xref: /openbmc/qemu/hw/i386/x86.c (revision feb6cb93)
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 "qemu/guest-random.h"
30 #include "qapi/error.h"
31 #include "qapi/qmp/qerror.h"
32 #include "qapi/qapi-visit-common.h"
33 #include "qapi/clone-visitor.h"
34 #include "qapi/qapi-visit-machine.h"
35 #include "qapi/visitor.h"
36 #include "sysemu/qtest.h"
37 #include "sysemu/whpx.h"
38 #include "sysemu/numa.h"
39 #include "sysemu/replay.h"
40 #include "sysemu/sysemu.h"
41 #include "sysemu/cpu-timers.h"
42 #include "sysemu/xen.h"
43 #include "trace.h"
44 
45 #include "hw/i386/x86.h"
46 #include "target/i386/cpu.h"
47 #include "hw/i386/topology.h"
48 #include "hw/i386/fw_cfg.h"
49 #include "hw/intc/i8259.h"
50 #include "hw/rtc/mc146818rtc.h"
51 #include "target/i386/sev.h"
52 
53 #include "hw/acpi/cpu_hotplug.h"
54 #include "hw/irq.h"
55 #include "hw/nmi.h"
56 #include "hw/loader.h"
57 #include "multiboot.h"
58 #include "elf.h"
59 #include "standard-headers/asm-x86/bootparam.h"
60 #include CONFIG_DEVICES
61 #include "kvm/kvm_i386.h"
62 
63 /* Physical Address of PVH entry point read from kernel ELF NOTE */
64 static size_t pvh_start_addr;
65 
66 inline void init_topo_info(X86CPUTopoInfo *topo_info,
67                            const X86MachineState *x86ms)
68 {
69     MachineState *ms = MACHINE(x86ms);
70 
71     topo_info->dies_per_pkg = ms->smp.dies;
72     topo_info->cores_per_die = ms->smp.cores;
73     topo_info->threads_per_core = ms->smp.threads;
74 }
75 
76 /*
77  * Calculates initial APIC ID for a specific CPU index
78  *
79  * Currently we need to be able to calculate the APIC ID from the CPU index
80  * alone (without requiring a CPU object), as the QEMU<->Seabios interfaces have
81  * no concept of "CPU index", and the NUMA tables on fw_cfg need the APIC ID of
82  * all CPUs up to max_cpus.
83  */
84 uint32_t x86_cpu_apic_id_from_index(X86MachineState *x86ms,
85                                     unsigned int cpu_index)
86 {
87     X86CPUTopoInfo topo_info;
88 
89     init_topo_info(&topo_info, x86ms);
90 
91     return x86_apicid_from_cpu_idx(&topo_info, cpu_index);
92 }
93 
94 
95 void x86_cpu_new(X86MachineState *x86ms, int64_t apic_id, Error **errp)
96 {
97     Object *cpu = object_new(MACHINE(x86ms)->cpu_type);
98 
99     if (!object_property_set_uint(cpu, "apic-id", apic_id, errp)) {
100         goto out;
101     }
102     qdev_realize(DEVICE(cpu), NULL, errp);
103 
104 out:
105     object_unref(cpu);
106 }
107 
108 void x86_cpus_init(X86MachineState *x86ms, int default_cpu_version)
109 {
110     int i;
111     const CPUArchIdList *possible_cpus;
112     MachineState *ms = MACHINE(x86ms);
113     MachineClass *mc = MACHINE_GET_CLASS(x86ms);
114 
115     x86_cpu_set_default_version(default_cpu_version);
116 
117     /*
118      * Calculates the limit to CPU APIC ID values
119      *
120      * Limit for the APIC ID value, so that all
121      * CPU APIC IDs are < x86ms->apic_id_limit.
122      *
123      * This is used for FW_CFG_MAX_CPUS. See comments on fw_cfg_arch_create().
124      */
125     x86ms->apic_id_limit = x86_cpu_apic_id_from_index(x86ms,
126                                                       ms->smp.max_cpus - 1) + 1;
127 
128     /*
129      * Can we support APIC ID 255 or higher?
130      *
131      * Under Xen: yes.
132      * With userspace emulated lapic: no
133      * With KVM's in-kernel lapic: only if X2APIC API is enabled.
134      */
135     if (x86ms->apic_id_limit > 255 && !xen_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     possible_cpus = mc->possible_cpu_arch_ids(ms);
143     for (i = 0; i < ms->smp.cpus; i++) {
144         x86_cpu_new(x86ms, possible_cpus->cpus[i].arch_id, &error_fatal);
145     }
146 }
147 
148 void x86_rtc_set_cpus_count(ISADevice *rtc, uint16_t cpus_count)
149 {
150     if (cpus_count > 0xff) {
151         /*
152          * If the number of CPUs can't be represented in 8 bits, the
153          * BIOS must use "FW_CFG_NB_CPUS". Set RTC field to 0 just
154          * to make old BIOSes fail more predictably.
155          */
156         rtc_set_memory(rtc, 0x5f, 0);
157     } else {
158         rtc_set_memory(rtc, 0x5f, cpus_count - 1);
159     }
160 }
161 
162 static int x86_apic_cmp(const void *a, const void *b)
163 {
164    CPUArchId *apic_a = (CPUArchId *)a;
165    CPUArchId *apic_b = (CPUArchId *)b;
166 
167    return apic_a->arch_id - apic_b->arch_id;
168 }
169 
170 /*
171  * returns pointer to CPUArchId descriptor that matches CPU's apic_id
172  * in ms->possible_cpus->cpus, if ms->possible_cpus->cpus has no
173  * entry corresponding to CPU's apic_id returns NULL.
174  */
175 CPUArchId *x86_find_cpu_slot(MachineState *ms, uint32_t id, int *idx)
176 {
177     CPUArchId apic_id, *found_cpu;
178 
179     apic_id.arch_id = id;
180     found_cpu = bsearch(&apic_id, ms->possible_cpus->cpus,
181         ms->possible_cpus->len, sizeof(*ms->possible_cpus->cpus),
182         x86_apic_cmp);
183     if (found_cpu && idx) {
184         *idx = found_cpu - ms->possible_cpus->cpus;
185     }
186     return found_cpu;
187 }
188 
189 void x86_cpu_plug(HotplugHandler *hotplug_dev,
190                   DeviceState *dev, Error **errp)
191 {
192     CPUArchId *found_cpu;
193     Error *local_err = NULL;
194     X86CPU *cpu = X86_CPU(dev);
195     X86MachineState *x86ms = X86_MACHINE(hotplug_dev);
196 
197     if (x86ms->acpi_dev) {
198         hotplug_handler_plug(x86ms->acpi_dev, dev, &local_err);
199         if (local_err) {
200             goto out;
201         }
202     }
203 
204     /* increment the number of CPUs */
205     x86ms->boot_cpus++;
206     if (x86ms->rtc) {
207         x86_rtc_set_cpus_count(x86ms->rtc, x86ms->boot_cpus);
208     }
209     if (x86ms->fw_cfg) {
210         fw_cfg_modify_i16(x86ms->fw_cfg, FW_CFG_NB_CPUS, x86ms->boot_cpus);
211     }
212 
213     found_cpu = x86_find_cpu_slot(MACHINE(x86ms), cpu->apic_id, NULL);
214     found_cpu->cpu = OBJECT(dev);
215 out:
216     error_propagate(errp, local_err);
217 }
218 
219 void x86_cpu_unplug_request_cb(HotplugHandler *hotplug_dev,
220                                DeviceState *dev, Error **errp)
221 {
222     int idx = -1;
223     X86CPU *cpu = X86_CPU(dev);
224     X86MachineState *x86ms = X86_MACHINE(hotplug_dev);
225 
226     if (!x86ms->acpi_dev) {
227         error_setg(errp, "CPU hot unplug not supported without ACPI");
228         return;
229     }
230 
231     x86_find_cpu_slot(MACHINE(x86ms), cpu->apic_id, &idx);
232     assert(idx != -1);
233     if (idx == 0) {
234         error_setg(errp, "Boot CPU is unpluggable");
235         return;
236     }
237 
238     hotplug_handler_unplug_request(x86ms->acpi_dev, dev,
239                                    errp);
240 }
241 
242 void x86_cpu_unplug_cb(HotplugHandler *hotplug_dev,
243                        DeviceState *dev, Error **errp)
244 {
245     CPUArchId *found_cpu;
246     Error *local_err = NULL;
247     X86CPU *cpu = X86_CPU(dev);
248     X86MachineState *x86ms = X86_MACHINE(hotplug_dev);
249 
250     hotplug_handler_unplug(x86ms->acpi_dev, dev, &local_err);
251     if (local_err) {
252         goto out;
253     }
254 
255     found_cpu = x86_find_cpu_slot(MACHINE(x86ms), cpu->apic_id, NULL);
256     found_cpu->cpu = NULL;
257     qdev_unrealize(dev);
258 
259     /* decrement the number of CPUs */
260     x86ms->boot_cpus--;
261     /* Update the number of CPUs in CMOS */
262     x86_rtc_set_cpus_count(x86ms->rtc, x86ms->boot_cpus);
263     fw_cfg_modify_i16(x86ms->fw_cfg, FW_CFG_NB_CPUS, x86ms->boot_cpus);
264  out:
265     error_propagate(errp, local_err);
266 }
267 
268 void x86_cpu_pre_plug(HotplugHandler *hotplug_dev,
269                       DeviceState *dev, Error **errp)
270 {
271     int idx;
272     CPUState *cs;
273     CPUArchId *cpu_slot;
274     X86CPUTopoIDs topo_ids;
275     X86CPU *cpu = X86_CPU(dev);
276     CPUX86State *env = &cpu->env;
277     MachineState *ms = MACHINE(hotplug_dev);
278     X86MachineState *x86ms = X86_MACHINE(hotplug_dev);
279     unsigned int smp_cores = ms->smp.cores;
280     unsigned int smp_threads = ms->smp.threads;
281     X86CPUTopoInfo topo_info;
282 
283     if (!object_dynamic_cast(OBJECT(cpu), ms->cpu_type)) {
284         error_setg(errp, "Invalid CPU type, expected cpu type: '%s'",
285                    ms->cpu_type);
286         return;
287     }
288 
289     if (x86ms->acpi_dev) {
290         Error *local_err = NULL;
291 
292         hotplug_handler_pre_plug(HOTPLUG_HANDLER(x86ms->acpi_dev), dev,
293                                  &local_err);
294         if (local_err) {
295             error_propagate(errp, local_err);
296             return;
297         }
298     }
299 
300     init_topo_info(&topo_info, x86ms);
301 
302     env->nr_dies = ms->smp.dies;
303 
304     /*
305      * If APIC ID is not set,
306      * set it based on socket/die/core/thread properties.
307      */
308     if (cpu->apic_id == UNASSIGNED_APIC_ID) {
309         int max_socket = (ms->smp.max_cpus - 1) /
310                                 smp_threads / smp_cores / ms->smp.dies;
311 
312         /*
313          * die-id was optional in QEMU 4.0 and older, so keep it optional
314          * if there's only one die per socket.
315          */
316         if (cpu->die_id < 0 && ms->smp.dies == 1) {
317             cpu->die_id = 0;
318         }
319 
320         if (cpu->socket_id < 0) {
321             error_setg(errp, "CPU socket-id is not set");
322             return;
323         } else if (cpu->socket_id > max_socket) {
324             error_setg(errp, "Invalid CPU socket-id: %u must be in range 0:%u",
325                        cpu->socket_id, max_socket);
326             return;
327         }
328         if (cpu->die_id < 0) {
329             error_setg(errp, "CPU die-id is not set");
330             return;
331         } else if (cpu->die_id > ms->smp.dies - 1) {
332             error_setg(errp, "Invalid CPU die-id: %u must be in range 0:%u",
333                        cpu->die_id, ms->smp.dies - 1);
334             return;
335         }
336         if (cpu->core_id < 0) {
337             error_setg(errp, "CPU core-id is not set");
338             return;
339         } else if (cpu->core_id > (smp_cores - 1)) {
340             error_setg(errp, "Invalid CPU core-id: %u must be in range 0:%u",
341                        cpu->core_id, smp_cores - 1);
342             return;
343         }
344         if (cpu->thread_id < 0) {
345             error_setg(errp, "CPU thread-id is not set");
346             return;
347         } else if (cpu->thread_id > (smp_threads - 1)) {
348             error_setg(errp, "Invalid CPU thread-id: %u must be in range 0:%u",
349                        cpu->thread_id, smp_threads - 1);
350             return;
351         }
352 
353         topo_ids.pkg_id = cpu->socket_id;
354         topo_ids.die_id = cpu->die_id;
355         topo_ids.core_id = cpu->core_id;
356         topo_ids.smt_id = cpu->thread_id;
357         cpu->apic_id = x86_apicid_from_topo_ids(&topo_info, &topo_ids);
358     }
359 
360     cpu_slot = x86_find_cpu_slot(MACHINE(x86ms), cpu->apic_id, &idx);
361     if (!cpu_slot) {
362         MachineState *ms = MACHINE(x86ms);
363 
364         x86_topo_ids_from_apicid(cpu->apic_id, &topo_info, &topo_ids);
365         error_setg(errp,
366             "Invalid CPU [socket: %u, die: %u, core: %u, thread: %u] with"
367             " APIC ID %" PRIu32 ", valid index range 0:%d",
368             topo_ids.pkg_id, topo_ids.die_id, topo_ids.core_id, topo_ids.smt_id,
369             cpu->apic_id, ms->possible_cpus->len - 1);
370         return;
371     }
372 
373     if (cpu_slot->cpu) {
374         error_setg(errp, "CPU[%d] with APIC ID %" PRIu32 " exists",
375                    idx, cpu->apic_id);
376         return;
377     }
378 
379     /* if 'address' properties socket-id/core-id/thread-id are not set, set them
380      * so that machine_query_hotpluggable_cpus would show correct values
381      */
382     /* TODO: move socket_id/core_id/thread_id checks into x86_cpu_realizefn()
383      * once -smp refactoring is complete and there will be CPU private
384      * CPUState::nr_cores and CPUState::nr_threads fields instead of globals */
385     x86_topo_ids_from_apicid(cpu->apic_id, &topo_info, &topo_ids);
386     if (cpu->socket_id != -1 && cpu->socket_id != topo_ids.pkg_id) {
387         error_setg(errp, "property socket-id: %u doesn't match set apic-id:"
388             " 0x%x (socket-id: %u)", cpu->socket_id, cpu->apic_id,
389             topo_ids.pkg_id);
390         return;
391     }
392     cpu->socket_id = topo_ids.pkg_id;
393 
394     if (cpu->die_id != -1 && cpu->die_id != topo_ids.die_id) {
395         error_setg(errp, "property die-id: %u doesn't match set apic-id:"
396             " 0x%x (die-id: %u)", cpu->die_id, cpu->apic_id, topo_ids.die_id);
397         return;
398     }
399     cpu->die_id = topo_ids.die_id;
400 
401     if (cpu->core_id != -1 && cpu->core_id != topo_ids.core_id) {
402         error_setg(errp, "property core-id: %u doesn't match set apic-id:"
403             " 0x%x (core-id: %u)", cpu->core_id, cpu->apic_id,
404             topo_ids.core_id);
405         return;
406     }
407     cpu->core_id = topo_ids.core_id;
408 
409     if (cpu->thread_id != -1 && cpu->thread_id != topo_ids.smt_id) {
410         error_setg(errp, "property thread-id: %u doesn't match set apic-id:"
411             " 0x%x (thread-id: %u)", cpu->thread_id, cpu->apic_id,
412             topo_ids.smt_id);
413         return;
414     }
415     cpu->thread_id = topo_ids.smt_id;
416 
417     if (hyperv_feat_enabled(cpu, HYPERV_FEAT_VPINDEX) &&
418         !kvm_hv_vpindex_settable()) {
419         error_setg(errp, "kernel doesn't allow setting HyperV VP_INDEX");
420         return;
421     }
422 
423     cs = CPU(cpu);
424     cs->cpu_index = idx;
425 
426     numa_cpu_pre_plug(cpu_slot, dev, errp);
427 }
428 
429 CpuInstanceProperties
430 x86_cpu_index_to_props(MachineState *ms, unsigned cpu_index)
431 {
432     MachineClass *mc = MACHINE_GET_CLASS(ms);
433     const CPUArchIdList *possible_cpus = mc->possible_cpu_arch_ids(ms);
434 
435     assert(cpu_index < possible_cpus->len);
436     return possible_cpus->cpus[cpu_index].props;
437 }
438 
439 int64_t x86_get_default_cpu_node_id(const MachineState *ms, int idx)
440 {
441    X86CPUTopoIDs topo_ids;
442    X86MachineState *x86ms = X86_MACHINE(ms);
443    X86CPUTopoInfo topo_info;
444 
445    init_topo_info(&topo_info, x86ms);
446 
447    assert(idx < ms->possible_cpus->len);
448    x86_topo_ids_from_apicid(ms->possible_cpus->cpus[idx].arch_id,
449                             &topo_info, &topo_ids);
450    return topo_ids.pkg_id % ms->numa_state->num_nodes;
451 }
452 
453 const CPUArchIdList *x86_possible_cpu_arch_ids(MachineState *ms)
454 {
455     X86MachineState *x86ms = X86_MACHINE(ms);
456     unsigned int max_cpus = ms->smp.max_cpus;
457     X86CPUTopoInfo topo_info;
458     int i;
459 
460     if (ms->possible_cpus) {
461         /*
462          * make sure that max_cpus hasn't changed since the first use, i.e.
463          * -smp hasn't been parsed after it
464          */
465         assert(ms->possible_cpus->len == max_cpus);
466         return ms->possible_cpus;
467     }
468 
469     ms->possible_cpus = g_malloc0(sizeof(CPUArchIdList) +
470                                   sizeof(CPUArchId) * max_cpus);
471     ms->possible_cpus->len = max_cpus;
472 
473     init_topo_info(&topo_info, x86ms);
474 
475     for (i = 0; i < ms->possible_cpus->len; i++) {
476         X86CPUTopoIDs topo_ids;
477 
478         ms->possible_cpus->cpus[i].type = ms->cpu_type;
479         ms->possible_cpus->cpus[i].vcpus_count = 1;
480         ms->possible_cpus->cpus[i].arch_id =
481             x86_cpu_apic_id_from_index(x86ms, i);
482         x86_topo_ids_from_apicid(ms->possible_cpus->cpus[i].arch_id,
483                                  &topo_info, &topo_ids);
484         ms->possible_cpus->cpus[i].props.has_socket_id = true;
485         ms->possible_cpus->cpus[i].props.socket_id = topo_ids.pkg_id;
486         if (ms->smp.dies > 1) {
487             ms->possible_cpus->cpus[i].props.has_die_id = true;
488             ms->possible_cpus->cpus[i].props.die_id = topo_ids.die_id;
489         }
490         ms->possible_cpus->cpus[i].props.has_core_id = true;
491         ms->possible_cpus->cpus[i].props.core_id = topo_ids.core_id;
492         ms->possible_cpus->cpus[i].props.has_thread_id = true;
493         ms->possible_cpus->cpus[i].props.thread_id = topo_ids.smt_id;
494     }
495     return ms->possible_cpus;
496 }
497 
498 static void x86_nmi(NMIState *n, int cpu_index, Error **errp)
499 {
500     /* cpu index isn't used */
501     CPUState *cs;
502 
503     CPU_FOREACH(cs) {
504         X86CPU *cpu = X86_CPU(cs);
505 
506         if (!cpu->apic_state) {
507             cpu_interrupt(cs, CPU_INTERRUPT_NMI);
508         } else {
509             apic_deliver_nmi(cpu->apic_state);
510         }
511     }
512 }
513 
514 static long get_file_size(FILE *f)
515 {
516     long where, size;
517 
518     /* XXX: on Unix systems, using fstat() probably makes more sense */
519 
520     where = ftell(f);
521     fseek(f, 0, SEEK_END);
522     size = ftell(f);
523     fseek(f, where, SEEK_SET);
524 
525     return size;
526 }
527 
528 /* TSC handling */
529 uint64_t cpu_get_tsc(CPUX86State *env)
530 {
531     return cpus_get_elapsed_ticks();
532 }
533 
534 /* IRQ handling */
535 static void pic_irq_request(void *opaque, int irq, int level)
536 {
537     CPUState *cs = first_cpu;
538     X86CPU *cpu = X86_CPU(cs);
539 
540     trace_x86_pic_interrupt(irq, level);
541     if (cpu->apic_state && !kvm_irqchip_in_kernel() &&
542         !whpx_apic_in_platform()) {
543         CPU_FOREACH(cs) {
544             cpu = X86_CPU(cs);
545             if (apic_accept_pic_intr(cpu->apic_state)) {
546                 apic_deliver_pic_intr(cpu->apic_state, level);
547             }
548         }
549     } else {
550         if (level) {
551             cpu_interrupt(cs, CPU_INTERRUPT_HARD);
552         } else {
553             cpu_reset_interrupt(cs, CPU_INTERRUPT_HARD);
554         }
555     }
556 }
557 
558 qemu_irq x86_allocate_cpu_irq(void)
559 {
560     return qemu_allocate_irq(pic_irq_request, NULL, 0);
561 }
562 
563 int cpu_get_pic_interrupt(CPUX86State *env)
564 {
565     X86CPU *cpu = env_archcpu(env);
566     int intno;
567 
568     if (!kvm_irqchip_in_kernel() && !whpx_apic_in_platform()) {
569         intno = apic_get_interrupt(cpu->apic_state);
570         if (intno >= 0) {
571             return intno;
572         }
573         /* read the irq from the PIC */
574         if (!apic_accept_pic_intr(cpu->apic_state)) {
575             return -1;
576         }
577     }
578 
579     intno = pic_read_irq(isa_pic);
580     return intno;
581 }
582 
583 DeviceState *cpu_get_current_apic(void)
584 {
585     if (current_cpu) {
586         X86CPU *cpu = X86_CPU(current_cpu);
587         return cpu->apic_state;
588     } else {
589         return NULL;
590     }
591 }
592 
593 void gsi_handler(void *opaque, int n, int level)
594 {
595     GSIState *s = opaque;
596 
597     trace_x86_gsi_interrupt(n, level);
598     switch (n) {
599     case 0 ... ISA_NUM_IRQS - 1:
600         if (s->i8259_irq[n]) {
601             /* Under KVM, Kernel will forward to both PIC and IOAPIC */
602             qemu_set_irq(s->i8259_irq[n], level);
603         }
604         /* fall through */
605     case ISA_NUM_IRQS ... IOAPIC_NUM_PINS - 1:
606         qemu_set_irq(s->ioapic_irq[n], level);
607         break;
608     case IO_APIC_SECONDARY_IRQBASE
609         ... IO_APIC_SECONDARY_IRQBASE + IOAPIC_NUM_PINS - 1:
610         qemu_set_irq(s->ioapic2_irq[n - IO_APIC_SECONDARY_IRQBASE], level);
611         break;
612     }
613 }
614 
615 void ioapic_init_gsi(GSIState *gsi_state, const char *parent_name)
616 {
617     DeviceState *dev;
618     SysBusDevice *d;
619     unsigned int i;
620 
621     assert(parent_name);
622     if (kvm_ioapic_in_kernel()) {
623         dev = qdev_new(TYPE_KVM_IOAPIC);
624     } else {
625         dev = qdev_new(TYPE_IOAPIC);
626     }
627     object_property_add_child(object_resolve_path(parent_name, NULL),
628                               "ioapic", OBJECT(dev));
629     d = SYS_BUS_DEVICE(dev);
630     sysbus_realize_and_unref(d, &error_fatal);
631     sysbus_mmio_map(d, 0, IO_APIC_DEFAULT_ADDRESS);
632 
633     for (i = 0; i < IOAPIC_NUM_PINS; i++) {
634         gsi_state->ioapic_irq[i] = qdev_get_gpio_in(dev, i);
635     }
636 }
637 
638 DeviceState *ioapic_init_secondary(GSIState *gsi_state)
639 {
640     DeviceState *dev;
641     SysBusDevice *d;
642     unsigned int i;
643 
644     dev = qdev_new(TYPE_IOAPIC);
645     d = SYS_BUS_DEVICE(dev);
646     sysbus_realize_and_unref(d, &error_fatal);
647     sysbus_mmio_map(d, 0, IO_APIC_SECONDARY_ADDRESS);
648 
649     for (i = 0; i < IOAPIC_NUM_PINS; i++) {
650         gsi_state->ioapic2_irq[i] = qdev_get_gpio_in(dev, i);
651     }
652     return dev;
653 }
654 
655 struct setup_data {
656     uint64_t next;
657     uint32_t type;
658     uint32_t len;
659     uint8_t data[];
660 } __attribute__((packed));
661 
662 
663 /*
664  * The entry point into the kernel for PVH boot is different from
665  * the native entry point.  The PVH entry is defined by the x86/HVM
666  * direct boot ABI and is available in an ELFNOTE in the kernel binary.
667  *
668  * This function is passed to load_elf() when it is called from
669  * load_elfboot() which then additionally checks for an ELF Note of
670  * type XEN_ELFNOTE_PHYS32_ENTRY and passes it to this function to
671  * parse the PVH entry address from the ELF Note.
672  *
673  * Due to trickery in elf_opts.h, load_elf() is actually available as
674  * load_elf32() or load_elf64() and this routine needs to be able
675  * to deal with being called as 32 or 64 bit.
676  *
677  * The address of the PVH entry point is saved to the 'pvh_start_addr'
678  * global variable.  (although the entry point is 32-bit, the kernel
679  * binary can be either 32-bit or 64-bit).
680  */
681 static uint64_t read_pvh_start_addr(void *arg1, void *arg2, bool is64)
682 {
683     size_t *elf_note_data_addr;
684 
685     /* Check if ELF Note header passed in is valid */
686     if (arg1 == NULL) {
687         return 0;
688     }
689 
690     if (is64) {
691         struct elf64_note *nhdr64 = (struct elf64_note *)arg1;
692         uint64_t nhdr_size64 = sizeof(struct elf64_note);
693         uint64_t phdr_align = *(uint64_t *)arg2;
694         uint64_t nhdr_namesz = nhdr64->n_namesz;
695 
696         elf_note_data_addr =
697             ((void *)nhdr64) + nhdr_size64 +
698             QEMU_ALIGN_UP(nhdr_namesz, phdr_align);
699 
700         pvh_start_addr = *elf_note_data_addr;
701     } else {
702         struct elf32_note *nhdr32 = (struct elf32_note *)arg1;
703         uint32_t nhdr_size32 = sizeof(struct elf32_note);
704         uint32_t phdr_align = *(uint32_t *)arg2;
705         uint32_t nhdr_namesz = nhdr32->n_namesz;
706 
707         elf_note_data_addr =
708             ((void *)nhdr32) + nhdr_size32 +
709             QEMU_ALIGN_UP(nhdr_namesz, phdr_align);
710 
711         pvh_start_addr = *(uint32_t *)elf_note_data_addr;
712     }
713 
714     return pvh_start_addr;
715 }
716 
717 static bool load_elfboot(const char *kernel_filename,
718                          int kernel_file_size,
719                          uint8_t *header,
720                          size_t pvh_xen_start_addr,
721                          FWCfgState *fw_cfg)
722 {
723     uint32_t flags = 0;
724     uint32_t mh_load_addr = 0;
725     uint32_t elf_kernel_size = 0;
726     uint64_t elf_entry;
727     uint64_t elf_low, elf_high;
728     int kernel_size;
729 
730     if (ldl_p(header) != 0x464c457f) {
731         return false; /* no elfboot */
732     }
733 
734     bool elf_is64 = header[EI_CLASS] == ELFCLASS64;
735     flags = elf_is64 ?
736         ((Elf64_Ehdr *)header)->e_flags : ((Elf32_Ehdr *)header)->e_flags;
737 
738     if (flags & 0x00010004) { /* LOAD_ELF_HEADER_HAS_ADDR */
739         error_report("elfboot unsupported flags = %x", flags);
740         exit(1);
741     }
742 
743     uint64_t elf_note_type = XEN_ELFNOTE_PHYS32_ENTRY;
744     kernel_size = load_elf(kernel_filename, read_pvh_start_addr,
745                            NULL, &elf_note_type, &elf_entry,
746                            &elf_low, &elf_high, NULL, 0, I386_ELF_MACHINE,
747                            0, 0);
748 
749     if (kernel_size < 0) {
750         error_report("Error while loading elf kernel");
751         exit(1);
752     }
753     mh_load_addr = elf_low;
754     elf_kernel_size = elf_high - elf_low;
755 
756     if (pvh_start_addr == 0) {
757         error_report("Error loading uncompressed kernel without PVH ELF Note");
758         exit(1);
759     }
760     fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_ENTRY, pvh_start_addr);
761     fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_ADDR, mh_load_addr);
762     fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_SIZE, elf_kernel_size);
763 
764     return true;
765 }
766 
767 void x86_load_linux(X86MachineState *x86ms,
768                     FWCfgState *fw_cfg,
769                     int acpi_data_size,
770                     bool pvh_enabled,
771                     bool legacy_no_rng_seed)
772 {
773     bool linuxboot_dma_enabled = X86_MACHINE_GET_CLASS(x86ms)->fwcfg_dma_enabled;
774     uint16_t protocol;
775     int setup_size, kernel_size, cmdline_size;
776     int dtb_size, setup_data_offset;
777     uint32_t initrd_max;
778     uint8_t header[8192], *setup, *kernel;
779     hwaddr real_addr, prot_addr, cmdline_addr, initrd_addr = 0, first_setup_data = 0;
780     FILE *f;
781     char *vmode;
782     MachineState *machine = MACHINE(x86ms);
783     struct setup_data *setup_data;
784     const char *kernel_filename = machine->kernel_filename;
785     const char *initrd_filename = machine->initrd_filename;
786     const char *dtb_filename = machine->dtb;
787     const char *kernel_cmdline = machine->kernel_cmdline;
788     SevKernelLoaderContext sev_load_ctx = {};
789     enum { RNG_SEED_LENGTH = 32 };
790 
791     /* Align to 16 bytes as a paranoia measure */
792     cmdline_size = (strlen(kernel_cmdline) + 16) & ~15;
793 
794     /* load the kernel header */
795     f = fopen(kernel_filename, "rb");
796     if (!f) {
797         fprintf(stderr, "qemu: could not open kernel file '%s': %s\n",
798                 kernel_filename, strerror(errno));
799         exit(1);
800     }
801 
802     kernel_size = get_file_size(f);
803     if (!kernel_size ||
804         fread(header, 1, MIN(ARRAY_SIZE(header), kernel_size), f) !=
805         MIN(ARRAY_SIZE(header), kernel_size)) {
806         fprintf(stderr, "qemu: could not load kernel '%s': %s\n",
807                 kernel_filename, strerror(errno));
808         exit(1);
809     }
810 
811     /* kernel protocol version */
812     if (ldl_p(header + 0x202) == 0x53726448) {
813         protocol = lduw_p(header + 0x206);
814     } else {
815         /*
816          * This could be a multiboot kernel. If it is, let's stop treating it
817          * like a Linux kernel.
818          * Note: some multiboot images could be in the ELF format (the same of
819          * PVH), so we try multiboot first since we check the multiboot magic
820          * header before to load it.
821          */
822         if (load_multiboot(x86ms, fw_cfg, f, kernel_filename, initrd_filename,
823                            kernel_cmdline, kernel_size, header)) {
824             return;
825         }
826         /*
827          * Check if the file is an uncompressed kernel file (ELF) and load it,
828          * saving the PVH entry point used by the x86/HVM direct boot ABI.
829          * If load_elfboot() is successful, populate the fw_cfg info.
830          */
831         if (pvh_enabled &&
832             load_elfboot(kernel_filename, kernel_size,
833                          header, pvh_start_addr, fw_cfg)) {
834             fclose(f);
835 
836             fw_cfg_add_i32(fw_cfg, FW_CFG_CMDLINE_SIZE,
837                 strlen(kernel_cmdline) + 1);
838             fw_cfg_add_string(fw_cfg, FW_CFG_CMDLINE_DATA, kernel_cmdline);
839 
840             fw_cfg_add_i32(fw_cfg, FW_CFG_SETUP_SIZE, sizeof(header));
841             fw_cfg_add_bytes(fw_cfg, FW_CFG_SETUP_DATA,
842                              header, sizeof(header));
843 
844             /* load initrd */
845             if (initrd_filename) {
846                 GMappedFile *mapped_file;
847                 gsize initrd_size;
848                 gchar *initrd_data;
849                 GError *gerr = NULL;
850 
851                 mapped_file = g_mapped_file_new(initrd_filename, false, &gerr);
852                 if (!mapped_file) {
853                     fprintf(stderr, "qemu: error reading initrd %s: %s\n",
854                             initrd_filename, gerr->message);
855                     exit(1);
856                 }
857                 x86ms->initrd_mapped_file = mapped_file;
858 
859                 initrd_data = g_mapped_file_get_contents(mapped_file);
860                 initrd_size = g_mapped_file_get_length(mapped_file);
861                 initrd_max = x86ms->below_4g_mem_size - acpi_data_size - 1;
862                 if (initrd_size >= initrd_max) {
863                     fprintf(stderr, "qemu: initrd is too large, cannot support."
864                             "(max: %"PRIu32", need %"PRId64")\n",
865                             initrd_max, (uint64_t)initrd_size);
866                     exit(1);
867                 }
868 
869                 initrd_addr = (initrd_max - initrd_size) & ~4095;
870 
871                 fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_ADDR, initrd_addr);
872                 fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_SIZE, initrd_size);
873                 fw_cfg_add_bytes(fw_cfg, FW_CFG_INITRD_DATA, initrd_data,
874                                  initrd_size);
875             }
876 
877             option_rom[nb_option_roms].bootindex = 0;
878             option_rom[nb_option_roms].name = "pvh.bin";
879             nb_option_roms++;
880 
881             return;
882         }
883         protocol = 0;
884     }
885 
886     if (protocol < 0x200 || !(header[0x211] & 0x01)) {
887         /* Low kernel */
888         real_addr    = 0x90000;
889         cmdline_addr = 0x9a000 - cmdline_size;
890         prot_addr    = 0x10000;
891     } else if (protocol < 0x202) {
892         /* High but ancient kernel */
893         real_addr    = 0x90000;
894         cmdline_addr = 0x9a000 - cmdline_size;
895         prot_addr    = 0x100000;
896     } else {
897         /* High and recent kernel */
898         real_addr    = 0x10000;
899         cmdline_addr = 0x20000;
900         prot_addr    = 0x100000;
901     }
902 
903     /* highest address for loading the initrd */
904     if (protocol >= 0x20c &&
905         lduw_p(header + 0x236) & XLF_CAN_BE_LOADED_ABOVE_4G) {
906         /*
907          * Linux has supported initrd up to 4 GB for a very long time (2007,
908          * long before XLF_CAN_BE_LOADED_ABOVE_4G which was added in 2013),
909          * though it only sets initrd_max to 2 GB to "work around bootloader
910          * bugs". Luckily, QEMU firmware(which does something like bootloader)
911          * has supported this.
912          *
913          * It's believed that if XLF_CAN_BE_LOADED_ABOVE_4G is set, initrd can
914          * be loaded into any address.
915          *
916          * In addition, initrd_max is uint32_t simply because QEMU doesn't
917          * support the 64-bit boot protocol (specifically the ext_ramdisk_image
918          * field).
919          *
920          * Therefore here just limit initrd_max to UINT32_MAX simply as well.
921          */
922         initrd_max = UINT32_MAX;
923     } else if (protocol >= 0x203) {
924         initrd_max = ldl_p(header + 0x22c);
925     } else {
926         initrd_max = 0x37ffffff;
927     }
928 
929     if (initrd_max >= x86ms->below_4g_mem_size - acpi_data_size) {
930         initrd_max = x86ms->below_4g_mem_size - acpi_data_size - 1;
931     }
932 
933     fw_cfg_add_i32(fw_cfg, FW_CFG_CMDLINE_ADDR, cmdline_addr);
934     fw_cfg_add_i32(fw_cfg, FW_CFG_CMDLINE_SIZE, strlen(kernel_cmdline) + 1);
935     fw_cfg_add_string(fw_cfg, FW_CFG_CMDLINE_DATA, kernel_cmdline);
936     sev_load_ctx.cmdline_data = (char *)kernel_cmdline;
937     sev_load_ctx.cmdline_size = strlen(kernel_cmdline) + 1;
938 
939     if (protocol >= 0x202) {
940         stl_p(header + 0x228, cmdline_addr);
941     } else {
942         stw_p(header + 0x20, 0xA33F);
943         stw_p(header + 0x22, cmdline_addr - real_addr);
944     }
945 
946     /* handle vga= parameter */
947     vmode = strstr(kernel_cmdline, "vga=");
948     if (vmode) {
949         unsigned int video_mode;
950         const char *end;
951         int ret;
952         /* skip "vga=" */
953         vmode += 4;
954         if (!strncmp(vmode, "normal", 6)) {
955             video_mode = 0xffff;
956         } else if (!strncmp(vmode, "ext", 3)) {
957             video_mode = 0xfffe;
958         } else if (!strncmp(vmode, "ask", 3)) {
959             video_mode = 0xfffd;
960         } else {
961             ret = qemu_strtoui(vmode, &end, 0, &video_mode);
962             if (ret != 0 || (*end && *end != ' ')) {
963                 fprintf(stderr, "qemu: invalid 'vga=' kernel parameter.\n");
964                 exit(1);
965             }
966         }
967         stw_p(header + 0x1fa, video_mode);
968     }
969 
970     /* loader type */
971     /*
972      * High nybble = B reserved for QEMU; low nybble is revision number.
973      * If this code is substantially changed, you may want to consider
974      * incrementing the revision.
975      */
976     if (protocol >= 0x200) {
977         header[0x210] = 0xB0;
978     }
979     /* heap */
980     if (protocol >= 0x201) {
981         header[0x211] |= 0x80; /* CAN_USE_HEAP */
982         stw_p(header + 0x224, cmdline_addr - real_addr - 0x200);
983     }
984 
985     /* load initrd */
986     if (initrd_filename) {
987         GMappedFile *mapped_file;
988         gsize initrd_size;
989         gchar *initrd_data;
990         GError *gerr = NULL;
991 
992         if (protocol < 0x200) {
993             fprintf(stderr, "qemu: linux kernel too old to load a ram disk\n");
994             exit(1);
995         }
996 
997         mapped_file = g_mapped_file_new(initrd_filename, false, &gerr);
998         if (!mapped_file) {
999             fprintf(stderr, "qemu: error reading initrd %s: %s\n",
1000                     initrd_filename, gerr->message);
1001             exit(1);
1002         }
1003         x86ms->initrd_mapped_file = mapped_file;
1004 
1005         initrd_data = g_mapped_file_get_contents(mapped_file);
1006         initrd_size = g_mapped_file_get_length(mapped_file);
1007         if (initrd_size >= initrd_max) {
1008             fprintf(stderr, "qemu: initrd is too large, cannot support."
1009                     "(max: %"PRIu32", need %"PRId64")\n",
1010                     initrd_max, (uint64_t)initrd_size);
1011             exit(1);
1012         }
1013 
1014         initrd_addr = (initrd_max - initrd_size) & ~4095;
1015 
1016         fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_ADDR, initrd_addr);
1017         fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_SIZE, initrd_size);
1018         fw_cfg_add_bytes(fw_cfg, FW_CFG_INITRD_DATA, initrd_data, initrd_size);
1019         sev_load_ctx.initrd_data = initrd_data;
1020         sev_load_ctx.initrd_size = initrd_size;
1021 
1022         stl_p(header + 0x218, initrd_addr);
1023         stl_p(header + 0x21c, initrd_size);
1024     }
1025 
1026     /* load kernel and setup */
1027     setup_size = header[0x1f1];
1028     if (setup_size == 0) {
1029         setup_size = 4;
1030     }
1031     setup_size = (setup_size + 1) * 512;
1032     if (setup_size > kernel_size) {
1033         fprintf(stderr, "qemu: invalid kernel header\n");
1034         exit(1);
1035     }
1036     kernel_size -= setup_size;
1037 
1038     setup  = g_malloc(setup_size);
1039     kernel = g_malloc(kernel_size);
1040     fseek(f, 0, SEEK_SET);
1041     if (fread(setup, 1, setup_size, f) != setup_size) {
1042         fprintf(stderr, "fread() failed\n");
1043         exit(1);
1044     }
1045     if (fread(kernel, 1, kernel_size, f) != kernel_size) {
1046         fprintf(stderr, "fread() failed\n");
1047         exit(1);
1048     }
1049     fclose(f);
1050 
1051     /* append dtb to kernel */
1052     if (dtb_filename) {
1053         if (protocol < 0x209) {
1054             fprintf(stderr, "qemu: Linux kernel too old to load a dtb\n");
1055             exit(1);
1056         }
1057 
1058         dtb_size = get_image_size(dtb_filename);
1059         if (dtb_size <= 0) {
1060             fprintf(stderr, "qemu: error reading dtb %s: %s\n",
1061                     dtb_filename, strerror(errno));
1062             exit(1);
1063         }
1064 
1065         setup_data_offset = QEMU_ALIGN_UP(kernel_size, 16);
1066         kernel_size = setup_data_offset + sizeof(struct setup_data) + dtb_size;
1067         kernel = g_realloc(kernel, kernel_size);
1068 
1069 
1070         setup_data = (struct setup_data *)(kernel + setup_data_offset);
1071         setup_data->next = cpu_to_le64(first_setup_data);
1072         first_setup_data = prot_addr + setup_data_offset;
1073         setup_data->type = cpu_to_le32(SETUP_DTB);
1074         setup_data->len = cpu_to_le32(dtb_size);
1075 
1076         load_image_size(dtb_filename, setup_data->data, dtb_size);
1077     }
1078 
1079     if (!legacy_no_rng_seed) {
1080         setup_data_offset = QEMU_ALIGN_UP(kernel_size, 16);
1081         kernel_size = setup_data_offset + sizeof(struct setup_data) + RNG_SEED_LENGTH;
1082         kernel = g_realloc(kernel, kernel_size);
1083         setup_data = (struct setup_data *)(kernel + setup_data_offset);
1084         setup_data->next = cpu_to_le64(first_setup_data);
1085         first_setup_data = prot_addr + setup_data_offset;
1086         setup_data->type = cpu_to_le32(SETUP_RNG_SEED);
1087         setup_data->len = cpu_to_le32(RNG_SEED_LENGTH);
1088         qemu_guest_getrandom_nofail(setup_data->data, RNG_SEED_LENGTH);
1089     }
1090 
1091     /* Offset 0x250 is a pointer to the first setup_data link. */
1092     stq_p(header + 0x250, first_setup_data);
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