xref: /openbmc/qemu/hw/core/machine.c (revision 6e6a6ca7d9097d1aa68089379cdaa7553dad7a25)
1 /*
2  * QEMU Machine
3  *
4  * Copyright (C) 2014 Red Hat Inc
5  *
6  * Authors:
7  *   Marcel Apfelbaum <marcel.a@redhat.com>
8  *
9  * This work is licensed under the terms of the GNU GPL, version 2 or later.
10  * See the COPYING file in the top-level directory.
11  */
12 
13 #include "qemu/osdep.h"
14 #include "qemu/option.h"
15 #include "qapi/qmp/qerror.h"
16 #include "sysemu/replay.h"
17 #include "qemu/units.h"
18 #include "hw/boards.h"
19 #include "hw/loader.h"
20 #include "qapi/error.h"
21 #include "qapi/qapi-visit-common.h"
22 #include "qapi/visitor.h"
23 #include "hw/sysbus.h"
24 #include "sysemu/cpus.h"
25 #include "sysemu/sysemu.h"
26 #include "sysemu/reset.h"
27 #include "sysemu/runstate.h"
28 #include "sysemu/numa.h"
29 #include "qemu/error-report.h"
30 #include "sysemu/qtest.h"
31 #include "hw/pci/pci.h"
32 #include "hw/mem/nvdimm.h"
33 #include "migration/global_state.h"
34 #include "migration/vmstate.h"
35 
36 GlobalProperty hw_compat_5_2[] = {};
37 const size_t hw_compat_5_2_len = G_N_ELEMENTS(hw_compat_5_2);
38 
39 GlobalProperty hw_compat_5_1[] = {
40     { "vhost-scsi", "num_queues", "1"},
41     { "vhost-user-blk", "num-queues", "1"},
42     { "vhost-user-scsi", "num_queues", "1"},
43     { "virtio-blk-device", "num-queues", "1"},
44     { "virtio-scsi-device", "num_queues", "1"},
45     { "nvme", "use-intel-id", "on"},
46     { "pvpanic", "events", "1"}, /* PVPANIC_PANICKED */
47 };
48 const size_t hw_compat_5_1_len = G_N_ELEMENTS(hw_compat_5_1);
49 
50 GlobalProperty hw_compat_5_0[] = {
51     { "pci-host-bridge", "x-config-reg-migration-enabled", "off" },
52     { "virtio-balloon-device", "page-poison", "false" },
53     { "vmport", "x-read-set-eax", "off" },
54     { "vmport", "x-signal-unsupported-cmd", "off" },
55     { "vmport", "x-report-vmx-type", "off" },
56     { "vmport", "x-cmds-v2", "off" },
57     { "virtio-device", "x-disable-legacy-check", "true" },
58 };
59 const size_t hw_compat_5_0_len = G_N_ELEMENTS(hw_compat_5_0);
60 
61 GlobalProperty hw_compat_4_2[] = {
62     { "virtio-blk-device", "queue-size", "128"},
63     { "virtio-scsi-device", "virtqueue_size", "128"},
64     { "virtio-blk-device", "x-enable-wce-if-config-wce", "off" },
65     { "virtio-blk-device", "seg-max-adjust", "off"},
66     { "virtio-scsi-device", "seg_max_adjust", "off"},
67     { "vhost-blk-device", "seg_max_adjust", "off"},
68     { "usb-host", "suppress-remote-wake", "off" },
69     { "usb-redir", "suppress-remote-wake", "off" },
70     { "qxl", "revision", "4" },
71     { "qxl-vga", "revision", "4" },
72     { "fw_cfg", "acpi-mr-restore", "false" },
73     { "virtio-device", "use-disabled-flag", "false" },
74 };
75 const size_t hw_compat_4_2_len = G_N_ELEMENTS(hw_compat_4_2);
76 
77 GlobalProperty hw_compat_4_1[] = {
78     { "virtio-pci", "x-pcie-flr-init", "off" },
79 };
80 const size_t hw_compat_4_1_len = G_N_ELEMENTS(hw_compat_4_1);
81 
82 GlobalProperty hw_compat_4_0[] = {
83     { "VGA",            "edid", "false" },
84     { "secondary-vga",  "edid", "false" },
85     { "bochs-display",  "edid", "false" },
86     { "virtio-vga",     "edid", "false" },
87     { "virtio-gpu-device", "edid", "false" },
88     { "virtio-device", "use-started", "false" },
89     { "virtio-balloon-device", "qemu-4-0-config-size", "true" },
90     { "pl031", "migrate-tick-offset", "false" },
91 };
92 const size_t hw_compat_4_0_len = G_N_ELEMENTS(hw_compat_4_0);
93 
94 GlobalProperty hw_compat_3_1[] = {
95     { "pcie-root-port", "x-speed", "2_5" },
96     { "pcie-root-port", "x-width", "1" },
97     { "memory-backend-file", "x-use-canonical-path-for-ramblock-id", "true" },
98     { "memory-backend-memfd", "x-use-canonical-path-for-ramblock-id", "true" },
99     { "tpm-crb", "ppi", "false" },
100     { "tpm-tis", "ppi", "false" },
101     { "usb-kbd", "serial", "42" },
102     { "usb-mouse", "serial", "42" },
103     { "usb-tablet", "serial", "42" },
104     { "virtio-blk-device", "discard", "false" },
105     { "virtio-blk-device", "write-zeroes", "false" },
106     { "virtio-balloon-device", "qemu-4-0-config-size", "false" },
107     { "pcie-root-port-base", "disable-acs", "true" }, /* Added in 4.1 */
108 };
109 const size_t hw_compat_3_1_len = G_N_ELEMENTS(hw_compat_3_1);
110 
111 GlobalProperty hw_compat_3_0[] = {};
112 const size_t hw_compat_3_0_len = G_N_ELEMENTS(hw_compat_3_0);
113 
114 GlobalProperty hw_compat_2_12[] = {
115     { "migration", "decompress-error-check", "off" },
116     { "hda-audio", "use-timer", "false" },
117     { "cirrus-vga", "global-vmstate", "true" },
118     { "VGA", "global-vmstate", "true" },
119     { "vmware-svga", "global-vmstate", "true" },
120     { "qxl-vga", "global-vmstate", "true" },
121 };
122 const size_t hw_compat_2_12_len = G_N_ELEMENTS(hw_compat_2_12);
123 
124 GlobalProperty hw_compat_2_11[] = {
125     { "hpet", "hpet-offset-saved", "false" },
126     { "virtio-blk-pci", "vectors", "2" },
127     { "vhost-user-blk-pci", "vectors", "2" },
128     { "e1000", "migrate_tso_props", "off" },
129 };
130 const size_t hw_compat_2_11_len = G_N_ELEMENTS(hw_compat_2_11);
131 
132 GlobalProperty hw_compat_2_10[] = {
133     { "virtio-mouse-device", "wheel-axis", "false" },
134     { "virtio-tablet-device", "wheel-axis", "false" },
135 };
136 const size_t hw_compat_2_10_len = G_N_ELEMENTS(hw_compat_2_10);
137 
138 GlobalProperty hw_compat_2_9[] = {
139     { "pci-bridge", "shpc", "off" },
140     { "intel-iommu", "pt", "off" },
141     { "virtio-net-device", "x-mtu-bypass-backend", "off" },
142     { "pcie-root-port", "x-migrate-msix", "false" },
143 };
144 const size_t hw_compat_2_9_len = G_N_ELEMENTS(hw_compat_2_9);
145 
146 GlobalProperty hw_compat_2_8[] = {
147     { "fw_cfg_mem", "x-file-slots", "0x10" },
148     { "fw_cfg_io", "x-file-slots", "0x10" },
149     { "pflash_cfi01", "old-multiple-chip-handling", "on" },
150     { "pci-bridge", "shpc", "on" },
151     { TYPE_PCI_DEVICE, "x-pcie-extcap-init", "off" },
152     { "virtio-pci", "x-pcie-deverr-init", "off" },
153     { "virtio-pci", "x-pcie-lnkctl-init", "off" },
154     { "virtio-pci", "x-pcie-pm-init", "off" },
155     { "cirrus-vga", "vgamem_mb", "8" },
156     { "isa-cirrus-vga", "vgamem_mb", "8" },
157 };
158 const size_t hw_compat_2_8_len = G_N_ELEMENTS(hw_compat_2_8);
159 
160 GlobalProperty hw_compat_2_7[] = {
161     { "virtio-pci", "page-per-vq", "on" },
162     { "virtio-serial-device", "emergency-write", "off" },
163     { "ioapic", "version", "0x11" },
164     { "intel-iommu", "x-buggy-eim", "true" },
165     { "virtio-pci", "x-ignore-backend-features", "on" },
166 };
167 const size_t hw_compat_2_7_len = G_N_ELEMENTS(hw_compat_2_7);
168 
169 GlobalProperty hw_compat_2_6[] = {
170     { "virtio-mmio", "format_transport_address", "off" },
171     /* Optional because not all virtio-pci devices support legacy mode */
172     { "virtio-pci", "disable-modern", "on",  .optional = true },
173     { "virtio-pci", "disable-legacy", "off", .optional = true },
174 };
175 const size_t hw_compat_2_6_len = G_N_ELEMENTS(hw_compat_2_6);
176 
177 GlobalProperty hw_compat_2_5[] = {
178     { "isa-fdc", "fallback", "144" },
179     { "pvscsi", "x-old-pci-configuration", "on" },
180     { "pvscsi", "x-disable-pcie", "on" },
181     { "vmxnet3", "x-old-msi-offsets", "on" },
182     { "vmxnet3", "x-disable-pcie", "on" },
183 };
184 const size_t hw_compat_2_5_len = G_N_ELEMENTS(hw_compat_2_5);
185 
186 GlobalProperty hw_compat_2_4[] = {
187     /* Optional because the 'scsi' property is Linux-only */
188     { "virtio-blk-device", "scsi", "true", .optional = true },
189     { "e1000", "extra_mac_registers", "off" },
190     { "virtio-pci", "x-disable-pcie", "on" },
191     { "virtio-pci", "migrate-extra", "off" },
192     { "fw_cfg_mem", "dma_enabled", "off" },
193     { "fw_cfg_io", "dma_enabled", "off" }
194 };
195 const size_t hw_compat_2_4_len = G_N_ELEMENTS(hw_compat_2_4);
196 
197 GlobalProperty hw_compat_2_3[] = {
198     { "virtio-blk-pci", "any_layout", "off" },
199     { "virtio-balloon-pci", "any_layout", "off" },
200     { "virtio-serial-pci", "any_layout", "off" },
201     { "virtio-9p-pci", "any_layout", "off" },
202     { "virtio-rng-pci", "any_layout", "off" },
203     { TYPE_PCI_DEVICE, "x-pcie-lnksta-dllla", "off" },
204     { "migration", "send-configuration", "off" },
205     { "migration", "send-section-footer", "off" },
206     { "migration", "store-global-state", "off" },
207 };
208 const size_t hw_compat_2_3_len = G_N_ELEMENTS(hw_compat_2_3);
209 
210 GlobalProperty hw_compat_2_2[] = {};
211 const size_t hw_compat_2_2_len = G_N_ELEMENTS(hw_compat_2_2);
212 
213 GlobalProperty hw_compat_2_1[] = {
214     { "intel-hda", "old_msi_addr", "on" },
215     { "VGA", "qemu-extended-regs", "off" },
216     { "secondary-vga", "qemu-extended-regs", "off" },
217     { "virtio-scsi-pci", "any_layout", "off" },
218     { "usb-mouse", "usb_version", "1" },
219     { "usb-kbd", "usb_version", "1" },
220     { "virtio-pci", "virtio-pci-bus-master-bug-migration", "on" },
221 };
222 const size_t hw_compat_2_1_len = G_N_ELEMENTS(hw_compat_2_1);
223 
224 MachineState *current_machine;
225 
226 static char *machine_get_kernel(Object *obj, Error **errp)
227 {
228     MachineState *ms = MACHINE(obj);
229 
230     return g_strdup(ms->kernel_filename);
231 }
232 
233 static void machine_set_kernel(Object *obj, const char *value, Error **errp)
234 {
235     MachineState *ms = MACHINE(obj);
236 
237     g_free(ms->kernel_filename);
238     ms->kernel_filename = g_strdup(value);
239 }
240 
241 static char *machine_get_initrd(Object *obj, Error **errp)
242 {
243     MachineState *ms = MACHINE(obj);
244 
245     return g_strdup(ms->initrd_filename);
246 }
247 
248 static void machine_set_initrd(Object *obj, const char *value, Error **errp)
249 {
250     MachineState *ms = MACHINE(obj);
251 
252     g_free(ms->initrd_filename);
253     ms->initrd_filename = g_strdup(value);
254 }
255 
256 static char *machine_get_append(Object *obj, Error **errp)
257 {
258     MachineState *ms = MACHINE(obj);
259 
260     return g_strdup(ms->kernel_cmdline);
261 }
262 
263 static void machine_set_append(Object *obj, const char *value, Error **errp)
264 {
265     MachineState *ms = MACHINE(obj);
266 
267     g_free(ms->kernel_cmdline);
268     ms->kernel_cmdline = g_strdup(value);
269 }
270 
271 static char *machine_get_dtb(Object *obj, Error **errp)
272 {
273     MachineState *ms = MACHINE(obj);
274 
275     return g_strdup(ms->dtb);
276 }
277 
278 static void machine_set_dtb(Object *obj, const char *value, Error **errp)
279 {
280     MachineState *ms = MACHINE(obj);
281 
282     g_free(ms->dtb);
283     ms->dtb = g_strdup(value);
284 }
285 
286 static char *machine_get_dumpdtb(Object *obj, Error **errp)
287 {
288     MachineState *ms = MACHINE(obj);
289 
290     return g_strdup(ms->dumpdtb);
291 }
292 
293 static void machine_set_dumpdtb(Object *obj, const char *value, Error **errp)
294 {
295     MachineState *ms = MACHINE(obj);
296 
297     g_free(ms->dumpdtb);
298     ms->dumpdtb = g_strdup(value);
299 }
300 
301 static void machine_get_phandle_start(Object *obj, Visitor *v,
302                                       const char *name, void *opaque,
303                                       Error **errp)
304 {
305     MachineState *ms = MACHINE(obj);
306     int64_t value = ms->phandle_start;
307 
308     visit_type_int(v, name, &value, errp);
309 }
310 
311 static void machine_set_phandle_start(Object *obj, Visitor *v,
312                                       const char *name, void *opaque,
313                                       Error **errp)
314 {
315     MachineState *ms = MACHINE(obj);
316     int64_t value;
317 
318     if (!visit_type_int(v, name, &value, errp)) {
319         return;
320     }
321 
322     ms->phandle_start = value;
323 }
324 
325 static char *machine_get_dt_compatible(Object *obj, Error **errp)
326 {
327     MachineState *ms = MACHINE(obj);
328 
329     return g_strdup(ms->dt_compatible);
330 }
331 
332 static void machine_set_dt_compatible(Object *obj, const char *value, Error **errp)
333 {
334     MachineState *ms = MACHINE(obj);
335 
336     g_free(ms->dt_compatible);
337     ms->dt_compatible = g_strdup(value);
338 }
339 
340 static bool machine_get_dump_guest_core(Object *obj, Error **errp)
341 {
342     MachineState *ms = MACHINE(obj);
343 
344     return ms->dump_guest_core;
345 }
346 
347 static void machine_set_dump_guest_core(Object *obj, bool value, Error **errp)
348 {
349     MachineState *ms = MACHINE(obj);
350 
351     ms->dump_guest_core = value;
352 }
353 
354 static bool machine_get_mem_merge(Object *obj, Error **errp)
355 {
356     MachineState *ms = MACHINE(obj);
357 
358     return ms->mem_merge;
359 }
360 
361 static void machine_set_mem_merge(Object *obj, bool value, Error **errp)
362 {
363     MachineState *ms = MACHINE(obj);
364 
365     ms->mem_merge = value;
366 }
367 
368 static bool machine_get_usb(Object *obj, Error **errp)
369 {
370     MachineState *ms = MACHINE(obj);
371 
372     return ms->usb;
373 }
374 
375 static void machine_set_usb(Object *obj, bool value, Error **errp)
376 {
377     MachineState *ms = MACHINE(obj);
378 
379     ms->usb = value;
380     ms->usb_disabled = !value;
381 }
382 
383 static bool machine_get_graphics(Object *obj, Error **errp)
384 {
385     MachineState *ms = MACHINE(obj);
386 
387     return ms->enable_graphics;
388 }
389 
390 static void machine_set_graphics(Object *obj, bool value, Error **errp)
391 {
392     MachineState *ms = MACHINE(obj);
393 
394     ms->enable_graphics = value;
395 }
396 
397 static char *machine_get_firmware(Object *obj, Error **errp)
398 {
399     MachineState *ms = MACHINE(obj);
400 
401     return g_strdup(ms->firmware);
402 }
403 
404 static void machine_set_firmware(Object *obj, const char *value, Error **errp)
405 {
406     MachineState *ms = MACHINE(obj);
407 
408     g_free(ms->firmware);
409     ms->firmware = g_strdup(value);
410 }
411 
412 static void machine_set_suppress_vmdesc(Object *obj, bool value, Error **errp)
413 {
414     MachineState *ms = MACHINE(obj);
415 
416     ms->suppress_vmdesc = value;
417 }
418 
419 static bool machine_get_suppress_vmdesc(Object *obj, Error **errp)
420 {
421     MachineState *ms = MACHINE(obj);
422 
423     return ms->suppress_vmdesc;
424 }
425 
426 static char *machine_get_memory_encryption(Object *obj, Error **errp)
427 {
428     MachineState *ms = MACHINE(obj);
429 
430     return g_strdup(ms->memory_encryption);
431 }
432 
433 static void machine_set_memory_encryption(Object *obj, const char *value,
434                                         Error **errp)
435 {
436     MachineState *ms = MACHINE(obj);
437 
438     g_free(ms->memory_encryption);
439     ms->memory_encryption = g_strdup(value);
440 }
441 
442 static bool machine_get_nvdimm(Object *obj, Error **errp)
443 {
444     MachineState *ms = MACHINE(obj);
445 
446     return ms->nvdimms_state->is_enabled;
447 }
448 
449 static void machine_set_nvdimm(Object *obj, bool value, Error **errp)
450 {
451     MachineState *ms = MACHINE(obj);
452 
453     ms->nvdimms_state->is_enabled = value;
454 }
455 
456 static bool machine_get_hmat(Object *obj, Error **errp)
457 {
458     MachineState *ms = MACHINE(obj);
459 
460     return ms->numa_state->hmat_enabled;
461 }
462 
463 static void machine_set_hmat(Object *obj, bool value, Error **errp)
464 {
465     MachineState *ms = MACHINE(obj);
466 
467     ms->numa_state->hmat_enabled = value;
468 }
469 
470 static char *machine_get_nvdimm_persistence(Object *obj, Error **errp)
471 {
472     MachineState *ms = MACHINE(obj);
473 
474     return g_strdup(ms->nvdimms_state->persistence_string);
475 }
476 
477 static void machine_set_nvdimm_persistence(Object *obj, const char *value,
478                                            Error **errp)
479 {
480     MachineState *ms = MACHINE(obj);
481     NVDIMMState *nvdimms_state = ms->nvdimms_state;
482 
483     if (strcmp(value, "cpu") == 0) {
484         nvdimms_state->persistence = 3;
485     } else if (strcmp(value, "mem-ctrl") == 0) {
486         nvdimms_state->persistence = 2;
487     } else {
488         error_setg(errp, "-machine nvdimm-persistence=%s: unsupported option",
489                    value);
490         return;
491     }
492 
493     g_free(nvdimms_state->persistence_string);
494     nvdimms_state->persistence_string = g_strdup(value);
495 }
496 
497 void machine_class_allow_dynamic_sysbus_dev(MachineClass *mc, const char *type)
498 {
499     QAPI_LIST_PREPEND(mc->allowed_dynamic_sysbus_devices, g_strdup(type));
500 }
501 
502 static void validate_sysbus_device(SysBusDevice *sbdev, void *opaque)
503 {
504     MachineState *machine = opaque;
505     MachineClass *mc = MACHINE_GET_CLASS(machine);
506     bool allowed = false;
507     strList *wl;
508 
509     for (wl = mc->allowed_dynamic_sysbus_devices;
510          !allowed && wl;
511          wl = wl->next) {
512         allowed |= !!object_dynamic_cast(OBJECT(sbdev), wl->value);
513     }
514 
515     if (!allowed) {
516         error_report("Option '-device %s' cannot be handled by this machine",
517                      object_class_get_name(object_get_class(OBJECT(sbdev))));
518         exit(1);
519     }
520 }
521 
522 static char *machine_get_memdev(Object *obj, Error **errp)
523 {
524     MachineState *ms = MACHINE(obj);
525 
526     return g_strdup(ms->ram_memdev_id);
527 }
528 
529 static void machine_set_memdev(Object *obj, const char *value, Error **errp)
530 {
531     MachineState *ms = MACHINE(obj);
532 
533     g_free(ms->ram_memdev_id);
534     ms->ram_memdev_id = g_strdup(value);
535 }
536 
537 
538 static void machine_init_notify(Notifier *notifier, void *data)
539 {
540     MachineState *machine = MACHINE(qdev_get_machine());
541 
542     /*
543      * Loop through all dynamically created sysbus devices and check if they are
544      * all allowed.  If a device is not allowed, error out.
545      */
546     foreach_dynamic_sysbus_device(validate_sysbus_device, machine);
547 }
548 
549 HotpluggableCPUList *machine_query_hotpluggable_cpus(MachineState *machine)
550 {
551     int i;
552     HotpluggableCPUList *head = NULL;
553     MachineClass *mc = MACHINE_GET_CLASS(machine);
554 
555     /* force board to initialize possible_cpus if it hasn't been done yet */
556     mc->possible_cpu_arch_ids(machine);
557 
558     for (i = 0; i < machine->possible_cpus->len; i++) {
559         Object *cpu;
560         HotpluggableCPU *cpu_item = g_new0(typeof(*cpu_item), 1);
561 
562         cpu_item->type = g_strdup(machine->possible_cpus->cpus[i].type);
563         cpu_item->vcpus_count = machine->possible_cpus->cpus[i].vcpus_count;
564         cpu_item->props = g_memdup(&machine->possible_cpus->cpus[i].props,
565                                    sizeof(*cpu_item->props));
566 
567         cpu = machine->possible_cpus->cpus[i].cpu;
568         if (cpu) {
569             cpu_item->has_qom_path = true;
570             cpu_item->qom_path = object_get_canonical_path(cpu);
571         }
572         QAPI_LIST_PREPEND(head, cpu_item);
573     }
574     return head;
575 }
576 
577 /**
578  * machine_set_cpu_numa_node:
579  * @machine: machine object to modify
580  * @props: specifies which cpu objects to assign to
581  *         numa node specified by @props.node_id
582  * @errp: if an error occurs, a pointer to an area to store the error
583  *
584  * Associate NUMA node specified by @props.node_id with cpu slots that
585  * match socket/core/thread-ids specified by @props. It's recommended to use
586  * query-hotpluggable-cpus.props values to specify affected cpu slots,
587  * which would lead to exact 1:1 mapping of cpu slots to NUMA node.
588  *
589  * However for CLI convenience it's possible to pass in subset of properties,
590  * which would affect all cpu slots that match it.
591  * Ex for pc machine:
592  *    -smp 4,cores=2,sockets=2 -numa node,nodeid=0 -numa node,nodeid=1 \
593  *    -numa cpu,node-id=0,socket_id=0 \
594  *    -numa cpu,node-id=1,socket_id=1
595  * will assign all child cores of socket 0 to node 0 and
596  * of socket 1 to node 1.
597  *
598  * On attempt of reassigning (already assigned) cpu slot to another NUMA node,
599  * return error.
600  * Empty subset is disallowed and function will return with error in this case.
601  */
602 void machine_set_cpu_numa_node(MachineState *machine,
603                                const CpuInstanceProperties *props, Error **errp)
604 {
605     MachineClass *mc = MACHINE_GET_CLASS(machine);
606     NodeInfo *numa_info = machine->numa_state->nodes;
607     bool match = false;
608     int i;
609 
610     if (!mc->possible_cpu_arch_ids) {
611         error_setg(errp, "mapping of CPUs to NUMA node is not supported");
612         return;
613     }
614 
615     /* disabling node mapping is not supported, forbid it */
616     assert(props->has_node_id);
617 
618     /* force board to initialize possible_cpus if it hasn't been done yet */
619     mc->possible_cpu_arch_ids(machine);
620 
621     for (i = 0; i < machine->possible_cpus->len; i++) {
622         CPUArchId *slot = &machine->possible_cpus->cpus[i];
623 
624         /* reject unsupported by board properties */
625         if (props->has_thread_id && !slot->props.has_thread_id) {
626             error_setg(errp, "thread-id is not supported");
627             return;
628         }
629 
630         if (props->has_core_id && !slot->props.has_core_id) {
631             error_setg(errp, "core-id is not supported");
632             return;
633         }
634 
635         if (props->has_socket_id && !slot->props.has_socket_id) {
636             error_setg(errp, "socket-id is not supported");
637             return;
638         }
639 
640         if (props->has_die_id && !slot->props.has_die_id) {
641             error_setg(errp, "die-id is not supported");
642             return;
643         }
644 
645         /* skip slots with explicit mismatch */
646         if (props->has_thread_id && props->thread_id != slot->props.thread_id) {
647                 continue;
648         }
649 
650         if (props->has_core_id && props->core_id != slot->props.core_id) {
651                 continue;
652         }
653 
654         if (props->has_die_id && props->die_id != slot->props.die_id) {
655                 continue;
656         }
657 
658         if (props->has_socket_id && props->socket_id != slot->props.socket_id) {
659                 continue;
660         }
661 
662         /* reject assignment if slot is already assigned, for compatibility
663          * of legacy cpu_index mapping with SPAPR core based mapping do not
664          * error out if cpu thread and matched core have the same node-id */
665         if (slot->props.has_node_id &&
666             slot->props.node_id != props->node_id) {
667             error_setg(errp, "CPU is already assigned to node-id: %" PRId64,
668                        slot->props.node_id);
669             return;
670         }
671 
672         /* assign slot to node as it's matched '-numa cpu' key */
673         match = true;
674         slot->props.node_id = props->node_id;
675         slot->props.has_node_id = props->has_node_id;
676 
677         if (machine->numa_state->hmat_enabled) {
678             if ((numa_info[props->node_id].initiator < MAX_NODES) &&
679                 (props->node_id != numa_info[props->node_id].initiator)) {
680                 error_setg(errp, "The initiator of CPU NUMA node %" PRId64
681                         " should be itself", props->node_id);
682                 return;
683             }
684             numa_info[props->node_id].has_cpu = true;
685             numa_info[props->node_id].initiator = props->node_id;
686         }
687     }
688 
689     if (!match) {
690         error_setg(errp, "no match found");
691     }
692 }
693 
694 static void smp_parse(MachineState *ms, QemuOpts *opts)
695 {
696     if (opts) {
697         unsigned cpus    = qemu_opt_get_number(opts, "cpus", 0);
698         unsigned sockets = qemu_opt_get_number(opts, "sockets", 0);
699         unsigned cores   = qemu_opt_get_number(opts, "cores", 0);
700         unsigned threads = qemu_opt_get_number(opts, "threads", 0);
701 
702         /* compute missing values, prefer sockets over cores over threads */
703         if (cpus == 0 || sockets == 0) {
704             cores = cores > 0 ? cores : 1;
705             threads = threads > 0 ? threads : 1;
706             if (cpus == 0) {
707                 sockets = sockets > 0 ? sockets : 1;
708                 cpus = cores * threads * sockets;
709             } else {
710                 ms->smp.max_cpus =
711                         qemu_opt_get_number(opts, "maxcpus", cpus);
712                 sockets = ms->smp.max_cpus / (cores * threads);
713             }
714         } else if (cores == 0) {
715             threads = threads > 0 ? threads : 1;
716             cores = cpus / (sockets * threads);
717             cores = cores > 0 ? cores : 1;
718         } else if (threads == 0) {
719             threads = cpus / (cores * sockets);
720             threads = threads > 0 ? threads : 1;
721         } else if (sockets * cores * threads < cpus) {
722             error_report("cpu topology: "
723                          "sockets (%u) * cores (%u) * threads (%u) < "
724                          "smp_cpus (%u)",
725                          sockets, cores, threads, cpus);
726             exit(1);
727         }
728 
729         ms->smp.max_cpus =
730                 qemu_opt_get_number(opts, "maxcpus", cpus);
731 
732         if (ms->smp.max_cpus < cpus) {
733             error_report("maxcpus must be equal to or greater than smp");
734             exit(1);
735         }
736 
737         if (sockets * cores * threads != ms->smp.max_cpus) {
738             error_report("Invalid CPU topology: "
739                          "sockets (%u) * cores (%u) * threads (%u) "
740                          "!= maxcpus (%u)",
741                          sockets, cores, threads,
742                          ms->smp.max_cpus);
743             exit(1);
744         }
745 
746         ms->smp.cpus = cpus;
747         ms->smp.cores = cores;
748         ms->smp.threads = threads;
749         ms->smp.sockets = sockets;
750     }
751 
752     if (ms->smp.cpus > 1) {
753         Error *blocker = NULL;
754         error_setg(&blocker, QERR_REPLAY_NOT_SUPPORTED, "smp");
755         replay_add_blocker(blocker);
756     }
757 }
758 
759 static void machine_class_init(ObjectClass *oc, void *data)
760 {
761     MachineClass *mc = MACHINE_CLASS(oc);
762 
763     /* Default 128 MB as guest ram size */
764     mc->default_ram_size = 128 * MiB;
765     mc->rom_file_has_mr = true;
766     mc->smp_parse = smp_parse;
767 
768     /* numa node memory size aligned on 8MB by default.
769      * On Linux, each node's border has to be 8MB aligned
770      */
771     mc->numa_mem_align_shift = 23;
772 
773     object_class_property_add_str(oc, "kernel",
774         machine_get_kernel, machine_set_kernel);
775     object_class_property_set_description(oc, "kernel",
776         "Linux kernel image file");
777 
778     object_class_property_add_str(oc, "initrd",
779         machine_get_initrd, machine_set_initrd);
780     object_class_property_set_description(oc, "initrd",
781         "Linux initial ramdisk file");
782 
783     object_class_property_add_str(oc, "append",
784         machine_get_append, machine_set_append);
785     object_class_property_set_description(oc, "append",
786         "Linux kernel command line");
787 
788     object_class_property_add_str(oc, "dtb",
789         machine_get_dtb, machine_set_dtb);
790     object_class_property_set_description(oc, "dtb",
791         "Linux kernel device tree file");
792 
793     object_class_property_add_str(oc, "dumpdtb",
794         machine_get_dumpdtb, machine_set_dumpdtb);
795     object_class_property_set_description(oc, "dumpdtb",
796         "Dump current dtb to a file and quit");
797 
798     object_class_property_add(oc, "phandle-start", "int",
799         machine_get_phandle_start, machine_set_phandle_start,
800         NULL, NULL);
801     object_class_property_set_description(oc, "phandle-start",
802         "The first phandle ID we may generate dynamically");
803 
804     object_class_property_add_str(oc, "dt-compatible",
805         machine_get_dt_compatible, machine_set_dt_compatible);
806     object_class_property_set_description(oc, "dt-compatible",
807         "Overrides the \"compatible\" property of the dt root node");
808 
809     object_class_property_add_bool(oc, "dump-guest-core",
810         machine_get_dump_guest_core, machine_set_dump_guest_core);
811     object_class_property_set_description(oc, "dump-guest-core",
812         "Include guest memory in a core dump");
813 
814     object_class_property_add_bool(oc, "mem-merge",
815         machine_get_mem_merge, machine_set_mem_merge);
816     object_class_property_set_description(oc, "mem-merge",
817         "Enable/disable memory merge support");
818 
819     object_class_property_add_bool(oc, "usb",
820         machine_get_usb, machine_set_usb);
821     object_class_property_set_description(oc, "usb",
822         "Set on/off to enable/disable usb");
823 
824     object_class_property_add_bool(oc, "graphics",
825         machine_get_graphics, machine_set_graphics);
826     object_class_property_set_description(oc, "graphics",
827         "Set on/off to enable/disable graphics emulation");
828 
829     object_class_property_add_str(oc, "firmware",
830         machine_get_firmware, machine_set_firmware);
831     object_class_property_set_description(oc, "firmware",
832         "Firmware image");
833 
834     object_class_property_add_bool(oc, "suppress-vmdesc",
835         machine_get_suppress_vmdesc, machine_set_suppress_vmdesc);
836     object_class_property_set_description(oc, "suppress-vmdesc",
837         "Set on to disable self-describing migration");
838 
839     object_class_property_add_str(oc, "memory-encryption",
840         machine_get_memory_encryption, machine_set_memory_encryption);
841     object_class_property_set_description(oc, "memory-encryption",
842         "Set memory encryption object to use");
843 
844     object_class_property_add_str(oc, "memory-backend",
845                                   machine_get_memdev, machine_set_memdev);
846     object_class_property_set_description(oc, "memory-backend",
847                                           "Set RAM backend"
848                                           "Valid value is ID of hostmem based backend");
849 }
850 
851 static void machine_class_base_init(ObjectClass *oc, void *data)
852 {
853     MachineClass *mc = MACHINE_CLASS(oc);
854     mc->max_cpus = mc->max_cpus ?: 1;
855     mc->min_cpus = mc->min_cpus ?: 1;
856     mc->default_cpus = mc->default_cpus ?: 1;
857 
858     if (!object_class_is_abstract(oc)) {
859         const char *cname = object_class_get_name(oc);
860         assert(g_str_has_suffix(cname, TYPE_MACHINE_SUFFIX));
861         mc->name = g_strndup(cname,
862                             strlen(cname) - strlen(TYPE_MACHINE_SUFFIX));
863         mc->compat_props = g_ptr_array_new();
864     }
865 }
866 
867 static void machine_initfn(Object *obj)
868 {
869     MachineState *ms = MACHINE(obj);
870     MachineClass *mc = MACHINE_GET_CLASS(obj);
871 
872     container_get(obj, "/peripheral");
873     container_get(obj, "/peripheral-anon");
874 
875     ms->dump_guest_core = true;
876     ms->mem_merge = true;
877     ms->enable_graphics = true;
878     ms->kernel_cmdline = g_strdup("");
879 
880     if (mc->nvdimm_supported) {
881         Object *obj = OBJECT(ms);
882 
883         ms->nvdimms_state = g_new0(NVDIMMState, 1);
884         object_property_add_bool(obj, "nvdimm",
885                                  machine_get_nvdimm, machine_set_nvdimm);
886         object_property_set_description(obj, "nvdimm",
887                                         "Set on/off to enable/disable "
888                                         "NVDIMM instantiation");
889 
890         object_property_add_str(obj, "nvdimm-persistence",
891                                 machine_get_nvdimm_persistence,
892                                 machine_set_nvdimm_persistence);
893         object_property_set_description(obj, "nvdimm-persistence",
894                                         "Set NVDIMM persistence"
895                                         "Valid values are cpu, mem-ctrl");
896     }
897 
898     if (mc->cpu_index_to_instance_props && mc->get_default_cpu_node_id) {
899         ms->numa_state = g_new0(NumaState, 1);
900         object_property_add_bool(obj, "hmat",
901                                  machine_get_hmat, machine_set_hmat);
902         object_property_set_description(obj, "hmat",
903                                         "Set on/off to enable/disable "
904                                         "ACPI Heterogeneous Memory Attribute "
905                                         "Table (HMAT)");
906     }
907 
908     /* Register notifier when init is done for sysbus sanity checks */
909     ms->sysbus_notifier.notify = machine_init_notify;
910     qemu_add_machine_init_done_notifier(&ms->sysbus_notifier);
911 
912     /* default to mc->default_cpus */
913     ms->smp.cpus = mc->default_cpus;
914     ms->smp.max_cpus = mc->default_cpus;
915     ms->smp.cores = 1;
916     ms->smp.threads = 1;
917     ms->smp.sockets = 1;
918 }
919 
920 static void machine_finalize(Object *obj)
921 {
922     MachineState *ms = MACHINE(obj);
923 
924     g_free(ms->kernel_filename);
925     g_free(ms->initrd_filename);
926     g_free(ms->kernel_cmdline);
927     g_free(ms->dtb);
928     g_free(ms->dumpdtb);
929     g_free(ms->dt_compatible);
930     g_free(ms->firmware);
931     g_free(ms->device_memory);
932     g_free(ms->nvdimms_state);
933     g_free(ms->numa_state);
934 }
935 
936 bool machine_usb(MachineState *machine)
937 {
938     return machine->usb;
939 }
940 
941 int machine_phandle_start(MachineState *machine)
942 {
943     return machine->phandle_start;
944 }
945 
946 bool machine_dump_guest_core(MachineState *machine)
947 {
948     return machine->dump_guest_core;
949 }
950 
951 bool machine_mem_merge(MachineState *machine)
952 {
953     return machine->mem_merge;
954 }
955 
956 static char *cpu_slot_to_string(const CPUArchId *cpu)
957 {
958     GString *s = g_string_new(NULL);
959     if (cpu->props.has_socket_id) {
960         g_string_append_printf(s, "socket-id: %"PRId64, cpu->props.socket_id);
961     }
962     if (cpu->props.has_die_id) {
963         g_string_append_printf(s, "die-id: %"PRId64, cpu->props.die_id);
964     }
965     if (cpu->props.has_core_id) {
966         if (s->len) {
967             g_string_append_printf(s, ", ");
968         }
969         g_string_append_printf(s, "core-id: %"PRId64, cpu->props.core_id);
970     }
971     if (cpu->props.has_thread_id) {
972         if (s->len) {
973             g_string_append_printf(s, ", ");
974         }
975         g_string_append_printf(s, "thread-id: %"PRId64, cpu->props.thread_id);
976     }
977     return g_string_free(s, false);
978 }
979 
980 static void numa_validate_initiator(NumaState *numa_state)
981 {
982     int i;
983     NodeInfo *numa_info = numa_state->nodes;
984 
985     for (i = 0; i < numa_state->num_nodes; i++) {
986         if (numa_info[i].initiator == MAX_NODES) {
987             error_report("The initiator of NUMA node %d is missing, use "
988                          "'-numa node,initiator' option to declare it", i);
989             exit(1);
990         }
991 
992         if (!numa_info[numa_info[i].initiator].present) {
993             error_report("NUMA node %" PRIu16 " is missing, use "
994                          "'-numa node' option to declare it first",
995                          numa_info[i].initiator);
996             exit(1);
997         }
998 
999         if (!numa_info[numa_info[i].initiator].has_cpu) {
1000             error_report("The initiator of NUMA node %d is invalid", i);
1001             exit(1);
1002         }
1003     }
1004 }
1005 
1006 static void machine_numa_finish_cpu_init(MachineState *machine)
1007 {
1008     int i;
1009     bool default_mapping;
1010     GString *s = g_string_new(NULL);
1011     MachineClass *mc = MACHINE_GET_CLASS(machine);
1012     const CPUArchIdList *possible_cpus = mc->possible_cpu_arch_ids(machine);
1013 
1014     assert(machine->numa_state->num_nodes);
1015     for (i = 0; i < possible_cpus->len; i++) {
1016         if (possible_cpus->cpus[i].props.has_node_id) {
1017             break;
1018         }
1019     }
1020     default_mapping = (i == possible_cpus->len);
1021 
1022     for (i = 0; i < possible_cpus->len; i++) {
1023         const CPUArchId *cpu_slot = &possible_cpus->cpus[i];
1024 
1025         if (!cpu_slot->props.has_node_id) {
1026             /* fetch default mapping from board and enable it */
1027             CpuInstanceProperties props = cpu_slot->props;
1028 
1029             props.node_id = mc->get_default_cpu_node_id(machine, i);
1030             if (!default_mapping) {
1031                 /* record slots with not set mapping,
1032                  * TODO: make it hard error in future */
1033                 char *cpu_str = cpu_slot_to_string(cpu_slot);
1034                 g_string_append_printf(s, "%sCPU %d [%s]",
1035                                        s->len ? ", " : "", i, cpu_str);
1036                 g_free(cpu_str);
1037 
1038                 /* non mapped cpus used to fallback to node 0 */
1039                 props.node_id = 0;
1040             }
1041 
1042             props.has_node_id = true;
1043             machine_set_cpu_numa_node(machine, &props, &error_fatal);
1044         }
1045     }
1046 
1047     if (machine->numa_state->hmat_enabled) {
1048         numa_validate_initiator(machine->numa_state);
1049     }
1050 
1051     if (s->len && !qtest_enabled()) {
1052         warn_report("CPU(s) not present in any NUMA nodes: %s",
1053                     s->str);
1054         warn_report("All CPU(s) up to maxcpus should be described "
1055                     "in NUMA config, ability to start up with partial NUMA "
1056                     "mappings is obsoleted and will be removed in future");
1057     }
1058     g_string_free(s, true);
1059 }
1060 
1061 MemoryRegion *machine_consume_memdev(MachineState *machine,
1062                                      HostMemoryBackend *backend)
1063 {
1064     MemoryRegion *ret = host_memory_backend_get_memory(backend);
1065 
1066     if (memory_region_is_mapped(ret)) {
1067         error_report("memory backend %s can't be used multiple times.",
1068                      object_get_canonical_path_component(OBJECT(backend)));
1069         exit(EXIT_FAILURE);
1070     }
1071     host_memory_backend_set_mapped(backend, true);
1072     vmstate_register_ram_global(ret);
1073     return ret;
1074 }
1075 
1076 bool machine_smp_parse(MachineState *ms, QemuOpts *opts, Error **errp)
1077 {
1078     MachineClass *mc = MACHINE_GET_CLASS(ms);
1079 
1080     mc->smp_parse(ms, opts);
1081 
1082     /* sanity-check smp_cpus and max_cpus against mc */
1083     if (ms->smp.cpus < mc->min_cpus) {
1084         error_setg(errp, "Invalid SMP CPUs %d. The min CPUs "
1085                    "supported by machine '%s' is %d",
1086                    ms->smp.cpus,
1087                    mc->name, mc->min_cpus);
1088         return false;
1089     } else if (ms->smp.max_cpus > mc->max_cpus) {
1090         error_setg(errp, "Invalid SMP CPUs %d. The max CPUs "
1091                    "supported by machine '%s' is %d",
1092                    current_machine->smp.max_cpus,
1093                    mc->name, mc->max_cpus);
1094         return false;
1095     }
1096     return true;
1097 }
1098 
1099 void machine_run_board_init(MachineState *machine)
1100 {
1101     MachineClass *machine_class = MACHINE_GET_CLASS(machine);
1102     ObjectClass *oc = object_class_by_name(machine->cpu_type);
1103     CPUClass *cc;
1104 
1105     /* This checkpoint is required by replay to separate prior clock
1106        reading from the other reads, because timer polling functions query
1107        clock values from the log. */
1108     replay_checkpoint(CHECKPOINT_INIT);
1109 
1110     if (machine->ram_memdev_id) {
1111         Object *o;
1112         o = object_resolve_path_type(machine->ram_memdev_id,
1113                                      TYPE_MEMORY_BACKEND, NULL);
1114         machine->ram = machine_consume_memdev(machine, MEMORY_BACKEND(o));
1115     }
1116 
1117     if (machine->numa_state) {
1118         numa_complete_configuration(machine);
1119         if (machine->numa_state->num_nodes) {
1120             machine_numa_finish_cpu_init(machine);
1121         }
1122     }
1123 
1124     /* If the machine supports the valid_cpu_types check and the user
1125      * specified a CPU with -cpu check here that the user CPU is supported.
1126      */
1127     if (machine_class->valid_cpu_types && machine->cpu_type) {
1128         int i;
1129 
1130         for (i = 0; machine_class->valid_cpu_types[i]; i++) {
1131             if (object_class_dynamic_cast(oc,
1132                                           machine_class->valid_cpu_types[i])) {
1133                 /* The user specificed CPU is in the valid field, we are
1134                  * good to go.
1135                  */
1136                 break;
1137             }
1138         }
1139 
1140         if (!machine_class->valid_cpu_types[i]) {
1141             /* The user specified CPU is not valid */
1142             error_report("Invalid CPU type: %s", machine->cpu_type);
1143             error_printf("The valid types are: %s",
1144                          machine_class->valid_cpu_types[0]);
1145             for (i = 1; machine_class->valid_cpu_types[i]; i++) {
1146                 error_printf(", %s", machine_class->valid_cpu_types[i]);
1147             }
1148             error_printf("\n");
1149 
1150             exit(1);
1151         }
1152     }
1153 
1154     /* Check if CPU type is deprecated and warn if so */
1155     cc = CPU_CLASS(oc);
1156     if (cc && cc->deprecation_note) {
1157         warn_report("CPU model %s is deprecated -- %s", machine->cpu_type,
1158                     cc->deprecation_note);
1159     }
1160 
1161     if (machine->memory_encryption) {
1162         /*
1163          * With memory encryption, the host can't see the real
1164          * contents of RAM, so there's no point in it trying to merge
1165          * areas.
1166          */
1167         machine_set_mem_merge(OBJECT(machine), false, &error_abort);
1168     }
1169 
1170     machine_class->init(machine);
1171     phase_advance(PHASE_MACHINE_INITIALIZED);
1172 }
1173 
1174 static NotifierList machine_init_done_notifiers =
1175     NOTIFIER_LIST_INITIALIZER(machine_init_done_notifiers);
1176 
1177 void qemu_add_machine_init_done_notifier(Notifier *notify)
1178 {
1179     notifier_list_add(&machine_init_done_notifiers, notify);
1180     if (phase_check(PHASE_MACHINE_READY)) {
1181         notify->notify(notify, NULL);
1182     }
1183 }
1184 
1185 void qemu_remove_machine_init_done_notifier(Notifier *notify)
1186 {
1187     notifier_remove(notify);
1188 }
1189 
1190 void qdev_machine_creation_done(void)
1191 {
1192     cpu_synchronize_all_post_init();
1193 
1194     if (current_machine->boot_once) {
1195         qemu_boot_set(current_machine->boot_once, &error_fatal);
1196         qemu_register_reset(restore_boot_order, g_strdup(current_machine->boot_order));
1197     }
1198 
1199     /*
1200      * ok, initial machine setup is done, starting from now we can
1201      * only create hotpluggable devices
1202      */
1203     phase_advance(PHASE_MACHINE_READY);
1204     qdev_assert_realized_properly();
1205 
1206     /* TODO: once all bus devices are qdevified, this should be done
1207      * when bus is created by qdev.c */
1208     /*
1209      * TODO: If we had a main 'reset container' that the whole system
1210      * lived in, we could reset that using the multi-phase reset
1211      * APIs. For the moment, we just reset the sysbus, which will cause
1212      * all devices hanging off it (and all their child buses, recursively)
1213      * to be reset. Note that this will *not* reset any Device objects
1214      * which are not attached to some part of the qbus tree!
1215      */
1216     qemu_register_reset(resettable_cold_reset_fn, sysbus_get_default());
1217 
1218     notifier_list_notify(&machine_init_done_notifiers, NULL);
1219 
1220     if (rom_check_and_register_reset() != 0) {
1221         exit(1);
1222     }
1223 
1224     replay_start();
1225 
1226     /* This checkpoint is required by replay to separate prior clock
1227        reading from the other reads, because timer polling functions query
1228        clock values from the log. */
1229     replay_checkpoint(CHECKPOINT_RESET);
1230     qemu_system_reset(SHUTDOWN_CAUSE_NONE);
1231     register_global_state();
1232 }
1233 
1234 static const TypeInfo machine_info = {
1235     .name = TYPE_MACHINE,
1236     .parent = TYPE_OBJECT,
1237     .abstract = true,
1238     .class_size = sizeof(MachineClass),
1239     .class_init    = machine_class_init,
1240     .class_base_init = machine_class_base_init,
1241     .instance_size = sizeof(MachineState),
1242     .instance_init = machine_initfn,
1243     .instance_finalize = machine_finalize,
1244 };
1245 
1246 static void machine_register_types(void)
1247 {
1248     type_register_static(&machine_info);
1249 }
1250 
1251 type_init(machine_register_types)
1252