xref: /openbmc/qemu/hw/i386/x86.c (revision 5242ef88)
1 /*
2  * Copyright (c) 2003-2004 Fabrice Bellard
3  * Copyright (c) 2019 Red Hat, Inc.
4  *
5  * Permission is hereby granted, free of charge, to any person obtaining a copy
6  * of this software and associated documentation files (the "Software"), to deal
7  * in the Software without restriction, including without limitation the rights
8  * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
9  * copies of the Software, and to permit persons to whom the Software is
10  * furnished to do so, subject to the following conditions:
11  *
12  * The above copyright notice and this permission notice shall be included in
13  * all copies or substantial portions of the Software.
14  *
15  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
16  * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
17  * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
18  * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
19  * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
20  * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
21  * THE SOFTWARE.
22  */
23 #include "qemu/osdep.h"
24 #include "qemu/error-report.h"
25 #include "qemu/option.h"
26 #include "qemu/cutils.h"
27 #include "qemu/units.h"
28 #include "qemu-common.h"
29 #include "qemu/datadir.h"
30 #include "qapi/error.h"
31 #include "qapi/qmp/qerror.h"
32 #include "qapi/qapi-visit-common.h"
33 #include "qapi/clone-visitor.h"
34 #include "qapi/qapi-visit-machine.h"
35 #include "qapi/visitor.h"
36 #include "sysemu/qtest.h"
37 #include "sysemu/whpx.h"
38 #include "sysemu/numa.h"
39 #include "sysemu/replay.h"
40 #include "sysemu/sysemu.h"
41 #include "sysemu/cpu-timers.h"
42 #include "trace.h"
43 
44 #include "hw/i386/x86.h"
45 #include "target/i386/cpu.h"
46 #include "hw/i386/topology.h"
47 #include "hw/i386/fw_cfg.h"
48 #include "hw/intc/i8259.h"
49 #include "hw/rtc/mc146818rtc.h"
50 #include "target/i386/sev.h"
51 
52 #include "hw/acpi/cpu_hotplug.h"
53 #include "hw/irq.h"
54 #include "hw/nmi.h"
55 #include "hw/loader.h"
56 #include "multiboot.h"
57 #include "elf.h"
58 #include "standard-headers/asm-x86/bootparam.h"
59 #include CONFIG_DEVICES
60 #include "kvm/kvm_i386.h"
61 
62 /* Physical Address of PVH entry point read from kernel ELF NOTE */
63 static size_t pvh_start_addr;
64 
65 inline void init_topo_info(X86CPUTopoInfo *topo_info,
66                            const X86MachineState *x86ms)
67 {
68     MachineState *ms = MACHINE(x86ms);
69 
70     topo_info->dies_per_pkg = ms->smp.dies;
71     topo_info->cores_per_die = ms->smp.cores;
72     topo_info->threads_per_core = ms->smp.threads;
73 }
74 
75 /*
76  * Calculates initial APIC ID for a specific CPU index
77  *
78  * Currently we need to be able to calculate the APIC ID from the CPU index
79  * alone (without requiring a CPU object), as the QEMU<->Seabios interfaces have
80  * no concept of "CPU index", and the NUMA tables on fw_cfg need the APIC ID of
81  * all CPUs up to max_cpus.
82  */
83 uint32_t x86_cpu_apic_id_from_index(X86MachineState *x86ms,
84                                     unsigned int cpu_index)
85 {
86     X86CPUTopoInfo topo_info;
87 
88     init_topo_info(&topo_info, x86ms);
89 
90     return x86_apicid_from_cpu_idx(&topo_info, cpu_index);
91 }
92 
93 
94 void x86_cpu_new(X86MachineState *x86ms, int64_t apic_id, Error **errp)
95 {
96     Object *cpu = object_new(MACHINE(x86ms)->cpu_type);
97 
98     if (!object_property_set_uint(cpu, "apic-id", apic_id, errp)) {
99         goto out;
100     }
101     qdev_realize(DEVICE(cpu), NULL, errp);
102 
103 out:
104     object_unref(cpu);
105 }
106 
107 void x86_cpus_init(X86MachineState *x86ms, int default_cpu_version)
108 {
109     int i;
110     const CPUArchIdList *possible_cpus;
111     MachineState *ms = MACHINE(x86ms);
112     MachineClass *mc = MACHINE_GET_CLASS(x86ms);
113 
114     x86_cpu_set_default_version(default_cpu_version);
115 
116     /*
117      * Calculates the limit to CPU APIC ID values
118      *
119      * Limit for the APIC ID value, so that all
120      * CPU APIC IDs are < x86ms->apic_id_limit.
121      *
122      * This is used for FW_CFG_MAX_CPUS. See comments on fw_cfg_arch_create().
123      */
124     x86ms->apic_id_limit = x86_cpu_apic_id_from_index(x86ms,
125                                                       ms->smp.max_cpus - 1) + 1;
126     possible_cpus = mc->possible_cpu_arch_ids(ms);
127     for (i = 0; i < ms->smp.cpus; i++) {
128         x86_cpu_new(x86ms, possible_cpus->cpus[i].arch_id, &error_fatal);
129     }
130 }
131 
132 void x86_rtc_set_cpus_count(ISADevice *rtc, uint16_t cpus_count)
133 {
134     if (cpus_count > 0xff) {
135         /*
136          * If the number of CPUs can't be represented in 8 bits, the
137          * BIOS must use "FW_CFG_NB_CPUS". Set RTC field to 0 just
138          * to make old BIOSes fail more predictably.
139          */
140         rtc_set_memory(rtc, 0x5f, 0);
141     } else {
142         rtc_set_memory(rtc, 0x5f, cpus_count - 1);
143     }
144 }
145 
146 static int x86_apic_cmp(const void *a, const void *b)
147 {
148    CPUArchId *apic_a = (CPUArchId *)a;
149    CPUArchId *apic_b = (CPUArchId *)b;
150 
151    return apic_a->arch_id - apic_b->arch_id;
152 }
153 
154 /*
155  * returns pointer to CPUArchId descriptor that matches CPU's apic_id
156  * in ms->possible_cpus->cpus, if ms->possible_cpus->cpus has no
157  * entry corresponding to CPU's apic_id returns NULL.
158  */
159 CPUArchId *x86_find_cpu_slot(MachineState *ms, uint32_t id, int *idx)
160 {
161     CPUArchId apic_id, *found_cpu;
162 
163     apic_id.arch_id = id;
164     found_cpu = bsearch(&apic_id, ms->possible_cpus->cpus,
165         ms->possible_cpus->len, sizeof(*ms->possible_cpus->cpus),
166         x86_apic_cmp);
167     if (found_cpu && idx) {
168         *idx = found_cpu - ms->possible_cpus->cpus;
169     }
170     return found_cpu;
171 }
172 
173 void x86_cpu_plug(HotplugHandler *hotplug_dev,
174                   DeviceState *dev, Error **errp)
175 {
176     CPUArchId *found_cpu;
177     Error *local_err = NULL;
178     X86CPU *cpu = X86_CPU(dev);
179     X86MachineState *x86ms = X86_MACHINE(hotplug_dev);
180 
181     if (x86ms->acpi_dev) {
182         hotplug_handler_plug(x86ms->acpi_dev, dev, &local_err);
183         if (local_err) {
184             goto out;
185         }
186     }
187 
188     /* increment the number of CPUs */
189     x86ms->boot_cpus++;
190     if (x86ms->rtc) {
191         x86_rtc_set_cpus_count(x86ms->rtc, x86ms->boot_cpus);
192     }
193     if (x86ms->fw_cfg) {
194         fw_cfg_modify_i16(x86ms->fw_cfg, FW_CFG_NB_CPUS, x86ms->boot_cpus);
195     }
196 
197     found_cpu = x86_find_cpu_slot(MACHINE(x86ms), cpu->apic_id, NULL);
198     found_cpu->cpu = OBJECT(dev);
199 out:
200     error_propagate(errp, local_err);
201 }
202 
203 void x86_cpu_unplug_request_cb(HotplugHandler *hotplug_dev,
204                                DeviceState *dev, Error **errp)
205 {
206     int idx = -1;
207     X86CPU *cpu = X86_CPU(dev);
208     X86MachineState *x86ms = X86_MACHINE(hotplug_dev);
209 
210     if (!x86ms->acpi_dev) {
211         error_setg(errp, "CPU hot unplug not supported without ACPI");
212         return;
213     }
214 
215     x86_find_cpu_slot(MACHINE(x86ms), cpu->apic_id, &idx);
216     assert(idx != -1);
217     if (idx == 0) {
218         error_setg(errp, "Boot CPU is unpluggable");
219         return;
220     }
221 
222     hotplug_handler_unplug_request(x86ms->acpi_dev, dev,
223                                    errp);
224 }
225 
226 void x86_cpu_unplug_cb(HotplugHandler *hotplug_dev,
227                        DeviceState *dev, Error **errp)
228 {
229     CPUArchId *found_cpu;
230     Error *local_err = NULL;
231     X86CPU *cpu = X86_CPU(dev);
232     X86MachineState *x86ms = X86_MACHINE(hotplug_dev);
233 
234     hotplug_handler_unplug(x86ms->acpi_dev, dev, &local_err);
235     if (local_err) {
236         goto out;
237     }
238 
239     found_cpu = x86_find_cpu_slot(MACHINE(x86ms), cpu->apic_id, NULL);
240     found_cpu->cpu = NULL;
241     qdev_unrealize(dev);
242 
243     /* decrement the number of CPUs */
244     x86ms->boot_cpus--;
245     /* Update the number of CPUs in CMOS */
246     x86_rtc_set_cpus_count(x86ms->rtc, x86ms->boot_cpus);
247     fw_cfg_modify_i16(x86ms->fw_cfg, FW_CFG_NB_CPUS, x86ms->boot_cpus);
248  out:
249     error_propagate(errp, local_err);
250 }
251 
252 void x86_cpu_pre_plug(HotplugHandler *hotplug_dev,
253                       DeviceState *dev, Error **errp)
254 {
255     int idx;
256     CPUState *cs;
257     CPUArchId *cpu_slot;
258     X86CPUTopoIDs topo_ids;
259     X86CPU *cpu = X86_CPU(dev);
260     CPUX86State *env = &cpu->env;
261     MachineState *ms = MACHINE(hotplug_dev);
262     X86MachineState *x86ms = X86_MACHINE(hotplug_dev);
263     unsigned int smp_cores = ms->smp.cores;
264     unsigned int smp_threads = ms->smp.threads;
265     X86CPUTopoInfo topo_info;
266 
267     if (!object_dynamic_cast(OBJECT(cpu), ms->cpu_type)) {
268         error_setg(errp, "Invalid CPU type, expected cpu type: '%s'",
269                    ms->cpu_type);
270         return;
271     }
272 
273     if (x86ms->acpi_dev) {
274         Error *local_err = NULL;
275 
276         hotplug_handler_pre_plug(HOTPLUG_HANDLER(x86ms->acpi_dev), dev,
277                                  &local_err);
278         if (local_err) {
279             error_propagate(errp, local_err);
280             return;
281         }
282     }
283 
284     init_topo_info(&topo_info, x86ms);
285 
286     env->nr_dies = ms->smp.dies;
287 
288     /*
289      * If APIC ID is not set,
290      * set it based on socket/die/core/thread properties.
291      */
292     if (cpu->apic_id == UNASSIGNED_APIC_ID) {
293         int max_socket = (ms->smp.max_cpus - 1) /
294                                 smp_threads / smp_cores / ms->smp.dies;
295 
296         /*
297          * die-id was optional in QEMU 4.0 and older, so keep it optional
298          * if there's only one die per socket.
299          */
300         if (cpu->die_id < 0 && ms->smp.dies == 1) {
301             cpu->die_id = 0;
302         }
303 
304         if (cpu->socket_id < 0) {
305             error_setg(errp, "CPU socket-id is not set");
306             return;
307         } else if (cpu->socket_id > max_socket) {
308             error_setg(errp, "Invalid CPU socket-id: %u must be in range 0:%u",
309                        cpu->socket_id, max_socket);
310             return;
311         }
312         if (cpu->die_id < 0) {
313             error_setg(errp, "CPU die-id is not set");
314             return;
315         } else if (cpu->die_id > ms->smp.dies - 1) {
316             error_setg(errp, "Invalid CPU die-id: %u must be in range 0:%u",
317                        cpu->die_id, ms->smp.dies - 1);
318             return;
319         }
320         if (cpu->core_id < 0) {
321             error_setg(errp, "CPU core-id is not set");
322             return;
323         } else if (cpu->core_id > (smp_cores - 1)) {
324             error_setg(errp, "Invalid CPU core-id: %u must be in range 0:%u",
325                        cpu->core_id, smp_cores - 1);
326             return;
327         }
328         if (cpu->thread_id < 0) {
329             error_setg(errp, "CPU thread-id is not set");
330             return;
331         } else if (cpu->thread_id > (smp_threads - 1)) {
332             error_setg(errp, "Invalid CPU thread-id: %u must be in range 0:%u",
333                        cpu->thread_id, smp_threads - 1);
334             return;
335         }
336 
337         topo_ids.pkg_id = cpu->socket_id;
338         topo_ids.die_id = cpu->die_id;
339         topo_ids.core_id = cpu->core_id;
340         topo_ids.smt_id = cpu->thread_id;
341         cpu->apic_id = x86_apicid_from_topo_ids(&topo_info, &topo_ids);
342     }
343 
344     cpu_slot = x86_find_cpu_slot(MACHINE(x86ms), cpu->apic_id, &idx);
345     if (!cpu_slot) {
346         MachineState *ms = MACHINE(x86ms);
347 
348         x86_topo_ids_from_apicid(cpu->apic_id, &topo_info, &topo_ids);
349         error_setg(errp,
350             "Invalid CPU [socket: %u, die: %u, core: %u, thread: %u] with"
351             " APIC ID %" PRIu32 ", valid index range 0:%d",
352             topo_ids.pkg_id, topo_ids.die_id, topo_ids.core_id, topo_ids.smt_id,
353             cpu->apic_id, ms->possible_cpus->len - 1);
354         return;
355     }
356 
357     if (cpu_slot->cpu) {
358         error_setg(errp, "CPU[%d] with APIC ID %" PRIu32 " exists",
359                    idx, cpu->apic_id);
360         return;
361     }
362 
363     /* if 'address' properties socket-id/core-id/thread-id are not set, set them
364      * so that machine_query_hotpluggable_cpus would show correct values
365      */
366     /* TODO: move socket_id/core_id/thread_id checks into x86_cpu_realizefn()
367      * once -smp refactoring is complete and there will be CPU private
368      * CPUState::nr_cores and CPUState::nr_threads fields instead of globals */
369     x86_topo_ids_from_apicid(cpu->apic_id, &topo_info, &topo_ids);
370     if (cpu->socket_id != -1 && cpu->socket_id != topo_ids.pkg_id) {
371         error_setg(errp, "property socket-id: %u doesn't match set apic-id:"
372             " 0x%x (socket-id: %u)", cpu->socket_id, cpu->apic_id,
373             topo_ids.pkg_id);
374         return;
375     }
376     cpu->socket_id = topo_ids.pkg_id;
377 
378     if (cpu->die_id != -1 && cpu->die_id != topo_ids.die_id) {
379         error_setg(errp, "property die-id: %u doesn't match set apic-id:"
380             " 0x%x (die-id: %u)", cpu->die_id, cpu->apic_id, topo_ids.die_id);
381         return;
382     }
383     cpu->die_id = topo_ids.die_id;
384 
385     if (cpu->core_id != -1 && cpu->core_id != topo_ids.core_id) {
386         error_setg(errp, "property core-id: %u doesn't match set apic-id:"
387             " 0x%x (core-id: %u)", cpu->core_id, cpu->apic_id,
388             topo_ids.core_id);
389         return;
390     }
391     cpu->core_id = topo_ids.core_id;
392 
393     if (cpu->thread_id != -1 && cpu->thread_id != topo_ids.smt_id) {
394         error_setg(errp, "property thread-id: %u doesn't match set apic-id:"
395             " 0x%x (thread-id: %u)", cpu->thread_id, cpu->apic_id,
396             topo_ids.smt_id);
397         return;
398     }
399     cpu->thread_id = topo_ids.smt_id;
400 
401     if (hyperv_feat_enabled(cpu, HYPERV_FEAT_VPINDEX) &&
402         !kvm_hv_vpindex_settable()) {
403         error_setg(errp, "kernel doesn't allow setting HyperV VP_INDEX");
404         return;
405     }
406 
407     cs = CPU(cpu);
408     cs->cpu_index = idx;
409 
410     numa_cpu_pre_plug(cpu_slot, dev, errp);
411 }
412 
413 CpuInstanceProperties
414 x86_cpu_index_to_props(MachineState *ms, unsigned cpu_index)
415 {
416     MachineClass *mc = MACHINE_GET_CLASS(ms);
417     const CPUArchIdList *possible_cpus = mc->possible_cpu_arch_ids(ms);
418 
419     assert(cpu_index < possible_cpus->len);
420     return possible_cpus->cpus[cpu_index].props;
421 }
422 
423 int64_t x86_get_default_cpu_node_id(const MachineState *ms, int idx)
424 {
425    X86CPUTopoIDs topo_ids;
426    X86MachineState *x86ms = X86_MACHINE(ms);
427    X86CPUTopoInfo topo_info;
428 
429    init_topo_info(&topo_info, x86ms);
430 
431    assert(idx < ms->possible_cpus->len);
432    x86_topo_ids_from_apicid(ms->possible_cpus->cpus[idx].arch_id,
433                             &topo_info, &topo_ids);
434    return topo_ids.pkg_id % ms->numa_state->num_nodes;
435 }
436 
437 const CPUArchIdList *x86_possible_cpu_arch_ids(MachineState *ms)
438 {
439     X86MachineState *x86ms = X86_MACHINE(ms);
440     unsigned int max_cpus = ms->smp.max_cpus;
441     X86CPUTopoInfo topo_info;
442     int i;
443 
444     if (ms->possible_cpus) {
445         /*
446          * make sure that max_cpus hasn't changed since the first use, i.e.
447          * -smp hasn't been parsed after it
448          */
449         assert(ms->possible_cpus->len == max_cpus);
450         return ms->possible_cpus;
451     }
452 
453     ms->possible_cpus = g_malloc0(sizeof(CPUArchIdList) +
454                                   sizeof(CPUArchId) * max_cpus);
455     ms->possible_cpus->len = max_cpus;
456 
457     init_topo_info(&topo_info, x86ms);
458 
459     for (i = 0; i < ms->possible_cpus->len; i++) {
460         X86CPUTopoIDs topo_ids;
461 
462         ms->possible_cpus->cpus[i].type = ms->cpu_type;
463         ms->possible_cpus->cpus[i].vcpus_count = 1;
464         ms->possible_cpus->cpus[i].arch_id =
465             x86_cpu_apic_id_from_index(x86ms, i);
466         x86_topo_ids_from_apicid(ms->possible_cpus->cpus[i].arch_id,
467                                  &topo_info, &topo_ids);
468         ms->possible_cpus->cpus[i].props.has_socket_id = true;
469         ms->possible_cpus->cpus[i].props.socket_id = topo_ids.pkg_id;
470         if (ms->smp.dies > 1) {
471             ms->possible_cpus->cpus[i].props.has_die_id = true;
472             ms->possible_cpus->cpus[i].props.die_id = topo_ids.die_id;
473         }
474         ms->possible_cpus->cpus[i].props.has_core_id = true;
475         ms->possible_cpus->cpus[i].props.core_id = topo_ids.core_id;
476         ms->possible_cpus->cpus[i].props.has_thread_id = true;
477         ms->possible_cpus->cpus[i].props.thread_id = topo_ids.smt_id;
478     }
479     return ms->possible_cpus;
480 }
481 
482 static void x86_nmi(NMIState *n, int cpu_index, Error **errp)
483 {
484     /* cpu index isn't used */
485     CPUState *cs;
486 
487     CPU_FOREACH(cs) {
488         X86CPU *cpu = X86_CPU(cs);
489 
490         if (!cpu->apic_state) {
491             cpu_interrupt(cs, CPU_INTERRUPT_NMI);
492         } else {
493             apic_deliver_nmi(cpu->apic_state);
494         }
495     }
496 }
497 
498 static long get_file_size(FILE *f)
499 {
500     long where, size;
501 
502     /* XXX: on Unix systems, using fstat() probably makes more sense */
503 
504     where = ftell(f);
505     fseek(f, 0, SEEK_END);
506     size = ftell(f);
507     fseek(f, where, SEEK_SET);
508 
509     return size;
510 }
511 
512 /* TSC handling */
513 uint64_t cpu_get_tsc(CPUX86State *env)
514 {
515     return cpus_get_elapsed_ticks();
516 }
517 
518 /* IRQ handling */
519 static void pic_irq_request(void *opaque, int irq, int level)
520 {
521     CPUState *cs = first_cpu;
522     X86CPU *cpu = X86_CPU(cs);
523 
524     trace_x86_pic_interrupt(irq, level);
525     if (cpu->apic_state && !kvm_irqchip_in_kernel() &&
526         !whpx_apic_in_platform()) {
527         CPU_FOREACH(cs) {
528             cpu = X86_CPU(cs);
529             if (apic_accept_pic_intr(cpu->apic_state)) {
530                 apic_deliver_pic_intr(cpu->apic_state, level);
531             }
532         }
533     } else {
534         if (level) {
535             cpu_interrupt(cs, CPU_INTERRUPT_HARD);
536         } else {
537             cpu_reset_interrupt(cs, CPU_INTERRUPT_HARD);
538         }
539     }
540 }
541 
542 qemu_irq x86_allocate_cpu_irq(void)
543 {
544     return qemu_allocate_irq(pic_irq_request, NULL, 0);
545 }
546 
547 int cpu_get_pic_interrupt(CPUX86State *env)
548 {
549     X86CPU *cpu = env_archcpu(env);
550     int intno;
551 
552     if (!kvm_irqchip_in_kernel() && !whpx_apic_in_platform()) {
553         intno = apic_get_interrupt(cpu->apic_state);
554         if (intno >= 0) {
555             return intno;
556         }
557         /* read the irq from the PIC */
558         if (!apic_accept_pic_intr(cpu->apic_state)) {
559             return -1;
560         }
561     }
562 
563     intno = pic_read_irq(isa_pic);
564     return intno;
565 }
566 
567 DeviceState *cpu_get_current_apic(void)
568 {
569     if (current_cpu) {
570         X86CPU *cpu = X86_CPU(current_cpu);
571         return cpu->apic_state;
572     } else {
573         return NULL;
574     }
575 }
576 
577 void gsi_handler(void *opaque, int n, int level)
578 {
579     GSIState *s = opaque;
580 
581     trace_x86_gsi_interrupt(n, level);
582     switch (n) {
583     case 0 ... ISA_NUM_IRQS - 1:
584         if (s->i8259_irq[n]) {
585             /* Under KVM, Kernel will forward to both PIC and IOAPIC */
586             qemu_set_irq(s->i8259_irq[n], level);
587         }
588         /* fall through */
589     case ISA_NUM_IRQS ... IOAPIC_NUM_PINS - 1:
590         qemu_set_irq(s->ioapic_irq[n], level);
591         break;
592     case IO_APIC_SECONDARY_IRQBASE
593         ... IO_APIC_SECONDARY_IRQBASE + IOAPIC_NUM_PINS - 1:
594         qemu_set_irq(s->ioapic2_irq[n - IO_APIC_SECONDARY_IRQBASE], level);
595         break;
596     }
597 }
598 
599 void ioapic_init_gsi(GSIState *gsi_state, const char *parent_name)
600 {
601     DeviceState *dev;
602     SysBusDevice *d;
603     unsigned int i;
604 
605     assert(parent_name);
606     if (kvm_ioapic_in_kernel()) {
607         dev = qdev_new(TYPE_KVM_IOAPIC);
608     } else {
609         dev = qdev_new(TYPE_IOAPIC);
610     }
611     object_property_add_child(object_resolve_path(parent_name, NULL),
612                               "ioapic", OBJECT(dev));
613     d = SYS_BUS_DEVICE(dev);
614     sysbus_realize_and_unref(d, &error_fatal);
615     sysbus_mmio_map(d, 0, IO_APIC_DEFAULT_ADDRESS);
616 
617     for (i = 0; i < IOAPIC_NUM_PINS; i++) {
618         gsi_state->ioapic_irq[i] = qdev_get_gpio_in(dev, i);
619     }
620 }
621 
622 DeviceState *ioapic_init_secondary(GSIState *gsi_state)
623 {
624     DeviceState *dev;
625     SysBusDevice *d;
626     unsigned int i;
627 
628     dev = qdev_new(TYPE_IOAPIC);
629     d = SYS_BUS_DEVICE(dev);
630     sysbus_realize_and_unref(d, &error_fatal);
631     sysbus_mmio_map(d, 0, IO_APIC_SECONDARY_ADDRESS);
632 
633     for (i = 0; i < IOAPIC_NUM_PINS; i++) {
634         gsi_state->ioapic2_irq[i] = qdev_get_gpio_in(dev, i);
635     }
636     return dev;
637 }
638 
639 struct setup_data {
640     uint64_t next;
641     uint32_t type;
642     uint32_t len;
643     uint8_t data[];
644 } __attribute__((packed));
645 
646 
647 /*
648  * The entry point into the kernel for PVH boot is different from
649  * the native entry point.  The PVH entry is defined by the x86/HVM
650  * direct boot ABI and is available in an ELFNOTE in the kernel binary.
651  *
652  * This function is passed to load_elf() when it is called from
653  * load_elfboot() which then additionally checks for an ELF Note of
654  * type XEN_ELFNOTE_PHYS32_ENTRY and passes it to this function to
655  * parse the PVH entry address from the ELF Note.
656  *
657  * Due to trickery in elf_opts.h, load_elf() is actually available as
658  * load_elf32() or load_elf64() and this routine needs to be able
659  * to deal with being called as 32 or 64 bit.
660  *
661  * The address of the PVH entry point is saved to the 'pvh_start_addr'
662  * global variable.  (although the entry point is 32-bit, the kernel
663  * binary can be either 32-bit or 64-bit).
664  */
665 static uint64_t read_pvh_start_addr(void *arg1, void *arg2, bool is64)
666 {
667     size_t *elf_note_data_addr;
668 
669     /* Check if ELF Note header passed in is valid */
670     if (arg1 == NULL) {
671         return 0;
672     }
673 
674     if (is64) {
675         struct elf64_note *nhdr64 = (struct elf64_note *)arg1;
676         uint64_t nhdr_size64 = sizeof(struct elf64_note);
677         uint64_t phdr_align = *(uint64_t *)arg2;
678         uint64_t nhdr_namesz = nhdr64->n_namesz;
679 
680         elf_note_data_addr =
681             ((void *)nhdr64) + nhdr_size64 +
682             QEMU_ALIGN_UP(nhdr_namesz, phdr_align);
683 
684         pvh_start_addr = *elf_note_data_addr;
685     } else {
686         struct elf32_note *nhdr32 = (struct elf32_note *)arg1;
687         uint32_t nhdr_size32 = sizeof(struct elf32_note);
688         uint32_t phdr_align = *(uint32_t *)arg2;
689         uint32_t nhdr_namesz = nhdr32->n_namesz;
690 
691         elf_note_data_addr =
692             ((void *)nhdr32) + nhdr_size32 +
693             QEMU_ALIGN_UP(nhdr_namesz, phdr_align);
694 
695         pvh_start_addr = *(uint32_t *)elf_note_data_addr;
696     }
697 
698     return pvh_start_addr;
699 }
700 
701 static bool load_elfboot(const char *kernel_filename,
702                          int kernel_file_size,
703                          uint8_t *header,
704                          size_t pvh_xen_start_addr,
705                          FWCfgState *fw_cfg)
706 {
707     uint32_t flags = 0;
708     uint32_t mh_load_addr = 0;
709     uint32_t elf_kernel_size = 0;
710     uint64_t elf_entry;
711     uint64_t elf_low, elf_high;
712     int kernel_size;
713 
714     if (ldl_p(header) != 0x464c457f) {
715         return false; /* no elfboot */
716     }
717 
718     bool elf_is64 = header[EI_CLASS] == ELFCLASS64;
719     flags = elf_is64 ?
720         ((Elf64_Ehdr *)header)->e_flags : ((Elf32_Ehdr *)header)->e_flags;
721 
722     if (flags & 0x00010004) { /* LOAD_ELF_HEADER_HAS_ADDR */
723         error_report("elfboot unsupported flags = %x", flags);
724         exit(1);
725     }
726 
727     uint64_t elf_note_type = XEN_ELFNOTE_PHYS32_ENTRY;
728     kernel_size = load_elf(kernel_filename, read_pvh_start_addr,
729                            NULL, &elf_note_type, &elf_entry,
730                            &elf_low, &elf_high, NULL, 0, I386_ELF_MACHINE,
731                            0, 0);
732 
733     if (kernel_size < 0) {
734         error_report("Error while loading elf kernel");
735         exit(1);
736     }
737     mh_load_addr = elf_low;
738     elf_kernel_size = elf_high - elf_low;
739 
740     if (pvh_start_addr == 0) {
741         error_report("Error loading uncompressed kernel without PVH ELF Note");
742         exit(1);
743     }
744     fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_ENTRY, pvh_start_addr);
745     fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_ADDR, mh_load_addr);
746     fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_SIZE, elf_kernel_size);
747 
748     return true;
749 }
750 
751 void x86_load_linux(X86MachineState *x86ms,
752                     FWCfgState *fw_cfg,
753                     int acpi_data_size,
754                     bool pvh_enabled)
755 {
756     bool linuxboot_dma_enabled = X86_MACHINE_GET_CLASS(x86ms)->fwcfg_dma_enabled;
757     uint16_t protocol;
758     int setup_size, kernel_size, cmdline_size;
759     int dtb_size, setup_data_offset;
760     uint32_t initrd_max;
761     uint8_t header[8192], *setup, *kernel;
762     hwaddr real_addr, prot_addr, cmdline_addr, initrd_addr = 0;
763     FILE *f;
764     char *vmode;
765     MachineState *machine = MACHINE(x86ms);
766     struct setup_data *setup_data;
767     const char *kernel_filename = machine->kernel_filename;
768     const char *initrd_filename = machine->initrd_filename;
769     const char *dtb_filename = machine->dtb;
770     const char *kernel_cmdline = machine->kernel_cmdline;
771     SevKernelLoaderContext sev_load_ctx = {};
772 
773     /* Align to 16 bytes as a paranoia measure */
774     cmdline_size = (strlen(kernel_cmdline) + 16) & ~15;
775 
776     /* load the kernel header */
777     f = fopen(kernel_filename, "rb");
778     if (!f) {
779         fprintf(stderr, "qemu: could not open kernel file '%s': %s\n",
780                 kernel_filename, strerror(errno));
781         exit(1);
782     }
783 
784     kernel_size = get_file_size(f);
785     if (!kernel_size ||
786         fread(header, 1, MIN(ARRAY_SIZE(header), kernel_size), f) !=
787         MIN(ARRAY_SIZE(header), kernel_size)) {
788         fprintf(stderr, "qemu: could not load kernel '%s': %s\n",
789                 kernel_filename, strerror(errno));
790         exit(1);
791     }
792 
793     /* kernel protocol version */
794     if (ldl_p(header + 0x202) == 0x53726448) {
795         protocol = lduw_p(header + 0x206);
796     } else {
797         /*
798          * This could be a multiboot kernel. If it is, let's stop treating it
799          * like a Linux kernel.
800          * Note: some multiboot images could be in the ELF format (the same of
801          * PVH), so we try multiboot first since we check the multiboot magic
802          * header before to load it.
803          */
804         if (load_multiboot(x86ms, fw_cfg, f, kernel_filename, initrd_filename,
805                            kernel_cmdline, kernel_size, header)) {
806             return;
807         }
808         /*
809          * Check if the file is an uncompressed kernel file (ELF) and load it,
810          * saving the PVH entry point used by the x86/HVM direct boot ABI.
811          * If load_elfboot() is successful, populate the fw_cfg info.
812          */
813         if (pvh_enabled &&
814             load_elfboot(kernel_filename, kernel_size,
815                          header, pvh_start_addr, fw_cfg)) {
816             fclose(f);
817 
818             fw_cfg_add_i32(fw_cfg, FW_CFG_CMDLINE_SIZE,
819                 strlen(kernel_cmdline) + 1);
820             fw_cfg_add_string(fw_cfg, FW_CFG_CMDLINE_DATA, kernel_cmdline);
821 
822             fw_cfg_add_i32(fw_cfg, FW_CFG_SETUP_SIZE, sizeof(header));
823             fw_cfg_add_bytes(fw_cfg, FW_CFG_SETUP_DATA,
824                              header, sizeof(header));
825 
826             /* load initrd */
827             if (initrd_filename) {
828                 GMappedFile *mapped_file;
829                 gsize initrd_size;
830                 gchar *initrd_data;
831                 GError *gerr = NULL;
832 
833                 mapped_file = g_mapped_file_new(initrd_filename, false, &gerr);
834                 if (!mapped_file) {
835                     fprintf(stderr, "qemu: error reading initrd %s: %s\n",
836                             initrd_filename, gerr->message);
837                     exit(1);
838                 }
839                 x86ms->initrd_mapped_file = mapped_file;
840 
841                 initrd_data = g_mapped_file_get_contents(mapped_file);
842                 initrd_size = g_mapped_file_get_length(mapped_file);
843                 initrd_max = x86ms->below_4g_mem_size - acpi_data_size - 1;
844                 if (initrd_size >= initrd_max) {
845                     fprintf(stderr, "qemu: initrd is too large, cannot support."
846                             "(max: %"PRIu32", need %"PRId64")\n",
847                             initrd_max, (uint64_t)initrd_size);
848                     exit(1);
849                 }
850 
851                 initrd_addr = (initrd_max - initrd_size) & ~4095;
852 
853                 fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_ADDR, initrd_addr);
854                 fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_SIZE, initrd_size);
855                 fw_cfg_add_bytes(fw_cfg, FW_CFG_INITRD_DATA, initrd_data,
856                                  initrd_size);
857             }
858 
859             option_rom[nb_option_roms].bootindex = 0;
860             option_rom[nb_option_roms].name = "pvh.bin";
861             nb_option_roms++;
862 
863             return;
864         }
865         protocol = 0;
866     }
867 
868     if (protocol < 0x200 || !(header[0x211] & 0x01)) {
869         /* Low kernel */
870         real_addr    = 0x90000;
871         cmdline_addr = 0x9a000 - cmdline_size;
872         prot_addr    = 0x10000;
873     } else if (protocol < 0x202) {
874         /* High but ancient kernel */
875         real_addr    = 0x90000;
876         cmdline_addr = 0x9a000 - cmdline_size;
877         prot_addr    = 0x100000;
878     } else {
879         /* High and recent kernel */
880         real_addr    = 0x10000;
881         cmdline_addr = 0x20000;
882         prot_addr    = 0x100000;
883     }
884 
885     /* highest address for loading the initrd */
886     if (protocol >= 0x20c &&
887         lduw_p(header + 0x236) & XLF_CAN_BE_LOADED_ABOVE_4G) {
888         /*
889          * Linux has supported initrd up to 4 GB for a very long time (2007,
890          * long before XLF_CAN_BE_LOADED_ABOVE_4G which was added in 2013),
891          * though it only sets initrd_max to 2 GB to "work around bootloader
892          * bugs". Luckily, QEMU firmware(which does something like bootloader)
893          * has supported this.
894          *
895          * It's believed that if XLF_CAN_BE_LOADED_ABOVE_4G is set, initrd can
896          * be loaded into any address.
897          *
898          * In addition, initrd_max is uint32_t simply because QEMU doesn't
899          * support the 64-bit boot protocol (specifically the ext_ramdisk_image
900          * field).
901          *
902          * Therefore here just limit initrd_max to UINT32_MAX simply as well.
903          */
904         initrd_max = UINT32_MAX;
905     } else if (protocol >= 0x203) {
906         initrd_max = ldl_p(header + 0x22c);
907     } else {
908         initrd_max = 0x37ffffff;
909     }
910 
911     if (initrd_max >= x86ms->below_4g_mem_size - acpi_data_size) {
912         initrd_max = x86ms->below_4g_mem_size - acpi_data_size - 1;
913     }
914 
915     fw_cfg_add_i32(fw_cfg, FW_CFG_CMDLINE_ADDR, cmdline_addr);
916     fw_cfg_add_i32(fw_cfg, FW_CFG_CMDLINE_SIZE, strlen(kernel_cmdline) + 1);
917     fw_cfg_add_string(fw_cfg, FW_CFG_CMDLINE_DATA, kernel_cmdline);
918     sev_load_ctx.cmdline_data = (char *)kernel_cmdline;
919     sev_load_ctx.cmdline_size = strlen(kernel_cmdline) + 1;
920 
921     if (protocol >= 0x202) {
922         stl_p(header + 0x228, cmdline_addr);
923     } else {
924         stw_p(header + 0x20, 0xA33F);
925         stw_p(header + 0x22, cmdline_addr - real_addr);
926     }
927 
928     /* handle vga= parameter */
929     vmode = strstr(kernel_cmdline, "vga=");
930     if (vmode) {
931         unsigned int video_mode;
932         const char *end;
933         int ret;
934         /* skip "vga=" */
935         vmode += 4;
936         if (!strncmp(vmode, "normal", 6)) {
937             video_mode = 0xffff;
938         } else if (!strncmp(vmode, "ext", 3)) {
939             video_mode = 0xfffe;
940         } else if (!strncmp(vmode, "ask", 3)) {
941             video_mode = 0xfffd;
942         } else {
943             ret = qemu_strtoui(vmode, &end, 0, &video_mode);
944             if (ret != 0 || (*end && *end != ' ')) {
945                 fprintf(stderr, "qemu: invalid 'vga=' kernel parameter.\n");
946                 exit(1);
947             }
948         }
949         stw_p(header + 0x1fa, video_mode);
950     }
951 
952     /* loader type */
953     /*
954      * High nybble = B reserved for QEMU; low nybble is revision number.
955      * If this code is substantially changed, you may want to consider
956      * incrementing the revision.
957      */
958     if (protocol >= 0x200) {
959         header[0x210] = 0xB0;
960     }
961     /* heap */
962     if (protocol >= 0x201) {
963         header[0x211] |= 0x80; /* CAN_USE_HEAP */
964         stw_p(header + 0x224, cmdline_addr - real_addr - 0x200);
965     }
966 
967     /* load initrd */
968     if (initrd_filename) {
969         GMappedFile *mapped_file;
970         gsize initrd_size;
971         gchar *initrd_data;
972         GError *gerr = NULL;
973 
974         if (protocol < 0x200) {
975             fprintf(stderr, "qemu: linux kernel too old to load a ram disk\n");
976             exit(1);
977         }
978 
979         mapped_file = g_mapped_file_new(initrd_filename, false, &gerr);
980         if (!mapped_file) {
981             fprintf(stderr, "qemu: error reading initrd %s: %s\n",
982                     initrd_filename, gerr->message);
983             exit(1);
984         }
985         x86ms->initrd_mapped_file = mapped_file;
986 
987         initrd_data = g_mapped_file_get_contents(mapped_file);
988         initrd_size = g_mapped_file_get_length(mapped_file);
989         if (initrd_size >= initrd_max) {
990             fprintf(stderr, "qemu: initrd is too large, cannot support."
991                     "(max: %"PRIu32", need %"PRId64")\n",
992                     initrd_max, (uint64_t)initrd_size);
993             exit(1);
994         }
995 
996         initrd_addr = (initrd_max - initrd_size) & ~4095;
997 
998         fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_ADDR, initrd_addr);
999         fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_SIZE, initrd_size);
1000         fw_cfg_add_bytes(fw_cfg, FW_CFG_INITRD_DATA, initrd_data, initrd_size);
1001         sev_load_ctx.initrd_data = initrd_data;
1002         sev_load_ctx.initrd_size = initrd_size;
1003 
1004         stl_p(header + 0x218, initrd_addr);
1005         stl_p(header + 0x21c, initrd_size);
1006     }
1007 
1008     /* load kernel and setup */
1009     setup_size = header[0x1f1];
1010     if (setup_size == 0) {
1011         setup_size = 4;
1012     }
1013     setup_size = (setup_size + 1) * 512;
1014     if (setup_size > kernel_size) {
1015         fprintf(stderr, "qemu: invalid kernel header\n");
1016         exit(1);
1017     }
1018     kernel_size -= setup_size;
1019 
1020     setup  = g_malloc(setup_size);
1021     kernel = g_malloc(kernel_size);
1022     fseek(f, 0, SEEK_SET);
1023     if (fread(setup, 1, setup_size, f) != setup_size) {
1024         fprintf(stderr, "fread() failed\n");
1025         exit(1);
1026     }
1027     if (fread(kernel, 1, kernel_size, f) != kernel_size) {
1028         fprintf(stderr, "fread() failed\n");
1029         exit(1);
1030     }
1031     fclose(f);
1032 
1033     /* append dtb to kernel */
1034     if (dtb_filename) {
1035         if (protocol < 0x209) {
1036             fprintf(stderr, "qemu: Linux kernel too old to load a dtb\n");
1037             exit(1);
1038         }
1039 
1040         dtb_size = get_image_size(dtb_filename);
1041         if (dtb_size <= 0) {
1042             fprintf(stderr, "qemu: error reading dtb %s: %s\n",
1043                     dtb_filename, strerror(errno));
1044             exit(1);
1045         }
1046 
1047         setup_data_offset = QEMU_ALIGN_UP(kernel_size, 16);
1048         kernel_size = setup_data_offset + sizeof(struct setup_data) + dtb_size;
1049         kernel = g_realloc(kernel, kernel_size);
1050 
1051         stq_p(header + 0x250, prot_addr + setup_data_offset);
1052 
1053         setup_data = (struct setup_data *)(kernel + setup_data_offset);
1054         setup_data->next = 0;
1055         setup_data->type = cpu_to_le32(SETUP_DTB);
1056         setup_data->len = cpu_to_le32(dtb_size);
1057 
1058         load_image_size(dtb_filename, setup_data->data, dtb_size);
1059     }
1060 
1061     /*
1062      * If we're starting an encrypted VM, it will be OVMF based, which uses the
1063      * efi stub for booting and doesn't require any values to be placed in the
1064      * kernel header.  We therefore don't update the header so the hash of the
1065      * kernel on the other side of the fw_cfg interface matches the hash of the
1066      * file the user passed in.
1067      */
1068     if (!sev_enabled()) {
1069         memcpy(setup, header, MIN(sizeof(header), setup_size));
1070     }
1071 
1072     fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_ADDR, prot_addr);
1073     fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_SIZE, kernel_size);
1074     fw_cfg_add_bytes(fw_cfg, FW_CFG_KERNEL_DATA, kernel, kernel_size);
1075     sev_load_ctx.kernel_data = (char *)kernel;
1076     sev_load_ctx.kernel_size = kernel_size;
1077 
1078     fw_cfg_add_i32(fw_cfg, FW_CFG_SETUP_ADDR, real_addr);
1079     fw_cfg_add_i32(fw_cfg, FW_CFG_SETUP_SIZE, setup_size);
1080     fw_cfg_add_bytes(fw_cfg, FW_CFG_SETUP_DATA, setup, setup_size);
1081     sev_load_ctx.setup_data = (char *)setup;
1082     sev_load_ctx.setup_size = setup_size;
1083 
1084     if (sev_enabled()) {
1085         sev_add_kernel_loader_hashes(&sev_load_ctx, &error_fatal);
1086     }
1087 
1088     option_rom[nb_option_roms].bootindex = 0;
1089     option_rom[nb_option_roms].name = "linuxboot.bin";
1090     if (linuxboot_dma_enabled && fw_cfg_dma_enabled(fw_cfg)) {
1091         option_rom[nb_option_roms].name = "linuxboot_dma.bin";
1092     }
1093     nb_option_roms++;
1094 }
1095 
1096 void x86_bios_rom_init(MachineState *ms, const char *default_firmware,
1097                        MemoryRegion *rom_memory, bool isapc_ram_fw)
1098 {
1099     const char *bios_name;
1100     char *filename;
1101     MemoryRegion *bios, *isa_bios;
1102     int bios_size, isa_bios_size;
1103     int ret;
1104 
1105     /* BIOS load */
1106     bios_name = ms->firmware ?: default_firmware;
1107     filename = qemu_find_file(QEMU_FILE_TYPE_BIOS, bios_name);
1108     if (filename) {
1109         bios_size = get_image_size(filename);
1110     } else {
1111         bios_size = -1;
1112     }
1113     if (bios_size <= 0 ||
1114         (bios_size % 65536) != 0) {
1115         goto bios_error;
1116     }
1117     bios = g_malloc(sizeof(*bios));
1118     memory_region_init_ram(bios, NULL, "pc.bios", bios_size, &error_fatal);
1119     if (!isapc_ram_fw) {
1120         memory_region_set_readonly(bios, true);
1121     }
1122     ret = rom_add_file_fixed(bios_name, (uint32_t)(-bios_size), -1);
1123     if (ret != 0) {
1124     bios_error:
1125         fprintf(stderr, "qemu: could not load PC BIOS '%s'\n", bios_name);
1126         exit(1);
1127     }
1128     g_free(filename);
1129 
1130     /* map the last 128KB of the BIOS in ISA space */
1131     isa_bios_size = MIN(bios_size, 128 * KiB);
1132     isa_bios = g_malloc(sizeof(*isa_bios));
1133     memory_region_init_alias(isa_bios, NULL, "isa-bios", bios,
1134                              bios_size - isa_bios_size, isa_bios_size);
1135     memory_region_add_subregion_overlap(rom_memory,
1136                                         0x100000 - isa_bios_size,
1137                                         isa_bios,
1138                                         1);
1139     if (!isapc_ram_fw) {
1140         memory_region_set_readonly(isa_bios, true);
1141     }
1142 
1143     /* map all the bios at the top of memory */
1144     memory_region_add_subregion(rom_memory,
1145                                 (uint32_t)(-bios_size),
1146                                 bios);
1147 }
1148 
1149 bool x86_machine_is_smm_enabled(const X86MachineState *x86ms)
1150 {
1151     bool smm_available = false;
1152 
1153     if (x86ms->smm == ON_OFF_AUTO_OFF) {
1154         return false;
1155     }
1156 
1157     if (tcg_enabled() || qtest_enabled()) {
1158         smm_available = true;
1159     } else if (kvm_enabled()) {
1160         smm_available = kvm_has_smm();
1161     }
1162 
1163     if (smm_available) {
1164         return true;
1165     }
1166 
1167     if (x86ms->smm == ON_OFF_AUTO_ON) {
1168         error_report("System Management Mode not supported by this hypervisor.");
1169         exit(1);
1170     }
1171     return false;
1172 }
1173 
1174 static void x86_machine_get_smm(Object *obj, Visitor *v, const char *name,
1175                                void *opaque, Error **errp)
1176 {
1177     X86MachineState *x86ms = X86_MACHINE(obj);
1178     OnOffAuto smm = x86ms->smm;
1179 
1180     visit_type_OnOffAuto(v, name, &smm, errp);
1181 }
1182 
1183 static void x86_machine_set_smm(Object *obj, Visitor *v, const char *name,
1184                                void *opaque, Error **errp)
1185 {
1186     X86MachineState *x86ms = X86_MACHINE(obj);
1187 
1188     visit_type_OnOffAuto(v, name, &x86ms->smm, errp);
1189 }
1190 
1191 bool x86_machine_is_acpi_enabled(const X86MachineState *x86ms)
1192 {
1193     if (x86ms->acpi == ON_OFF_AUTO_OFF) {
1194         return false;
1195     }
1196     return true;
1197 }
1198 
1199 static void x86_machine_get_acpi(Object *obj, Visitor *v, const char *name,
1200                                  void *opaque, Error **errp)
1201 {
1202     X86MachineState *x86ms = X86_MACHINE(obj);
1203     OnOffAuto acpi = x86ms->acpi;
1204 
1205     visit_type_OnOffAuto(v, name, &acpi, errp);
1206 }
1207 
1208 static void x86_machine_set_acpi(Object *obj, Visitor *v, const char *name,
1209                                  void *opaque, Error **errp)
1210 {
1211     X86MachineState *x86ms = X86_MACHINE(obj);
1212 
1213     visit_type_OnOffAuto(v, name, &x86ms->acpi, errp);
1214 }
1215 
1216 static char *x86_machine_get_oem_id(Object *obj, Error **errp)
1217 {
1218     X86MachineState *x86ms = X86_MACHINE(obj);
1219 
1220     return g_strdup(x86ms->oem_id);
1221 }
1222 
1223 static void x86_machine_set_oem_id(Object *obj, const char *value, Error **errp)
1224 {
1225     X86MachineState *x86ms = X86_MACHINE(obj);
1226     size_t len = strlen(value);
1227 
1228     if (len > 6) {
1229         error_setg(errp,
1230                    "User specified "X86_MACHINE_OEM_ID" value is bigger than "
1231                    "6 bytes in size");
1232         return;
1233     }
1234 
1235     strncpy(x86ms->oem_id, value, 6);
1236 }
1237 
1238 static char *x86_machine_get_oem_table_id(Object *obj, Error **errp)
1239 {
1240     X86MachineState *x86ms = X86_MACHINE(obj);
1241 
1242     return g_strdup(x86ms->oem_table_id);
1243 }
1244 
1245 static void x86_machine_set_oem_table_id(Object *obj, const char *value,
1246                                          Error **errp)
1247 {
1248     X86MachineState *x86ms = X86_MACHINE(obj);
1249     size_t len = strlen(value);
1250 
1251     if (len > 8) {
1252         error_setg(errp,
1253                    "User specified "X86_MACHINE_OEM_TABLE_ID
1254                    " value is bigger than "
1255                    "8 bytes in size");
1256         return;
1257     }
1258     strncpy(x86ms->oem_table_id, value, 8);
1259 }
1260 
1261 static void x86_machine_get_bus_lock_ratelimit(Object *obj, Visitor *v,
1262                                 const char *name, void *opaque, Error **errp)
1263 {
1264     X86MachineState *x86ms = X86_MACHINE(obj);
1265     uint64_t bus_lock_ratelimit = x86ms->bus_lock_ratelimit;
1266 
1267     visit_type_uint64(v, name, &bus_lock_ratelimit, errp);
1268 }
1269 
1270 static void x86_machine_set_bus_lock_ratelimit(Object *obj, Visitor *v,
1271                                const char *name, void *opaque, Error **errp)
1272 {
1273     X86MachineState *x86ms = X86_MACHINE(obj);
1274 
1275     visit_type_uint64(v, name, &x86ms->bus_lock_ratelimit, errp);
1276 }
1277 
1278 static void machine_get_sgx_epc(Object *obj, Visitor *v, const char *name,
1279                                 void *opaque, Error **errp)
1280 {
1281     X86MachineState *x86ms = X86_MACHINE(obj);
1282     SgxEPCList *list = x86ms->sgx_epc_list;
1283 
1284     visit_type_SgxEPCList(v, name, &list, errp);
1285 }
1286 
1287 static void machine_set_sgx_epc(Object *obj, Visitor *v, const char *name,
1288                                 void *opaque, Error **errp)
1289 {
1290     X86MachineState *x86ms = X86_MACHINE(obj);
1291     SgxEPCList *list;
1292 
1293     list = x86ms->sgx_epc_list;
1294     visit_type_SgxEPCList(v, name, &x86ms->sgx_epc_list, errp);
1295 
1296     qapi_free_SgxEPCList(list);
1297 }
1298 
1299 static void x86_machine_initfn(Object *obj)
1300 {
1301     X86MachineState *x86ms = X86_MACHINE(obj);
1302 
1303     x86ms->smm = ON_OFF_AUTO_AUTO;
1304     x86ms->acpi = ON_OFF_AUTO_AUTO;
1305     x86ms->pci_irq_mask = ACPI_BUILD_PCI_IRQS;
1306     x86ms->oem_id = g_strndup(ACPI_BUILD_APPNAME6, 6);
1307     x86ms->oem_table_id = g_strndup(ACPI_BUILD_APPNAME8, 8);
1308     x86ms->bus_lock_ratelimit = 0;
1309 }
1310 
1311 static void x86_machine_class_init(ObjectClass *oc, void *data)
1312 {
1313     MachineClass *mc = MACHINE_CLASS(oc);
1314     X86MachineClass *x86mc = X86_MACHINE_CLASS(oc);
1315     NMIClass *nc = NMI_CLASS(oc);
1316 
1317     mc->cpu_index_to_instance_props = x86_cpu_index_to_props;
1318     mc->get_default_cpu_node_id = x86_get_default_cpu_node_id;
1319     mc->possible_cpu_arch_ids = x86_possible_cpu_arch_ids;
1320     x86mc->save_tsc_khz = true;
1321     x86mc->fwcfg_dma_enabled = true;
1322     nc->nmi_monitor_handler = x86_nmi;
1323 
1324     object_class_property_add(oc, X86_MACHINE_SMM, "OnOffAuto",
1325         x86_machine_get_smm, x86_machine_set_smm,
1326         NULL, NULL);
1327     object_class_property_set_description(oc, X86_MACHINE_SMM,
1328         "Enable SMM");
1329 
1330     object_class_property_add(oc, X86_MACHINE_ACPI, "OnOffAuto",
1331         x86_machine_get_acpi, x86_machine_set_acpi,
1332         NULL, NULL);
1333     object_class_property_set_description(oc, X86_MACHINE_ACPI,
1334         "Enable ACPI");
1335 
1336     object_class_property_add_str(oc, X86_MACHINE_OEM_ID,
1337                                   x86_machine_get_oem_id,
1338                                   x86_machine_set_oem_id);
1339     object_class_property_set_description(oc, X86_MACHINE_OEM_ID,
1340                                           "Override the default value of field OEMID "
1341                                           "in ACPI table header."
1342                                           "The string may be up to 6 bytes in size");
1343 
1344 
1345     object_class_property_add_str(oc, X86_MACHINE_OEM_TABLE_ID,
1346                                   x86_machine_get_oem_table_id,
1347                                   x86_machine_set_oem_table_id);
1348     object_class_property_set_description(oc, X86_MACHINE_OEM_TABLE_ID,
1349                                           "Override the default value of field OEM Table ID "
1350                                           "in ACPI table header."
1351                                           "The string may be up to 8 bytes in size");
1352 
1353     object_class_property_add(oc, X86_MACHINE_BUS_LOCK_RATELIMIT, "uint64_t",
1354                                 x86_machine_get_bus_lock_ratelimit,
1355                                 x86_machine_set_bus_lock_ratelimit, NULL, NULL);
1356     object_class_property_set_description(oc, X86_MACHINE_BUS_LOCK_RATELIMIT,
1357             "Set the ratelimit for the bus locks acquired in VMs");
1358 
1359     object_class_property_add(oc, "sgx-epc", "SgxEPC",
1360         machine_get_sgx_epc, machine_set_sgx_epc,
1361         NULL, NULL);
1362     object_class_property_set_description(oc, "sgx-epc",
1363         "SGX EPC device");
1364 }
1365 
1366 static const TypeInfo x86_machine_info = {
1367     .name = TYPE_X86_MACHINE,
1368     .parent = TYPE_MACHINE,
1369     .abstract = true,
1370     .instance_size = sizeof(X86MachineState),
1371     .instance_init = x86_machine_initfn,
1372     .class_size = sizeof(X86MachineClass),
1373     .class_init = x86_machine_class_init,
1374     .interfaces = (InterfaceInfo[]) {
1375          { TYPE_NMI },
1376          { }
1377     },
1378 };
1379 
1380 static void x86_machine_register_types(void)
1381 {
1382     type_register_static(&x86_machine_info);
1383 }
1384 
1385 type_init(x86_machine_register_types)
1386