xref: /openbmc/qemu/hw/i386/x86.c (revision 88693ab2)
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?
133      *
134      * Under Xen: yes.
135      * With userspace emulated lapic: no
136      * With KVM's in-kernel lapic: only if X2APIC API is enabled.
137      */
138     if (x86ms->apic_id_limit > 255 && !xen_enabled() &&
139         (!kvm_irqchip_in_kernel() || !kvm_enable_x2apic())) {
140         error_report("current -smp configuration requires kernel "
141                      "irqchip and X2APIC API support.");
142         exit(EXIT_FAILURE);
143     }
144 
145     if (kvm_enabled()) {
146         kvm_set_max_apic_id(x86ms->apic_id_limit);
147     }
148 
149     possible_cpus = mc->possible_cpu_arch_ids(ms);
150     for (i = 0; i < ms->smp.cpus; i++) {
151         x86_cpu_new(x86ms, possible_cpus->cpus[i].arch_id, &error_fatal);
152     }
153 }
154 
155 void x86_rtc_set_cpus_count(ISADevice *s, uint16_t cpus_count)
156 {
157     MC146818RtcState *rtc = MC146818_RTC(s);
158 
159     if (cpus_count > 0xff) {
160         /*
161          * If the number of CPUs can't be represented in 8 bits, the
162          * BIOS must use "FW_CFG_NB_CPUS". Set RTC field to 0 just
163          * to make old BIOSes fail more predictably.
164          */
165         mc146818rtc_set_cmos_data(rtc, 0x5f, 0);
166     } else {
167         mc146818rtc_set_cmos_data(rtc, 0x5f, cpus_count - 1);
168     }
169 }
170 
171 static int x86_apic_cmp(const void *a, const void *b)
172 {
173    CPUArchId *apic_a = (CPUArchId *)a;
174    CPUArchId *apic_b = (CPUArchId *)b;
175 
176    return apic_a->arch_id - apic_b->arch_id;
177 }
178 
179 /*
180  * returns pointer to CPUArchId descriptor that matches CPU's apic_id
181  * in ms->possible_cpus->cpus, if ms->possible_cpus->cpus has no
182  * entry corresponding to CPU's apic_id returns NULL.
183  */
184 CPUArchId *x86_find_cpu_slot(MachineState *ms, uint32_t id, int *idx)
185 {
186     CPUArchId apic_id, *found_cpu;
187 
188     apic_id.arch_id = id;
189     found_cpu = bsearch(&apic_id, ms->possible_cpus->cpus,
190         ms->possible_cpus->len, sizeof(*ms->possible_cpus->cpus),
191         x86_apic_cmp);
192     if (found_cpu && idx) {
193         *idx = found_cpu - ms->possible_cpus->cpus;
194     }
195     return found_cpu;
196 }
197 
198 void x86_cpu_plug(HotplugHandler *hotplug_dev,
199                   DeviceState *dev, Error **errp)
200 {
201     CPUArchId *found_cpu;
202     Error *local_err = NULL;
203     X86CPU *cpu = X86_CPU(dev);
204     X86MachineState *x86ms = X86_MACHINE(hotplug_dev);
205 
206     if (x86ms->acpi_dev) {
207         hotplug_handler_plug(x86ms->acpi_dev, dev, &local_err);
208         if (local_err) {
209             goto out;
210         }
211     }
212 
213     /* increment the number of CPUs */
214     x86ms->boot_cpus++;
215     if (x86ms->rtc) {
216         x86_rtc_set_cpus_count(x86ms->rtc, x86ms->boot_cpus);
217     }
218     if (x86ms->fw_cfg) {
219         fw_cfg_modify_i16(x86ms->fw_cfg, FW_CFG_NB_CPUS, x86ms->boot_cpus);
220     }
221 
222     found_cpu = x86_find_cpu_slot(MACHINE(x86ms), cpu->apic_id, NULL);
223     found_cpu->cpu = OBJECT(dev);
224 out:
225     error_propagate(errp, local_err);
226 }
227 
228 void x86_cpu_unplug_request_cb(HotplugHandler *hotplug_dev,
229                                DeviceState *dev, Error **errp)
230 {
231     int idx = -1;
232     X86CPU *cpu = X86_CPU(dev);
233     X86MachineState *x86ms = X86_MACHINE(hotplug_dev);
234 
235     if (!x86ms->acpi_dev) {
236         error_setg(errp, "CPU hot unplug not supported without ACPI");
237         return;
238     }
239 
240     x86_find_cpu_slot(MACHINE(x86ms), cpu->apic_id, &idx);
241     assert(idx != -1);
242     if (idx == 0) {
243         error_setg(errp, "Boot CPU is unpluggable");
244         return;
245     }
246 
247     hotplug_handler_unplug_request(x86ms->acpi_dev, dev,
248                                    errp);
249 }
250 
251 void x86_cpu_unplug_cb(HotplugHandler *hotplug_dev,
252                        DeviceState *dev, Error **errp)
253 {
254     CPUArchId *found_cpu;
255     Error *local_err = NULL;
256     X86CPU *cpu = X86_CPU(dev);
257     X86MachineState *x86ms = X86_MACHINE(hotplug_dev);
258 
259     hotplug_handler_unplug(x86ms->acpi_dev, dev, &local_err);
260     if (local_err) {
261         goto out;
262     }
263 
264     found_cpu = x86_find_cpu_slot(MACHINE(x86ms), cpu->apic_id, NULL);
265     found_cpu->cpu = NULL;
266     qdev_unrealize(dev);
267 
268     /* decrement the number of CPUs */
269     x86ms->boot_cpus--;
270     /* Update the number of CPUs in CMOS */
271     x86_rtc_set_cpus_count(x86ms->rtc, x86ms->boot_cpus);
272     fw_cfg_modify_i16(x86ms->fw_cfg, FW_CFG_NB_CPUS, x86ms->boot_cpus);
273  out:
274     error_propagate(errp, local_err);
275 }
276 
277 void x86_cpu_pre_plug(HotplugHandler *hotplug_dev,
278                       DeviceState *dev, Error **errp)
279 {
280     int idx;
281     CPUState *cs;
282     CPUArchId *cpu_slot;
283     X86CPUTopoIDs topo_ids;
284     X86CPU *cpu = X86_CPU(dev);
285     CPUX86State *env = &cpu->env;
286     MachineState *ms = MACHINE(hotplug_dev);
287     X86MachineState *x86ms = X86_MACHINE(hotplug_dev);
288     unsigned int smp_cores = ms->smp.cores;
289     unsigned int smp_threads = ms->smp.threads;
290     X86CPUTopoInfo topo_info;
291 
292     if (!object_dynamic_cast(OBJECT(cpu), ms->cpu_type)) {
293         error_setg(errp, "Invalid CPU type, expected cpu type: '%s'",
294                    ms->cpu_type);
295         return;
296     }
297 
298     if (x86ms->acpi_dev) {
299         Error *local_err = NULL;
300 
301         hotplug_handler_pre_plug(HOTPLUG_HANDLER(x86ms->acpi_dev), dev,
302                                  &local_err);
303         if (local_err) {
304             error_propagate(errp, local_err);
305             return;
306         }
307     }
308 
309     init_topo_info(&topo_info, x86ms);
310 
311     env->nr_dies = ms->smp.dies;
312 
313     /*
314      * If APIC ID is not set,
315      * set it based on socket/die/core/thread properties.
316      */
317     if (cpu->apic_id == UNASSIGNED_APIC_ID) {
318         int max_socket = (ms->smp.max_cpus - 1) /
319                                 smp_threads / smp_cores / ms->smp.dies;
320 
321         /*
322          * die-id was optional in QEMU 4.0 and older, so keep it optional
323          * if there's only one die per socket.
324          */
325         if (cpu->die_id < 0 && ms->smp.dies == 1) {
326             cpu->die_id = 0;
327         }
328 
329         if (cpu->socket_id < 0) {
330             error_setg(errp, "CPU socket-id is not set");
331             return;
332         } else if (cpu->socket_id > max_socket) {
333             error_setg(errp, "Invalid CPU socket-id: %u must be in range 0:%u",
334                        cpu->socket_id, max_socket);
335             return;
336         }
337         if (cpu->die_id < 0) {
338             error_setg(errp, "CPU die-id is not set");
339             return;
340         } else if (cpu->die_id > ms->smp.dies - 1) {
341             error_setg(errp, "Invalid CPU die-id: %u must be in range 0:%u",
342                        cpu->die_id, ms->smp.dies - 1);
343             return;
344         }
345         if (cpu->core_id < 0) {
346             error_setg(errp, "CPU core-id is not set");
347             return;
348         } else if (cpu->core_id > (smp_cores - 1)) {
349             error_setg(errp, "Invalid CPU core-id: %u must be in range 0:%u",
350                        cpu->core_id, smp_cores - 1);
351             return;
352         }
353         if (cpu->thread_id < 0) {
354             error_setg(errp, "CPU thread-id is not set");
355             return;
356         } else if (cpu->thread_id > (smp_threads - 1)) {
357             error_setg(errp, "Invalid CPU thread-id: %u must be in range 0:%u",
358                        cpu->thread_id, smp_threads - 1);
359             return;
360         }
361 
362         topo_ids.pkg_id = cpu->socket_id;
363         topo_ids.die_id = cpu->die_id;
364         topo_ids.core_id = cpu->core_id;
365         topo_ids.smt_id = cpu->thread_id;
366         cpu->apic_id = x86_apicid_from_topo_ids(&topo_info, &topo_ids);
367     }
368 
369     cpu_slot = x86_find_cpu_slot(MACHINE(x86ms), cpu->apic_id, &idx);
370     if (!cpu_slot) {
371         MachineState *ms = MACHINE(x86ms);
372 
373         x86_topo_ids_from_apicid(cpu->apic_id, &topo_info, &topo_ids);
374         error_setg(errp,
375             "Invalid CPU [socket: %u, die: %u, core: %u, thread: %u] with"
376             " APIC ID %" PRIu32 ", valid index range 0:%d",
377             topo_ids.pkg_id, topo_ids.die_id, topo_ids.core_id, topo_ids.smt_id,
378             cpu->apic_id, ms->possible_cpus->len - 1);
379         return;
380     }
381 
382     if (cpu_slot->cpu) {
383         error_setg(errp, "CPU[%d] with APIC ID %" PRIu32 " exists",
384                    idx, cpu->apic_id);
385         return;
386     }
387 
388     /* if 'address' properties socket-id/core-id/thread-id are not set, set them
389      * so that machine_query_hotpluggable_cpus would show correct values
390      */
391     /* TODO: move socket_id/core_id/thread_id checks into x86_cpu_realizefn()
392      * once -smp refactoring is complete and there will be CPU private
393      * CPUState::nr_cores and CPUState::nr_threads fields instead of globals */
394     x86_topo_ids_from_apicid(cpu->apic_id, &topo_info, &topo_ids);
395     if (cpu->socket_id != -1 && cpu->socket_id != topo_ids.pkg_id) {
396         error_setg(errp, "property socket-id: %u doesn't match set apic-id:"
397             " 0x%x (socket-id: %u)", cpu->socket_id, cpu->apic_id,
398             topo_ids.pkg_id);
399         return;
400     }
401     cpu->socket_id = topo_ids.pkg_id;
402 
403     if (cpu->die_id != -1 && cpu->die_id != topo_ids.die_id) {
404         error_setg(errp, "property die-id: %u doesn't match set apic-id:"
405             " 0x%x (die-id: %u)", cpu->die_id, cpu->apic_id, topo_ids.die_id);
406         return;
407     }
408     cpu->die_id = topo_ids.die_id;
409 
410     if (cpu->core_id != -1 && cpu->core_id != topo_ids.core_id) {
411         error_setg(errp, "property core-id: %u doesn't match set apic-id:"
412             " 0x%x (core-id: %u)", cpu->core_id, cpu->apic_id,
413             topo_ids.core_id);
414         return;
415     }
416     cpu->core_id = topo_ids.core_id;
417 
418     if (cpu->thread_id != -1 && cpu->thread_id != topo_ids.smt_id) {
419         error_setg(errp, "property thread-id: %u doesn't match set apic-id:"
420             " 0x%x (thread-id: %u)", cpu->thread_id, cpu->apic_id,
421             topo_ids.smt_id);
422         return;
423     }
424     cpu->thread_id = topo_ids.smt_id;
425 
426     if (hyperv_feat_enabled(cpu, HYPERV_FEAT_VPINDEX) &&
427         !kvm_hv_vpindex_settable()) {
428         error_setg(errp, "kernel doesn't allow setting HyperV VP_INDEX");
429         return;
430     }
431 
432     cs = CPU(cpu);
433     cs->cpu_index = idx;
434 
435     numa_cpu_pre_plug(cpu_slot, dev, errp);
436 }
437 
438 CpuInstanceProperties
439 x86_cpu_index_to_props(MachineState *ms, unsigned cpu_index)
440 {
441     MachineClass *mc = MACHINE_GET_CLASS(ms);
442     const CPUArchIdList *possible_cpus = mc->possible_cpu_arch_ids(ms);
443 
444     assert(cpu_index < possible_cpus->len);
445     return possible_cpus->cpus[cpu_index].props;
446 }
447 
448 int64_t x86_get_default_cpu_node_id(const MachineState *ms, int idx)
449 {
450    X86CPUTopoIDs topo_ids;
451    X86MachineState *x86ms = X86_MACHINE(ms);
452    X86CPUTopoInfo topo_info;
453 
454    init_topo_info(&topo_info, x86ms);
455 
456    assert(idx < ms->possible_cpus->len);
457    x86_topo_ids_from_apicid(ms->possible_cpus->cpus[idx].arch_id,
458                             &topo_info, &topo_ids);
459    return topo_ids.pkg_id % ms->numa_state->num_nodes;
460 }
461 
462 const CPUArchIdList *x86_possible_cpu_arch_ids(MachineState *ms)
463 {
464     X86MachineState *x86ms = X86_MACHINE(ms);
465     unsigned int max_cpus = ms->smp.max_cpus;
466     X86CPUTopoInfo topo_info;
467     int i;
468 
469     if (ms->possible_cpus) {
470         /*
471          * make sure that max_cpus hasn't changed since the first use, i.e.
472          * -smp hasn't been parsed after it
473          */
474         assert(ms->possible_cpus->len == max_cpus);
475         return ms->possible_cpus;
476     }
477 
478     ms->possible_cpus = g_malloc0(sizeof(CPUArchIdList) +
479                                   sizeof(CPUArchId) * max_cpus);
480     ms->possible_cpus->len = max_cpus;
481 
482     init_topo_info(&topo_info, x86ms);
483 
484     for (i = 0; i < ms->possible_cpus->len; i++) {
485         X86CPUTopoIDs topo_ids;
486 
487         ms->possible_cpus->cpus[i].type = ms->cpu_type;
488         ms->possible_cpus->cpus[i].vcpus_count = 1;
489         ms->possible_cpus->cpus[i].arch_id =
490             x86_cpu_apic_id_from_index(x86ms, i);
491         x86_topo_ids_from_apicid(ms->possible_cpus->cpus[i].arch_id,
492                                  &topo_info, &topo_ids);
493         ms->possible_cpus->cpus[i].props.has_socket_id = true;
494         ms->possible_cpus->cpus[i].props.socket_id = topo_ids.pkg_id;
495         if (ms->smp.dies > 1) {
496             ms->possible_cpus->cpus[i].props.has_die_id = true;
497             ms->possible_cpus->cpus[i].props.die_id = topo_ids.die_id;
498         }
499         ms->possible_cpus->cpus[i].props.has_core_id = true;
500         ms->possible_cpus->cpus[i].props.core_id = topo_ids.core_id;
501         ms->possible_cpus->cpus[i].props.has_thread_id = true;
502         ms->possible_cpus->cpus[i].props.thread_id = topo_ids.smt_id;
503     }
504     return ms->possible_cpus;
505 }
506 
507 static void x86_nmi(NMIState *n, int cpu_index, Error **errp)
508 {
509     /* cpu index isn't used */
510     CPUState *cs;
511 
512     CPU_FOREACH(cs) {
513         X86CPU *cpu = X86_CPU(cs);
514 
515         if (!cpu->apic_state) {
516             cpu_interrupt(cs, CPU_INTERRUPT_NMI);
517         } else {
518             apic_deliver_nmi(cpu->apic_state);
519         }
520     }
521 }
522 
523 static long get_file_size(FILE *f)
524 {
525     long where, size;
526 
527     /* XXX: on Unix systems, using fstat() probably makes more sense */
528 
529     where = ftell(f);
530     fseek(f, 0, SEEK_END);
531     size = ftell(f);
532     fseek(f, where, SEEK_SET);
533 
534     return size;
535 }
536 
537 /* TSC handling */
538 uint64_t cpu_get_tsc(CPUX86State *env)
539 {
540     return cpus_get_elapsed_ticks();
541 }
542 
543 /* IRQ handling */
544 static void pic_irq_request(void *opaque, int irq, int level)
545 {
546     CPUState *cs = first_cpu;
547     X86CPU *cpu = X86_CPU(cs);
548 
549     trace_x86_pic_interrupt(irq, level);
550     if (cpu->apic_state && !kvm_irqchip_in_kernel() &&
551         !whpx_apic_in_platform()) {
552         CPU_FOREACH(cs) {
553             cpu = X86_CPU(cs);
554             if (apic_accept_pic_intr(cpu->apic_state)) {
555                 apic_deliver_pic_intr(cpu->apic_state, level);
556             }
557         }
558     } else {
559         if (level) {
560             cpu_interrupt(cs, CPU_INTERRUPT_HARD);
561         } else {
562             cpu_reset_interrupt(cs, CPU_INTERRUPT_HARD);
563         }
564     }
565 }
566 
567 qemu_irq x86_allocate_cpu_irq(void)
568 {
569     return qemu_allocate_irq(pic_irq_request, NULL, 0);
570 }
571 
572 int cpu_get_pic_interrupt(CPUX86State *env)
573 {
574     X86CPU *cpu = env_archcpu(env);
575     int intno;
576 
577     if (!kvm_irqchip_in_kernel() && !whpx_apic_in_platform()) {
578         intno = apic_get_interrupt(cpu->apic_state);
579         if (intno >= 0) {
580             return intno;
581         }
582         /* read the irq from the PIC */
583         if (!apic_accept_pic_intr(cpu->apic_state)) {
584             return -1;
585         }
586     }
587 
588     intno = pic_read_irq(isa_pic);
589     return intno;
590 }
591 
592 DeviceState *cpu_get_current_apic(void)
593 {
594     if (current_cpu) {
595         X86CPU *cpu = X86_CPU(current_cpu);
596         return cpu->apic_state;
597     } else {
598         return NULL;
599     }
600 }
601 
602 void gsi_handler(void *opaque, int n, int level)
603 {
604     GSIState *s = opaque;
605 
606     trace_x86_gsi_interrupt(n, level);
607     switch (n) {
608     case 0 ... ISA_NUM_IRQS - 1:
609         if (s->i8259_irq[n]) {
610             /* Under KVM, Kernel will forward to both PIC and IOAPIC */
611             qemu_set_irq(s->i8259_irq[n], level);
612         }
613         /* fall through */
614     case ISA_NUM_IRQS ... IOAPIC_NUM_PINS - 1:
615 #ifdef CONFIG_XEN_EMU
616         /*
617          * Xen delivers the GSI to the Legacy PIC (not that Legacy PIC
618          * routing actually works properly under Xen). And then to
619          * *either* the PIRQ handling or the I/OAPIC depending on
620          * whether the former wants it.
621          */
622         if (xen_mode == XEN_EMULATE && xen_evtchn_set_gsi(n, level)) {
623             break;
624         }
625 #endif
626         qemu_set_irq(s->ioapic_irq[n], level);
627         break;
628     case IO_APIC_SECONDARY_IRQBASE
629         ... IO_APIC_SECONDARY_IRQBASE + IOAPIC_NUM_PINS - 1:
630         qemu_set_irq(s->ioapic2_irq[n - IO_APIC_SECONDARY_IRQBASE], level);
631         break;
632     }
633 }
634 
635 void ioapic_init_gsi(GSIState *gsi_state, const char *parent_name)
636 {
637     DeviceState *dev;
638     SysBusDevice *d;
639     unsigned int i;
640 
641     assert(parent_name);
642     if (kvm_ioapic_in_kernel()) {
643         dev = qdev_new(TYPE_KVM_IOAPIC);
644     } else {
645         dev = qdev_new(TYPE_IOAPIC);
646     }
647     object_property_add_child(object_resolve_path(parent_name, NULL),
648                               "ioapic", OBJECT(dev));
649     d = SYS_BUS_DEVICE(dev);
650     sysbus_realize_and_unref(d, &error_fatal);
651     sysbus_mmio_map(d, 0, IO_APIC_DEFAULT_ADDRESS);
652 
653     for (i = 0; i < IOAPIC_NUM_PINS; i++) {
654         gsi_state->ioapic_irq[i] = qdev_get_gpio_in(dev, i);
655     }
656 }
657 
658 DeviceState *ioapic_init_secondary(GSIState *gsi_state)
659 {
660     DeviceState *dev;
661     SysBusDevice *d;
662     unsigned int i;
663 
664     dev = qdev_new(TYPE_IOAPIC);
665     d = SYS_BUS_DEVICE(dev);
666     sysbus_realize_and_unref(d, &error_fatal);
667     sysbus_mmio_map(d, 0, IO_APIC_SECONDARY_ADDRESS);
668 
669     for (i = 0; i < IOAPIC_NUM_PINS; i++) {
670         gsi_state->ioapic2_irq[i] = qdev_get_gpio_in(dev, i);
671     }
672     return dev;
673 }
674 
675 struct setup_data {
676     uint64_t next;
677     uint32_t type;
678     uint32_t len;
679     uint8_t data[];
680 } __attribute__((packed));
681 
682 
683 /*
684  * The entry point into the kernel for PVH boot is different from
685  * the native entry point.  The PVH entry is defined by the x86/HVM
686  * direct boot ABI and is available in an ELFNOTE in the kernel binary.
687  *
688  * This function is passed to load_elf() when it is called from
689  * load_elfboot() which then additionally checks for an ELF Note of
690  * type XEN_ELFNOTE_PHYS32_ENTRY and passes it to this function to
691  * parse the PVH entry address from the ELF Note.
692  *
693  * Due to trickery in elf_opts.h, load_elf() is actually available as
694  * load_elf32() or load_elf64() and this routine needs to be able
695  * to deal with being called as 32 or 64 bit.
696  *
697  * The address of the PVH entry point is saved to the 'pvh_start_addr'
698  * global variable.  (although the entry point is 32-bit, the kernel
699  * binary can be either 32-bit or 64-bit).
700  */
701 static uint64_t read_pvh_start_addr(void *arg1, void *arg2, bool is64)
702 {
703     size_t *elf_note_data_addr;
704 
705     /* Check if ELF Note header passed in is valid */
706     if (arg1 == NULL) {
707         return 0;
708     }
709 
710     if (is64) {
711         struct elf64_note *nhdr64 = (struct elf64_note *)arg1;
712         uint64_t nhdr_size64 = sizeof(struct elf64_note);
713         uint64_t phdr_align = *(uint64_t *)arg2;
714         uint64_t nhdr_namesz = nhdr64->n_namesz;
715 
716         elf_note_data_addr =
717             ((void *)nhdr64) + nhdr_size64 +
718             QEMU_ALIGN_UP(nhdr_namesz, phdr_align);
719 
720         pvh_start_addr = *elf_note_data_addr;
721     } else {
722         struct elf32_note *nhdr32 = (struct elf32_note *)arg1;
723         uint32_t nhdr_size32 = sizeof(struct elf32_note);
724         uint32_t phdr_align = *(uint32_t *)arg2;
725         uint32_t nhdr_namesz = nhdr32->n_namesz;
726 
727         elf_note_data_addr =
728             ((void *)nhdr32) + nhdr_size32 +
729             QEMU_ALIGN_UP(nhdr_namesz, phdr_align);
730 
731         pvh_start_addr = *(uint32_t *)elf_note_data_addr;
732     }
733 
734     return pvh_start_addr;
735 }
736 
737 static bool load_elfboot(const char *kernel_filename,
738                          int kernel_file_size,
739                          uint8_t *header,
740                          size_t pvh_xen_start_addr,
741                          FWCfgState *fw_cfg)
742 {
743     uint32_t flags = 0;
744     uint32_t mh_load_addr = 0;
745     uint32_t elf_kernel_size = 0;
746     uint64_t elf_entry;
747     uint64_t elf_low, elf_high;
748     int kernel_size;
749 
750     if (ldl_p(header) != 0x464c457f) {
751         return false; /* no elfboot */
752     }
753 
754     bool elf_is64 = header[EI_CLASS] == ELFCLASS64;
755     flags = elf_is64 ?
756         ((Elf64_Ehdr *)header)->e_flags : ((Elf32_Ehdr *)header)->e_flags;
757 
758     if (flags & 0x00010004) { /* LOAD_ELF_HEADER_HAS_ADDR */
759         error_report("elfboot unsupported flags = %x", flags);
760         exit(1);
761     }
762 
763     uint64_t elf_note_type = XEN_ELFNOTE_PHYS32_ENTRY;
764     kernel_size = load_elf(kernel_filename, read_pvh_start_addr,
765                            NULL, &elf_note_type, &elf_entry,
766                            &elf_low, &elf_high, NULL, 0, I386_ELF_MACHINE,
767                            0, 0);
768 
769     if (kernel_size < 0) {
770         error_report("Error while loading elf kernel");
771         exit(1);
772     }
773     mh_load_addr = elf_low;
774     elf_kernel_size = elf_high - elf_low;
775 
776     if (pvh_start_addr == 0) {
777         error_report("Error loading uncompressed kernel without PVH ELF Note");
778         exit(1);
779     }
780     fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_ENTRY, pvh_start_addr);
781     fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_ADDR, mh_load_addr);
782     fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_SIZE, elf_kernel_size);
783 
784     return true;
785 }
786 
787 void x86_load_linux(X86MachineState *x86ms,
788                     FWCfgState *fw_cfg,
789                     int acpi_data_size,
790                     bool pvh_enabled)
791 {
792     bool linuxboot_dma_enabled = X86_MACHINE_GET_CLASS(x86ms)->fwcfg_dma_enabled;
793     uint16_t protocol;
794     int setup_size, kernel_size, cmdline_size;
795     int dtb_size, setup_data_offset;
796     uint32_t initrd_max;
797     uint8_t header[8192], *setup, *kernel;
798     hwaddr real_addr, prot_addr, cmdline_addr, initrd_addr = 0;
799     FILE *f;
800     char *vmode;
801     MachineState *machine = MACHINE(x86ms);
802     struct setup_data *setup_data;
803     const char *kernel_filename = machine->kernel_filename;
804     const char *initrd_filename = machine->initrd_filename;
805     const char *dtb_filename = machine->dtb;
806     const char *kernel_cmdline = machine->kernel_cmdline;
807     SevKernelLoaderContext sev_load_ctx = {};
808 
809     /* Align to 16 bytes as a paranoia measure */
810     cmdline_size = (strlen(kernel_cmdline) + 16) & ~15;
811 
812     /* load the kernel header */
813     f = fopen(kernel_filename, "rb");
814     if (!f) {
815         fprintf(stderr, "qemu: could not open kernel file '%s': %s\n",
816                 kernel_filename, strerror(errno));
817         exit(1);
818     }
819 
820     kernel_size = get_file_size(f);
821     if (!kernel_size ||
822         fread(header, 1, MIN(ARRAY_SIZE(header), kernel_size), f) !=
823         MIN(ARRAY_SIZE(header), kernel_size)) {
824         fprintf(stderr, "qemu: could not load kernel '%s': %s\n",
825                 kernel_filename, strerror(errno));
826         exit(1);
827     }
828 
829     /* kernel protocol version */
830     if (ldl_p(header + 0x202) == 0x53726448) {
831         protocol = lduw_p(header + 0x206);
832     } else {
833         /*
834          * This could be a multiboot kernel. If it is, let's stop treating it
835          * like a Linux kernel.
836          * Note: some multiboot images could be in the ELF format (the same of
837          * PVH), so we try multiboot first since we check the multiboot magic
838          * header before to load it.
839          */
840         if (load_multiboot(x86ms, fw_cfg, f, kernel_filename, initrd_filename,
841                            kernel_cmdline, kernel_size, header)) {
842             return;
843         }
844         /*
845          * Check if the file is an uncompressed kernel file (ELF) and load it,
846          * saving the PVH entry point used by the x86/HVM direct boot ABI.
847          * If load_elfboot() is successful, populate the fw_cfg info.
848          */
849         if (pvh_enabled &&
850             load_elfboot(kernel_filename, kernel_size,
851                          header, pvh_start_addr, fw_cfg)) {
852             fclose(f);
853 
854             fw_cfg_add_i32(fw_cfg, FW_CFG_CMDLINE_SIZE,
855                 strlen(kernel_cmdline) + 1);
856             fw_cfg_add_string(fw_cfg, FW_CFG_CMDLINE_DATA, kernel_cmdline);
857 
858             fw_cfg_add_i32(fw_cfg, FW_CFG_SETUP_SIZE, sizeof(header));
859             fw_cfg_add_bytes(fw_cfg, FW_CFG_SETUP_DATA,
860                              header, sizeof(header));
861 
862             /* load initrd */
863             if (initrd_filename) {
864                 GMappedFile *mapped_file;
865                 gsize initrd_size;
866                 gchar *initrd_data;
867                 GError *gerr = NULL;
868 
869                 mapped_file = g_mapped_file_new(initrd_filename, false, &gerr);
870                 if (!mapped_file) {
871                     fprintf(stderr, "qemu: error reading initrd %s: %s\n",
872                             initrd_filename, gerr->message);
873                     exit(1);
874                 }
875                 x86ms->initrd_mapped_file = mapped_file;
876 
877                 initrd_data = g_mapped_file_get_contents(mapped_file);
878                 initrd_size = g_mapped_file_get_length(mapped_file);
879                 initrd_max = x86ms->below_4g_mem_size - acpi_data_size - 1;
880                 if (initrd_size >= initrd_max) {
881                     fprintf(stderr, "qemu: initrd is too large, cannot support."
882                             "(max: %"PRIu32", need %"PRId64")\n",
883                             initrd_max, (uint64_t)initrd_size);
884                     exit(1);
885                 }
886 
887                 initrd_addr = (initrd_max - initrd_size) & ~4095;
888 
889                 fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_ADDR, initrd_addr);
890                 fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_SIZE, initrd_size);
891                 fw_cfg_add_bytes(fw_cfg, FW_CFG_INITRD_DATA, initrd_data,
892                                  initrd_size);
893             }
894 
895             option_rom[nb_option_roms].bootindex = 0;
896             option_rom[nb_option_roms].name = "pvh.bin";
897             nb_option_roms++;
898 
899             return;
900         }
901         protocol = 0;
902     }
903 
904     if (protocol < 0x200 || !(header[0x211] & 0x01)) {
905         /* Low kernel */
906         real_addr    = 0x90000;
907         cmdline_addr = 0x9a000 - cmdline_size;
908         prot_addr    = 0x10000;
909     } else if (protocol < 0x202) {
910         /* High but ancient kernel */
911         real_addr    = 0x90000;
912         cmdline_addr = 0x9a000 - cmdline_size;
913         prot_addr    = 0x100000;
914     } else {
915         /* High and recent kernel */
916         real_addr    = 0x10000;
917         cmdline_addr = 0x20000;
918         prot_addr    = 0x100000;
919     }
920 
921     /* highest address for loading the initrd */
922     if (protocol >= 0x20c &&
923         lduw_p(header + 0x236) & XLF_CAN_BE_LOADED_ABOVE_4G) {
924         /*
925          * Linux has supported initrd up to 4 GB for a very long time (2007,
926          * long before XLF_CAN_BE_LOADED_ABOVE_4G which was added in 2013),
927          * though it only sets initrd_max to 2 GB to "work around bootloader
928          * bugs". Luckily, QEMU firmware(which does something like bootloader)
929          * has supported this.
930          *
931          * It's believed that if XLF_CAN_BE_LOADED_ABOVE_4G is set, initrd can
932          * be loaded into any address.
933          *
934          * In addition, initrd_max is uint32_t simply because QEMU doesn't
935          * support the 64-bit boot protocol (specifically the ext_ramdisk_image
936          * field).
937          *
938          * Therefore here just limit initrd_max to UINT32_MAX simply as well.
939          */
940         initrd_max = UINT32_MAX;
941     } else if (protocol >= 0x203) {
942         initrd_max = ldl_p(header + 0x22c);
943     } else {
944         initrd_max = 0x37ffffff;
945     }
946 
947     if (initrd_max >= x86ms->below_4g_mem_size - acpi_data_size) {
948         initrd_max = x86ms->below_4g_mem_size - acpi_data_size - 1;
949     }
950 
951     fw_cfg_add_i32(fw_cfg, FW_CFG_CMDLINE_ADDR, cmdline_addr);
952     fw_cfg_add_i32(fw_cfg, FW_CFG_CMDLINE_SIZE, strlen(kernel_cmdline) + 1);
953     fw_cfg_add_string(fw_cfg, FW_CFG_CMDLINE_DATA, kernel_cmdline);
954     sev_load_ctx.cmdline_data = (char *)kernel_cmdline;
955     sev_load_ctx.cmdline_size = strlen(kernel_cmdline) + 1;
956 
957     if (protocol >= 0x202) {
958         stl_p(header + 0x228, cmdline_addr);
959     } else {
960         stw_p(header + 0x20, 0xA33F);
961         stw_p(header + 0x22, cmdline_addr - real_addr);
962     }
963 
964     /* handle vga= parameter */
965     vmode = strstr(kernel_cmdline, "vga=");
966     if (vmode) {
967         unsigned int video_mode;
968         const char *end;
969         int ret;
970         /* skip "vga=" */
971         vmode += 4;
972         if (!strncmp(vmode, "normal", 6)) {
973             video_mode = 0xffff;
974         } else if (!strncmp(vmode, "ext", 3)) {
975             video_mode = 0xfffe;
976         } else if (!strncmp(vmode, "ask", 3)) {
977             video_mode = 0xfffd;
978         } else {
979             ret = qemu_strtoui(vmode, &end, 0, &video_mode);
980             if (ret != 0 || (*end && *end != ' ')) {
981                 fprintf(stderr, "qemu: invalid 'vga=' kernel parameter.\n");
982                 exit(1);
983             }
984         }
985         stw_p(header + 0x1fa, video_mode);
986     }
987 
988     /* loader type */
989     /*
990      * High nybble = B reserved for QEMU; low nybble is revision number.
991      * If this code is substantially changed, you may want to consider
992      * incrementing the revision.
993      */
994     if (protocol >= 0x200) {
995         header[0x210] = 0xB0;
996     }
997     /* heap */
998     if (protocol >= 0x201) {
999         header[0x211] |= 0x80; /* CAN_USE_HEAP */
1000         stw_p(header + 0x224, cmdline_addr - real_addr - 0x200);
1001     }
1002 
1003     /* load initrd */
1004     if (initrd_filename) {
1005         GMappedFile *mapped_file;
1006         gsize initrd_size;
1007         gchar *initrd_data;
1008         GError *gerr = NULL;
1009 
1010         if (protocol < 0x200) {
1011             fprintf(stderr, "qemu: linux kernel too old to load a ram disk\n");
1012             exit(1);
1013         }
1014 
1015         mapped_file = g_mapped_file_new(initrd_filename, false, &gerr);
1016         if (!mapped_file) {
1017             fprintf(stderr, "qemu: error reading initrd %s: %s\n",
1018                     initrd_filename, gerr->message);
1019             exit(1);
1020         }
1021         x86ms->initrd_mapped_file = mapped_file;
1022 
1023         initrd_data = g_mapped_file_get_contents(mapped_file);
1024         initrd_size = g_mapped_file_get_length(mapped_file);
1025         if (initrd_size >= initrd_max) {
1026             fprintf(stderr, "qemu: initrd is too large, cannot support."
1027                     "(max: %"PRIu32", need %"PRId64")\n",
1028                     initrd_max, (uint64_t)initrd_size);
1029             exit(1);
1030         }
1031 
1032         initrd_addr = (initrd_max - initrd_size) & ~4095;
1033 
1034         fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_ADDR, initrd_addr);
1035         fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_SIZE, initrd_size);
1036         fw_cfg_add_bytes(fw_cfg, FW_CFG_INITRD_DATA, initrd_data, initrd_size);
1037         sev_load_ctx.initrd_data = initrd_data;
1038         sev_load_ctx.initrd_size = initrd_size;
1039 
1040         stl_p(header + 0x218, initrd_addr);
1041         stl_p(header + 0x21c, initrd_size);
1042     }
1043 
1044     /* load kernel and setup */
1045     setup_size = header[0x1f1];
1046     if (setup_size == 0) {
1047         setup_size = 4;
1048     }
1049     setup_size = (setup_size + 1) * 512;
1050     if (setup_size > kernel_size) {
1051         fprintf(stderr, "qemu: invalid kernel header\n");
1052         exit(1);
1053     }
1054     kernel_size -= setup_size;
1055 
1056     setup  = g_malloc(setup_size);
1057     kernel = g_malloc(kernel_size);
1058     fseek(f, 0, SEEK_SET);
1059     if (fread(setup, 1, setup_size, f) != setup_size) {
1060         fprintf(stderr, "fread() failed\n");
1061         exit(1);
1062     }
1063     if (fread(kernel, 1, kernel_size, f) != kernel_size) {
1064         fprintf(stderr, "fread() failed\n");
1065         exit(1);
1066     }
1067     fclose(f);
1068 
1069     /* append dtb to kernel */
1070     if (dtb_filename) {
1071         if (protocol < 0x209) {
1072             fprintf(stderr, "qemu: Linux kernel too old to load a dtb\n");
1073             exit(1);
1074         }
1075 
1076         dtb_size = get_image_size(dtb_filename);
1077         if (dtb_size <= 0) {
1078             fprintf(stderr, "qemu: error reading dtb %s: %s\n",
1079                     dtb_filename, strerror(errno));
1080             exit(1);
1081         }
1082 
1083         setup_data_offset = QEMU_ALIGN_UP(kernel_size, 16);
1084         kernel_size = setup_data_offset + sizeof(struct setup_data) + dtb_size;
1085         kernel = g_realloc(kernel, kernel_size);
1086 
1087         stq_p(header + 0x250, prot_addr + setup_data_offset);
1088 
1089         setup_data = (struct setup_data *)(kernel + setup_data_offset);
1090         setup_data->next = 0;
1091         setup_data->type = cpu_to_le32(SETUP_DTB);
1092         setup_data->len = cpu_to_le32(dtb_size);
1093 
1094         load_image_size(dtb_filename, setup_data->data, dtb_size);
1095     }
1096 
1097     /*
1098      * If we're starting an encrypted VM, it will be OVMF based, which uses the
1099      * efi stub for booting and doesn't require any values to be placed in the
1100      * kernel header.  We therefore don't update the header so the hash of the
1101      * kernel on the other side of the fw_cfg interface matches the hash of the
1102      * file the user passed in.
1103      */
1104     if (!sev_enabled()) {
1105         memcpy(setup, header, MIN(sizeof(header), setup_size));
1106     }
1107 
1108     fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_ADDR, prot_addr);
1109     fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_SIZE, kernel_size);
1110     fw_cfg_add_bytes(fw_cfg, FW_CFG_KERNEL_DATA, kernel, kernel_size);
1111     sev_load_ctx.kernel_data = (char *)kernel;
1112     sev_load_ctx.kernel_size = kernel_size;
1113 
1114     fw_cfg_add_i32(fw_cfg, FW_CFG_SETUP_ADDR, real_addr);
1115     fw_cfg_add_i32(fw_cfg, FW_CFG_SETUP_SIZE, setup_size);
1116     fw_cfg_add_bytes(fw_cfg, FW_CFG_SETUP_DATA, setup, setup_size);
1117     sev_load_ctx.setup_data = (char *)setup;
1118     sev_load_ctx.setup_size = setup_size;
1119 
1120     if (sev_enabled()) {
1121         sev_add_kernel_loader_hashes(&sev_load_ctx, &error_fatal);
1122     }
1123 
1124     option_rom[nb_option_roms].bootindex = 0;
1125     option_rom[nb_option_roms].name = "linuxboot.bin";
1126     if (linuxboot_dma_enabled && fw_cfg_dma_enabled(fw_cfg)) {
1127         option_rom[nb_option_roms].name = "linuxboot_dma.bin";
1128     }
1129     nb_option_roms++;
1130 }
1131 
1132 void x86_bios_rom_init(MachineState *ms, const char *default_firmware,
1133                        MemoryRegion *rom_memory, bool isapc_ram_fw)
1134 {
1135     const char *bios_name;
1136     char *filename;
1137     MemoryRegion *bios, *isa_bios;
1138     int bios_size, isa_bios_size;
1139     ssize_t ret;
1140 
1141     /* BIOS load */
1142     bios_name = ms->firmware ?: default_firmware;
1143     filename = qemu_find_file(QEMU_FILE_TYPE_BIOS, bios_name);
1144     if (filename) {
1145         bios_size = get_image_size(filename);
1146     } else {
1147         bios_size = -1;
1148     }
1149     if (bios_size <= 0 ||
1150         (bios_size % 65536) != 0) {
1151         goto bios_error;
1152     }
1153     bios = g_malloc(sizeof(*bios));
1154     memory_region_init_ram(bios, NULL, "pc.bios", bios_size, &error_fatal);
1155     if (sev_enabled()) {
1156         /*
1157          * The concept of a "reset" simply doesn't exist for
1158          * confidential computing guests, we have to destroy and
1159          * re-launch them instead.  So there is no need to register
1160          * the firmware as rom to properly re-initialize on reset.
1161          * Just go for a straight file load instead.
1162          */
1163         void *ptr = memory_region_get_ram_ptr(bios);
1164         load_image_size(filename, ptr, bios_size);
1165         x86_firmware_configure(ptr, bios_size);
1166     } else {
1167         if (!isapc_ram_fw) {
1168             memory_region_set_readonly(bios, true);
1169         }
1170         ret = rom_add_file_fixed(bios_name, (uint32_t)(-bios_size), -1);
1171         if (ret != 0) {
1172             goto bios_error;
1173         }
1174     }
1175     g_free(filename);
1176 
1177     /* map the last 128KB of the BIOS in ISA space */
1178     isa_bios_size = MIN(bios_size, 128 * KiB);
1179     isa_bios = g_malloc(sizeof(*isa_bios));
1180     memory_region_init_alias(isa_bios, NULL, "isa-bios", bios,
1181                              bios_size - isa_bios_size, isa_bios_size);
1182     memory_region_add_subregion_overlap(rom_memory,
1183                                         0x100000 - isa_bios_size,
1184                                         isa_bios,
1185                                         1);
1186     if (!isapc_ram_fw) {
1187         memory_region_set_readonly(isa_bios, true);
1188     }
1189 
1190     /* map all the bios at the top of memory */
1191     memory_region_add_subregion(rom_memory,
1192                                 (uint32_t)(-bios_size),
1193                                 bios);
1194     return;
1195 
1196 bios_error:
1197     fprintf(stderr, "qemu: could not load PC BIOS '%s'\n", bios_name);
1198     exit(1);
1199 }
1200 
1201 bool x86_machine_is_smm_enabled(const X86MachineState *x86ms)
1202 {
1203     bool smm_available = false;
1204 
1205     if (x86ms->smm == ON_OFF_AUTO_OFF) {
1206         return false;
1207     }
1208 
1209     if (tcg_enabled() || qtest_enabled()) {
1210         smm_available = true;
1211     } else if (kvm_enabled()) {
1212         smm_available = kvm_has_smm();
1213     }
1214 
1215     if (smm_available) {
1216         return true;
1217     }
1218 
1219     if (x86ms->smm == ON_OFF_AUTO_ON) {
1220         error_report("System Management Mode not supported by this hypervisor.");
1221         exit(1);
1222     }
1223     return false;
1224 }
1225 
1226 static void x86_machine_get_smm(Object *obj, Visitor *v, const char *name,
1227                                void *opaque, Error **errp)
1228 {
1229     X86MachineState *x86ms = X86_MACHINE(obj);
1230     OnOffAuto smm = x86ms->smm;
1231 
1232     visit_type_OnOffAuto(v, name, &smm, errp);
1233 }
1234 
1235 static void x86_machine_set_smm(Object *obj, Visitor *v, const char *name,
1236                                void *opaque, Error **errp)
1237 {
1238     X86MachineState *x86ms = X86_MACHINE(obj);
1239 
1240     visit_type_OnOffAuto(v, name, &x86ms->smm, errp);
1241 }
1242 
1243 bool x86_machine_is_acpi_enabled(const X86MachineState *x86ms)
1244 {
1245     if (x86ms->acpi == ON_OFF_AUTO_OFF) {
1246         return false;
1247     }
1248     return true;
1249 }
1250 
1251 static void x86_machine_get_acpi(Object *obj, Visitor *v, const char *name,
1252                                  void *opaque, Error **errp)
1253 {
1254     X86MachineState *x86ms = X86_MACHINE(obj);
1255     OnOffAuto acpi = x86ms->acpi;
1256 
1257     visit_type_OnOffAuto(v, name, &acpi, errp);
1258 }
1259 
1260 static void x86_machine_set_acpi(Object *obj, Visitor *v, const char *name,
1261                                  void *opaque, Error **errp)
1262 {
1263     X86MachineState *x86ms = X86_MACHINE(obj);
1264 
1265     visit_type_OnOffAuto(v, name, &x86ms->acpi, errp);
1266 }
1267 
1268 static void x86_machine_get_pit(Object *obj, Visitor *v, const char *name,
1269                                     void *opaque, Error **errp)
1270 {
1271     X86MachineState *x86ms = X86_MACHINE(obj);
1272     OnOffAuto pit = x86ms->pit;
1273 
1274     visit_type_OnOffAuto(v, name, &pit, errp);
1275 }
1276 
1277 static void x86_machine_set_pit(Object *obj, Visitor *v, const char *name,
1278                                     void *opaque, Error **errp)
1279 {
1280     X86MachineState *x86ms = X86_MACHINE(obj);;
1281 
1282     visit_type_OnOffAuto(v, name, &x86ms->pit, errp);
1283 }
1284 
1285 static void x86_machine_get_pic(Object *obj, Visitor *v, const char *name,
1286                                 void *opaque, Error **errp)
1287 {
1288     X86MachineState *x86ms = X86_MACHINE(obj);
1289     OnOffAuto pic = x86ms->pic;
1290 
1291     visit_type_OnOffAuto(v, name, &pic, errp);
1292 }
1293 
1294 static void x86_machine_set_pic(Object *obj, Visitor *v, const char *name,
1295                                 void *opaque, Error **errp)
1296 {
1297     X86MachineState *x86ms = X86_MACHINE(obj);
1298 
1299     visit_type_OnOffAuto(v, name, &x86ms->pic, errp);
1300 }
1301 
1302 static char *x86_machine_get_oem_id(Object *obj, Error **errp)
1303 {
1304     X86MachineState *x86ms = X86_MACHINE(obj);
1305 
1306     return g_strdup(x86ms->oem_id);
1307 }
1308 
1309 static void x86_machine_set_oem_id(Object *obj, const char *value, Error **errp)
1310 {
1311     X86MachineState *x86ms = X86_MACHINE(obj);
1312     size_t len = strlen(value);
1313 
1314     if (len > 6) {
1315         error_setg(errp,
1316                    "User specified "X86_MACHINE_OEM_ID" value is bigger than "
1317                    "6 bytes in size");
1318         return;
1319     }
1320 
1321     strncpy(x86ms->oem_id, value, 6);
1322 }
1323 
1324 static char *x86_machine_get_oem_table_id(Object *obj, Error **errp)
1325 {
1326     X86MachineState *x86ms = X86_MACHINE(obj);
1327 
1328     return g_strdup(x86ms->oem_table_id);
1329 }
1330 
1331 static void x86_machine_set_oem_table_id(Object *obj, const char *value,
1332                                          Error **errp)
1333 {
1334     X86MachineState *x86ms = X86_MACHINE(obj);
1335     size_t len = strlen(value);
1336 
1337     if (len > 8) {
1338         error_setg(errp,
1339                    "User specified "X86_MACHINE_OEM_TABLE_ID
1340                    " value is bigger than "
1341                    "8 bytes in size");
1342         return;
1343     }
1344     strncpy(x86ms->oem_table_id, value, 8);
1345 }
1346 
1347 static void x86_machine_get_bus_lock_ratelimit(Object *obj, Visitor *v,
1348                                 const char *name, void *opaque, Error **errp)
1349 {
1350     X86MachineState *x86ms = X86_MACHINE(obj);
1351     uint64_t bus_lock_ratelimit = x86ms->bus_lock_ratelimit;
1352 
1353     visit_type_uint64(v, name, &bus_lock_ratelimit, errp);
1354 }
1355 
1356 static void x86_machine_set_bus_lock_ratelimit(Object *obj, Visitor *v,
1357                                const char *name, void *opaque, Error **errp)
1358 {
1359     X86MachineState *x86ms = X86_MACHINE(obj);
1360 
1361     visit_type_uint64(v, name, &x86ms->bus_lock_ratelimit, errp);
1362 }
1363 
1364 static void machine_get_sgx_epc(Object *obj, Visitor *v, const char *name,
1365                                 void *opaque, Error **errp)
1366 {
1367     X86MachineState *x86ms = X86_MACHINE(obj);
1368     SgxEPCList *list = x86ms->sgx_epc_list;
1369 
1370     visit_type_SgxEPCList(v, name, &list, errp);
1371 }
1372 
1373 static void machine_set_sgx_epc(Object *obj, Visitor *v, const char *name,
1374                                 void *opaque, Error **errp)
1375 {
1376     X86MachineState *x86ms = X86_MACHINE(obj);
1377     SgxEPCList *list;
1378 
1379     list = x86ms->sgx_epc_list;
1380     visit_type_SgxEPCList(v, name, &x86ms->sgx_epc_list, errp);
1381 
1382     qapi_free_SgxEPCList(list);
1383 }
1384 
1385 static void x86_machine_initfn(Object *obj)
1386 {
1387     X86MachineState *x86ms = X86_MACHINE(obj);
1388 
1389     x86ms->smm = ON_OFF_AUTO_AUTO;
1390     x86ms->acpi = ON_OFF_AUTO_AUTO;
1391     x86ms->pit = ON_OFF_AUTO_AUTO;
1392     x86ms->pic = ON_OFF_AUTO_AUTO;
1393     x86ms->pci_irq_mask = ACPI_BUILD_PCI_IRQS;
1394     x86ms->oem_id = g_strndup(ACPI_BUILD_APPNAME6, 6);
1395     x86ms->oem_table_id = g_strndup(ACPI_BUILD_APPNAME8, 8);
1396     x86ms->bus_lock_ratelimit = 0;
1397     x86ms->above_4g_mem_start = 4 * GiB;
1398 }
1399 
1400 static void x86_machine_class_init(ObjectClass *oc, void *data)
1401 {
1402     MachineClass *mc = MACHINE_CLASS(oc);
1403     X86MachineClass *x86mc = X86_MACHINE_CLASS(oc);
1404     NMIClass *nc = NMI_CLASS(oc);
1405 
1406     mc->cpu_index_to_instance_props = x86_cpu_index_to_props;
1407     mc->get_default_cpu_node_id = x86_get_default_cpu_node_id;
1408     mc->possible_cpu_arch_ids = x86_possible_cpu_arch_ids;
1409     x86mc->save_tsc_khz = true;
1410     x86mc->fwcfg_dma_enabled = true;
1411     nc->nmi_monitor_handler = x86_nmi;
1412 
1413     object_class_property_add(oc, X86_MACHINE_SMM, "OnOffAuto",
1414         x86_machine_get_smm, x86_machine_set_smm,
1415         NULL, NULL);
1416     object_class_property_set_description(oc, X86_MACHINE_SMM,
1417         "Enable SMM");
1418 
1419     object_class_property_add(oc, X86_MACHINE_ACPI, "OnOffAuto",
1420         x86_machine_get_acpi, x86_machine_set_acpi,
1421         NULL, NULL);
1422     object_class_property_set_description(oc, X86_MACHINE_ACPI,
1423         "Enable ACPI");
1424 
1425     object_class_property_add(oc, X86_MACHINE_PIT, "OnOffAuto",
1426                               x86_machine_get_pit,
1427                               x86_machine_set_pit,
1428                               NULL, NULL);
1429     object_class_property_set_description(oc, X86_MACHINE_PIT,
1430         "Enable i8254 PIT");
1431 
1432     object_class_property_add(oc, X86_MACHINE_PIC, "OnOffAuto",
1433                               x86_machine_get_pic,
1434                               x86_machine_set_pic,
1435                               NULL, NULL);
1436     object_class_property_set_description(oc, X86_MACHINE_PIC,
1437         "Enable i8259 PIC");
1438 
1439     object_class_property_add_str(oc, X86_MACHINE_OEM_ID,
1440                                   x86_machine_get_oem_id,
1441                                   x86_machine_set_oem_id);
1442     object_class_property_set_description(oc, X86_MACHINE_OEM_ID,
1443                                           "Override the default value of field OEMID "
1444                                           "in ACPI table header."
1445                                           "The string may be up to 6 bytes in size");
1446 
1447 
1448     object_class_property_add_str(oc, X86_MACHINE_OEM_TABLE_ID,
1449                                   x86_machine_get_oem_table_id,
1450                                   x86_machine_set_oem_table_id);
1451     object_class_property_set_description(oc, X86_MACHINE_OEM_TABLE_ID,
1452                                           "Override the default value of field OEM Table ID "
1453                                           "in ACPI table header."
1454                                           "The string may be up to 8 bytes in size");
1455 
1456     object_class_property_add(oc, X86_MACHINE_BUS_LOCK_RATELIMIT, "uint64_t",
1457                                 x86_machine_get_bus_lock_ratelimit,
1458                                 x86_machine_set_bus_lock_ratelimit, NULL, NULL);
1459     object_class_property_set_description(oc, X86_MACHINE_BUS_LOCK_RATELIMIT,
1460             "Set the ratelimit for the bus locks acquired in VMs");
1461 
1462     object_class_property_add(oc, "sgx-epc", "SgxEPC",
1463         machine_get_sgx_epc, machine_set_sgx_epc,
1464         NULL, NULL);
1465     object_class_property_set_description(oc, "sgx-epc",
1466         "SGX EPC device");
1467 }
1468 
1469 static const TypeInfo x86_machine_info = {
1470     .name = TYPE_X86_MACHINE,
1471     .parent = TYPE_MACHINE,
1472     .abstract = true,
1473     .instance_size = sizeof(X86MachineState),
1474     .instance_init = x86_machine_initfn,
1475     .class_size = sizeof(X86MachineClass),
1476     .class_init = x86_machine_class_init,
1477     .interfaces = (InterfaceInfo[]) {
1478          { TYPE_NMI },
1479          { }
1480     },
1481 };
1482 
1483 static void x86_machine_register_types(void)
1484 {
1485     type_register_static(&x86_machine_info);
1486 }
1487 
1488 type_init(x86_machine_register_types)
1489