xref: /openbmc/qemu/hw/i386/x86.c (revision f6476697)
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/visitor.h"
34 #include "sysemu/qtest.h"
35 #include "sysemu/whpx.h"
36 #include "sysemu/numa.h"
37 #include "sysemu/replay.h"
38 #include "sysemu/sysemu.h"
39 #include "sysemu/cpu-timers.h"
40 #include "trace.h"
41 
42 #include "hw/i386/x86.h"
43 #include "target/i386/cpu.h"
44 #include "hw/i386/topology.h"
45 #include "hw/i386/fw_cfg.h"
46 #include "hw/intc/i8259.h"
47 #include "hw/rtc/mc146818rtc.h"
48 
49 #include "hw/acpi/cpu_hotplug.h"
50 #include "hw/irq.h"
51 #include "hw/nmi.h"
52 #include "hw/loader.h"
53 #include "multiboot.h"
54 #include "elf.h"
55 #include "standard-headers/asm-x86/bootparam.h"
56 #include CONFIG_DEVICES
57 #include "kvm/kvm_i386.h"
58 
59 /* Physical Address of PVH entry point read from kernel ELF NOTE */
60 static size_t pvh_start_addr;
61 
62 inline void init_topo_info(X86CPUTopoInfo *topo_info,
63                            const X86MachineState *x86ms)
64 {
65     MachineState *ms = MACHINE(x86ms);
66 
67     topo_info->dies_per_pkg = x86ms->smp_dies;
68     topo_info->cores_per_die = ms->smp.cores;
69     topo_info->threads_per_core = ms->smp.threads;
70 }
71 
72 /*
73  * Calculates initial APIC ID for a specific CPU index
74  *
75  * Currently we need to be able to calculate the APIC ID from the CPU index
76  * alone (without requiring a CPU object), as the QEMU<->Seabios interfaces have
77  * no concept of "CPU index", and the NUMA tables on fw_cfg need the APIC ID of
78  * all CPUs up to max_cpus.
79  */
80 uint32_t x86_cpu_apic_id_from_index(X86MachineState *x86ms,
81                                     unsigned int cpu_index)
82 {
83     X86MachineClass *x86mc = X86_MACHINE_GET_CLASS(x86ms);
84     X86CPUTopoInfo topo_info;
85     uint32_t correct_id;
86     static bool warned;
87 
88     init_topo_info(&topo_info, x86ms);
89 
90     correct_id = x86_apicid_from_cpu_idx(&topo_info, cpu_index);
91     if (x86mc->compat_apic_id_mode) {
92         if (cpu_index != correct_id && !warned && !qtest_enabled()) {
93             error_report("APIC IDs set in compatibility mode, "
94                          "CPU topology won't match the configuration");
95             warned = true;
96         }
97         return cpu_index;
98     } else {
99         return correct_id;
100     }
101 }
102 
103 
104 void x86_cpu_new(X86MachineState *x86ms, int64_t apic_id, Error **errp)
105 {
106     Object *cpu = object_new(MACHINE(x86ms)->cpu_type);
107 
108     if (!object_property_set_uint(cpu, "apic-id", apic_id, errp)) {
109         goto out;
110     }
111     qdev_realize(DEVICE(cpu), NULL, errp);
112 
113 out:
114     object_unref(cpu);
115 }
116 
117 void x86_cpus_init(X86MachineState *x86ms, int default_cpu_version)
118 {
119     int i;
120     const CPUArchIdList *possible_cpus;
121     MachineState *ms = MACHINE(x86ms);
122     MachineClass *mc = MACHINE_GET_CLASS(x86ms);
123 
124     x86_cpu_set_default_version(default_cpu_version);
125 
126     /*
127      * Calculates the limit to CPU APIC ID values
128      *
129      * Limit for the APIC ID value, so that all
130      * CPU APIC IDs are < x86ms->apic_id_limit.
131      *
132      * This is used for FW_CFG_MAX_CPUS. See comments on fw_cfg_arch_create().
133      */
134     x86ms->apic_id_limit = x86_cpu_apic_id_from_index(x86ms,
135                                                       ms->smp.max_cpus - 1) + 1;
136     possible_cpus = mc->possible_cpu_arch_ids(ms);
137     for (i = 0; i < ms->smp.cpus; i++) {
138         x86_cpu_new(x86ms, possible_cpus->cpus[i].arch_id, &error_fatal);
139     }
140 }
141 
142 void x86_rtc_set_cpus_count(ISADevice *rtc, uint16_t cpus_count)
143 {
144     if (cpus_count > 0xff) {
145         /*
146          * If the number of CPUs can't be represented in 8 bits, the
147          * BIOS must use "FW_CFG_NB_CPUS". Set RTC field to 0 just
148          * to make old BIOSes fail more predictably.
149          */
150         rtc_set_memory(rtc, 0x5f, 0);
151     } else {
152         rtc_set_memory(rtc, 0x5f, cpus_count - 1);
153     }
154 }
155 
156 static int x86_apic_cmp(const void *a, const void *b)
157 {
158    CPUArchId *apic_a = (CPUArchId *)a;
159    CPUArchId *apic_b = (CPUArchId *)b;
160 
161    return apic_a->arch_id - apic_b->arch_id;
162 }
163 
164 /*
165  * returns pointer to CPUArchId descriptor that matches CPU's apic_id
166  * in ms->possible_cpus->cpus, if ms->possible_cpus->cpus has no
167  * entry corresponding to CPU's apic_id returns NULL.
168  */
169 CPUArchId *x86_find_cpu_slot(MachineState *ms, uint32_t id, int *idx)
170 {
171     CPUArchId apic_id, *found_cpu;
172 
173     apic_id.arch_id = id;
174     found_cpu = bsearch(&apic_id, ms->possible_cpus->cpus,
175         ms->possible_cpus->len, sizeof(*ms->possible_cpus->cpus),
176         x86_apic_cmp);
177     if (found_cpu && idx) {
178         *idx = found_cpu - ms->possible_cpus->cpus;
179     }
180     return found_cpu;
181 }
182 
183 void x86_cpu_plug(HotplugHandler *hotplug_dev,
184                   DeviceState *dev, Error **errp)
185 {
186     CPUArchId *found_cpu;
187     Error *local_err = NULL;
188     X86CPU *cpu = X86_CPU(dev);
189     X86MachineState *x86ms = X86_MACHINE(hotplug_dev);
190 
191     if (x86ms->acpi_dev) {
192         hotplug_handler_plug(x86ms->acpi_dev, dev, &local_err);
193         if (local_err) {
194             goto out;
195         }
196     }
197 
198     /* increment the number of CPUs */
199     x86ms->boot_cpus++;
200     if (x86ms->rtc) {
201         x86_rtc_set_cpus_count(x86ms->rtc, x86ms->boot_cpus);
202     }
203     if (x86ms->fw_cfg) {
204         fw_cfg_modify_i16(x86ms->fw_cfg, FW_CFG_NB_CPUS, x86ms->boot_cpus);
205     }
206 
207     found_cpu = x86_find_cpu_slot(MACHINE(x86ms), cpu->apic_id, NULL);
208     found_cpu->cpu = OBJECT(dev);
209 out:
210     error_propagate(errp, local_err);
211 }
212 
213 void x86_cpu_unplug_request_cb(HotplugHandler *hotplug_dev,
214                                DeviceState *dev, Error **errp)
215 {
216     int idx = -1;
217     X86CPU *cpu = X86_CPU(dev);
218     X86MachineState *x86ms = X86_MACHINE(hotplug_dev);
219 
220     if (!x86ms->acpi_dev) {
221         error_setg(errp, "CPU hot unplug not supported without ACPI");
222         return;
223     }
224 
225     x86_find_cpu_slot(MACHINE(x86ms), cpu->apic_id, &idx);
226     assert(idx != -1);
227     if (idx == 0) {
228         error_setg(errp, "Boot CPU is unpluggable");
229         return;
230     }
231 
232     hotplug_handler_unplug_request(x86ms->acpi_dev, dev,
233                                    errp);
234 }
235 
236 void x86_cpu_unplug_cb(HotplugHandler *hotplug_dev,
237                        DeviceState *dev, Error **errp)
238 {
239     CPUArchId *found_cpu;
240     Error *local_err = NULL;
241     X86CPU *cpu = X86_CPU(dev);
242     X86MachineState *x86ms = X86_MACHINE(hotplug_dev);
243 
244     hotplug_handler_unplug(x86ms->acpi_dev, dev, &local_err);
245     if (local_err) {
246         goto out;
247     }
248 
249     found_cpu = x86_find_cpu_slot(MACHINE(x86ms), cpu->apic_id, NULL);
250     found_cpu->cpu = NULL;
251     qdev_unrealize(dev);
252 
253     /* decrement the number of CPUs */
254     x86ms->boot_cpus--;
255     /* Update the number of CPUs in CMOS */
256     x86_rtc_set_cpus_count(x86ms->rtc, x86ms->boot_cpus);
257     fw_cfg_modify_i16(x86ms->fw_cfg, FW_CFG_NB_CPUS, x86ms->boot_cpus);
258  out:
259     error_propagate(errp, local_err);
260 }
261 
262 void x86_cpu_pre_plug(HotplugHandler *hotplug_dev,
263                       DeviceState *dev, Error **errp)
264 {
265     int idx;
266     CPUState *cs;
267     CPUArchId *cpu_slot;
268     X86CPUTopoIDs topo_ids;
269     X86CPU *cpu = X86_CPU(dev);
270     CPUX86State *env = &cpu->env;
271     MachineState *ms = MACHINE(hotplug_dev);
272     X86MachineState *x86ms = X86_MACHINE(hotplug_dev);
273     unsigned int smp_cores = ms->smp.cores;
274     unsigned int smp_threads = ms->smp.threads;
275     X86CPUTopoInfo topo_info;
276 
277     if (!object_dynamic_cast(OBJECT(cpu), ms->cpu_type)) {
278         error_setg(errp, "Invalid CPU type, expected cpu type: '%s'",
279                    ms->cpu_type);
280         return;
281     }
282 
283     if (x86ms->acpi_dev) {
284         Error *local_err = NULL;
285 
286         hotplug_handler_pre_plug(HOTPLUG_HANDLER(x86ms->acpi_dev), dev,
287                                  &local_err);
288         if (local_err) {
289             error_propagate(errp, local_err);
290             return;
291         }
292     }
293 
294     init_topo_info(&topo_info, x86ms);
295 
296     env->nr_dies = x86ms->smp_dies;
297 
298     /*
299      * If APIC ID is not set,
300      * set it based on socket/die/core/thread properties.
301      */
302     if (cpu->apic_id == UNASSIGNED_APIC_ID) {
303         int max_socket = (ms->smp.max_cpus - 1) /
304                                 smp_threads / smp_cores / x86ms->smp_dies;
305 
306         /*
307          * die-id was optional in QEMU 4.0 and older, so keep it optional
308          * if there's only one die per socket.
309          */
310         if (cpu->die_id < 0 && x86ms->smp_dies == 1) {
311             cpu->die_id = 0;
312         }
313 
314         if (cpu->socket_id < 0) {
315             error_setg(errp, "CPU socket-id is not set");
316             return;
317         } else if (cpu->socket_id > max_socket) {
318             error_setg(errp, "Invalid CPU socket-id: %u must be in range 0:%u",
319                        cpu->socket_id, max_socket);
320             return;
321         }
322         if (cpu->die_id < 0) {
323             error_setg(errp, "CPU die-id is not set");
324             return;
325         } else if (cpu->die_id > x86ms->smp_dies - 1) {
326             error_setg(errp, "Invalid CPU die-id: %u must be in range 0:%u",
327                        cpu->die_id, x86ms->smp_dies - 1);
328             return;
329         }
330         if (cpu->core_id < 0) {
331             error_setg(errp, "CPU core-id is not set");
332             return;
333         } else if (cpu->core_id > (smp_cores - 1)) {
334             error_setg(errp, "Invalid CPU core-id: %u must be in range 0:%u",
335                        cpu->core_id, smp_cores - 1);
336             return;
337         }
338         if (cpu->thread_id < 0) {
339             error_setg(errp, "CPU thread-id is not set");
340             return;
341         } else if (cpu->thread_id > (smp_threads - 1)) {
342             error_setg(errp, "Invalid CPU thread-id: %u must be in range 0:%u",
343                        cpu->thread_id, smp_threads - 1);
344             return;
345         }
346 
347         topo_ids.pkg_id = cpu->socket_id;
348         topo_ids.die_id = cpu->die_id;
349         topo_ids.core_id = cpu->core_id;
350         topo_ids.smt_id = cpu->thread_id;
351         cpu->apic_id = x86_apicid_from_topo_ids(&topo_info, &topo_ids);
352     }
353 
354     cpu_slot = x86_find_cpu_slot(MACHINE(x86ms), cpu->apic_id, &idx);
355     if (!cpu_slot) {
356         MachineState *ms = MACHINE(x86ms);
357 
358         x86_topo_ids_from_apicid(cpu->apic_id, &topo_info, &topo_ids);
359         error_setg(errp,
360             "Invalid CPU [socket: %u, die: %u, core: %u, thread: %u] with"
361             " APIC ID %" PRIu32 ", valid index range 0:%d",
362             topo_ids.pkg_id, topo_ids.die_id, topo_ids.core_id, topo_ids.smt_id,
363             cpu->apic_id, ms->possible_cpus->len - 1);
364         return;
365     }
366 
367     if (cpu_slot->cpu) {
368         error_setg(errp, "CPU[%d] with APIC ID %" PRIu32 " exists",
369                    idx, cpu->apic_id);
370         return;
371     }
372 
373     /* if 'address' properties socket-id/core-id/thread-id are not set, set them
374      * so that machine_query_hotpluggable_cpus would show correct values
375      */
376     /* TODO: move socket_id/core_id/thread_id checks into x86_cpu_realizefn()
377      * once -smp refactoring is complete and there will be CPU private
378      * CPUState::nr_cores and CPUState::nr_threads fields instead of globals */
379     x86_topo_ids_from_apicid(cpu->apic_id, &topo_info, &topo_ids);
380     if (cpu->socket_id != -1 && cpu->socket_id != topo_ids.pkg_id) {
381         error_setg(errp, "property socket-id: %u doesn't match set apic-id:"
382             " 0x%x (socket-id: %u)", cpu->socket_id, cpu->apic_id,
383             topo_ids.pkg_id);
384         return;
385     }
386     cpu->socket_id = topo_ids.pkg_id;
387 
388     if (cpu->die_id != -1 && cpu->die_id != topo_ids.die_id) {
389         error_setg(errp, "property die-id: %u doesn't match set apic-id:"
390             " 0x%x (die-id: %u)", cpu->die_id, cpu->apic_id, topo_ids.die_id);
391         return;
392     }
393     cpu->die_id = topo_ids.die_id;
394 
395     if (cpu->core_id != -1 && cpu->core_id != topo_ids.core_id) {
396         error_setg(errp, "property core-id: %u doesn't match set apic-id:"
397             " 0x%x (core-id: %u)", cpu->core_id, cpu->apic_id,
398             topo_ids.core_id);
399         return;
400     }
401     cpu->core_id = topo_ids.core_id;
402 
403     if (cpu->thread_id != -1 && cpu->thread_id != topo_ids.smt_id) {
404         error_setg(errp, "property thread-id: %u doesn't match set apic-id:"
405             " 0x%x (thread-id: %u)", cpu->thread_id, cpu->apic_id,
406             topo_ids.smt_id);
407         return;
408     }
409     cpu->thread_id = topo_ids.smt_id;
410 
411     if (hyperv_feat_enabled(cpu, HYPERV_FEAT_VPINDEX) &&
412         !kvm_hv_vpindex_settable()) {
413         error_setg(errp, "kernel doesn't allow setting HyperV VP_INDEX");
414         return;
415     }
416 
417     cs = CPU(cpu);
418     cs->cpu_index = idx;
419 
420     numa_cpu_pre_plug(cpu_slot, dev, errp);
421 }
422 
423 CpuInstanceProperties
424 x86_cpu_index_to_props(MachineState *ms, unsigned cpu_index)
425 {
426     MachineClass *mc = MACHINE_GET_CLASS(ms);
427     const CPUArchIdList *possible_cpus = mc->possible_cpu_arch_ids(ms);
428 
429     assert(cpu_index < possible_cpus->len);
430     return possible_cpus->cpus[cpu_index].props;
431 }
432 
433 int64_t x86_get_default_cpu_node_id(const MachineState *ms, int idx)
434 {
435    X86CPUTopoIDs topo_ids;
436    X86MachineState *x86ms = X86_MACHINE(ms);
437    X86CPUTopoInfo topo_info;
438 
439    init_topo_info(&topo_info, x86ms);
440 
441    assert(idx < ms->possible_cpus->len);
442    x86_topo_ids_from_apicid(ms->possible_cpus->cpus[idx].arch_id,
443                             &topo_info, &topo_ids);
444    return topo_ids.pkg_id % ms->numa_state->num_nodes;
445 }
446 
447 const CPUArchIdList *x86_possible_cpu_arch_ids(MachineState *ms)
448 {
449     X86MachineState *x86ms = X86_MACHINE(ms);
450     unsigned int max_cpus = ms->smp.max_cpus;
451     X86CPUTopoInfo topo_info;
452     int i;
453 
454     if (ms->possible_cpus) {
455         /*
456          * make sure that max_cpus hasn't changed since the first use, i.e.
457          * -smp hasn't been parsed after it
458          */
459         assert(ms->possible_cpus->len == max_cpus);
460         return ms->possible_cpus;
461     }
462 
463     ms->possible_cpus = g_malloc0(sizeof(CPUArchIdList) +
464                                   sizeof(CPUArchId) * max_cpus);
465     ms->possible_cpus->len = max_cpus;
466 
467     init_topo_info(&topo_info, x86ms);
468 
469     for (i = 0; i < ms->possible_cpus->len; i++) {
470         X86CPUTopoIDs topo_ids;
471 
472         ms->possible_cpus->cpus[i].type = ms->cpu_type;
473         ms->possible_cpus->cpus[i].vcpus_count = 1;
474         ms->possible_cpus->cpus[i].arch_id =
475             x86_cpu_apic_id_from_index(x86ms, i);
476         x86_topo_ids_from_apicid(ms->possible_cpus->cpus[i].arch_id,
477                                  &topo_info, &topo_ids);
478         ms->possible_cpus->cpus[i].props.has_socket_id = true;
479         ms->possible_cpus->cpus[i].props.socket_id = topo_ids.pkg_id;
480         if (x86ms->smp_dies > 1) {
481             ms->possible_cpus->cpus[i].props.has_die_id = true;
482             ms->possible_cpus->cpus[i].props.die_id = topo_ids.die_id;
483         }
484         ms->possible_cpus->cpus[i].props.has_core_id = true;
485         ms->possible_cpus->cpus[i].props.core_id = topo_ids.core_id;
486         ms->possible_cpus->cpus[i].props.has_thread_id = true;
487         ms->possible_cpus->cpus[i].props.thread_id = topo_ids.smt_id;
488     }
489     return ms->possible_cpus;
490 }
491 
492 static void x86_nmi(NMIState *n, int cpu_index, Error **errp)
493 {
494     /* cpu index isn't used */
495     CPUState *cs;
496 
497     CPU_FOREACH(cs) {
498         X86CPU *cpu = X86_CPU(cs);
499 
500         if (!cpu->apic_state) {
501             cpu_interrupt(cs, CPU_INTERRUPT_NMI);
502         } else {
503             apic_deliver_nmi(cpu->apic_state);
504         }
505     }
506 }
507 
508 static long get_file_size(FILE *f)
509 {
510     long where, size;
511 
512     /* XXX: on Unix systems, using fstat() probably makes more sense */
513 
514     where = ftell(f);
515     fseek(f, 0, SEEK_END);
516     size = ftell(f);
517     fseek(f, where, SEEK_SET);
518 
519     return size;
520 }
521 
522 /* TSC handling */
523 uint64_t cpu_get_tsc(CPUX86State *env)
524 {
525     return cpus_get_elapsed_ticks();
526 }
527 
528 /* IRQ handling */
529 static void pic_irq_request(void *opaque, int irq, int level)
530 {
531     CPUState *cs = first_cpu;
532     X86CPU *cpu = X86_CPU(cs);
533 
534     trace_x86_pic_interrupt(irq, level);
535     if (cpu->apic_state && !kvm_irqchip_in_kernel() &&
536         !whpx_apic_in_platform()) {
537         CPU_FOREACH(cs) {
538             cpu = X86_CPU(cs);
539             if (apic_accept_pic_intr(cpu->apic_state)) {
540                 apic_deliver_pic_intr(cpu->apic_state, level);
541             }
542         }
543     } else {
544         if (level) {
545             cpu_interrupt(cs, CPU_INTERRUPT_HARD);
546         } else {
547             cpu_reset_interrupt(cs, CPU_INTERRUPT_HARD);
548         }
549     }
550 }
551 
552 qemu_irq x86_allocate_cpu_irq(void)
553 {
554     return qemu_allocate_irq(pic_irq_request, NULL, 0);
555 }
556 
557 int cpu_get_pic_interrupt(CPUX86State *env)
558 {
559     X86CPU *cpu = env_archcpu(env);
560     int intno;
561 
562     if (!kvm_irqchip_in_kernel() && !whpx_apic_in_platform()) {
563         intno = apic_get_interrupt(cpu->apic_state);
564         if (intno >= 0) {
565             return intno;
566         }
567         /* read the irq from the PIC */
568         if (!apic_accept_pic_intr(cpu->apic_state)) {
569             return -1;
570         }
571     }
572 
573     intno = pic_read_irq(isa_pic);
574     return intno;
575 }
576 
577 DeviceState *cpu_get_current_apic(void)
578 {
579     if (current_cpu) {
580         X86CPU *cpu = X86_CPU(current_cpu);
581         return cpu->apic_state;
582     } else {
583         return NULL;
584     }
585 }
586 
587 void gsi_handler(void *opaque, int n, int level)
588 {
589     GSIState *s = opaque;
590 
591     trace_x86_gsi_interrupt(n, level);
592     switch (n) {
593     case 0 ... ISA_NUM_IRQS - 1:
594         if (s->i8259_irq[n]) {
595             /* Under KVM, Kernel will forward to both PIC and IOAPIC */
596             qemu_set_irq(s->i8259_irq[n], level);
597         }
598         /* fall through */
599     case ISA_NUM_IRQS ... IOAPIC_NUM_PINS - 1:
600         qemu_set_irq(s->ioapic_irq[n], level);
601         break;
602     case IO_APIC_SECONDARY_IRQBASE
603         ... IO_APIC_SECONDARY_IRQBASE + IOAPIC_NUM_PINS - 1:
604         qemu_set_irq(s->ioapic2_irq[n - IO_APIC_SECONDARY_IRQBASE], level);
605         break;
606     }
607 }
608 
609 void ioapic_init_gsi(GSIState *gsi_state, const char *parent_name)
610 {
611     DeviceState *dev;
612     SysBusDevice *d;
613     unsigned int i;
614 
615     assert(parent_name);
616     if (kvm_ioapic_in_kernel()) {
617         dev = qdev_new(TYPE_KVM_IOAPIC);
618     } else {
619         dev = qdev_new(TYPE_IOAPIC);
620     }
621     object_property_add_child(object_resolve_path(parent_name, NULL),
622                               "ioapic", OBJECT(dev));
623     d = SYS_BUS_DEVICE(dev);
624     sysbus_realize_and_unref(d, &error_fatal);
625     sysbus_mmio_map(d, 0, IO_APIC_DEFAULT_ADDRESS);
626 
627     for (i = 0; i < IOAPIC_NUM_PINS; i++) {
628         gsi_state->ioapic_irq[i] = qdev_get_gpio_in(dev, i);
629     }
630 }
631 
632 DeviceState *ioapic_init_secondary(GSIState *gsi_state)
633 {
634     DeviceState *dev;
635     SysBusDevice *d;
636     unsigned int i;
637 
638     dev = qdev_new(TYPE_IOAPIC);
639     d = SYS_BUS_DEVICE(dev);
640     sysbus_realize_and_unref(d, &error_fatal);
641     sysbus_mmio_map(d, 0, IO_APIC_SECONDARY_ADDRESS);
642 
643     for (i = 0; i < IOAPIC_NUM_PINS; i++) {
644         gsi_state->ioapic2_irq[i] = qdev_get_gpio_in(dev, i);
645     }
646     return dev;
647 }
648 
649 struct setup_data {
650     uint64_t next;
651     uint32_t type;
652     uint32_t len;
653     uint8_t data[];
654 } __attribute__((packed));
655 
656 
657 /*
658  * The entry point into the kernel for PVH boot is different from
659  * the native entry point.  The PVH entry is defined by the x86/HVM
660  * direct boot ABI and is available in an ELFNOTE in the kernel binary.
661  *
662  * This function is passed to load_elf() when it is called from
663  * load_elfboot() which then additionally checks for an ELF Note of
664  * type XEN_ELFNOTE_PHYS32_ENTRY and passes it to this function to
665  * parse the PVH entry address from the ELF Note.
666  *
667  * Due to trickery in elf_opts.h, load_elf() is actually available as
668  * load_elf32() or load_elf64() and this routine needs to be able
669  * to deal with being called as 32 or 64 bit.
670  *
671  * The address of the PVH entry point is saved to the 'pvh_start_addr'
672  * global variable.  (although the entry point is 32-bit, the kernel
673  * binary can be either 32-bit or 64-bit).
674  */
675 static uint64_t read_pvh_start_addr(void *arg1, void *arg2, bool is64)
676 {
677     size_t *elf_note_data_addr;
678 
679     /* Check if ELF Note header passed in is valid */
680     if (arg1 == NULL) {
681         return 0;
682     }
683 
684     if (is64) {
685         struct elf64_note *nhdr64 = (struct elf64_note *)arg1;
686         uint64_t nhdr_size64 = sizeof(struct elf64_note);
687         uint64_t phdr_align = *(uint64_t *)arg2;
688         uint64_t nhdr_namesz = nhdr64->n_namesz;
689 
690         elf_note_data_addr =
691             ((void *)nhdr64) + nhdr_size64 +
692             QEMU_ALIGN_UP(nhdr_namesz, phdr_align);
693 
694         pvh_start_addr = *elf_note_data_addr;
695     } else {
696         struct elf32_note *nhdr32 = (struct elf32_note *)arg1;
697         uint32_t nhdr_size32 = sizeof(struct elf32_note);
698         uint32_t phdr_align = *(uint32_t *)arg2;
699         uint32_t nhdr_namesz = nhdr32->n_namesz;
700 
701         elf_note_data_addr =
702             ((void *)nhdr32) + nhdr_size32 +
703             QEMU_ALIGN_UP(nhdr_namesz, phdr_align);
704 
705         pvh_start_addr = *(uint32_t *)elf_note_data_addr;
706     }
707 
708     return pvh_start_addr;
709 }
710 
711 static bool load_elfboot(const char *kernel_filename,
712                          int kernel_file_size,
713                          uint8_t *header,
714                          size_t pvh_xen_start_addr,
715                          FWCfgState *fw_cfg)
716 {
717     uint32_t flags = 0;
718     uint32_t mh_load_addr = 0;
719     uint32_t elf_kernel_size = 0;
720     uint64_t elf_entry;
721     uint64_t elf_low, elf_high;
722     int kernel_size;
723 
724     if (ldl_p(header) != 0x464c457f) {
725         return false; /* no elfboot */
726     }
727 
728     bool elf_is64 = header[EI_CLASS] == ELFCLASS64;
729     flags = elf_is64 ?
730         ((Elf64_Ehdr *)header)->e_flags : ((Elf32_Ehdr *)header)->e_flags;
731 
732     if (flags & 0x00010004) { /* LOAD_ELF_HEADER_HAS_ADDR */
733         error_report("elfboot unsupported flags = %x", flags);
734         exit(1);
735     }
736 
737     uint64_t elf_note_type = XEN_ELFNOTE_PHYS32_ENTRY;
738     kernel_size = load_elf(kernel_filename, read_pvh_start_addr,
739                            NULL, &elf_note_type, &elf_entry,
740                            &elf_low, &elf_high, NULL, 0, I386_ELF_MACHINE,
741                            0, 0);
742 
743     if (kernel_size < 0) {
744         error_report("Error while loading elf kernel");
745         exit(1);
746     }
747     mh_load_addr = elf_low;
748     elf_kernel_size = elf_high - elf_low;
749 
750     if (pvh_start_addr == 0) {
751         error_report("Error loading uncompressed kernel without PVH ELF Note");
752         exit(1);
753     }
754     fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_ENTRY, pvh_start_addr);
755     fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_ADDR, mh_load_addr);
756     fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_SIZE, elf_kernel_size);
757 
758     return true;
759 }
760 
761 void x86_load_linux(X86MachineState *x86ms,
762                     FWCfgState *fw_cfg,
763                     int acpi_data_size,
764                     bool pvh_enabled,
765                     bool linuxboot_dma_enabled)
766 {
767     uint16_t protocol;
768     int setup_size, kernel_size, cmdline_size;
769     int dtb_size, setup_data_offset;
770     uint32_t initrd_max;
771     uint8_t header[8192], *setup, *kernel;
772     hwaddr real_addr, prot_addr, cmdline_addr, initrd_addr = 0;
773     FILE *f;
774     char *vmode;
775     MachineState *machine = MACHINE(x86ms);
776     struct setup_data *setup_data;
777     const char *kernel_filename = machine->kernel_filename;
778     const char *initrd_filename = machine->initrd_filename;
779     const char *dtb_filename = machine->dtb;
780     const char *kernel_cmdline = machine->kernel_cmdline;
781 
782     /* Align to 16 bytes as a paranoia measure */
783     cmdline_size = (strlen(kernel_cmdline) + 16) & ~15;
784 
785     /* load the kernel header */
786     f = fopen(kernel_filename, "rb");
787     if (!f) {
788         fprintf(stderr, "qemu: could not open kernel file '%s': %s\n",
789                 kernel_filename, strerror(errno));
790         exit(1);
791     }
792 
793     kernel_size = get_file_size(f);
794     if (!kernel_size ||
795         fread(header, 1, MIN(ARRAY_SIZE(header), kernel_size), f) !=
796         MIN(ARRAY_SIZE(header), kernel_size)) {
797         fprintf(stderr, "qemu: could not load kernel '%s': %s\n",
798                 kernel_filename, strerror(errno));
799         exit(1);
800     }
801 
802     /* kernel protocol version */
803     if (ldl_p(header + 0x202) == 0x53726448) {
804         protocol = lduw_p(header + 0x206);
805     } else {
806         /*
807          * This could be a multiboot kernel. If it is, let's stop treating it
808          * like a Linux kernel.
809          * Note: some multiboot images could be in the ELF format (the same of
810          * PVH), so we try multiboot first since we check the multiboot magic
811          * header before to load it.
812          */
813         if (load_multiboot(fw_cfg, f, kernel_filename, initrd_filename,
814                            kernel_cmdline, kernel_size, header)) {
815             return;
816         }
817         /*
818          * Check if the file is an uncompressed kernel file (ELF) and load it,
819          * saving the PVH entry point used by the x86/HVM direct boot ABI.
820          * If load_elfboot() is successful, populate the fw_cfg info.
821          */
822         if (pvh_enabled &&
823             load_elfboot(kernel_filename, kernel_size,
824                          header, pvh_start_addr, fw_cfg)) {
825             fclose(f);
826 
827             fw_cfg_add_i32(fw_cfg, FW_CFG_CMDLINE_SIZE,
828                 strlen(kernel_cmdline) + 1);
829             fw_cfg_add_string(fw_cfg, FW_CFG_CMDLINE_DATA, kernel_cmdline);
830 
831             fw_cfg_add_i32(fw_cfg, FW_CFG_SETUP_SIZE, sizeof(header));
832             fw_cfg_add_bytes(fw_cfg, FW_CFG_SETUP_DATA,
833                              header, sizeof(header));
834 
835             /* load initrd */
836             if (initrd_filename) {
837                 GMappedFile *mapped_file;
838                 gsize initrd_size;
839                 gchar *initrd_data;
840                 GError *gerr = NULL;
841 
842                 mapped_file = g_mapped_file_new(initrd_filename, false, &gerr);
843                 if (!mapped_file) {
844                     fprintf(stderr, "qemu: error reading initrd %s: %s\n",
845                             initrd_filename, gerr->message);
846                     exit(1);
847                 }
848                 x86ms->initrd_mapped_file = mapped_file;
849 
850                 initrd_data = g_mapped_file_get_contents(mapped_file);
851                 initrd_size = g_mapped_file_get_length(mapped_file);
852                 initrd_max = x86ms->below_4g_mem_size - acpi_data_size - 1;
853                 if (initrd_size >= initrd_max) {
854                     fprintf(stderr, "qemu: initrd is too large, cannot support."
855                             "(max: %"PRIu32", need %"PRId64")\n",
856                             initrd_max, (uint64_t)initrd_size);
857                     exit(1);
858                 }
859 
860                 initrd_addr = (initrd_max - initrd_size) & ~4095;
861 
862                 fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_ADDR, initrd_addr);
863                 fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_SIZE, initrd_size);
864                 fw_cfg_add_bytes(fw_cfg, FW_CFG_INITRD_DATA, initrd_data,
865                                  initrd_size);
866             }
867 
868             option_rom[nb_option_roms].bootindex = 0;
869             option_rom[nb_option_roms].name = "pvh.bin";
870             nb_option_roms++;
871 
872             return;
873         }
874         protocol = 0;
875     }
876 
877     if (protocol < 0x200 || !(header[0x211] & 0x01)) {
878         /* Low kernel */
879         real_addr    = 0x90000;
880         cmdline_addr = 0x9a000 - cmdline_size;
881         prot_addr    = 0x10000;
882     } else if (protocol < 0x202) {
883         /* High but ancient kernel */
884         real_addr    = 0x90000;
885         cmdline_addr = 0x9a000 - cmdline_size;
886         prot_addr    = 0x100000;
887     } else {
888         /* High and recent kernel */
889         real_addr    = 0x10000;
890         cmdline_addr = 0x20000;
891         prot_addr    = 0x100000;
892     }
893 
894     /* highest address for loading the initrd */
895     if (protocol >= 0x20c &&
896         lduw_p(header + 0x236) & XLF_CAN_BE_LOADED_ABOVE_4G) {
897         /*
898          * Linux has supported initrd up to 4 GB for a very long time (2007,
899          * long before XLF_CAN_BE_LOADED_ABOVE_4G which was added in 2013),
900          * though it only sets initrd_max to 2 GB to "work around bootloader
901          * bugs". Luckily, QEMU firmware(which does something like bootloader)
902          * has supported this.
903          *
904          * It's believed that if XLF_CAN_BE_LOADED_ABOVE_4G is set, initrd can
905          * be loaded into any address.
906          *
907          * In addition, initrd_max is uint32_t simply because QEMU doesn't
908          * support the 64-bit boot protocol (specifically the ext_ramdisk_image
909          * field).
910          *
911          * Therefore here just limit initrd_max to UINT32_MAX simply as well.
912          */
913         initrd_max = UINT32_MAX;
914     } else if (protocol >= 0x203) {
915         initrd_max = ldl_p(header + 0x22c);
916     } else {
917         initrd_max = 0x37ffffff;
918     }
919 
920     if (initrd_max >= x86ms->below_4g_mem_size - acpi_data_size) {
921         initrd_max = x86ms->below_4g_mem_size - acpi_data_size - 1;
922     }
923 
924     fw_cfg_add_i32(fw_cfg, FW_CFG_CMDLINE_ADDR, cmdline_addr);
925     fw_cfg_add_i32(fw_cfg, FW_CFG_CMDLINE_SIZE, strlen(kernel_cmdline) + 1);
926     fw_cfg_add_string(fw_cfg, FW_CFG_CMDLINE_DATA, kernel_cmdline);
927 
928     if (protocol >= 0x202) {
929         stl_p(header + 0x228, cmdline_addr);
930     } else {
931         stw_p(header + 0x20, 0xA33F);
932         stw_p(header + 0x22, cmdline_addr - real_addr);
933     }
934 
935     /* handle vga= parameter */
936     vmode = strstr(kernel_cmdline, "vga=");
937     if (vmode) {
938         unsigned int video_mode;
939         const char *end;
940         int ret;
941         /* skip "vga=" */
942         vmode += 4;
943         if (!strncmp(vmode, "normal", 6)) {
944             video_mode = 0xffff;
945         } else if (!strncmp(vmode, "ext", 3)) {
946             video_mode = 0xfffe;
947         } else if (!strncmp(vmode, "ask", 3)) {
948             video_mode = 0xfffd;
949         } else {
950             ret = qemu_strtoui(vmode, &end, 0, &video_mode);
951             if (ret != 0 || (*end && *end != ' ')) {
952                 fprintf(stderr, "qemu: invalid 'vga=' kernel parameter.\n");
953                 exit(1);
954             }
955         }
956         stw_p(header + 0x1fa, video_mode);
957     }
958 
959     /* loader type */
960     /*
961      * High nybble = B reserved for QEMU; low nybble is revision number.
962      * If this code is substantially changed, you may want to consider
963      * incrementing the revision.
964      */
965     if (protocol >= 0x200) {
966         header[0x210] = 0xB0;
967     }
968     /* heap */
969     if (protocol >= 0x201) {
970         header[0x211] |= 0x80; /* CAN_USE_HEAP */
971         stw_p(header + 0x224, cmdline_addr - real_addr - 0x200);
972     }
973 
974     /* load initrd */
975     if (initrd_filename) {
976         GMappedFile *mapped_file;
977         gsize initrd_size;
978         gchar *initrd_data;
979         GError *gerr = NULL;
980 
981         if (protocol < 0x200) {
982             fprintf(stderr, "qemu: linux kernel too old to load a ram disk\n");
983             exit(1);
984         }
985 
986         mapped_file = g_mapped_file_new(initrd_filename, false, &gerr);
987         if (!mapped_file) {
988             fprintf(stderr, "qemu: error reading initrd %s: %s\n",
989                     initrd_filename, gerr->message);
990             exit(1);
991         }
992         x86ms->initrd_mapped_file = mapped_file;
993 
994         initrd_data = g_mapped_file_get_contents(mapped_file);
995         initrd_size = g_mapped_file_get_length(mapped_file);
996         if (initrd_size >= initrd_max) {
997             fprintf(stderr, "qemu: initrd is too large, cannot support."
998                     "(max: %"PRIu32", need %"PRId64")\n",
999                     initrd_max, (uint64_t)initrd_size);
1000             exit(1);
1001         }
1002 
1003         initrd_addr = (initrd_max - initrd_size) & ~4095;
1004 
1005         fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_ADDR, initrd_addr);
1006         fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_SIZE, initrd_size);
1007         fw_cfg_add_bytes(fw_cfg, FW_CFG_INITRD_DATA, initrd_data, initrd_size);
1008 
1009         stl_p(header + 0x218, initrd_addr);
1010         stl_p(header + 0x21c, initrd_size);
1011     }
1012 
1013     /* load kernel and setup */
1014     setup_size = header[0x1f1];
1015     if (setup_size == 0) {
1016         setup_size = 4;
1017     }
1018     setup_size = (setup_size + 1) * 512;
1019     if (setup_size > kernel_size) {
1020         fprintf(stderr, "qemu: invalid kernel header\n");
1021         exit(1);
1022     }
1023     kernel_size -= setup_size;
1024 
1025     setup  = g_malloc(setup_size);
1026     kernel = g_malloc(kernel_size);
1027     fseek(f, 0, SEEK_SET);
1028     if (fread(setup, 1, setup_size, f) != setup_size) {
1029         fprintf(stderr, "fread() failed\n");
1030         exit(1);
1031     }
1032     if (fread(kernel, 1, kernel_size, f) != kernel_size) {
1033         fprintf(stderr, "fread() failed\n");
1034         exit(1);
1035     }
1036     fclose(f);
1037 
1038     /* append dtb to kernel */
1039     if (dtb_filename) {
1040         if (protocol < 0x209) {
1041             fprintf(stderr, "qemu: Linux kernel too old to load a dtb\n");
1042             exit(1);
1043         }
1044 
1045         dtb_size = get_image_size(dtb_filename);
1046         if (dtb_size <= 0) {
1047             fprintf(stderr, "qemu: error reading dtb %s: %s\n",
1048                     dtb_filename, strerror(errno));
1049             exit(1);
1050         }
1051 
1052         setup_data_offset = QEMU_ALIGN_UP(kernel_size, 16);
1053         kernel_size = setup_data_offset + sizeof(struct setup_data) + dtb_size;
1054         kernel = g_realloc(kernel, kernel_size);
1055 
1056         stq_p(header + 0x250, prot_addr + setup_data_offset);
1057 
1058         setup_data = (struct setup_data *)(kernel + setup_data_offset);
1059         setup_data->next = 0;
1060         setup_data->type = cpu_to_le32(SETUP_DTB);
1061         setup_data->len = cpu_to_le32(dtb_size);
1062 
1063         load_image_size(dtb_filename, setup_data->data, dtb_size);
1064     }
1065 
1066     memcpy(setup, header, MIN(sizeof(header), setup_size));
1067 
1068     fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_ADDR, prot_addr);
1069     fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_SIZE, kernel_size);
1070     fw_cfg_add_bytes(fw_cfg, FW_CFG_KERNEL_DATA, kernel, kernel_size);
1071 
1072     fw_cfg_add_i32(fw_cfg, FW_CFG_SETUP_ADDR, real_addr);
1073     fw_cfg_add_i32(fw_cfg, FW_CFG_SETUP_SIZE, setup_size);
1074     fw_cfg_add_bytes(fw_cfg, FW_CFG_SETUP_DATA, setup, setup_size);
1075 
1076     option_rom[nb_option_roms].bootindex = 0;
1077     option_rom[nb_option_roms].name = "linuxboot.bin";
1078     if (linuxboot_dma_enabled && fw_cfg_dma_enabled(fw_cfg)) {
1079         option_rom[nb_option_roms].name = "linuxboot_dma.bin";
1080     }
1081     nb_option_roms++;
1082 }
1083 
1084 void x86_bios_rom_init(MachineState *ms, const char *default_firmware,
1085                        MemoryRegion *rom_memory, bool isapc_ram_fw)
1086 {
1087     const char *bios_name;
1088     char *filename;
1089     MemoryRegion *bios, *isa_bios;
1090     int bios_size, isa_bios_size;
1091     int ret;
1092 
1093     /* BIOS load */
1094     bios_name = ms->firmware ?: default_firmware;
1095     filename = qemu_find_file(QEMU_FILE_TYPE_BIOS, bios_name);
1096     if (filename) {
1097         bios_size = get_image_size(filename);
1098     } else {
1099         bios_size = -1;
1100     }
1101     if (bios_size <= 0 ||
1102         (bios_size % 65536) != 0) {
1103         goto bios_error;
1104     }
1105     bios = g_malloc(sizeof(*bios));
1106     memory_region_init_ram(bios, NULL, "pc.bios", bios_size, &error_fatal);
1107     if (!isapc_ram_fw) {
1108         memory_region_set_readonly(bios, true);
1109     }
1110     ret = rom_add_file_fixed(bios_name, (uint32_t)(-bios_size), -1);
1111     if (ret != 0) {
1112     bios_error:
1113         fprintf(stderr, "qemu: could not load PC BIOS '%s'\n", bios_name);
1114         exit(1);
1115     }
1116     g_free(filename);
1117 
1118     /* map the last 128KB of the BIOS in ISA space */
1119     isa_bios_size = MIN(bios_size, 128 * KiB);
1120     isa_bios = g_malloc(sizeof(*isa_bios));
1121     memory_region_init_alias(isa_bios, NULL, "isa-bios", bios,
1122                              bios_size - isa_bios_size, isa_bios_size);
1123     memory_region_add_subregion_overlap(rom_memory,
1124                                         0x100000 - isa_bios_size,
1125                                         isa_bios,
1126                                         1);
1127     if (!isapc_ram_fw) {
1128         memory_region_set_readonly(isa_bios, true);
1129     }
1130 
1131     /* map all the bios at the top of memory */
1132     memory_region_add_subregion(rom_memory,
1133                                 (uint32_t)(-bios_size),
1134                                 bios);
1135 }
1136 
1137 bool x86_machine_is_smm_enabled(const X86MachineState *x86ms)
1138 {
1139     bool smm_available = false;
1140 
1141     if (x86ms->smm == ON_OFF_AUTO_OFF) {
1142         return false;
1143     }
1144 
1145     if (tcg_enabled() || qtest_enabled()) {
1146         smm_available = true;
1147     } else if (kvm_enabled()) {
1148         smm_available = kvm_has_smm();
1149     }
1150 
1151     if (smm_available) {
1152         return true;
1153     }
1154 
1155     if (x86ms->smm == ON_OFF_AUTO_ON) {
1156         error_report("System Management Mode not supported by this hypervisor.");
1157         exit(1);
1158     }
1159     return false;
1160 }
1161 
1162 static void x86_machine_get_smm(Object *obj, Visitor *v, const char *name,
1163                                void *opaque, Error **errp)
1164 {
1165     X86MachineState *x86ms = X86_MACHINE(obj);
1166     OnOffAuto smm = x86ms->smm;
1167 
1168     visit_type_OnOffAuto(v, name, &smm, errp);
1169 }
1170 
1171 static void x86_machine_set_smm(Object *obj, Visitor *v, const char *name,
1172                                void *opaque, Error **errp)
1173 {
1174     X86MachineState *x86ms = X86_MACHINE(obj);
1175 
1176     visit_type_OnOffAuto(v, name, &x86ms->smm, errp);
1177 }
1178 
1179 bool x86_machine_is_acpi_enabled(const X86MachineState *x86ms)
1180 {
1181     if (x86ms->acpi == ON_OFF_AUTO_OFF) {
1182         return false;
1183     }
1184     return true;
1185 }
1186 
1187 static void x86_machine_get_acpi(Object *obj, Visitor *v, const char *name,
1188                                  void *opaque, Error **errp)
1189 {
1190     X86MachineState *x86ms = X86_MACHINE(obj);
1191     OnOffAuto acpi = x86ms->acpi;
1192 
1193     visit_type_OnOffAuto(v, name, &acpi, errp);
1194 }
1195 
1196 static void x86_machine_set_acpi(Object *obj, Visitor *v, const char *name,
1197                                  void *opaque, Error **errp)
1198 {
1199     X86MachineState *x86ms = X86_MACHINE(obj);
1200 
1201     visit_type_OnOffAuto(v, name, &x86ms->acpi, errp);
1202 }
1203 
1204 static char *x86_machine_get_oem_id(Object *obj, Error **errp)
1205 {
1206     X86MachineState *x86ms = X86_MACHINE(obj);
1207 
1208     return g_strdup(x86ms->oem_id);
1209 }
1210 
1211 static void x86_machine_set_oem_id(Object *obj, const char *value, Error **errp)
1212 {
1213     X86MachineState *x86ms = X86_MACHINE(obj);
1214     size_t len = strlen(value);
1215 
1216     if (len > 6) {
1217         error_setg(errp,
1218                    "User specified "X86_MACHINE_OEM_ID" value is bigger than "
1219                    "6 bytes in size");
1220         return;
1221     }
1222 
1223     strncpy(x86ms->oem_id, value, 6);
1224 }
1225 
1226 static char *x86_machine_get_oem_table_id(Object *obj, Error **errp)
1227 {
1228     X86MachineState *x86ms = X86_MACHINE(obj);
1229 
1230     return g_strdup(x86ms->oem_table_id);
1231 }
1232 
1233 static void x86_machine_set_oem_table_id(Object *obj, const char *value,
1234                                          Error **errp)
1235 {
1236     X86MachineState *x86ms = X86_MACHINE(obj);
1237     size_t len = strlen(value);
1238 
1239     if (len > 8) {
1240         error_setg(errp,
1241                    "User specified "X86_MACHINE_OEM_TABLE_ID
1242                    " value is bigger than "
1243                    "8 bytes in size");
1244         return;
1245     }
1246     strncpy(x86ms->oem_table_id, value, 8);
1247 }
1248 
1249 static void x86_machine_get_bus_lock_ratelimit(Object *obj, Visitor *v,
1250                                 const char *name, void *opaque, Error **errp)
1251 {
1252     X86MachineState *x86ms = X86_MACHINE(obj);
1253     uint64_t bus_lock_ratelimit = x86ms->bus_lock_ratelimit;
1254 
1255     visit_type_uint64(v, name, &bus_lock_ratelimit, errp);
1256 }
1257 
1258 static void x86_machine_set_bus_lock_ratelimit(Object *obj, Visitor *v,
1259                                const char *name, void *opaque, Error **errp)
1260 {
1261     X86MachineState *x86ms = X86_MACHINE(obj);
1262 
1263     visit_type_uint64(v, name, &x86ms->bus_lock_ratelimit, errp);
1264 }
1265 
1266 static void x86_machine_initfn(Object *obj)
1267 {
1268     X86MachineState *x86ms = X86_MACHINE(obj);
1269 
1270     x86ms->smm = ON_OFF_AUTO_AUTO;
1271     x86ms->acpi = ON_OFF_AUTO_AUTO;
1272     x86ms->smp_dies = 1;
1273     x86ms->pci_irq_mask = ACPI_BUILD_PCI_IRQS;
1274     x86ms->oem_id = g_strndup(ACPI_BUILD_APPNAME6, 6);
1275     x86ms->oem_table_id = g_strndup(ACPI_BUILD_APPNAME8, 8);
1276     x86ms->bus_lock_ratelimit = 0;
1277 }
1278 
1279 static void x86_machine_class_init(ObjectClass *oc, void *data)
1280 {
1281     MachineClass *mc = MACHINE_CLASS(oc);
1282     X86MachineClass *x86mc = X86_MACHINE_CLASS(oc);
1283     NMIClass *nc = NMI_CLASS(oc);
1284 
1285     mc->cpu_index_to_instance_props = x86_cpu_index_to_props;
1286     mc->get_default_cpu_node_id = x86_get_default_cpu_node_id;
1287     mc->possible_cpu_arch_ids = x86_possible_cpu_arch_ids;
1288     x86mc->compat_apic_id_mode = false;
1289     x86mc->save_tsc_khz = true;
1290     nc->nmi_monitor_handler = x86_nmi;
1291 
1292     object_class_property_add(oc, X86_MACHINE_SMM, "OnOffAuto",
1293         x86_machine_get_smm, x86_machine_set_smm,
1294         NULL, NULL);
1295     object_class_property_set_description(oc, X86_MACHINE_SMM,
1296         "Enable SMM");
1297 
1298     object_class_property_add(oc, X86_MACHINE_ACPI, "OnOffAuto",
1299         x86_machine_get_acpi, x86_machine_set_acpi,
1300         NULL, NULL);
1301     object_class_property_set_description(oc, X86_MACHINE_ACPI,
1302         "Enable ACPI");
1303 
1304     object_class_property_add_str(oc, X86_MACHINE_OEM_ID,
1305                                   x86_machine_get_oem_id,
1306                                   x86_machine_set_oem_id);
1307     object_class_property_set_description(oc, X86_MACHINE_OEM_ID,
1308                                           "Override the default value of field OEMID "
1309                                           "in ACPI table header."
1310                                           "The string may be up to 6 bytes in size");
1311 
1312 
1313     object_class_property_add_str(oc, X86_MACHINE_OEM_TABLE_ID,
1314                                   x86_machine_get_oem_table_id,
1315                                   x86_machine_set_oem_table_id);
1316     object_class_property_set_description(oc, X86_MACHINE_OEM_TABLE_ID,
1317                                           "Override the default value of field OEM Table ID "
1318                                           "in ACPI table header."
1319                                           "The string may be up to 8 bytes in size");
1320 
1321     object_class_property_add(oc, X86_MACHINE_BUS_LOCK_RATELIMIT, "uint64_t",
1322                                 x86_machine_get_bus_lock_ratelimit,
1323                                 x86_machine_set_bus_lock_ratelimit, NULL, NULL);
1324     object_class_property_set_description(oc, X86_MACHINE_BUS_LOCK_RATELIMIT,
1325             "Set the ratelimit for the bus locks acquired in VMs");
1326 }
1327 
1328 static const TypeInfo x86_machine_info = {
1329     .name = TYPE_X86_MACHINE,
1330     .parent = TYPE_MACHINE,
1331     .abstract = true,
1332     .instance_size = sizeof(X86MachineState),
1333     .instance_init = x86_machine_initfn,
1334     .class_size = sizeof(X86MachineClass),
1335     .class_init = x86_machine_class_init,
1336     .interfaces = (InterfaceInfo[]) {
1337          { TYPE_NMI },
1338          { }
1339     },
1340 };
1341 
1342 static void x86_machine_register_types(void)
1343 {
1344     type_register_static(&x86_machine_info);
1345 }
1346 
1347 type_init(x86_machine_register_types)
1348