xref: /openbmc/qemu/hw/i386/x86.c (revision bad5cfcd)
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/datadir.h"
29 #include "qapi/error.h"
30 #include "qapi/qapi-visit-common.h"
31 #include "qapi/clone-visitor.h"
32 #include "qapi/qapi-visit-machine.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 "sysemu/xen.h"
41 #include "trace.h"
42 
43 #include "hw/i386/x86.h"
44 #include "target/i386/cpu.h"
45 #include "hw/i386/topology.h"
46 #include "hw/i386/fw_cfg.h"
47 #include "hw/intc/i8259.h"
48 #include "hw/rtc/mc146818rtc.h"
49 #include "target/i386/sev.h"
50 
51 #include "hw/acpi/cpu_hotplug.h"
52 #include "hw/irq.h"
53 #include "hw/nmi.h"
54 #include "hw/loader.h"
55 #include "multiboot.h"
56 #include "elf.h"
57 #include "standard-headers/asm-x86/bootparam.h"
58 #include CONFIG_DEVICES
59 #include "kvm/kvm_i386.h"
60 
61 #ifdef CONFIG_XEN_EMU
62 #include "hw/xen/xen.h"
63 #include "hw/i386/kvm/xen_evtchn.h"
64 #endif
65 
66 /* Physical Address of PVH entry point read from kernel ELF NOTE */
67 static size_t pvh_start_addr;
68 
69 static void init_topo_info(X86CPUTopoInfo *topo_info,
70                            const X86MachineState *x86ms)
71 {
72     MachineState *ms = MACHINE(x86ms);
73 
74     topo_info->dies_per_pkg = ms->smp.dies;
75     topo_info->cores_per_die = ms->smp.cores;
76     topo_info->threads_per_core = ms->smp.threads;
77 }
78 
79 /*
80  * Calculates initial APIC ID for a specific CPU index
81  *
82  * Currently we need to be able to calculate the APIC ID from the CPU index
83  * alone (without requiring a CPU object), as the QEMU<->Seabios interfaces have
84  * no concept of "CPU index", and the NUMA tables on fw_cfg need the APIC ID of
85  * all CPUs up to max_cpus.
86  */
87 uint32_t x86_cpu_apic_id_from_index(X86MachineState *x86ms,
88                                     unsigned int cpu_index)
89 {
90     X86CPUTopoInfo topo_info;
91 
92     init_topo_info(&topo_info, x86ms);
93 
94     return x86_apicid_from_cpu_idx(&topo_info, cpu_index);
95 }
96 
97 
98 void x86_cpu_new(X86MachineState *x86ms, int64_t apic_id, Error **errp)
99 {
100     Object *cpu = object_new(MACHINE(x86ms)->cpu_type);
101 
102     if (!object_property_set_uint(cpu, "apic-id", apic_id, errp)) {
103         goto out;
104     }
105     qdev_realize(DEVICE(cpu), NULL, errp);
106 
107 out:
108     object_unref(cpu);
109 }
110 
111 void x86_cpus_init(X86MachineState *x86ms, int default_cpu_version)
112 {
113     int i;
114     const CPUArchIdList *possible_cpus;
115     MachineState *ms = MACHINE(x86ms);
116     MachineClass *mc = MACHINE_GET_CLASS(x86ms);
117 
118     x86_cpu_set_default_version(default_cpu_version);
119 
120     /*
121      * Calculates the limit to CPU APIC ID values
122      *
123      * Limit for the APIC ID value, so that all
124      * CPU APIC IDs are < x86ms->apic_id_limit.
125      *
126      * This is used for FW_CFG_MAX_CPUS. See comments on fw_cfg_arch_create().
127      */
128     x86ms->apic_id_limit = x86_cpu_apic_id_from_index(x86ms,
129                                                       ms->smp.max_cpus - 1) + 1;
130 
131     /*
132      * Can we support APIC ID 255 or higher?  With KVM, that requires
133      * both in-kernel lapic and X2APIC userspace API.
134      */
135     if (x86ms->apic_id_limit > 255 && kvm_enabled() &&
136         (!kvm_irqchip_in_kernel() || !kvm_enable_x2apic())) {
137         error_report("current -smp configuration requires kernel "
138                      "irqchip and X2APIC API support.");
139         exit(EXIT_FAILURE);
140     }
141 
142     if (kvm_enabled()) {
143         kvm_set_max_apic_id(x86ms->apic_id_limit);
144     }
145 
146     possible_cpus = mc->possible_cpu_arch_ids(ms);
147     for (i = 0; i < ms->smp.cpus; i++) {
148         x86_cpu_new(x86ms, possible_cpus->cpus[i].arch_id, &error_fatal);
149     }
150 }
151 
152 void x86_rtc_set_cpus_count(ISADevice *s, uint16_t cpus_count)
153 {
154     MC146818RtcState *rtc = MC146818_RTC(s);
155 
156     if (cpus_count > 0xff) {
157         /*
158          * If the number of CPUs can't be represented in 8 bits, the
159          * BIOS must use "FW_CFG_NB_CPUS". Set RTC field to 0 just
160          * to make old BIOSes fail more predictably.
161          */
162         mc146818rtc_set_cmos_data(rtc, 0x5f, 0);
163     } else {
164         mc146818rtc_set_cmos_data(rtc, 0x5f, cpus_count - 1);
165     }
166 }
167 
168 static int x86_apic_cmp(const void *a, const void *b)
169 {
170    CPUArchId *apic_a = (CPUArchId *)a;
171    CPUArchId *apic_b = (CPUArchId *)b;
172 
173    return apic_a->arch_id - apic_b->arch_id;
174 }
175 
176 /*
177  * returns pointer to CPUArchId descriptor that matches CPU's apic_id
178  * in ms->possible_cpus->cpus, if ms->possible_cpus->cpus has no
179  * entry corresponding to CPU's apic_id returns NULL.
180  */
181 CPUArchId *x86_find_cpu_slot(MachineState *ms, uint32_t id, int *idx)
182 {
183     CPUArchId apic_id, *found_cpu;
184 
185     apic_id.arch_id = id;
186     found_cpu = bsearch(&apic_id, ms->possible_cpus->cpus,
187         ms->possible_cpus->len, sizeof(*ms->possible_cpus->cpus),
188         x86_apic_cmp);
189     if (found_cpu && idx) {
190         *idx = found_cpu - ms->possible_cpus->cpus;
191     }
192     return found_cpu;
193 }
194 
195 void x86_cpu_plug(HotplugHandler *hotplug_dev,
196                   DeviceState *dev, Error **errp)
197 {
198     CPUArchId *found_cpu;
199     Error *local_err = NULL;
200     X86CPU *cpu = X86_CPU(dev);
201     X86MachineState *x86ms = X86_MACHINE(hotplug_dev);
202 
203     if (x86ms->acpi_dev) {
204         hotplug_handler_plug(x86ms->acpi_dev, dev, &local_err);
205         if (local_err) {
206             goto out;
207         }
208     }
209 
210     /* increment the number of CPUs */
211     x86ms->boot_cpus++;
212     if (x86ms->rtc) {
213         x86_rtc_set_cpus_count(x86ms->rtc, x86ms->boot_cpus);
214     }
215     if (x86ms->fw_cfg) {
216         fw_cfg_modify_i16(x86ms->fw_cfg, FW_CFG_NB_CPUS, x86ms->boot_cpus);
217     }
218 
219     found_cpu = x86_find_cpu_slot(MACHINE(x86ms), cpu->apic_id, NULL);
220     found_cpu->cpu = OBJECT(dev);
221 out:
222     error_propagate(errp, local_err);
223 }
224 
225 void x86_cpu_unplug_request_cb(HotplugHandler *hotplug_dev,
226                                DeviceState *dev, Error **errp)
227 {
228     int idx = -1;
229     X86CPU *cpu = X86_CPU(dev);
230     X86MachineState *x86ms = X86_MACHINE(hotplug_dev);
231 
232     if (!x86ms->acpi_dev) {
233         error_setg(errp, "CPU hot unplug not supported without ACPI");
234         return;
235     }
236 
237     x86_find_cpu_slot(MACHINE(x86ms), cpu->apic_id, &idx);
238     assert(idx != -1);
239     if (idx == 0) {
240         error_setg(errp, "Boot CPU is unpluggable");
241         return;
242     }
243 
244     hotplug_handler_unplug_request(x86ms->acpi_dev, dev,
245                                    errp);
246 }
247 
248 void x86_cpu_unplug_cb(HotplugHandler *hotplug_dev,
249                        DeviceState *dev, Error **errp)
250 {
251     CPUArchId *found_cpu;
252     Error *local_err = NULL;
253     X86CPU *cpu = X86_CPU(dev);
254     X86MachineState *x86ms = X86_MACHINE(hotplug_dev);
255 
256     hotplug_handler_unplug(x86ms->acpi_dev, dev, &local_err);
257     if (local_err) {
258         goto out;
259     }
260 
261     found_cpu = x86_find_cpu_slot(MACHINE(x86ms), cpu->apic_id, NULL);
262     found_cpu->cpu = NULL;
263     qdev_unrealize(dev);
264 
265     /* decrement the number of CPUs */
266     x86ms->boot_cpus--;
267     /* Update the number of CPUs in CMOS */
268     x86_rtc_set_cpus_count(x86ms->rtc, x86ms->boot_cpus);
269     fw_cfg_modify_i16(x86ms->fw_cfg, FW_CFG_NB_CPUS, x86ms->boot_cpus);
270  out:
271     error_propagate(errp, local_err);
272 }
273 
274 void x86_cpu_pre_plug(HotplugHandler *hotplug_dev,
275                       DeviceState *dev, Error **errp)
276 {
277     int idx;
278     CPUState *cs;
279     CPUArchId *cpu_slot;
280     X86CPUTopoIDs topo_ids;
281     X86CPU *cpu = X86_CPU(dev);
282     CPUX86State *env = &cpu->env;
283     MachineState *ms = MACHINE(hotplug_dev);
284     X86MachineState *x86ms = X86_MACHINE(hotplug_dev);
285     unsigned int smp_cores = ms->smp.cores;
286     unsigned int smp_threads = ms->smp.threads;
287     X86CPUTopoInfo topo_info;
288 
289     if (!object_dynamic_cast(OBJECT(cpu), ms->cpu_type)) {
290         error_setg(errp, "Invalid CPU type, expected cpu type: '%s'",
291                    ms->cpu_type);
292         return;
293     }
294 
295     if (x86ms->acpi_dev) {
296         Error *local_err = NULL;
297 
298         hotplug_handler_pre_plug(HOTPLUG_HANDLER(x86ms->acpi_dev), dev,
299                                  &local_err);
300         if (local_err) {
301             error_propagate(errp, local_err);
302             return;
303         }
304     }
305 
306     init_topo_info(&topo_info, x86ms);
307 
308     env->nr_dies = ms->smp.dies;
309 
310     /*
311      * If APIC ID is not set,
312      * set it based on socket/die/core/thread properties.
313      */
314     if (cpu->apic_id == UNASSIGNED_APIC_ID) {
315         int max_socket = (ms->smp.max_cpus - 1) /
316                                 smp_threads / smp_cores / ms->smp.dies;
317 
318         /*
319          * die-id was optional in QEMU 4.0 and older, so keep it optional
320          * if there's only one die per socket.
321          */
322         if (cpu->die_id < 0 && ms->smp.dies == 1) {
323             cpu->die_id = 0;
324         }
325 
326         if (cpu->socket_id < 0) {
327             error_setg(errp, "CPU socket-id is not set");
328             return;
329         } else if (cpu->socket_id > max_socket) {
330             error_setg(errp, "Invalid CPU socket-id: %u must be in range 0:%u",
331                        cpu->socket_id, max_socket);
332             return;
333         }
334         if (cpu->die_id < 0) {
335             error_setg(errp, "CPU die-id is not set");
336             return;
337         } else if (cpu->die_id > ms->smp.dies - 1) {
338             error_setg(errp, "Invalid CPU die-id: %u must be in range 0:%u",
339                        cpu->die_id, ms->smp.dies - 1);
340             return;
341         }
342         if (cpu->core_id < 0) {
343             error_setg(errp, "CPU core-id is not set");
344             return;
345         } else if (cpu->core_id > (smp_cores - 1)) {
346             error_setg(errp, "Invalid CPU core-id: %u must be in range 0:%u",
347                        cpu->core_id, smp_cores - 1);
348             return;
349         }
350         if (cpu->thread_id < 0) {
351             error_setg(errp, "CPU thread-id is not set");
352             return;
353         } else if (cpu->thread_id > (smp_threads - 1)) {
354             error_setg(errp, "Invalid CPU thread-id: %u must be in range 0:%u",
355                        cpu->thread_id, smp_threads - 1);
356             return;
357         }
358 
359         topo_ids.pkg_id = cpu->socket_id;
360         topo_ids.die_id = cpu->die_id;
361         topo_ids.core_id = cpu->core_id;
362         topo_ids.smt_id = cpu->thread_id;
363         cpu->apic_id = x86_apicid_from_topo_ids(&topo_info, &topo_ids);
364     }
365 
366     cpu_slot = x86_find_cpu_slot(MACHINE(x86ms), cpu->apic_id, &idx);
367     if (!cpu_slot) {
368         MachineState *ms = MACHINE(x86ms);
369 
370         x86_topo_ids_from_apicid(cpu->apic_id, &topo_info, &topo_ids);
371         error_setg(errp,
372             "Invalid CPU [socket: %u, die: %u, core: %u, thread: %u] with"
373             " APIC ID %" PRIu32 ", valid index range 0:%d",
374             topo_ids.pkg_id, topo_ids.die_id, topo_ids.core_id, topo_ids.smt_id,
375             cpu->apic_id, ms->possible_cpus->len - 1);
376         return;
377     }
378 
379     if (cpu_slot->cpu) {
380         error_setg(errp, "CPU[%d] with APIC ID %" PRIu32 " exists",
381                    idx, cpu->apic_id);
382         return;
383     }
384 
385     /* if 'address' properties socket-id/core-id/thread-id are not set, set them
386      * so that machine_query_hotpluggable_cpus would show correct values
387      */
388     /* TODO: move socket_id/core_id/thread_id checks into x86_cpu_realizefn()
389      * once -smp refactoring is complete and there will be CPU private
390      * CPUState::nr_cores and CPUState::nr_threads fields instead of globals */
391     x86_topo_ids_from_apicid(cpu->apic_id, &topo_info, &topo_ids);
392     if (cpu->socket_id != -1 && cpu->socket_id != topo_ids.pkg_id) {
393         error_setg(errp, "property socket-id: %u doesn't match set apic-id:"
394             " 0x%x (socket-id: %u)", cpu->socket_id, cpu->apic_id,
395             topo_ids.pkg_id);
396         return;
397     }
398     cpu->socket_id = topo_ids.pkg_id;
399 
400     if (cpu->die_id != -1 && cpu->die_id != topo_ids.die_id) {
401         error_setg(errp, "property die-id: %u doesn't match set apic-id:"
402             " 0x%x (die-id: %u)", cpu->die_id, cpu->apic_id, topo_ids.die_id);
403         return;
404     }
405     cpu->die_id = topo_ids.die_id;
406 
407     if (cpu->core_id != -1 && cpu->core_id != topo_ids.core_id) {
408         error_setg(errp, "property core-id: %u doesn't match set apic-id:"
409             " 0x%x (core-id: %u)", cpu->core_id, cpu->apic_id,
410             topo_ids.core_id);
411         return;
412     }
413     cpu->core_id = topo_ids.core_id;
414 
415     if (cpu->thread_id != -1 && cpu->thread_id != topo_ids.smt_id) {
416         error_setg(errp, "property thread-id: %u doesn't match set apic-id:"
417             " 0x%x (thread-id: %u)", cpu->thread_id, cpu->apic_id,
418             topo_ids.smt_id);
419         return;
420     }
421     cpu->thread_id = topo_ids.smt_id;
422 
423     if (hyperv_feat_enabled(cpu, HYPERV_FEAT_VPINDEX) &&
424         kvm_enabled() && !kvm_hv_vpindex_settable()) {
425         error_setg(errp, "kernel doesn't allow setting HyperV VP_INDEX");
426         return;
427     }
428 
429     cs = CPU(cpu);
430     cs->cpu_index = idx;
431 
432     numa_cpu_pre_plug(cpu_slot, dev, errp);
433 }
434 
435 CpuInstanceProperties
436 x86_cpu_index_to_props(MachineState *ms, unsigned cpu_index)
437 {
438     MachineClass *mc = MACHINE_GET_CLASS(ms);
439     const CPUArchIdList *possible_cpus = mc->possible_cpu_arch_ids(ms);
440 
441     assert(cpu_index < possible_cpus->len);
442     return possible_cpus->cpus[cpu_index].props;
443 }
444 
445 int64_t x86_get_default_cpu_node_id(const MachineState *ms, int idx)
446 {
447    X86CPUTopoIDs topo_ids;
448    X86MachineState *x86ms = X86_MACHINE(ms);
449    X86CPUTopoInfo topo_info;
450 
451    init_topo_info(&topo_info, x86ms);
452 
453    assert(idx < ms->possible_cpus->len);
454    x86_topo_ids_from_apicid(ms->possible_cpus->cpus[idx].arch_id,
455                             &topo_info, &topo_ids);
456    return topo_ids.pkg_id % ms->numa_state->num_nodes;
457 }
458 
459 const CPUArchIdList *x86_possible_cpu_arch_ids(MachineState *ms)
460 {
461     X86MachineState *x86ms = X86_MACHINE(ms);
462     unsigned int max_cpus = ms->smp.max_cpus;
463     X86CPUTopoInfo topo_info;
464     int i;
465 
466     if (ms->possible_cpus) {
467         /*
468          * make sure that max_cpus hasn't changed since the first use, i.e.
469          * -smp hasn't been parsed after it
470          */
471         assert(ms->possible_cpus->len == max_cpus);
472         return ms->possible_cpus;
473     }
474 
475     ms->possible_cpus = g_malloc0(sizeof(CPUArchIdList) +
476                                   sizeof(CPUArchId) * max_cpus);
477     ms->possible_cpus->len = max_cpus;
478 
479     init_topo_info(&topo_info, x86ms);
480 
481     for (i = 0; i < ms->possible_cpus->len; i++) {
482         X86CPUTopoIDs topo_ids;
483 
484         ms->possible_cpus->cpus[i].type = ms->cpu_type;
485         ms->possible_cpus->cpus[i].vcpus_count = 1;
486         ms->possible_cpus->cpus[i].arch_id =
487             x86_cpu_apic_id_from_index(x86ms, i);
488         x86_topo_ids_from_apicid(ms->possible_cpus->cpus[i].arch_id,
489                                  &topo_info, &topo_ids);
490         ms->possible_cpus->cpus[i].props.has_socket_id = true;
491         ms->possible_cpus->cpus[i].props.socket_id = topo_ids.pkg_id;
492         if (ms->smp.dies > 1) {
493             ms->possible_cpus->cpus[i].props.has_die_id = true;
494             ms->possible_cpus->cpus[i].props.die_id = topo_ids.die_id;
495         }
496         ms->possible_cpus->cpus[i].props.has_core_id = true;
497         ms->possible_cpus->cpus[i].props.core_id = topo_ids.core_id;
498         ms->possible_cpus->cpus[i].props.has_thread_id = true;
499         ms->possible_cpus->cpus[i].props.thread_id = topo_ids.smt_id;
500     }
501     return ms->possible_cpus;
502 }
503 
504 static void x86_nmi(NMIState *n, int cpu_index, Error **errp)
505 {
506     /* cpu index isn't used */
507     CPUState *cs;
508 
509     CPU_FOREACH(cs) {
510         X86CPU *cpu = X86_CPU(cs);
511 
512         if (!cpu->apic_state) {
513             cpu_interrupt(cs, CPU_INTERRUPT_NMI);
514         } else {
515             apic_deliver_nmi(cpu->apic_state);
516         }
517     }
518 }
519 
520 static long get_file_size(FILE *f)
521 {
522     long where, size;
523 
524     /* XXX: on Unix systems, using fstat() probably makes more sense */
525 
526     where = ftell(f);
527     fseek(f, 0, SEEK_END);
528     size = ftell(f);
529     fseek(f, where, SEEK_SET);
530 
531     return size;
532 }
533 
534 /* TSC handling */
535 uint64_t cpu_get_tsc(CPUX86State *env)
536 {
537     return cpus_get_elapsed_ticks();
538 }
539 
540 /* IRQ handling */
541 static void pic_irq_request(void *opaque, int irq, int level)
542 {
543     CPUState *cs = first_cpu;
544     X86CPU *cpu = X86_CPU(cs);
545 
546     trace_x86_pic_interrupt(irq, level);
547     if (cpu->apic_state && !kvm_irqchip_in_kernel() &&
548         !whpx_apic_in_platform()) {
549         CPU_FOREACH(cs) {
550             cpu = X86_CPU(cs);
551             if (apic_accept_pic_intr(cpu->apic_state)) {
552                 apic_deliver_pic_intr(cpu->apic_state, level);
553             }
554         }
555     } else {
556         if (level) {
557             cpu_interrupt(cs, CPU_INTERRUPT_HARD);
558         } else {
559             cpu_reset_interrupt(cs, CPU_INTERRUPT_HARD);
560         }
561     }
562 }
563 
564 qemu_irq x86_allocate_cpu_irq(void)
565 {
566     return qemu_allocate_irq(pic_irq_request, NULL, 0);
567 }
568 
569 int cpu_get_pic_interrupt(CPUX86State *env)
570 {
571     X86CPU *cpu = env_archcpu(env);
572     int intno;
573 
574     if (!kvm_irqchip_in_kernel() && !whpx_apic_in_platform()) {
575         intno = apic_get_interrupt(cpu->apic_state);
576         if (intno >= 0) {
577             return intno;
578         }
579         /* read the irq from the PIC */
580         if (!apic_accept_pic_intr(cpu->apic_state)) {
581             return -1;
582         }
583     }
584 
585     intno = pic_read_irq(isa_pic);
586     return intno;
587 }
588 
589 DeviceState *cpu_get_current_apic(void)
590 {
591     if (current_cpu) {
592         X86CPU *cpu = X86_CPU(current_cpu);
593         return cpu->apic_state;
594     } else {
595         return NULL;
596     }
597 }
598 
599 void gsi_handler(void *opaque, int n, int level)
600 {
601     GSIState *s = opaque;
602 
603     trace_x86_gsi_interrupt(n, level);
604     switch (n) {
605     case 0 ... ISA_NUM_IRQS - 1:
606         if (s->i8259_irq[n]) {
607             /* Under KVM, Kernel will forward to both PIC and IOAPIC */
608             qemu_set_irq(s->i8259_irq[n], level);
609         }
610         /* fall through */
611     case ISA_NUM_IRQS ... IOAPIC_NUM_PINS - 1:
612 #ifdef CONFIG_XEN_EMU
613         /*
614          * Xen delivers the GSI to the Legacy PIC (not that Legacy PIC
615          * routing actually works properly under Xen). And then to
616          * *either* the PIRQ handling or the I/OAPIC depending on
617          * whether the former wants it.
618          */
619         if (xen_mode == XEN_EMULATE && xen_evtchn_set_gsi(n, level)) {
620             break;
621         }
622 #endif
623         qemu_set_irq(s->ioapic_irq[n], level);
624         break;
625     case IO_APIC_SECONDARY_IRQBASE
626         ... IO_APIC_SECONDARY_IRQBASE + IOAPIC_NUM_PINS - 1:
627         qemu_set_irq(s->ioapic2_irq[n - IO_APIC_SECONDARY_IRQBASE], level);
628         break;
629     }
630 }
631 
632 void ioapic_init_gsi(GSIState *gsi_state, const char *parent_name)
633 {
634     DeviceState *dev;
635     SysBusDevice *d;
636     unsigned int i;
637 
638     assert(parent_name);
639     if (kvm_ioapic_in_kernel()) {
640         dev = qdev_new(TYPE_KVM_IOAPIC);
641     } else {
642         dev = qdev_new(TYPE_IOAPIC);
643     }
644     object_property_add_child(object_resolve_path(parent_name, NULL),
645                               "ioapic", OBJECT(dev));
646     d = SYS_BUS_DEVICE(dev);
647     sysbus_realize_and_unref(d, &error_fatal);
648     sysbus_mmio_map(d, 0, IO_APIC_DEFAULT_ADDRESS);
649 
650     for (i = 0; i < IOAPIC_NUM_PINS; i++) {
651         gsi_state->ioapic_irq[i] = qdev_get_gpio_in(dev, i);
652     }
653 }
654 
655 DeviceState *ioapic_init_secondary(GSIState *gsi_state)
656 {
657     DeviceState *dev;
658     SysBusDevice *d;
659     unsigned int i;
660 
661     dev = qdev_new(TYPE_IOAPIC);
662     d = SYS_BUS_DEVICE(dev);
663     sysbus_realize_and_unref(d, &error_fatal);
664     sysbus_mmio_map(d, 0, IO_APIC_SECONDARY_ADDRESS);
665 
666     for (i = 0; i < IOAPIC_NUM_PINS; i++) {
667         gsi_state->ioapic2_irq[i] = qdev_get_gpio_in(dev, i);
668     }
669     return dev;
670 }
671 
672 struct setup_data {
673     uint64_t next;
674     uint32_t type;
675     uint32_t len;
676     uint8_t data[];
677 } __attribute__((packed));
678 
679 
680 /*
681  * The entry point into the kernel for PVH boot is different from
682  * the native entry point.  The PVH entry is defined by the x86/HVM
683  * direct boot ABI and is available in an ELFNOTE in the kernel binary.
684  *
685  * This function is passed to load_elf() when it is called from
686  * load_elfboot() which then additionally checks for an ELF Note of
687  * type XEN_ELFNOTE_PHYS32_ENTRY and passes it to this function to
688  * parse the PVH entry address from the ELF Note.
689  *
690  * Due to trickery in elf_opts.h, load_elf() is actually available as
691  * load_elf32() or load_elf64() and this routine needs to be able
692  * to deal with being called as 32 or 64 bit.
693  *
694  * The address of the PVH entry point is saved to the 'pvh_start_addr'
695  * global variable.  (although the entry point is 32-bit, the kernel
696  * binary can be either 32-bit or 64-bit).
697  */
698 static uint64_t read_pvh_start_addr(void *arg1, void *arg2, bool is64)
699 {
700     size_t *elf_note_data_addr;
701 
702     /* Check if ELF Note header passed in is valid */
703     if (arg1 == NULL) {
704         return 0;
705     }
706 
707     if (is64) {
708         struct elf64_note *nhdr64 = (struct elf64_note *)arg1;
709         uint64_t nhdr_size64 = sizeof(struct elf64_note);
710         uint64_t phdr_align = *(uint64_t *)arg2;
711         uint64_t nhdr_namesz = nhdr64->n_namesz;
712 
713         elf_note_data_addr =
714             ((void *)nhdr64) + nhdr_size64 +
715             QEMU_ALIGN_UP(nhdr_namesz, phdr_align);
716 
717         pvh_start_addr = *elf_note_data_addr;
718     } else {
719         struct elf32_note *nhdr32 = (struct elf32_note *)arg1;
720         uint32_t nhdr_size32 = sizeof(struct elf32_note);
721         uint32_t phdr_align = *(uint32_t *)arg2;
722         uint32_t nhdr_namesz = nhdr32->n_namesz;
723 
724         elf_note_data_addr =
725             ((void *)nhdr32) + nhdr_size32 +
726             QEMU_ALIGN_UP(nhdr_namesz, phdr_align);
727 
728         pvh_start_addr = *(uint32_t *)elf_note_data_addr;
729     }
730 
731     return pvh_start_addr;
732 }
733 
734 static bool load_elfboot(const char *kernel_filename,
735                          int kernel_file_size,
736                          uint8_t *header,
737                          size_t pvh_xen_start_addr,
738                          FWCfgState *fw_cfg)
739 {
740     uint32_t flags = 0;
741     uint32_t mh_load_addr = 0;
742     uint32_t elf_kernel_size = 0;
743     uint64_t elf_entry;
744     uint64_t elf_low, elf_high;
745     int kernel_size;
746 
747     if (ldl_p(header) != 0x464c457f) {
748         return false; /* no elfboot */
749     }
750 
751     bool elf_is64 = header[EI_CLASS] == ELFCLASS64;
752     flags = elf_is64 ?
753         ((Elf64_Ehdr *)header)->e_flags : ((Elf32_Ehdr *)header)->e_flags;
754 
755     if (flags & 0x00010004) { /* LOAD_ELF_HEADER_HAS_ADDR */
756         error_report("elfboot unsupported flags = %x", flags);
757         exit(1);
758     }
759 
760     uint64_t elf_note_type = XEN_ELFNOTE_PHYS32_ENTRY;
761     kernel_size = load_elf(kernel_filename, read_pvh_start_addr,
762                            NULL, &elf_note_type, &elf_entry,
763                            &elf_low, &elf_high, NULL, 0, I386_ELF_MACHINE,
764                            0, 0);
765 
766     if (kernel_size < 0) {
767         error_report("Error while loading elf kernel");
768         exit(1);
769     }
770     mh_load_addr = elf_low;
771     elf_kernel_size = elf_high - elf_low;
772 
773     if (pvh_start_addr == 0) {
774         error_report("Error loading uncompressed kernel without PVH ELF Note");
775         exit(1);
776     }
777     fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_ENTRY, pvh_start_addr);
778     fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_ADDR, mh_load_addr);
779     fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_SIZE, elf_kernel_size);
780 
781     return true;
782 }
783 
784 void x86_load_linux(X86MachineState *x86ms,
785                     FWCfgState *fw_cfg,
786                     int acpi_data_size,
787                     bool pvh_enabled)
788 {
789     bool linuxboot_dma_enabled = X86_MACHINE_GET_CLASS(x86ms)->fwcfg_dma_enabled;
790     uint16_t protocol;
791     int setup_size, kernel_size, cmdline_size;
792     int dtb_size, setup_data_offset;
793     uint32_t initrd_max;
794     uint8_t header[8192], *setup, *kernel;
795     hwaddr real_addr, prot_addr, cmdline_addr, initrd_addr = 0;
796     FILE *f;
797     char *vmode;
798     MachineState *machine = MACHINE(x86ms);
799     struct setup_data *setup_data;
800     const char *kernel_filename = machine->kernel_filename;
801     const char *initrd_filename = machine->initrd_filename;
802     const char *dtb_filename = machine->dtb;
803     const char *kernel_cmdline = machine->kernel_cmdline;
804     SevKernelLoaderContext sev_load_ctx = {};
805 
806     /* Align to 16 bytes as a paranoia measure */
807     cmdline_size = (strlen(kernel_cmdline) + 16) & ~15;
808 
809     /* load the kernel header */
810     f = fopen(kernel_filename, "rb");
811     if (!f) {
812         fprintf(stderr, "qemu: could not open kernel file '%s': %s\n",
813                 kernel_filename, strerror(errno));
814         exit(1);
815     }
816 
817     kernel_size = get_file_size(f);
818     if (!kernel_size ||
819         fread(header, 1, MIN(ARRAY_SIZE(header), kernel_size), f) !=
820         MIN(ARRAY_SIZE(header), kernel_size)) {
821         fprintf(stderr, "qemu: could not load kernel '%s': %s\n",
822                 kernel_filename, strerror(errno));
823         exit(1);
824     }
825 
826     /* kernel protocol version */
827     if (ldl_p(header + 0x202) == 0x53726448) {
828         protocol = lduw_p(header + 0x206);
829     } else {
830         /*
831          * This could be a multiboot kernel. If it is, let's stop treating it
832          * like a Linux kernel.
833          * Note: some multiboot images could be in the ELF format (the same of
834          * PVH), so we try multiboot first since we check the multiboot magic
835          * header before to load it.
836          */
837         if (load_multiboot(x86ms, fw_cfg, f, kernel_filename, initrd_filename,
838                            kernel_cmdline, kernel_size, header)) {
839             return;
840         }
841         /*
842          * Check if the file is an uncompressed kernel file (ELF) and load it,
843          * saving the PVH entry point used by the x86/HVM direct boot ABI.
844          * If load_elfboot() is successful, populate the fw_cfg info.
845          */
846         if (pvh_enabled &&
847             load_elfboot(kernel_filename, kernel_size,
848                          header, pvh_start_addr, fw_cfg)) {
849             fclose(f);
850 
851             fw_cfg_add_i32(fw_cfg, FW_CFG_CMDLINE_SIZE,
852                 strlen(kernel_cmdline) + 1);
853             fw_cfg_add_string(fw_cfg, FW_CFG_CMDLINE_DATA, kernel_cmdline);
854 
855             fw_cfg_add_i32(fw_cfg, FW_CFG_SETUP_SIZE, sizeof(header));
856             fw_cfg_add_bytes(fw_cfg, FW_CFG_SETUP_DATA,
857                              header, sizeof(header));
858 
859             /* load initrd */
860             if (initrd_filename) {
861                 GMappedFile *mapped_file;
862                 gsize initrd_size;
863                 gchar *initrd_data;
864                 GError *gerr = NULL;
865 
866                 mapped_file = g_mapped_file_new(initrd_filename, false, &gerr);
867                 if (!mapped_file) {
868                     fprintf(stderr, "qemu: error reading initrd %s: %s\n",
869                             initrd_filename, gerr->message);
870                     exit(1);
871                 }
872                 x86ms->initrd_mapped_file = mapped_file;
873 
874                 initrd_data = g_mapped_file_get_contents(mapped_file);
875                 initrd_size = g_mapped_file_get_length(mapped_file);
876                 initrd_max = x86ms->below_4g_mem_size - acpi_data_size - 1;
877                 if (initrd_size >= initrd_max) {
878                     fprintf(stderr, "qemu: initrd is too large, cannot support."
879                             "(max: %"PRIu32", need %"PRId64")\n",
880                             initrd_max, (uint64_t)initrd_size);
881                     exit(1);
882                 }
883 
884                 initrd_addr = (initrd_max - initrd_size) & ~4095;
885 
886                 fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_ADDR, initrd_addr);
887                 fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_SIZE, initrd_size);
888                 fw_cfg_add_bytes(fw_cfg, FW_CFG_INITRD_DATA, initrd_data,
889                                  initrd_size);
890             }
891 
892             option_rom[nb_option_roms].bootindex = 0;
893             option_rom[nb_option_roms].name = "pvh.bin";
894             nb_option_roms++;
895 
896             return;
897         }
898         protocol = 0;
899     }
900 
901     if (protocol < 0x200 || !(header[0x211] & 0x01)) {
902         /* Low kernel */
903         real_addr    = 0x90000;
904         cmdline_addr = 0x9a000 - cmdline_size;
905         prot_addr    = 0x10000;
906     } else if (protocol < 0x202) {
907         /* High but ancient kernel */
908         real_addr    = 0x90000;
909         cmdline_addr = 0x9a000 - cmdline_size;
910         prot_addr    = 0x100000;
911     } else {
912         /* High and recent kernel */
913         real_addr    = 0x10000;
914         cmdline_addr = 0x20000;
915         prot_addr    = 0x100000;
916     }
917 
918     /* highest address for loading the initrd */
919     if (protocol >= 0x20c &&
920         lduw_p(header + 0x236) & XLF_CAN_BE_LOADED_ABOVE_4G) {
921         /*
922          * Linux has supported initrd up to 4 GB for a very long time (2007,
923          * long before XLF_CAN_BE_LOADED_ABOVE_4G which was added in 2013),
924          * though it only sets initrd_max to 2 GB to "work around bootloader
925          * bugs". Luckily, QEMU firmware(which does something like bootloader)
926          * has supported this.
927          *
928          * It's believed that if XLF_CAN_BE_LOADED_ABOVE_4G is set, initrd can
929          * be loaded into any address.
930          *
931          * In addition, initrd_max is uint32_t simply because QEMU doesn't
932          * support the 64-bit boot protocol (specifically the ext_ramdisk_image
933          * field).
934          *
935          * Therefore here just limit initrd_max to UINT32_MAX simply as well.
936          */
937         initrd_max = UINT32_MAX;
938     } else if (protocol >= 0x203) {
939         initrd_max = ldl_p(header + 0x22c);
940     } else {
941         initrd_max = 0x37ffffff;
942     }
943 
944     if (initrd_max >= x86ms->below_4g_mem_size - acpi_data_size) {
945         initrd_max = x86ms->below_4g_mem_size - acpi_data_size - 1;
946     }
947 
948     fw_cfg_add_i32(fw_cfg, FW_CFG_CMDLINE_ADDR, cmdline_addr);
949     fw_cfg_add_i32(fw_cfg, FW_CFG_CMDLINE_SIZE, strlen(kernel_cmdline) + 1);
950     fw_cfg_add_string(fw_cfg, FW_CFG_CMDLINE_DATA, kernel_cmdline);
951     sev_load_ctx.cmdline_data = (char *)kernel_cmdline;
952     sev_load_ctx.cmdline_size = strlen(kernel_cmdline) + 1;
953 
954     if (protocol >= 0x202) {
955         stl_p(header + 0x228, cmdline_addr);
956     } else {
957         stw_p(header + 0x20, 0xA33F);
958         stw_p(header + 0x22, cmdline_addr - real_addr);
959     }
960 
961     /* handle vga= parameter */
962     vmode = strstr(kernel_cmdline, "vga=");
963     if (vmode) {
964         unsigned int video_mode;
965         const char *end;
966         int ret;
967         /* skip "vga=" */
968         vmode += 4;
969         if (!strncmp(vmode, "normal", 6)) {
970             video_mode = 0xffff;
971         } else if (!strncmp(vmode, "ext", 3)) {
972             video_mode = 0xfffe;
973         } else if (!strncmp(vmode, "ask", 3)) {
974             video_mode = 0xfffd;
975         } else {
976             ret = qemu_strtoui(vmode, &end, 0, &video_mode);
977             if (ret != 0 || (*end && *end != ' ')) {
978                 fprintf(stderr, "qemu: invalid 'vga=' kernel parameter.\n");
979                 exit(1);
980             }
981         }
982         stw_p(header + 0x1fa, video_mode);
983     }
984 
985     /* loader type */
986     /*
987      * High nybble = B reserved for QEMU; low nybble is revision number.
988      * If this code is substantially changed, you may want to consider
989      * incrementing the revision.
990      */
991     if (protocol >= 0x200) {
992         header[0x210] = 0xB0;
993     }
994     /* heap */
995     if (protocol >= 0x201) {
996         header[0x211] |= 0x80; /* CAN_USE_HEAP */
997         stw_p(header + 0x224, cmdline_addr - real_addr - 0x200);
998     }
999 
1000     /* load initrd */
1001     if (initrd_filename) {
1002         GMappedFile *mapped_file;
1003         gsize initrd_size;
1004         gchar *initrd_data;
1005         GError *gerr = NULL;
1006 
1007         if (protocol < 0x200) {
1008             fprintf(stderr, "qemu: linux kernel too old to load a ram disk\n");
1009             exit(1);
1010         }
1011 
1012         mapped_file = g_mapped_file_new(initrd_filename, false, &gerr);
1013         if (!mapped_file) {
1014             fprintf(stderr, "qemu: error reading initrd %s: %s\n",
1015                     initrd_filename, gerr->message);
1016             exit(1);
1017         }
1018         x86ms->initrd_mapped_file = mapped_file;
1019 
1020         initrd_data = g_mapped_file_get_contents(mapped_file);
1021         initrd_size = g_mapped_file_get_length(mapped_file);
1022         if (initrd_size >= initrd_max) {
1023             fprintf(stderr, "qemu: initrd is too large, cannot support."
1024                     "(max: %"PRIu32", need %"PRId64")\n",
1025                     initrd_max, (uint64_t)initrd_size);
1026             exit(1);
1027         }
1028 
1029         initrd_addr = (initrd_max - initrd_size) & ~4095;
1030 
1031         fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_ADDR, initrd_addr);
1032         fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_SIZE, initrd_size);
1033         fw_cfg_add_bytes(fw_cfg, FW_CFG_INITRD_DATA, initrd_data, initrd_size);
1034         sev_load_ctx.initrd_data = initrd_data;
1035         sev_load_ctx.initrd_size = initrd_size;
1036 
1037         stl_p(header + 0x218, initrd_addr);
1038         stl_p(header + 0x21c, initrd_size);
1039     }
1040 
1041     /* load kernel and setup */
1042     setup_size = header[0x1f1];
1043     if (setup_size == 0) {
1044         setup_size = 4;
1045     }
1046     setup_size = (setup_size + 1) * 512;
1047     if (setup_size > kernel_size) {
1048         fprintf(stderr, "qemu: invalid kernel header\n");
1049         exit(1);
1050     }
1051     kernel_size -= setup_size;
1052 
1053     setup  = g_malloc(setup_size);
1054     kernel = g_malloc(kernel_size);
1055     fseek(f, 0, SEEK_SET);
1056     if (fread(setup, 1, setup_size, f) != setup_size) {
1057         fprintf(stderr, "fread() failed\n");
1058         exit(1);
1059     }
1060     if (fread(kernel, 1, kernel_size, f) != kernel_size) {
1061         fprintf(stderr, "fread() failed\n");
1062         exit(1);
1063     }
1064     fclose(f);
1065 
1066     /* append dtb to kernel */
1067     if (dtb_filename) {
1068         if (protocol < 0x209) {
1069             fprintf(stderr, "qemu: Linux kernel too old to load a dtb\n");
1070             exit(1);
1071         }
1072 
1073         dtb_size = get_image_size(dtb_filename);
1074         if (dtb_size <= 0) {
1075             fprintf(stderr, "qemu: error reading dtb %s: %s\n",
1076                     dtb_filename, strerror(errno));
1077             exit(1);
1078         }
1079 
1080         setup_data_offset = QEMU_ALIGN_UP(kernel_size, 16);
1081         kernel_size = setup_data_offset + sizeof(struct setup_data) + dtb_size;
1082         kernel = g_realloc(kernel, kernel_size);
1083 
1084         stq_p(header + 0x250, prot_addr + setup_data_offset);
1085 
1086         setup_data = (struct setup_data *)(kernel + setup_data_offset);
1087         setup_data->next = 0;
1088         setup_data->type = cpu_to_le32(SETUP_DTB);
1089         setup_data->len = cpu_to_le32(dtb_size);
1090 
1091         load_image_size(dtb_filename, setup_data->data, dtb_size);
1092     }
1093 
1094     /*
1095      * If we're starting an encrypted VM, it will be OVMF based, which uses the
1096      * efi stub for booting and doesn't require any values to be placed in the
1097      * kernel header.  We therefore don't update the header so the hash of the
1098      * kernel on the other side of the fw_cfg interface matches the hash of the
1099      * file the user passed in.
1100      */
1101     if (!sev_enabled()) {
1102         memcpy(setup, header, MIN(sizeof(header), setup_size));
1103     }
1104 
1105     fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_ADDR, prot_addr);
1106     fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_SIZE, kernel_size);
1107     fw_cfg_add_bytes(fw_cfg, FW_CFG_KERNEL_DATA, kernel, kernel_size);
1108     sev_load_ctx.kernel_data = (char *)kernel;
1109     sev_load_ctx.kernel_size = kernel_size;
1110 
1111     fw_cfg_add_i32(fw_cfg, FW_CFG_SETUP_ADDR, real_addr);
1112     fw_cfg_add_i32(fw_cfg, FW_CFG_SETUP_SIZE, setup_size);
1113     fw_cfg_add_bytes(fw_cfg, FW_CFG_SETUP_DATA, setup, setup_size);
1114     sev_load_ctx.setup_data = (char *)setup;
1115     sev_load_ctx.setup_size = setup_size;
1116 
1117     if (sev_enabled()) {
1118         sev_add_kernel_loader_hashes(&sev_load_ctx, &error_fatal);
1119     }
1120 
1121     option_rom[nb_option_roms].bootindex = 0;
1122     option_rom[nb_option_roms].name = "linuxboot.bin";
1123     if (linuxboot_dma_enabled && fw_cfg_dma_enabled(fw_cfg)) {
1124         option_rom[nb_option_roms].name = "linuxboot_dma.bin";
1125     }
1126     nb_option_roms++;
1127 }
1128 
1129 void x86_bios_rom_init(MachineState *ms, const char *default_firmware,
1130                        MemoryRegion *rom_memory, bool isapc_ram_fw)
1131 {
1132     const char *bios_name;
1133     char *filename;
1134     MemoryRegion *bios, *isa_bios;
1135     int bios_size, isa_bios_size;
1136     ssize_t ret;
1137 
1138     /* BIOS load */
1139     bios_name = ms->firmware ?: default_firmware;
1140     filename = qemu_find_file(QEMU_FILE_TYPE_BIOS, bios_name);
1141     if (filename) {
1142         bios_size = get_image_size(filename);
1143     } else {
1144         bios_size = -1;
1145     }
1146     if (bios_size <= 0 ||
1147         (bios_size % 65536) != 0) {
1148         goto bios_error;
1149     }
1150     bios = g_malloc(sizeof(*bios));
1151     memory_region_init_ram(bios, NULL, "pc.bios", bios_size, &error_fatal);
1152     if (sev_enabled()) {
1153         /*
1154          * The concept of a "reset" simply doesn't exist for
1155          * confidential computing guests, we have to destroy and
1156          * re-launch them instead.  So there is no need to register
1157          * the firmware as rom to properly re-initialize on reset.
1158          * Just go for a straight file load instead.
1159          */
1160         void *ptr = memory_region_get_ram_ptr(bios);
1161         load_image_size(filename, ptr, bios_size);
1162         x86_firmware_configure(ptr, bios_size);
1163     } else {
1164         if (!isapc_ram_fw) {
1165             memory_region_set_readonly(bios, true);
1166         }
1167         ret = rom_add_file_fixed(bios_name, (uint32_t)(-bios_size), -1);
1168         if (ret != 0) {
1169             goto bios_error;
1170         }
1171     }
1172     g_free(filename);
1173 
1174     /* map the last 128KB of the BIOS in ISA space */
1175     isa_bios_size = MIN(bios_size, 128 * KiB);
1176     isa_bios = g_malloc(sizeof(*isa_bios));
1177     memory_region_init_alias(isa_bios, NULL, "isa-bios", bios,
1178                              bios_size - isa_bios_size, isa_bios_size);
1179     memory_region_add_subregion_overlap(rom_memory,
1180                                         0x100000 - isa_bios_size,
1181                                         isa_bios,
1182                                         1);
1183     if (!isapc_ram_fw) {
1184         memory_region_set_readonly(isa_bios, true);
1185     }
1186 
1187     /* map all the bios at the top of memory */
1188     memory_region_add_subregion(rom_memory,
1189                                 (uint32_t)(-bios_size),
1190                                 bios);
1191     return;
1192 
1193 bios_error:
1194     fprintf(stderr, "qemu: could not load PC BIOS '%s'\n", bios_name);
1195     exit(1);
1196 }
1197 
1198 bool x86_machine_is_smm_enabled(const X86MachineState *x86ms)
1199 {
1200     bool smm_available = false;
1201 
1202     if (x86ms->smm == ON_OFF_AUTO_OFF) {
1203         return false;
1204     }
1205 
1206     if (tcg_enabled() || qtest_enabled()) {
1207         smm_available = true;
1208     } else if (kvm_enabled()) {
1209         smm_available = kvm_has_smm();
1210     }
1211 
1212     if (smm_available) {
1213         return true;
1214     }
1215 
1216     if (x86ms->smm == ON_OFF_AUTO_ON) {
1217         error_report("System Management Mode not supported by this hypervisor.");
1218         exit(1);
1219     }
1220     return false;
1221 }
1222 
1223 static void x86_machine_get_smm(Object *obj, Visitor *v, const char *name,
1224                                void *opaque, Error **errp)
1225 {
1226     X86MachineState *x86ms = X86_MACHINE(obj);
1227     OnOffAuto smm = x86ms->smm;
1228 
1229     visit_type_OnOffAuto(v, name, &smm, errp);
1230 }
1231 
1232 static void x86_machine_set_smm(Object *obj, Visitor *v, const char *name,
1233                                void *opaque, Error **errp)
1234 {
1235     X86MachineState *x86ms = X86_MACHINE(obj);
1236 
1237     visit_type_OnOffAuto(v, name, &x86ms->smm, errp);
1238 }
1239 
1240 bool x86_machine_is_acpi_enabled(const X86MachineState *x86ms)
1241 {
1242     if (x86ms->acpi == ON_OFF_AUTO_OFF) {
1243         return false;
1244     }
1245     return true;
1246 }
1247 
1248 static void x86_machine_get_acpi(Object *obj, Visitor *v, const char *name,
1249                                  void *opaque, Error **errp)
1250 {
1251     X86MachineState *x86ms = X86_MACHINE(obj);
1252     OnOffAuto acpi = x86ms->acpi;
1253 
1254     visit_type_OnOffAuto(v, name, &acpi, errp);
1255 }
1256 
1257 static void x86_machine_set_acpi(Object *obj, Visitor *v, const char *name,
1258                                  void *opaque, Error **errp)
1259 {
1260     X86MachineState *x86ms = X86_MACHINE(obj);
1261 
1262     visit_type_OnOffAuto(v, name, &x86ms->acpi, errp);
1263 }
1264 
1265 static void x86_machine_get_pit(Object *obj, Visitor *v, const char *name,
1266                                     void *opaque, Error **errp)
1267 {
1268     X86MachineState *x86ms = X86_MACHINE(obj);
1269     OnOffAuto pit = x86ms->pit;
1270 
1271     visit_type_OnOffAuto(v, name, &pit, errp);
1272 }
1273 
1274 static void x86_machine_set_pit(Object *obj, Visitor *v, const char *name,
1275                                     void *opaque, Error **errp)
1276 {
1277     X86MachineState *x86ms = X86_MACHINE(obj);;
1278 
1279     visit_type_OnOffAuto(v, name, &x86ms->pit, errp);
1280 }
1281 
1282 static void x86_machine_get_pic(Object *obj, Visitor *v, const char *name,
1283                                 void *opaque, Error **errp)
1284 {
1285     X86MachineState *x86ms = X86_MACHINE(obj);
1286     OnOffAuto pic = x86ms->pic;
1287 
1288     visit_type_OnOffAuto(v, name, &pic, errp);
1289 }
1290 
1291 static void x86_machine_set_pic(Object *obj, Visitor *v, const char *name,
1292                                 void *opaque, Error **errp)
1293 {
1294     X86MachineState *x86ms = X86_MACHINE(obj);
1295 
1296     visit_type_OnOffAuto(v, name, &x86ms->pic, errp);
1297 }
1298 
1299 static char *x86_machine_get_oem_id(Object *obj, Error **errp)
1300 {
1301     X86MachineState *x86ms = X86_MACHINE(obj);
1302 
1303     return g_strdup(x86ms->oem_id);
1304 }
1305 
1306 static void x86_machine_set_oem_id(Object *obj, const char *value, Error **errp)
1307 {
1308     X86MachineState *x86ms = X86_MACHINE(obj);
1309     size_t len = strlen(value);
1310 
1311     if (len > 6) {
1312         error_setg(errp,
1313                    "User specified "X86_MACHINE_OEM_ID" value is bigger than "
1314                    "6 bytes in size");
1315         return;
1316     }
1317 
1318     strncpy(x86ms->oem_id, value, 6);
1319 }
1320 
1321 static char *x86_machine_get_oem_table_id(Object *obj, Error **errp)
1322 {
1323     X86MachineState *x86ms = X86_MACHINE(obj);
1324 
1325     return g_strdup(x86ms->oem_table_id);
1326 }
1327 
1328 static void x86_machine_set_oem_table_id(Object *obj, const char *value,
1329                                          Error **errp)
1330 {
1331     X86MachineState *x86ms = X86_MACHINE(obj);
1332     size_t len = strlen(value);
1333 
1334     if (len > 8) {
1335         error_setg(errp,
1336                    "User specified "X86_MACHINE_OEM_TABLE_ID
1337                    " value is bigger than "
1338                    "8 bytes in size");
1339         return;
1340     }
1341     strncpy(x86ms->oem_table_id, value, 8);
1342 }
1343 
1344 static void x86_machine_get_bus_lock_ratelimit(Object *obj, Visitor *v,
1345                                 const char *name, void *opaque, Error **errp)
1346 {
1347     X86MachineState *x86ms = X86_MACHINE(obj);
1348     uint64_t bus_lock_ratelimit = x86ms->bus_lock_ratelimit;
1349 
1350     visit_type_uint64(v, name, &bus_lock_ratelimit, errp);
1351 }
1352 
1353 static void x86_machine_set_bus_lock_ratelimit(Object *obj, Visitor *v,
1354                                const char *name, void *opaque, Error **errp)
1355 {
1356     X86MachineState *x86ms = X86_MACHINE(obj);
1357 
1358     visit_type_uint64(v, name, &x86ms->bus_lock_ratelimit, errp);
1359 }
1360 
1361 static void machine_get_sgx_epc(Object *obj, Visitor *v, const char *name,
1362                                 void *opaque, Error **errp)
1363 {
1364     X86MachineState *x86ms = X86_MACHINE(obj);
1365     SgxEPCList *list = x86ms->sgx_epc_list;
1366 
1367     visit_type_SgxEPCList(v, name, &list, errp);
1368 }
1369 
1370 static void machine_set_sgx_epc(Object *obj, Visitor *v, const char *name,
1371                                 void *opaque, Error **errp)
1372 {
1373     X86MachineState *x86ms = X86_MACHINE(obj);
1374     SgxEPCList *list;
1375 
1376     list = x86ms->sgx_epc_list;
1377     visit_type_SgxEPCList(v, name, &x86ms->sgx_epc_list, errp);
1378 
1379     qapi_free_SgxEPCList(list);
1380 }
1381 
1382 static void x86_machine_initfn(Object *obj)
1383 {
1384     X86MachineState *x86ms = X86_MACHINE(obj);
1385 
1386     x86ms->smm = ON_OFF_AUTO_AUTO;
1387     x86ms->acpi = ON_OFF_AUTO_AUTO;
1388     x86ms->pit = ON_OFF_AUTO_AUTO;
1389     x86ms->pic = ON_OFF_AUTO_AUTO;
1390     x86ms->pci_irq_mask = ACPI_BUILD_PCI_IRQS;
1391     x86ms->oem_id = g_strndup(ACPI_BUILD_APPNAME6, 6);
1392     x86ms->oem_table_id = g_strndup(ACPI_BUILD_APPNAME8, 8);
1393     x86ms->bus_lock_ratelimit = 0;
1394     x86ms->above_4g_mem_start = 4 * GiB;
1395 }
1396 
1397 static void x86_machine_class_init(ObjectClass *oc, void *data)
1398 {
1399     MachineClass *mc = MACHINE_CLASS(oc);
1400     X86MachineClass *x86mc = X86_MACHINE_CLASS(oc);
1401     NMIClass *nc = NMI_CLASS(oc);
1402 
1403     mc->cpu_index_to_instance_props = x86_cpu_index_to_props;
1404     mc->get_default_cpu_node_id = x86_get_default_cpu_node_id;
1405     mc->possible_cpu_arch_ids = x86_possible_cpu_arch_ids;
1406     x86mc->save_tsc_khz = true;
1407     x86mc->fwcfg_dma_enabled = true;
1408     nc->nmi_monitor_handler = x86_nmi;
1409 
1410     object_class_property_add(oc, X86_MACHINE_SMM, "OnOffAuto",
1411         x86_machine_get_smm, x86_machine_set_smm,
1412         NULL, NULL);
1413     object_class_property_set_description(oc, X86_MACHINE_SMM,
1414         "Enable SMM");
1415 
1416     object_class_property_add(oc, X86_MACHINE_ACPI, "OnOffAuto",
1417         x86_machine_get_acpi, x86_machine_set_acpi,
1418         NULL, NULL);
1419     object_class_property_set_description(oc, X86_MACHINE_ACPI,
1420         "Enable ACPI");
1421 
1422     object_class_property_add(oc, X86_MACHINE_PIT, "OnOffAuto",
1423                               x86_machine_get_pit,
1424                               x86_machine_set_pit,
1425                               NULL, NULL);
1426     object_class_property_set_description(oc, X86_MACHINE_PIT,
1427         "Enable i8254 PIT");
1428 
1429     object_class_property_add(oc, X86_MACHINE_PIC, "OnOffAuto",
1430                               x86_machine_get_pic,
1431                               x86_machine_set_pic,
1432                               NULL, NULL);
1433     object_class_property_set_description(oc, X86_MACHINE_PIC,
1434         "Enable i8259 PIC");
1435 
1436     object_class_property_add_str(oc, X86_MACHINE_OEM_ID,
1437                                   x86_machine_get_oem_id,
1438                                   x86_machine_set_oem_id);
1439     object_class_property_set_description(oc, X86_MACHINE_OEM_ID,
1440                                           "Override the default value of field OEMID "
1441                                           "in ACPI table header."
1442                                           "The string may be up to 6 bytes in size");
1443 
1444 
1445     object_class_property_add_str(oc, X86_MACHINE_OEM_TABLE_ID,
1446                                   x86_machine_get_oem_table_id,
1447                                   x86_machine_set_oem_table_id);
1448     object_class_property_set_description(oc, X86_MACHINE_OEM_TABLE_ID,
1449                                           "Override the default value of field OEM Table ID "
1450                                           "in ACPI table header."
1451                                           "The string may be up to 8 bytes in size");
1452 
1453     object_class_property_add(oc, X86_MACHINE_BUS_LOCK_RATELIMIT, "uint64_t",
1454                                 x86_machine_get_bus_lock_ratelimit,
1455                                 x86_machine_set_bus_lock_ratelimit, NULL, NULL);
1456     object_class_property_set_description(oc, X86_MACHINE_BUS_LOCK_RATELIMIT,
1457             "Set the ratelimit for the bus locks acquired in VMs");
1458 
1459     object_class_property_add(oc, "sgx-epc", "SgxEPC",
1460         machine_get_sgx_epc, machine_set_sgx_epc,
1461         NULL, NULL);
1462     object_class_property_set_description(oc, "sgx-epc",
1463         "SGX EPC device");
1464 }
1465 
1466 static const TypeInfo x86_machine_info = {
1467     .name = TYPE_X86_MACHINE,
1468     .parent = TYPE_MACHINE,
1469     .abstract = true,
1470     .instance_size = sizeof(X86MachineState),
1471     .instance_init = x86_machine_initfn,
1472     .class_size = sizeof(X86MachineClass),
1473     .class_init = x86_machine_class_init,
1474     .interfaces = (InterfaceInfo[]) {
1475          { TYPE_NMI },
1476          { }
1477     },
1478 };
1479 
1480 static void x86_machine_register_types(void)
1481 {
1482     type_register_static(&x86_machine_info);
1483 }
1484 
1485 type_init(x86_machine_register_types)
1486