s = g_string_new(NULL);
1271     MachineClass *mc = MACHINE_GET_CLASS(machine);
1272     const CPUArchIdList *possible_cpus = mc->possible_cpu_arch_ids(machine);
1273 
1274     assert(machine->numa_state->num_nodes);
1275     for (i = 0; i < possible_cpus->len; i++) {
1276         if (possible_cpus->cpus[i].props.has_node_id) {
1277             break;
1278         }
1279     }
1280     default_mapping = (i == possible_cpus->len);
1281 
1282     for (i = 0; i < possible_cpus->len; i++) {
1283         const CPUArchId *cpu_slot = &possible_cpus->cpus[i];
1284 
1285         if (!cpu_slot->props.has_node_id) {
1286             /* fetch default mapping from board and enable it */
1287             CpuInstanceProperties props = cpu_slot->props;
1288 
1289             props.node_id = mc->get_default_cpu_node_id(machine, i);
1290             if (!default_mapping) {
1291                 /* record slots with not set mapping,
1292                  * TODO: make it hard error in future */
1293                 char *cpu_str = cpu_slot_to_string(cpu_slot);
1294                 g_string_append_printf(s, "%sCPU %d [%s]",
1295                                        s->len ? ", " : "", i, cpu_str);
1296                 g_free(cpu_str);
1297 
1298                 /* non mapped cpus used to fallback to node 0 */
1299                 props.node_id = 0;
1300             }
1301 
1302             props.has_node_id = true;
1303             machine_set_cpu_numa_node(machine, &props, &error_fatal);
1304         }
1305     }
1306 
1307     if (machine->numa_state->hmat_enabled) {
1308         numa_validate_initiator(machine->numa_state);
1309     }
1310 
1311     if (s->len && !qtest_enabled()) {
1312         warn_report("CPU(s) not present in any NUMA nodes: %s",
1313                     s->str);
1314         warn_report("All CPU(s) up to maxcpus should be described "
1315                     "in NUMA config, ability to start up with partial NUMA "
1316                     "mappings is obsoleted and will be removed in future");
1317     }
1318     g_string_free(s, true);
1319 }
1320 
1321 static void validate_cpu_cluster_to_numa_boundary(MachineState *ms)
1322 {
1323     MachineClass *mc = MACHINE_GET_CLASS(ms);
1324     NumaState *state = ms->numa_state;
1325     const CPUArchIdList *possible_cpus = mc->possible_cpu_arch_ids(ms);
1326     const CPUArchId *cpus = possible_cpus->cpus;
1327     int i, j;
1328 
1329     if (qtest_enabled() || state->num_nodes <= 1 || possible_cpus->len <= 1) {
1330         return;
1331     }
1332 
1333     /*
1334      * The Linux scheduling domain can't be parsed when the multiple CPUs
1335      * in one cluster have been associated with different NUMA nodes. However,
1336      * it's fine to associate one NUMA node with CPUs in different clusters.
1337      */
1338     for (i = 0; i < possible_cpus->len; i++) {
1339         for (j = i + 1; j < possible_cpus->len; j++) {
1340             if (cpus[i].props.has_socket_id &&
1341                 cpus[i].props.has_cluster_id &&
1342                 cpus[i].props.has_node_id &&
1343                 cpus[j].props.has_socket_id &&
1344                 cpus[j].props.has_cluster_id &&
1345                 cpus[j].props.has_node_id &&
1346                 cpus[i].props.socket_id == cpus[j].props.socket_id &&
1347                 cpus[i].props.cluster_id == cpus[j].props.cluster_id &&
1348                 cpus[i].props.node_id != cpus[j].props.node_id) {
1349                 warn_report("CPU-%d and CPU-%d in socket-%" PRId64 "-cluster-%" PRId64
1350                              " have been associated with node-%" PRId64 " and node-%" PRId64
1351                              " respectively. It can cause OSes like Linux to"
1352                              " misbehave", i, j, cpus[i].props.socket_id,
1353                              cpus[i].props.cluster_id, cpus[i].props.node_id,
1354                              cpus[j].props.node_id);
1355             }
1356         }
1357     }
1358 }
1359 
1360 MemoryRegion *machine_consume_memdev(MachineState *machine,
1361                                      HostMemoryBackend *backend)
1362 {
1363     MemoryRegion *ret = host_memory_backend_get_memory(backend);
1364 
1365     if (host_memory_backend_is_mapped(backend)) {
1366         error_report("memory backend %s can't be used multiple times.",
1367                      object_get_canonical_path_component(OBJECT(backend)));
1368         exit(EXIT_FAILURE);
1369     }
1370     host_memory_backend_set_mapped(backend, true);
1371     vmstate_register_ram_global(ret);
1372     return ret;
1373 }
1374 
1375 static bool create_default_memdev(MachineState *ms, const char *path, Error **errp)
1376 {
1377     Object *obj;
1378     MachineClass *mc = MACHINE_GET_CLASS(ms);
1379     bool r = false;
1380 
1381     obj = object_new(path ? TYPE_MEMORY_BACKEND_FILE : TYPE_MEMORY_BACKEND_RAM);
1382     if (path) {
1383         if (!object_property_set_str(obj, "mem-path", path, errp)) {
1384             goto out;
1385         }
1386     }
1387     if (!object_property_set_int(obj, "size", ms->ram_size, errp)) {
1388         goto out;
1389     }
1390     object_property_add_child(object_get_objects_root(), mc->default_ram_id,
1391                               obj);
1392     /* Ensure backend's memory region name is equal to mc->default_ram_id */
1393     if (!object_property_set_bool(obj, "x-use-canonical-path-for-ramblock-id",
1394                              false, errp)) {
1395         goto out;
1396     }
1397     if (!user_creatable_complete(USER_CREATABLE(obj), errp)) {
1398         goto out;
1399     }
1400     r = object_property_set_link(OBJECT(ms), "memory-backend", obj, errp);
1401 
1402 out:
1403     object_unref(obj);
1404     return r;
1405 }
1406 
1407 const char *machine_class_default_cpu_type(MachineClass *mc)
1408 {
1409     if (mc->valid_cpu_types && !mc->valid_cpu_types[1]) {
1410         /* Only a single CPU type allowed: use it as default. */
1411         return mc->valid_cpu_types[0];
1412     }
1413     return mc->default_cpu_type;
1414 }
1415 
1416 static bool is_cpu_type_supported(const MachineState *machine, Error **errp)
1417 {
1418     MachineClass *mc = MACHINE_GET_CLASS(machine);
1419     ObjectClass *oc = object_class_by_name(machine->cpu_type);
1420     CPUClass *cc;
1421     int i;
1422 
1423     /*
1424      * Check if the user specified CPU type is supported when the valid
1425      * CPU types have been determined. Note that the user specified CPU
1426      * type is provided through '-cpu' option.
1427      */
1428     if (mc->valid_cpu_types) {
1429         assert(mc->valid_cpu_types[0] != NULL);
1430         for (i = 0; mc->valid_cpu_types[i]; i++) {
1431             if (object_class_dynamic_cast(oc, mc->valid_cpu_types[i])) {
1432                 break;
1433             }
1434         }
1435 
1436         /* The user specified CPU type isn't valid */
1437         if (!mc->valid_cpu_types[i]) {
1438             g_autofree char *requested = cpu_model_from_type(machine->cpu_type);
1439             error_setg(errp, "Invalid CPU model: %s", requested);
1440             if (!mc->valid_cpu_types[1]) {
1441                 g_autofree char *model = cpu_model_from_type(
1442                                                  mc->valid_cpu_types[0]);
1443                 error_append_hint(errp, "The only valid type is: %s\n", model);
1444             } else {
1445                 error_append_hint(errp, "The valid models are: ");
1446                 for (i = 0; mc->valid_cpu_types[i]; i++) {
1447                     g_autofree char *model = cpu_model_from_type(
1448                                                  mc->valid_cpu_types[i]);
1449                     error_append_hint(errp, "%s%s",
1450                                       model,
1451                                       mc->valid_cpu_types[i + 1] ? ", " : "");
1452                 }
1453                 error_append_hint(errp, "\n");
1454             }
1455 
1456             return false;
1457         }
1458     }
1459 
1460     /* Check if CPU type is deprecated and warn if so */
1461     cc = CPU_CLASS(oc);
1462     assert(cc != NULL);
1463     if (cc->deprecation_note) {
1464         warn_report("CPU model %s is deprecated -- %s",
1465                     machine->cpu_type, cc->deprecation_note);
1466     }
1467 
1468     return true;
1469 }
1470 
1471 void machine_run_board_init(MachineState *machine, const char *mem_path, Error **errp)
1472 {
1473     ERRP_GUARD();
1474     MachineClass *machine_class = MACHINE_GET_CLASS(machine);
1475 
1476     /* This checkpoint is required by replay to separate prior clock
1477        reading from the other reads, because timer polling functions query
1478        clock values from the log. */
1479     replay_checkpoint(CHECKPOINT_INIT);
1480 
1481     if (!xen_enabled()) {
1482         /* On 32-bit hosts, QEMU is limited by virtual address space */
1483         if (machine->ram_size > (2047 << 20) && HOST_LONG_BITS == 32) {
1484             error_setg(errp, "at most 2047 MB RAM can be simulated");
1485             return;
1486         }
1487     }
1488 
1489     if (machine->memdev) {
1490         ram_addr_t backend_size = object_property_get_uint(OBJECT(machine->memdev),
1491                                                            "size",  &error_abort);
1492         if (backend_size != machine->ram_size) {
1493             error_setg(errp, "Machine memory size does not match the size of the memory backend");
1494             return;
1495         }
1496     } else if (machine_class->default_ram_id && machine->ram_size &&
1497                numa_uses_legacy_mem()) {
1498         if (object_property_find(object_get_objects_root(),
1499                                  machine_class->default_ram_id)) {
1500             error_setg(errp, "object's id '%s' is reserved for the default"
1501                 " RAM backend, it can't be used for any other purposes",
1502                 machine_class->default_ram_id);
1503             error_append_hint(errp,
1504                 "Change the object's 'id' to something else or disable"
1505                 " automatic creation of the default RAM backend by setting"
1506                 " 'memory-backend=%s' with '-machine'.\n",
1507                 machine_class->default_ram_id);
1508             return;
1509         }
1510         if (!create_default_memdev(current_machine, mem_path, errp)) {
1511             return;
1512         }
1513     }
1514 
1515     if (machine->numa_state) {
1516         numa_complete_configuration(machine);
1517         if (machine->numa_state->num_nodes) {
1518             machine_numa_finish_cpu_init(machine);
1519             if (machine_class->cpu_cluster_has_numa_boundary) {
1520                 validate_cpu_cluster_to_numa_boundary(machine);
1521             }
1522         }
1523     }
1524 
1525     if (!machine->ram && machine->memdev) {
1526         machine->ram = machine_consume_memdev(machine, machine->memdev);
1527     }
1528 
1529     /* Check if the CPU type is supported */
1530     if (machine->cpu_type && !is_cpu_type_supported(machine, errp)) {
1531         return;
1532     }
1533 
1534     if (machine->cgs) {
1535         /*
1536          * With confidential guests, the host can't see the real
1537          * contents of RAM, so there's no point in it trying to merge
1538          * areas.
1539          */
1540         machine_set_mem_merge(OBJECT(machine), false, &error_abort);
1541 
1542         /*
1543          * Virtio devices can't count on directly accessing guest
1544          * memory, so they need iommu_platform=on to use normal DMA
1545          * mechanisms.  That requires also disabling legacy virtio
1546          * support for those virtio pci devices which allow it.
1547          */
1548         object_register_sugar_prop(TYPE_VIRTIO_PCI, "disable-legacy",
1549                                    "on", true);
1550         object_register_sugar_prop(TYPE_VIRTIO_DEVICE, "iommu_platform",
1551                                    "on", false);
1552     }
1553 
1554     accel_init_interfaces(ACCEL_GET_CLASS(machine->accelerator));
1555     machine_class->init(machine);
1556     phase_advance(PHASE_MACHINE_INITIALIZED);
1557 }
1558 
1559 static NotifierList machine_init_done_notifiers =
1560     NOTIFIER_LIST_INITIALIZER(machine_init_done_notifiers);
1561 
1562 void qemu_add_machine_init_done_notifier(Notifier *notify)
1563 {
1564     notifier_list_add(&machine_init_done_notifiers, notify);
1565     if (phase_check(PHASE_MACHINE_READY)) {
1566         notify->notify(notify, NULL);
1567     }
1568 }
1569 
1570 void qemu_remove_machine_init_done_notifier(Notifier *notify)
1571 {
1572     notifier_remove(notify);
1573 }
1574 
1575 void qdev_machine_creation_done(void)
1576 {
1577     cpu_synchronize_all_post_init();
1578 
1579     if (current_machine->boot_config.once) {
1580         qemu_boot_set(current_machine->boot_config.once, &error_fatal);
1581         qemu_register_reset(restore_boot_order, g_strdup(current_machine->boot_config.order));
1582     }
1583 
1584     /*
1585      * ok, initial machine setup is done, starting from now we can
1586      * only create hotpluggable devices
1587      */
1588     phase_advance(PHASE_MACHINE_READY);
1589     qdev_assert_realized_properly();
1590 
1591     /* TODO: once all bus devices are qdevified, this should be done
1592      * when bus is created by qdev.c */
1593     /*
1594      * This is where we arrange for the sysbus to be reset when the
1595      * whole simulation is reset. In turn, resetting the sysbus will cause
1596      * all devices hanging off it (and all their child buses, recursively)
1597      * to be reset. Note that this will *not* reset any Device objects
1598      * which are not attached to some part of the qbus tree!
1599      */
1600     qemu_register_resettable(OBJECT(sysbus_get_default()));
1601 
1602     notifier_list_notify(&machine_init_done_notifiers, NULL);
1603 
1604     if (rom_check_and_register_reset() != 0) {
1605         exit(1);
1606     }
1607 
1608     replay_start();
1609 
1610     /* This checkpoint is required by replay to separate prior clock
1611        reading from the other reads, because timer polling functions query
1612        clock values from the log. */
1613     replay_checkpoint(CHECKPOINT_RESET);
1614     qemu_system_reset(SHUTDOWN_CAUSE_NONE);
1615     register_global_state();
1616 }
1617 
1618 static const TypeInfo machine_info = {
1619     .name = TYPE_MACHINE,
1620     .parent = TYPE_OBJECT,
1621     .abstract = true,
1622     .class_size = sizeof(MachineClass),
1623     .class_init    = machine_class_init,
1624     .class_base_init = machine_class_base_init,
1625     .instance_size = sizeof(MachineState),
1626     .instance_init = machine_initfn,
1627     .instance_finalize = machine_finalize,
1628 };
1629 
1630 static void machine_register_types(void)
1631 {
1632     type_register_static(&machine_info);
1633 }
1634 
1635 type_init(machine_register_types)
1636