xref: /openbmc/qemu/hw/i386/pc.c (revision 91bfcdb0)
1 /*
2  * QEMU PC System Emulator
3  *
4  * Copyright (c) 2003-2004 Fabrice Bellard
5  *
6  * Permission is hereby granted, free of charge, to any person obtaining a copy
7  * of this software and associated documentation files (the "Software"), to deal
8  * in the Software without restriction, including without limitation the rights
9  * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
10  * copies of the Software, and to permit persons to whom the Software is
11  * furnished to do so, subject to the following conditions:
12  *
13  * The above copyright notice and this permission notice shall be included in
14  * all copies or substantial portions of the Software.
15  *
16  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
17  * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
18  * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
19  * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
20  * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
21  * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
22  * THE SOFTWARE.
23  */
24 #include "hw/hw.h"
25 #include "hw/i386/pc.h"
26 #include "hw/char/serial.h"
27 #include "hw/i386/apic.h"
28 #include "hw/i386/topology.h"
29 #include "sysemu/cpus.h"
30 #include "hw/block/fdc.h"
31 #include "hw/ide.h"
32 #include "hw/pci/pci.h"
33 #include "hw/pci/pci_bus.h"
34 #include "hw/nvram/fw_cfg.h"
35 #include "hw/timer/hpet.h"
36 #include "hw/smbios/smbios.h"
37 #include "hw/loader.h"
38 #include "elf.h"
39 #include "multiboot.h"
40 #include "hw/timer/mc146818rtc.h"
41 #include "hw/timer/i8254.h"
42 #include "hw/audio/pcspk.h"
43 #include "hw/pci/msi.h"
44 #include "hw/sysbus.h"
45 #include "sysemu/sysemu.h"
46 #include "sysemu/numa.h"
47 #include "sysemu/kvm.h"
48 #include "sysemu/qtest.h"
49 #include "kvm_i386.h"
50 #include "hw/xen/xen.h"
51 #include "sysemu/block-backend.h"
52 #include "hw/block/block.h"
53 #include "ui/qemu-spice.h"
54 #include "exec/memory.h"
55 #include "exec/address-spaces.h"
56 #include "sysemu/arch_init.h"
57 #include "qemu/bitmap.h"
58 #include "qemu/config-file.h"
59 #include "qemu/error-report.h"
60 #include "hw/acpi/acpi.h"
61 #include "hw/acpi/cpu_hotplug.h"
62 #include "hw/boards.h"
63 #include "hw/pci/pci_host.h"
64 #include "acpi-build.h"
65 #include "hw/mem/pc-dimm.h"
66 #include "qapi/visitor.h"
67 #include "qapi-visit.h"
68 
69 /* debug PC/ISA interrupts */
70 //#define DEBUG_IRQ
71 
72 #ifdef DEBUG_IRQ
73 #define DPRINTF(fmt, ...)                                       \
74     do { printf("CPUIRQ: " fmt , ## __VA_ARGS__); } while (0)
75 #else
76 #define DPRINTF(fmt, ...)
77 #endif
78 
79 /* Leave a chunk of memory at the top of RAM for the BIOS ACPI tables
80  * (128K) and other BIOS datastructures (less than 4K reported to be used at
81  * the moment, 32K should be enough for a while).  */
82 static unsigned acpi_data_size = 0x20000 + 0x8000;
83 void pc_set_legacy_acpi_data_size(void)
84 {
85     acpi_data_size = 0x10000;
86 }
87 
88 #define BIOS_CFG_IOPORT 0x510
89 #define FW_CFG_ACPI_TABLES (FW_CFG_ARCH_LOCAL + 0)
90 #define FW_CFG_SMBIOS_ENTRIES (FW_CFG_ARCH_LOCAL + 1)
91 #define FW_CFG_IRQ0_OVERRIDE (FW_CFG_ARCH_LOCAL + 2)
92 #define FW_CFG_E820_TABLE (FW_CFG_ARCH_LOCAL + 3)
93 #define FW_CFG_HPET (FW_CFG_ARCH_LOCAL + 4)
94 
95 #define E820_NR_ENTRIES		16
96 
97 struct e820_entry {
98     uint64_t address;
99     uint64_t length;
100     uint32_t type;
101 } QEMU_PACKED __attribute((__aligned__(4)));
102 
103 struct e820_table {
104     uint32_t count;
105     struct e820_entry entry[E820_NR_ENTRIES];
106 } QEMU_PACKED __attribute((__aligned__(4)));
107 
108 static struct e820_table e820_reserve;
109 static struct e820_entry *e820_table;
110 static unsigned e820_entries;
111 struct hpet_fw_config hpet_cfg = {.count = UINT8_MAX};
112 
113 void gsi_handler(void *opaque, int n, int level)
114 {
115     GSIState *s = opaque;
116 
117     DPRINTF("pc: %s GSI %d\n", level ? "raising" : "lowering", n);
118     if (n < ISA_NUM_IRQS) {
119         qemu_set_irq(s->i8259_irq[n], level);
120     }
121     qemu_set_irq(s->ioapic_irq[n], level);
122 }
123 
124 static void ioport80_write(void *opaque, hwaddr addr, uint64_t data,
125                            unsigned size)
126 {
127 }
128 
129 static uint64_t ioport80_read(void *opaque, hwaddr addr, unsigned size)
130 {
131     return 0xffffffffffffffffULL;
132 }
133 
134 /* MSDOS compatibility mode FPU exception support */
135 static qemu_irq ferr_irq;
136 
137 void pc_register_ferr_irq(qemu_irq irq)
138 {
139     ferr_irq = irq;
140 }
141 
142 /* XXX: add IGNNE support */
143 void cpu_set_ferr(CPUX86State *s)
144 {
145     qemu_irq_raise(ferr_irq);
146 }
147 
148 static void ioportF0_write(void *opaque, hwaddr addr, uint64_t data,
149                            unsigned size)
150 {
151     qemu_irq_lower(ferr_irq);
152 }
153 
154 static uint64_t ioportF0_read(void *opaque, hwaddr addr, unsigned size)
155 {
156     return 0xffffffffffffffffULL;
157 }
158 
159 /* TSC handling */
160 uint64_t cpu_get_tsc(CPUX86State *env)
161 {
162     return cpu_get_ticks();
163 }
164 
165 /* IRQ handling */
166 int cpu_get_pic_interrupt(CPUX86State *env)
167 {
168     X86CPU *cpu = x86_env_get_cpu(env);
169     int intno;
170 
171     intno = apic_get_interrupt(cpu->apic_state);
172     if (intno >= 0) {
173         return intno;
174     }
175     /* read the irq from the PIC */
176     if (!apic_accept_pic_intr(cpu->apic_state)) {
177         return -1;
178     }
179 
180     intno = pic_read_irq(isa_pic);
181     return intno;
182 }
183 
184 static void pic_irq_request(void *opaque, int irq, int level)
185 {
186     CPUState *cs = first_cpu;
187     X86CPU *cpu = X86_CPU(cs);
188 
189     DPRINTF("pic_irqs: %s irq %d\n", level? "raise" : "lower", irq);
190     if (cpu->apic_state) {
191         CPU_FOREACH(cs) {
192             cpu = X86_CPU(cs);
193             if (apic_accept_pic_intr(cpu->apic_state)) {
194                 apic_deliver_pic_intr(cpu->apic_state, level);
195             }
196         }
197     } else {
198         if (level) {
199             cpu_interrupt(cs, CPU_INTERRUPT_HARD);
200         } else {
201             cpu_reset_interrupt(cs, CPU_INTERRUPT_HARD);
202         }
203     }
204 }
205 
206 /* PC cmos mappings */
207 
208 #define REG_EQUIPMENT_BYTE          0x14
209 
210 static int cmos_get_fd_drive_type(FDriveType fd0)
211 {
212     int val;
213 
214     switch (fd0) {
215     case FDRIVE_DRV_144:
216         /* 1.44 Mb 3"5 drive */
217         val = 4;
218         break;
219     case FDRIVE_DRV_288:
220         /* 2.88 Mb 3"5 drive */
221         val = 5;
222         break;
223     case FDRIVE_DRV_120:
224         /* 1.2 Mb 5"5 drive */
225         val = 2;
226         break;
227     case FDRIVE_DRV_NONE:
228     default:
229         val = 0;
230         break;
231     }
232     return val;
233 }
234 
235 static void cmos_init_hd(ISADevice *s, int type_ofs, int info_ofs,
236                          int16_t cylinders, int8_t heads, int8_t sectors)
237 {
238     rtc_set_memory(s, type_ofs, 47);
239     rtc_set_memory(s, info_ofs, cylinders);
240     rtc_set_memory(s, info_ofs + 1, cylinders >> 8);
241     rtc_set_memory(s, info_ofs + 2, heads);
242     rtc_set_memory(s, info_ofs + 3, 0xff);
243     rtc_set_memory(s, info_ofs + 4, 0xff);
244     rtc_set_memory(s, info_ofs + 5, 0xc0 | ((heads > 8) << 3));
245     rtc_set_memory(s, info_ofs + 6, cylinders);
246     rtc_set_memory(s, info_ofs + 7, cylinders >> 8);
247     rtc_set_memory(s, info_ofs + 8, sectors);
248 }
249 
250 /* convert boot_device letter to something recognizable by the bios */
251 static int boot_device2nibble(char boot_device)
252 {
253     switch(boot_device) {
254     case 'a':
255     case 'b':
256         return 0x01; /* floppy boot */
257     case 'c':
258         return 0x02; /* hard drive boot */
259     case 'd':
260         return 0x03; /* CD-ROM boot */
261     case 'n':
262         return 0x04; /* Network boot */
263     }
264     return 0;
265 }
266 
267 static void set_boot_dev(ISADevice *s, const char *boot_device, Error **errp)
268 {
269 #define PC_MAX_BOOT_DEVICES 3
270     int nbds, bds[3] = { 0, };
271     int i;
272 
273     nbds = strlen(boot_device);
274     if (nbds > PC_MAX_BOOT_DEVICES) {
275         error_setg(errp, "Too many boot devices for PC");
276         return;
277     }
278     for (i = 0; i < nbds; i++) {
279         bds[i] = boot_device2nibble(boot_device[i]);
280         if (bds[i] == 0) {
281             error_setg(errp, "Invalid boot device for PC: '%c'",
282                        boot_device[i]);
283             return;
284         }
285     }
286     rtc_set_memory(s, 0x3d, (bds[1] << 4) | bds[0]);
287     rtc_set_memory(s, 0x38, (bds[2] << 4) | (fd_bootchk ? 0x0 : 0x1));
288 }
289 
290 static void pc_boot_set(void *opaque, const char *boot_device, Error **errp)
291 {
292     set_boot_dev(opaque, boot_device, errp);
293 }
294 
295 static void pc_cmos_init_floppy(ISADevice *rtc_state, ISADevice *floppy)
296 {
297     int val, nb, i;
298     FDriveType fd_type[2] = { FDRIVE_DRV_NONE, FDRIVE_DRV_NONE };
299 
300     /* floppy type */
301     if (floppy) {
302         for (i = 0; i < 2; i++) {
303             fd_type[i] = isa_fdc_get_drive_type(floppy, i);
304         }
305     }
306     val = (cmos_get_fd_drive_type(fd_type[0]) << 4) |
307         cmos_get_fd_drive_type(fd_type[1]);
308     rtc_set_memory(rtc_state, 0x10, val);
309 
310     val = rtc_get_memory(rtc_state, REG_EQUIPMENT_BYTE);
311     nb = 0;
312     if (fd_type[0] < FDRIVE_DRV_NONE) {
313         nb++;
314     }
315     if (fd_type[1] < FDRIVE_DRV_NONE) {
316         nb++;
317     }
318     switch (nb) {
319     case 0:
320         break;
321     case 1:
322         val |= 0x01; /* 1 drive, ready for boot */
323         break;
324     case 2:
325         val |= 0x41; /* 2 drives, ready for boot */
326         break;
327     }
328     rtc_set_memory(rtc_state, REG_EQUIPMENT_BYTE, val);
329 }
330 
331 typedef struct pc_cmos_init_late_arg {
332     ISADevice *rtc_state;
333     BusState *idebus[2];
334 } pc_cmos_init_late_arg;
335 
336 typedef struct check_fdc_state {
337     ISADevice *floppy;
338     bool multiple;
339 } CheckFdcState;
340 
341 static int check_fdc(Object *obj, void *opaque)
342 {
343     CheckFdcState *state = opaque;
344     Object *fdc;
345     uint32_t iobase;
346     Error *local_err = NULL;
347 
348     fdc = object_dynamic_cast(obj, TYPE_ISA_FDC);
349     if (!fdc) {
350         return 0;
351     }
352 
353     iobase = object_property_get_int(obj, "iobase", &local_err);
354     if (local_err || iobase != 0x3f0) {
355         error_free(local_err);
356         return 0;
357     }
358 
359     if (state->floppy) {
360         state->multiple = true;
361     } else {
362         state->floppy = ISA_DEVICE(obj);
363     }
364     return 0;
365 }
366 
367 static const char * const fdc_container_path[] = {
368     "/unattached", "/peripheral", "/peripheral-anon"
369 };
370 
371 static void pc_cmos_init_late(void *opaque)
372 {
373     pc_cmos_init_late_arg *arg = opaque;
374     ISADevice *s = arg->rtc_state;
375     int16_t cylinders;
376     int8_t heads, sectors;
377     int val;
378     int i, trans;
379     Object *container;
380     CheckFdcState state = { 0 };
381 
382     val = 0;
383     if (ide_get_geometry(arg->idebus[0], 0,
384                          &cylinders, &heads, &sectors) >= 0) {
385         cmos_init_hd(s, 0x19, 0x1b, cylinders, heads, sectors);
386         val |= 0xf0;
387     }
388     if (ide_get_geometry(arg->idebus[0], 1,
389                          &cylinders, &heads, &sectors) >= 0) {
390         cmos_init_hd(s, 0x1a, 0x24, cylinders, heads, sectors);
391         val |= 0x0f;
392     }
393     rtc_set_memory(s, 0x12, val);
394 
395     val = 0;
396     for (i = 0; i < 4; i++) {
397         /* NOTE: ide_get_geometry() returns the physical
398            geometry.  It is always such that: 1 <= sects <= 63, 1
399            <= heads <= 16, 1 <= cylinders <= 16383. The BIOS
400            geometry can be different if a translation is done. */
401         if (ide_get_geometry(arg->idebus[i / 2], i % 2,
402                              &cylinders, &heads, &sectors) >= 0) {
403             trans = ide_get_bios_chs_trans(arg->idebus[i / 2], i % 2) - 1;
404             assert((trans & ~3) == 0);
405             val |= trans << (i * 2);
406         }
407     }
408     rtc_set_memory(s, 0x39, val);
409 
410     /*
411      * Locate the FDC at IO address 0x3f0, and configure the CMOS registers
412      * accordingly.
413      */
414     for (i = 0; i < ARRAY_SIZE(fdc_container_path); i++) {
415         container = container_get(qdev_get_machine(), fdc_container_path[i]);
416         object_child_foreach(container, check_fdc, &state);
417     }
418 
419     if (state.multiple) {
420         error_report("warning: multiple floppy disk controllers with "
421                      "iobase=0x3f0 have been found;\n"
422                      "the one being picked for CMOS setup might not reflect "
423                      "your intent");
424     }
425     pc_cmos_init_floppy(s, state.floppy);
426 
427     qemu_unregister_reset(pc_cmos_init_late, opaque);
428 }
429 
430 void pc_cmos_init(PCMachineState *pcms,
431                   BusState *idebus0, BusState *idebus1,
432                   ISADevice *s)
433 {
434     int val;
435     static pc_cmos_init_late_arg arg;
436     Error *local_err = NULL;
437 
438     /* various important CMOS locations needed by PC/Bochs bios */
439 
440     /* memory size */
441     /* base memory (first MiB) */
442     val = MIN(pcms->below_4g_mem_size / 1024, 640);
443     rtc_set_memory(s, 0x15, val);
444     rtc_set_memory(s, 0x16, val >> 8);
445     /* extended memory (next 64MiB) */
446     if (pcms->below_4g_mem_size > 1024 * 1024) {
447         val = (pcms->below_4g_mem_size - 1024 * 1024) / 1024;
448     } else {
449         val = 0;
450     }
451     if (val > 65535)
452         val = 65535;
453     rtc_set_memory(s, 0x17, val);
454     rtc_set_memory(s, 0x18, val >> 8);
455     rtc_set_memory(s, 0x30, val);
456     rtc_set_memory(s, 0x31, val >> 8);
457     /* memory between 16MiB and 4GiB */
458     if (pcms->below_4g_mem_size > 16 * 1024 * 1024) {
459         val = (pcms->below_4g_mem_size - 16 * 1024 * 1024) / 65536;
460     } else {
461         val = 0;
462     }
463     if (val > 65535)
464         val = 65535;
465     rtc_set_memory(s, 0x34, val);
466     rtc_set_memory(s, 0x35, val >> 8);
467     /* memory above 4GiB */
468     val = pcms->above_4g_mem_size / 65536;
469     rtc_set_memory(s, 0x5b, val);
470     rtc_set_memory(s, 0x5c, val >> 8);
471     rtc_set_memory(s, 0x5d, val >> 16);
472 
473     /* set the number of CPU */
474     rtc_set_memory(s, 0x5f, smp_cpus - 1);
475 
476     object_property_add_link(OBJECT(pcms), "rtc_state",
477                              TYPE_ISA_DEVICE,
478                              (Object **)&pcms->rtc,
479                              object_property_allow_set_link,
480                              OBJ_PROP_LINK_UNREF_ON_RELEASE, &error_abort);
481     object_property_set_link(OBJECT(pcms), OBJECT(s),
482                              "rtc_state", &error_abort);
483 
484     set_boot_dev(s, MACHINE(pcms)->boot_order, &local_err);
485     if (local_err) {
486         error_report_err(local_err);
487         exit(1);
488     }
489 
490     val = 0;
491     val |= 0x02; /* FPU is there */
492     val |= 0x04; /* PS/2 mouse installed */
493     rtc_set_memory(s, REG_EQUIPMENT_BYTE, val);
494 
495     /* hard drives and FDC */
496     arg.rtc_state = s;
497     arg.idebus[0] = idebus0;
498     arg.idebus[1] = idebus1;
499     qemu_register_reset(pc_cmos_init_late, &arg);
500 }
501 
502 #define TYPE_PORT92 "port92"
503 #define PORT92(obj) OBJECT_CHECK(Port92State, (obj), TYPE_PORT92)
504 
505 /* port 92 stuff: could be split off */
506 typedef struct Port92State {
507     ISADevice parent_obj;
508 
509     MemoryRegion io;
510     uint8_t outport;
511     qemu_irq *a20_out;
512 } Port92State;
513 
514 static void port92_write(void *opaque, hwaddr addr, uint64_t val,
515                          unsigned size)
516 {
517     Port92State *s = opaque;
518     int oldval = s->outport;
519 
520     DPRINTF("port92: write 0x%02" PRIx64 "\n", val);
521     s->outport = val;
522     qemu_set_irq(*s->a20_out, (val >> 1) & 1);
523     if ((val & 1) && !(oldval & 1)) {
524         qemu_system_reset_request();
525     }
526 }
527 
528 static uint64_t port92_read(void *opaque, hwaddr addr,
529                             unsigned size)
530 {
531     Port92State *s = opaque;
532     uint32_t ret;
533 
534     ret = s->outport;
535     DPRINTF("port92: read 0x%02x\n", ret);
536     return ret;
537 }
538 
539 static void port92_init(ISADevice *dev, qemu_irq *a20_out)
540 {
541     Port92State *s = PORT92(dev);
542 
543     s->a20_out = a20_out;
544 }
545 
546 static const VMStateDescription vmstate_port92_isa = {
547     .name = "port92",
548     .version_id = 1,
549     .minimum_version_id = 1,
550     .fields = (VMStateField[]) {
551         VMSTATE_UINT8(outport, Port92State),
552         VMSTATE_END_OF_LIST()
553     }
554 };
555 
556 static void port92_reset(DeviceState *d)
557 {
558     Port92State *s = PORT92(d);
559 
560     s->outport &= ~1;
561 }
562 
563 static const MemoryRegionOps port92_ops = {
564     .read = port92_read,
565     .write = port92_write,
566     .impl = {
567         .min_access_size = 1,
568         .max_access_size = 1,
569     },
570     .endianness = DEVICE_LITTLE_ENDIAN,
571 };
572 
573 static void port92_initfn(Object *obj)
574 {
575     Port92State *s = PORT92(obj);
576 
577     memory_region_init_io(&s->io, OBJECT(s), &port92_ops, s, "port92", 1);
578 
579     s->outport = 0;
580 }
581 
582 static void port92_realizefn(DeviceState *dev, Error **errp)
583 {
584     ISADevice *isadev = ISA_DEVICE(dev);
585     Port92State *s = PORT92(dev);
586 
587     isa_register_ioport(isadev, &s->io, 0x92);
588 }
589 
590 static void port92_class_initfn(ObjectClass *klass, void *data)
591 {
592     DeviceClass *dc = DEVICE_CLASS(klass);
593 
594     dc->realize = port92_realizefn;
595     dc->reset = port92_reset;
596     dc->vmsd = &vmstate_port92_isa;
597     /*
598      * Reason: unlike ordinary ISA devices, this one needs additional
599      * wiring: its A20 output line needs to be wired up by
600      * port92_init().
601      */
602     dc->cannot_instantiate_with_device_add_yet = true;
603 }
604 
605 static const TypeInfo port92_info = {
606     .name          = TYPE_PORT92,
607     .parent        = TYPE_ISA_DEVICE,
608     .instance_size = sizeof(Port92State),
609     .instance_init = port92_initfn,
610     .class_init    = port92_class_initfn,
611 };
612 
613 static void port92_register_types(void)
614 {
615     type_register_static(&port92_info);
616 }
617 
618 type_init(port92_register_types)
619 
620 static void handle_a20_line_change(void *opaque, int irq, int level)
621 {
622     X86CPU *cpu = opaque;
623 
624     /* XXX: send to all CPUs ? */
625     /* XXX: add logic to handle multiple A20 line sources */
626     x86_cpu_set_a20(cpu, level);
627 }
628 
629 int e820_add_entry(uint64_t address, uint64_t length, uint32_t type)
630 {
631     int index = le32_to_cpu(e820_reserve.count);
632     struct e820_entry *entry;
633 
634     if (type != E820_RAM) {
635         /* old FW_CFG_E820_TABLE entry -- reservations only */
636         if (index >= E820_NR_ENTRIES) {
637             return -EBUSY;
638         }
639         entry = &e820_reserve.entry[index++];
640 
641         entry->address = cpu_to_le64(address);
642         entry->length = cpu_to_le64(length);
643         entry->type = cpu_to_le32(type);
644 
645         e820_reserve.count = cpu_to_le32(index);
646     }
647 
648     /* new "etc/e820" file -- include ram too */
649     e820_table = g_renew(struct e820_entry, e820_table, e820_entries + 1);
650     e820_table[e820_entries].address = cpu_to_le64(address);
651     e820_table[e820_entries].length = cpu_to_le64(length);
652     e820_table[e820_entries].type = cpu_to_le32(type);
653     e820_entries++;
654 
655     return e820_entries;
656 }
657 
658 int e820_get_num_entries(void)
659 {
660     return e820_entries;
661 }
662 
663 bool e820_get_entry(int idx, uint32_t type, uint64_t *address, uint64_t *length)
664 {
665     if (idx < e820_entries && e820_table[idx].type == cpu_to_le32(type)) {
666         *address = le64_to_cpu(e820_table[idx].address);
667         *length = le64_to_cpu(e820_table[idx].length);
668         return true;
669     }
670     return false;
671 }
672 
673 /* Enables contiguous-apic-ID mode, for compatibility */
674 static bool compat_apic_id_mode;
675 
676 void enable_compat_apic_id_mode(void)
677 {
678     compat_apic_id_mode = true;
679 }
680 
681 /* Calculates initial APIC ID for a specific CPU index
682  *
683  * Currently we need to be able to calculate the APIC ID from the CPU index
684  * alone (without requiring a CPU object), as the QEMU<->Seabios interfaces have
685  * no concept of "CPU index", and the NUMA tables on fw_cfg need the APIC ID of
686  * all CPUs up to max_cpus.
687  */
688 static uint32_t x86_cpu_apic_id_from_index(unsigned int cpu_index)
689 {
690     uint32_t correct_id;
691     static bool warned;
692 
693     correct_id = x86_apicid_from_cpu_idx(smp_cores, smp_threads, cpu_index);
694     if (compat_apic_id_mode) {
695         if (cpu_index != correct_id && !warned && !qtest_enabled()) {
696             error_report("APIC IDs set in compatibility mode, "
697                          "CPU topology won't match the configuration");
698             warned = true;
699         }
700         return cpu_index;
701     } else {
702         return correct_id;
703     }
704 }
705 
706 /* Calculates the limit to CPU APIC ID values
707  *
708  * This function returns the limit for the APIC ID value, so that all
709  * CPU APIC IDs are < pc_apic_id_limit().
710  *
711  * This is used for FW_CFG_MAX_CPUS. See comments on bochs_bios_init().
712  */
713 static unsigned int pc_apic_id_limit(unsigned int max_cpus)
714 {
715     return x86_cpu_apic_id_from_index(max_cpus - 1) + 1;
716 }
717 
718 static void pc_build_smbios(FWCfgState *fw_cfg)
719 {
720     uint8_t *smbios_tables, *smbios_anchor;
721     size_t smbios_tables_len, smbios_anchor_len;
722     struct smbios_phys_mem_area *mem_array;
723     unsigned i, array_count;
724 
725     smbios_tables = smbios_get_table_legacy(&smbios_tables_len);
726     if (smbios_tables) {
727         fw_cfg_add_bytes(fw_cfg, FW_CFG_SMBIOS_ENTRIES,
728                          smbios_tables, smbios_tables_len);
729     }
730 
731     /* build the array of physical mem area from e820 table */
732     mem_array = g_malloc0(sizeof(*mem_array) * e820_get_num_entries());
733     for (i = 0, array_count = 0; i < e820_get_num_entries(); i++) {
734         uint64_t addr, len;
735 
736         if (e820_get_entry(i, E820_RAM, &addr, &len)) {
737             mem_array[array_count].address = addr;
738             mem_array[array_count].length = len;
739             array_count++;
740         }
741     }
742     smbios_get_tables(mem_array, array_count,
743                       &smbios_tables, &smbios_tables_len,
744                       &smbios_anchor, &smbios_anchor_len);
745     g_free(mem_array);
746 
747     if (smbios_anchor) {
748         fw_cfg_add_file(fw_cfg, "etc/smbios/smbios-tables",
749                         smbios_tables, smbios_tables_len);
750         fw_cfg_add_file(fw_cfg, "etc/smbios/smbios-anchor",
751                         smbios_anchor, smbios_anchor_len);
752     }
753 }
754 
755 static FWCfgState *bochs_bios_init(AddressSpace *as)
756 {
757     FWCfgState *fw_cfg;
758     uint64_t *numa_fw_cfg;
759     int i, j;
760     unsigned int apic_id_limit = pc_apic_id_limit(max_cpus);
761 
762     fw_cfg = fw_cfg_init_io_dma(BIOS_CFG_IOPORT, BIOS_CFG_IOPORT + 4, as);
763 
764     /* FW_CFG_MAX_CPUS is a bit confusing/problematic on x86:
765      *
766      * SeaBIOS needs FW_CFG_MAX_CPUS for CPU hotplug, but the CPU hotplug
767      * QEMU<->SeaBIOS interface is not based on the "CPU index", but on the APIC
768      * ID of hotplugged CPUs[1]. This means that FW_CFG_MAX_CPUS is not the
769      * "maximum number of CPUs", but the "limit to the APIC ID values SeaBIOS
770      * may see".
771      *
772      * So, this means we must not use max_cpus, here, but the maximum possible
773      * APIC ID value, plus one.
774      *
775      * [1] The only kind of "CPU identifier" used between SeaBIOS and QEMU is
776      *     the APIC ID, not the "CPU index"
777      */
778     fw_cfg_add_i16(fw_cfg, FW_CFG_MAX_CPUS, (uint16_t)apic_id_limit);
779     fw_cfg_add_i64(fw_cfg, FW_CFG_RAM_SIZE, (uint64_t)ram_size);
780     fw_cfg_add_bytes(fw_cfg, FW_CFG_ACPI_TABLES,
781                      acpi_tables, acpi_tables_len);
782     fw_cfg_add_i32(fw_cfg, FW_CFG_IRQ0_OVERRIDE, kvm_allows_irq0_override());
783 
784     pc_build_smbios(fw_cfg);
785 
786     fw_cfg_add_bytes(fw_cfg, FW_CFG_E820_TABLE,
787                      &e820_reserve, sizeof(e820_reserve));
788     fw_cfg_add_file(fw_cfg, "etc/e820", e820_table,
789                     sizeof(struct e820_entry) * e820_entries);
790 
791     fw_cfg_add_bytes(fw_cfg, FW_CFG_HPET, &hpet_cfg, sizeof(hpet_cfg));
792     /* allocate memory for the NUMA channel: one (64bit) word for the number
793      * of nodes, one word for each VCPU->node and one word for each node to
794      * hold the amount of memory.
795      */
796     numa_fw_cfg = g_new0(uint64_t, 1 + apic_id_limit + nb_numa_nodes);
797     numa_fw_cfg[0] = cpu_to_le64(nb_numa_nodes);
798     for (i = 0; i < max_cpus; i++) {
799         unsigned int apic_id = x86_cpu_apic_id_from_index(i);
800         assert(apic_id < apic_id_limit);
801         for (j = 0; j < nb_numa_nodes; j++) {
802             if (test_bit(i, numa_info[j].node_cpu)) {
803                 numa_fw_cfg[apic_id + 1] = cpu_to_le64(j);
804                 break;
805             }
806         }
807     }
808     for (i = 0; i < nb_numa_nodes; i++) {
809         numa_fw_cfg[apic_id_limit + 1 + i] = cpu_to_le64(numa_info[i].node_mem);
810     }
811     fw_cfg_add_bytes(fw_cfg, FW_CFG_NUMA, numa_fw_cfg,
812                      (1 + apic_id_limit + nb_numa_nodes) *
813                      sizeof(*numa_fw_cfg));
814 
815     return fw_cfg;
816 }
817 
818 static long get_file_size(FILE *f)
819 {
820     long where, size;
821 
822     /* XXX: on Unix systems, using fstat() probably makes more sense */
823 
824     where = ftell(f);
825     fseek(f, 0, SEEK_END);
826     size = ftell(f);
827     fseek(f, where, SEEK_SET);
828 
829     return size;
830 }
831 
832 static void load_linux(PCMachineState *pcms,
833                        FWCfgState *fw_cfg)
834 {
835     uint16_t protocol;
836     int setup_size, kernel_size, initrd_size = 0, cmdline_size;
837     uint32_t initrd_max;
838     uint8_t header[8192], *setup, *kernel, *initrd_data;
839     hwaddr real_addr, prot_addr, cmdline_addr, initrd_addr = 0;
840     FILE *f;
841     char *vmode;
842     MachineState *machine = MACHINE(pcms);
843     const char *kernel_filename = machine->kernel_filename;
844     const char *initrd_filename = machine->initrd_filename;
845     const char *kernel_cmdline = machine->kernel_cmdline;
846 
847     /* Align to 16 bytes as a paranoia measure */
848     cmdline_size = (strlen(kernel_cmdline)+16) & ~15;
849 
850     /* load the kernel header */
851     f = fopen(kernel_filename, "rb");
852     if (!f || !(kernel_size = get_file_size(f)) ||
853         fread(header, 1, MIN(ARRAY_SIZE(header), kernel_size), f) !=
854         MIN(ARRAY_SIZE(header), kernel_size)) {
855         fprintf(stderr, "qemu: could not load kernel '%s': %s\n",
856                 kernel_filename, strerror(errno));
857         exit(1);
858     }
859 
860     /* kernel protocol version */
861 #if 0
862     fprintf(stderr, "header magic: %#x\n", ldl_p(header+0x202));
863 #endif
864     if (ldl_p(header+0x202) == 0x53726448) {
865         protocol = lduw_p(header+0x206);
866     } else {
867         /* This looks like a multiboot kernel. If it is, let's stop
868            treating it like a Linux kernel. */
869         if (load_multiboot(fw_cfg, f, kernel_filename, initrd_filename,
870                            kernel_cmdline, kernel_size, header)) {
871             return;
872         }
873         protocol = 0;
874     }
875 
876     if (protocol < 0x200 || !(header[0x211] & 0x01)) {
877         /* Low kernel */
878         real_addr    = 0x90000;
879         cmdline_addr = 0x9a000 - cmdline_size;
880         prot_addr    = 0x10000;
881     } else if (protocol < 0x202) {
882         /* High but ancient kernel */
883         real_addr    = 0x90000;
884         cmdline_addr = 0x9a000 - cmdline_size;
885         prot_addr    = 0x100000;
886     } else {
887         /* High and recent kernel */
888         real_addr    = 0x10000;
889         cmdline_addr = 0x20000;
890         prot_addr    = 0x100000;
891     }
892 
893 #if 0
894     fprintf(stderr,
895             "qemu: real_addr     = 0x" TARGET_FMT_plx "\n"
896             "qemu: cmdline_addr  = 0x" TARGET_FMT_plx "\n"
897             "qemu: prot_addr     = 0x" TARGET_FMT_plx "\n",
898             real_addr,
899             cmdline_addr,
900             prot_addr);
901 #endif
902 
903     /* highest address for loading the initrd */
904     if (protocol >= 0x203) {
905         initrd_max = ldl_p(header+0x22c);
906     } else {
907         initrd_max = 0x37ffffff;
908     }
909 
910     if (initrd_max >= pcms->below_4g_mem_size - acpi_data_size) {
911         initrd_max = pcms->below_4g_mem_size - acpi_data_size - 1;
912     }
913 
914     fw_cfg_add_i32(fw_cfg, FW_CFG_CMDLINE_ADDR, cmdline_addr);
915     fw_cfg_add_i32(fw_cfg, FW_CFG_CMDLINE_SIZE, strlen(kernel_cmdline)+1);
916     fw_cfg_add_string(fw_cfg, FW_CFG_CMDLINE_DATA, kernel_cmdline);
917 
918     if (protocol >= 0x202) {
919         stl_p(header+0x228, cmdline_addr);
920     } else {
921         stw_p(header+0x20, 0xA33F);
922         stw_p(header+0x22, cmdline_addr-real_addr);
923     }
924 
925     /* handle vga= parameter */
926     vmode = strstr(kernel_cmdline, "vga=");
927     if (vmode) {
928         unsigned int video_mode;
929         /* skip "vga=" */
930         vmode += 4;
931         if (!strncmp(vmode, "normal", 6)) {
932             video_mode = 0xffff;
933         } else if (!strncmp(vmode, "ext", 3)) {
934             video_mode = 0xfffe;
935         } else if (!strncmp(vmode, "ask", 3)) {
936             video_mode = 0xfffd;
937         } else {
938             video_mode = strtol(vmode, NULL, 0);
939         }
940         stw_p(header+0x1fa, video_mode);
941     }
942 
943     /* loader type */
944     /* High nybble = B reserved for QEMU; low nybble is revision number.
945        If this code is substantially changed, you may want to consider
946        incrementing the revision. */
947     if (protocol >= 0x200) {
948         header[0x210] = 0xB0;
949     }
950     /* heap */
951     if (protocol >= 0x201) {
952         header[0x211] |= 0x80;	/* CAN_USE_HEAP */
953         stw_p(header+0x224, cmdline_addr-real_addr-0x200);
954     }
955 
956     /* load initrd */
957     if (initrd_filename) {
958         if (protocol < 0x200) {
959             fprintf(stderr, "qemu: linux kernel too old to load a ram disk\n");
960             exit(1);
961         }
962 
963         initrd_size = get_image_size(initrd_filename);
964         if (initrd_size < 0) {
965             fprintf(stderr, "qemu: error reading initrd %s: %s\n",
966                     initrd_filename, strerror(errno));
967             exit(1);
968         }
969 
970         initrd_addr = (initrd_max-initrd_size) & ~4095;
971 
972         initrd_data = g_malloc(initrd_size);
973         load_image(initrd_filename, initrd_data);
974 
975         fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_ADDR, initrd_addr);
976         fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_SIZE, initrd_size);
977         fw_cfg_add_bytes(fw_cfg, FW_CFG_INITRD_DATA, initrd_data, initrd_size);
978 
979         stl_p(header+0x218, initrd_addr);
980         stl_p(header+0x21c, initrd_size);
981     }
982 
983     /* load kernel and setup */
984     setup_size = header[0x1f1];
985     if (setup_size == 0) {
986         setup_size = 4;
987     }
988     setup_size = (setup_size+1)*512;
989     if (setup_size > kernel_size) {
990         fprintf(stderr, "qemu: invalid kernel header\n");
991         exit(1);
992     }
993     kernel_size -= setup_size;
994 
995     setup  = g_malloc(setup_size);
996     kernel = g_malloc(kernel_size);
997     fseek(f, 0, SEEK_SET);
998     if (fread(setup, 1, setup_size, f) != setup_size) {
999         fprintf(stderr, "fread() failed\n");
1000         exit(1);
1001     }
1002     if (fread(kernel, 1, kernel_size, f) != kernel_size) {
1003         fprintf(stderr, "fread() failed\n");
1004         exit(1);
1005     }
1006     fclose(f);
1007     memcpy(setup, header, MIN(sizeof(header), setup_size));
1008 
1009     fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_ADDR, prot_addr);
1010     fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_SIZE, kernel_size);
1011     fw_cfg_add_bytes(fw_cfg, FW_CFG_KERNEL_DATA, kernel, kernel_size);
1012 
1013     fw_cfg_add_i32(fw_cfg, FW_CFG_SETUP_ADDR, real_addr);
1014     fw_cfg_add_i32(fw_cfg, FW_CFG_SETUP_SIZE, setup_size);
1015     fw_cfg_add_bytes(fw_cfg, FW_CFG_SETUP_DATA, setup, setup_size);
1016 
1017     option_rom[nb_option_roms].name = "linuxboot.bin";
1018     option_rom[nb_option_roms].bootindex = 0;
1019     nb_option_roms++;
1020 }
1021 
1022 #define NE2000_NB_MAX 6
1023 
1024 static const int ne2000_io[NE2000_NB_MAX] = { 0x300, 0x320, 0x340, 0x360,
1025                                               0x280, 0x380 };
1026 static const int ne2000_irq[NE2000_NB_MAX] = { 9, 10, 11, 3, 4, 5 };
1027 
1028 void pc_init_ne2k_isa(ISABus *bus, NICInfo *nd)
1029 {
1030     static int nb_ne2k = 0;
1031 
1032     if (nb_ne2k == NE2000_NB_MAX)
1033         return;
1034     isa_ne2000_init(bus, ne2000_io[nb_ne2k],
1035                     ne2000_irq[nb_ne2k], nd);
1036     nb_ne2k++;
1037 }
1038 
1039 DeviceState *cpu_get_current_apic(void)
1040 {
1041     if (current_cpu) {
1042         X86CPU *cpu = X86_CPU(current_cpu);
1043         return cpu->apic_state;
1044     } else {
1045         return NULL;
1046     }
1047 }
1048 
1049 void pc_acpi_smi_interrupt(void *opaque, int irq, int level)
1050 {
1051     X86CPU *cpu = opaque;
1052 
1053     if (level) {
1054         cpu_interrupt(CPU(cpu), CPU_INTERRUPT_SMI);
1055     }
1056 }
1057 
1058 static X86CPU *pc_new_cpu(const char *cpu_model, int64_t apic_id,
1059                           Error **errp)
1060 {
1061     X86CPU *cpu = NULL;
1062     Error *local_err = NULL;
1063 
1064     cpu = cpu_x86_create(cpu_model, &local_err);
1065     if (local_err != NULL) {
1066         goto out;
1067     }
1068 
1069     object_property_set_int(OBJECT(cpu), apic_id, "apic-id", &local_err);
1070     object_property_set_bool(OBJECT(cpu), true, "realized", &local_err);
1071 
1072 out:
1073     if (local_err) {
1074         error_propagate(errp, local_err);
1075         object_unref(OBJECT(cpu));
1076         cpu = NULL;
1077     }
1078     return cpu;
1079 }
1080 
1081 void pc_hot_add_cpu(const int64_t id, Error **errp)
1082 {
1083     X86CPU *cpu;
1084     MachineState *machine = MACHINE(qdev_get_machine());
1085     int64_t apic_id = x86_cpu_apic_id_from_index(id);
1086     Error *local_err = NULL;
1087 
1088     if (id < 0) {
1089         error_setg(errp, "Invalid CPU id: %" PRIi64, id);
1090         return;
1091     }
1092 
1093     if (cpu_exists(apic_id)) {
1094         error_setg(errp, "Unable to add CPU: %" PRIi64
1095                    ", it already exists", id);
1096         return;
1097     }
1098 
1099     if (id >= max_cpus) {
1100         error_setg(errp, "Unable to add CPU: %" PRIi64
1101                    ", max allowed: %d", id, max_cpus - 1);
1102         return;
1103     }
1104 
1105     if (apic_id >= ACPI_CPU_HOTPLUG_ID_LIMIT) {
1106         error_setg(errp, "Unable to add CPU: %" PRIi64
1107                    ", resulting APIC ID (%" PRIi64 ") is too large",
1108                    id, apic_id);
1109         return;
1110     }
1111 
1112     cpu = pc_new_cpu(machine->cpu_model, apic_id, &local_err);
1113     if (local_err) {
1114         error_propagate(errp, local_err);
1115         return;
1116     }
1117     object_unref(OBJECT(cpu));
1118 }
1119 
1120 void pc_cpus_init(PCMachineState *pcms)
1121 {
1122     int i;
1123     X86CPU *cpu = NULL;
1124     MachineState *machine = MACHINE(pcms);
1125     Error *error = NULL;
1126     unsigned long apic_id_limit;
1127 
1128     /* init CPUs */
1129     if (machine->cpu_model == NULL) {
1130 #ifdef TARGET_X86_64
1131         machine->cpu_model = "qemu64";
1132 #else
1133         machine->cpu_model = "qemu32";
1134 #endif
1135     }
1136 
1137     apic_id_limit = pc_apic_id_limit(max_cpus);
1138     if (apic_id_limit > ACPI_CPU_HOTPLUG_ID_LIMIT) {
1139         error_report("max_cpus is too large. APIC ID of last CPU is %lu",
1140                      apic_id_limit - 1);
1141         exit(1);
1142     }
1143 
1144     for (i = 0; i < smp_cpus; i++) {
1145         cpu = pc_new_cpu(machine->cpu_model, x86_cpu_apic_id_from_index(i),
1146                          &error);
1147         if (error) {
1148             error_report_err(error);
1149             exit(1);
1150         }
1151         object_unref(OBJECT(cpu));
1152     }
1153 
1154     /* tell smbios about cpuid version and features */
1155     smbios_set_cpuid(cpu->env.cpuid_version, cpu->env.features[FEAT_1_EDX]);
1156 }
1157 
1158 /* pci-info ROM file. Little endian format */
1159 typedef struct PcRomPciInfo {
1160     uint64_t w32_min;
1161     uint64_t w32_max;
1162     uint64_t w64_min;
1163     uint64_t w64_max;
1164 } PcRomPciInfo;
1165 
1166 typedef struct PcGuestInfoState {
1167     PcGuestInfo info;
1168     Notifier machine_done;
1169 } PcGuestInfoState;
1170 
1171 static
1172 void pc_guest_info_machine_done(Notifier *notifier, void *data)
1173 {
1174     PcGuestInfoState *guest_info_state = container_of(notifier,
1175                                                       PcGuestInfoState,
1176                                                       machine_done);
1177     PCIBus *bus = find_i440fx();
1178 
1179     if (bus) {
1180         int extra_hosts = 0;
1181 
1182         QLIST_FOREACH(bus, &bus->child, sibling) {
1183             /* look for expander root buses */
1184             if (pci_bus_is_root(bus)) {
1185                 extra_hosts++;
1186             }
1187         }
1188         if (extra_hosts && guest_info_state->info.fw_cfg) {
1189             uint64_t *val = g_malloc(sizeof(*val));
1190             *val = cpu_to_le64(extra_hosts);
1191             fw_cfg_add_file(guest_info_state->info.fw_cfg,
1192                     "etc/extra-pci-roots", val, sizeof(*val));
1193         }
1194     }
1195 
1196     acpi_setup(&guest_info_state->info);
1197 }
1198 
1199 PcGuestInfo *pc_guest_info_init(PCMachineState *pcms)
1200 {
1201     PcGuestInfoState *guest_info_state = g_malloc0(sizeof *guest_info_state);
1202     PcGuestInfo *guest_info = &guest_info_state->info;
1203     int i, j;
1204 
1205     guest_info->ram_size_below_4g = pcms->below_4g_mem_size;
1206     guest_info->ram_size = pcms->below_4g_mem_size + pcms->above_4g_mem_size;
1207     guest_info->apic_id_limit = pc_apic_id_limit(max_cpus);
1208     guest_info->apic_xrupt_override = kvm_allows_irq0_override();
1209     guest_info->numa_nodes = nb_numa_nodes;
1210     guest_info->node_mem = g_malloc0(guest_info->numa_nodes *
1211                                     sizeof *guest_info->node_mem);
1212     for (i = 0; i < nb_numa_nodes; i++) {
1213         guest_info->node_mem[i] = numa_info[i].node_mem;
1214     }
1215 
1216     guest_info->node_cpu = g_malloc0(guest_info->apic_id_limit *
1217                                      sizeof *guest_info->node_cpu);
1218 
1219     for (i = 0; i < max_cpus; i++) {
1220         unsigned int apic_id = x86_cpu_apic_id_from_index(i);
1221         assert(apic_id < guest_info->apic_id_limit);
1222         for (j = 0; j < nb_numa_nodes; j++) {
1223             if (test_bit(i, numa_info[j].node_cpu)) {
1224                 guest_info->node_cpu[apic_id] = j;
1225                 break;
1226             }
1227         }
1228     }
1229 
1230     guest_info_state->machine_done.notify = pc_guest_info_machine_done;
1231     qemu_add_machine_init_done_notifier(&guest_info_state->machine_done);
1232     return guest_info;
1233 }
1234 
1235 /* setup pci memory address space mapping into system address space */
1236 void pc_pci_as_mapping_init(Object *owner, MemoryRegion *system_memory,
1237                             MemoryRegion *pci_address_space)
1238 {
1239     /* Set to lower priority than RAM */
1240     memory_region_add_subregion_overlap(system_memory, 0x0,
1241                                         pci_address_space, -1);
1242 }
1243 
1244 void pc_acpi_init(const char *default_dsdt)
1245 {
1246     char *filename;
1247 
1248     if (acpi_tables != NULL) {
1249         /* manually set via -acpitable, leave it alone */
1250         return;
1251     }
1252 
1253     filename = qemu_find_file(QEMU_FILE_TYPE_BIOS, default_dsdt);
1254     if (filename == NULL) {
1255         fprintf(stderr, "WARNING: failed to find %s\n", default_dsdt);
1256     } else {
1257         QemuOpts *opts = qemu_opts_create(qemu_find_opts("acpi"), NULL, 0,
1258                                           &error_abort);
1259         Error *err = NULL;
1260 
1261         qemu_opt_set(opts, "file", filename, &error_abort);
1262 
1263         acpi_table_add_builtin(opts, &err);
1264         if (err) {
1265             error_report("WARNING: failed to load %s: %s", filename,
1266                          error_get_pretty(err));
1267             error_free(err);
1268         }
1269         g_free(filename);
1270     }
1271 }
1272 
1273 FWCfgState *xen_load_linux(PCMachineState *pcms,
1274                            PcGuestInfo *guest_info)
1275 {
1276     int i;
1277     FWCfgState *fw_cfg;
1278 
1279     assert(MACHINE(pcms)->kernel_filename != NULL);
1280 
1281     fw_cfg = fw_cfg_init_io(BIOS_CFG_IOPORT);
1282     rom_set_fw(fw_cfg);
1283 
1284     load_linux(pcms, fw_cfg);
1285     for (i = 0; i < nb_option_roms; i++) {
1286         assert(!strcmp(option_rom[i].name, "linuxboot.bin") ||
1287                !strcmp(option_rom[i].name, "multiboot.bin"));
1288         rom_add_option(option_rom[i].name, option_rom[i].bootindex);
1289     }
1290     guest_info->fw_cfg = fw_cfg;
1291     return fw_cfg;
1292 }
1293 
1294 FWCfgState *pc_memory_init(PCMachineState *pcms,
1295                            MemoryRegion *system_memory,
1296                            MemoryRegion *rom_memory,
1297                            MemoryRegion **ram_memory,
1298                            PcGuestInfo *guest_info)
1299 {
1300     int linux_boot, i;
1301     MemoryRegion *ram, *option_rom_mr;
1302     MemoryRegion *ram_below_4g, *ram_above_4g;
1303     FWCfgState *fw_cfg;
1304     MachineState *machine = MACHINE(pcms);
1305 
1306     assert(machine->ram_size == pcms->below_4g_mem_size +
1307                                 pcms->above_4g_mem_size);
1308 
1309     linux_boot = (machine->kernel_filename != NULL);
1310 
1311     /* Allocate RAM.  We allocate it as a single memory region and use
1312      * aliases to address portions of it, mostly for backwards compatibility
1313      * with older qemus that used qemu_ram_alloc().
1314      */
1315     ram = g_malloc(sizeof(*ram));
1316     memory_region_allocate_system_memory(ram, NULL, "pc.ram",
1317                                          machine->ram_size);
1318     *ram_memory = ram;
1319     ram_below_4g = g_malloc(sizeof(*ram_below_4g));
1320     memory_region_init_alias(ram_below_4g, NULL, "ram-below-4g", ram,
1321                              0, pcms->below_4g_mem_size);
1322     memory_region_add_subregion(system_memory, 0, ram_below_4g);
1323     e820_add_entry(0, pcms->below_4g_mem_size, E820_RAM);
1324     if (pcms->above_4g_mem_size > 0) {
1325         ram_above_4g = g_malloc(sizeof(*ram_above_4g));
1326         memory_region_init_alias(ram_above_4g, NULL, "ram-above-4g", ram,
1327                                  pcms->below_4g_mem_size,
1328                                  pcms->above_4g_mem_size);
1329         memory_region_add_subregion(system_memory, 0x100000000ULL,
1330                                     ram_above_4g);
1331         e820_add_entry(0x100000000ULL, pcms->above_4g_mem_size, E820_RAM);
1332     }
1333 
1334     if (!guest_info->has_reserved_memory &&
1335         (machine->ram_slots ||
1336          (machine->maxram_size > machine->ram_size))) {
1337         MachineClass *mc = MACHINE_GET_CLASS(machine);
1338 
1339         error_report("\"-memory 'slots|maxmem'\" is not supported by: %s",
1340                      mc->name);
1341         exit(EXIT_FAILURE);
1342     }
1343 
1344     /* initialize hotplug memory address space */
1345     if (guest_info->has_reserved_memory &&
1346         (machine->ram_size < machine->maxram_size)) {
1347         ram_addr_t hotplug_mem_size =
1348             machine->maxram_size - machine->ram_size;
1349 
1350         if (machine->ram_slots > ACPI_MAX_RAM_SLOTS) {
1351             error_report("unsupported amount of memory slots: %"PRIu64,
1352                          machine->ram_slots);
1353             exit(EXIT_FAILURE);
1354         }
1355 
1356         if (QEMU_ALIGN_UP(machine->maxram_size,
1357                           TARGET_PAGE_SIZE) != machine->maxram_size) {
1358             error_report("maximum memory size must by aligned to multiple of "
1359                          "%d bytes", TARGET_PAGE_SIZE);
1360             exit(EXIT_FAILURE);
1361         }
1362 
1363         pcms->hotplug_memory.base =
1364             ROUND_UP(0x100000000ULL + pcms->above_4g_mem_size, 1ULL << 30);
1365 
1366         if (pcms->enforce_aligned_dimm) {
1367             /* size hotplug region assuming 1G page max alignment per slot */
1368             hotplug_mem_size += (1ULL << 30) * machine->ram_slots;
1369         }
1370 
1371         if ((pcms->hotplug_memory.base + hotplug_mem_size) <
1372             hotplug_mem_size) {
1373             error_report("unsupported amount of maximum memory: " RAM_ADDR_FMT,
1374                          machine->maxram_size);
1375             exit(EXIT_FAILURE);
1376         }
1377 
1378         memory_region_init(&pcms->hotplug_memory.mr, OBJECT(pcms),
1379                            "hotplug-memory", hotplug_mem_size);
1380         memory_region_add_subregion(system_memory, pcms->hotplug_memory.base,
1381                                     &pcms->hotplug_memory.mr);
1382     }
1383 
1384     /* Initialize PC system firmware */
1385     pc_system_firmware_init(rom_memory, guest_info->isapc_ram_fw);
1386 
1387     option_rom_mr = g_malloc(sizeof(*option_rom_mr));
1388     memory_region_init_ram(option_rom_mr, NULL, "pc.rom", PC_ROM_SIZE,
1389                            &error_fatal);
1390     vmstate_register_ram_global(option_rom_mr);
1391     memory_region_add_subregion_overlap(rom_memory,
1392                                         PC_ROM_MIN_VGA,
1393                                         option_rom_mr,
1394                                         1);
1395 
1396     fw_cfg = bochs_bios_init(&address_space_memory);
1397 
1398     rom_set_fw(fw_cfg);
1399 
1400     if (guest_info->has_reserved_memory && pcms->hotplug_memory.base) {
1401         uint64_t *val = g_malloc(sizeof(*val));
1402         PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
1403         uint64_t res_mem_end = pcms->hotplug_memory.base;
1404 
1405         if (!pcmc->broken_reserved_end) {
1406             res_mem_end += memory_region_size(&pcms->hotplug_memory.mr);
1407         }
1408         *val = cpu_to_le64(ROUND_UP(res_mem_end, 0x1ULL << 30));
1409         fw_cfg_add_file(fw_cfg, "etc/reserved-memory-end", val, sizeof(*val));
1410     }
1411 
1412     if (linux_boot) {
1413         load_linux(pcms, fw_cfg);
1414     }
1415 
1416     for (i = 0; i < nb_option_roms; i++) {
1417         rom_add_option(option_rom[i].name, option_rom[i].bootindex);
1418     }
1419     guest_info->fw_cfg = fw_cfg;
1420     return fw_cfg;
1421 }
1422 
1423 qemu_irq pc_allocate_cpu_irq(void)
1424 {
1425     return qemu_allocate_irq(pic_irq_request, NULL, 0);
1426 }
1427 
1428 DeviceState *pc_vga_init(ISABus *isa_bus, PCIBus *pci_bus)
1429 {
1430     DeviceState *dev = NULL;
1431 
1432     if (pci_bus) {
1433         PCIDevice *pcidev = pci_vga_init(pci_bus);
1434         dev = pcidev ? &pcidev->qdev : NULL;
1435     } else if (isa_bus) {
1436         ISADevice *isadev = isa_vga_init(isa_bus);
1437         dev = isadev ? DEVICE(isadev) : NULL;
1438     }
1439     return dev;
1440 }
1441 
1442 static const MemoryRegionOps ioport80_io_ops = {
1443     .write = ioport80_write,
1444     .read = ioport80_read,
1445     .endianness = DEVICE_NATIVE_ENDIAN,
1446     .impl = {
1447         .min_access_size = 1,
1448         .max_access_size = 1,
1449     },
1450 };
1451 
1452 static const MemoryRegionOps ioportF0_io_ops = {
1453     .write = ioportF0_write,
1454     .read = ioportF0_read,
1455     .endianness = DEVICE_NATIVE_ENDIAN,
1456     .impl = {
1457         .min_access_size = 1,
1458         .max_access_size = 1,
1459     },
1460 };
1461 
1462 void pc_basic_device_init(ISABus *isa_bus, qemu_irq *gsi,
1463                           ISADevice **rtc_state,
1464                           bool create_fdctrl,
1465                           bool no_vmport,
1466                           uint32 hpet_irqs)
1467 {
1468     int i;
1469     DriveInfo *fd[MAX_FD];
1470     DeviceState *hpet = NULL;
1471     int pit_isa_irq = 0;
1472     qemu_irq pit_alt_irq = NULL;
1473     qemu_irq rtc_irq = NULL;
1474     qemu_irq *a20_line;
1475     ISADevice *i8042, *port92, *vmmouse, *pit = NULL;
1476     MemoryRegion *ioport80_io = g_new(MemoryRegion, 1);
1477     MemoryRegion *ioportF0_io = g_new(MemoryRegion, 1);
1478 
1479     memory_region_init_io(ioport80_io, NULL, &ioport80_io_ops, NULL, "ioport80", 1);
1480     memory_region_add_subregion(isa_bus->address_space_io, 0x80, ioport80_io);
1481 
1482     memory_region_init_io(ioportF0_io, NULL, &ioportF0_io_ops, NULL, "ioportF0", 1);
1483     memory_region_add_subregion(isa_bus->address_space_io, 0xf0, ioportF0_io);
1484 
1485     /*
1486      * Check if an HPET shall be created.
1487      *
1488      * Without KVM_CAP_PIT_STATE2, we cannot switch off the in-kernel PIT
1489      * when the HPET wants to take over. Thus we have to disable the latter.
1490      */
1491     if (!no_hpet && (!kvm_irqchip_in_kernel() || kvm_has_pit_state2())) {
1492         /* In order to set property, here not using sysbus_try_create_simple */
1493         hpet = qdev_try_create(NULL, TYPE_HPET);
1494         if (hpet) {
1495             /* For pc-piix-*, hpet's intcap is always IRQ2. For pc-q35-1.7
1496              * and earlier, use IRQ2 for compat. Otherwise, use IRQ16~23,
1497              * IRQ8 and IRQ2.
1498              */
1499             uint8_t compat = object_property_get_int(OBJECT(hpet),
1500                     HPET_INTCAP, NULL);
1501             if (!compat) {
1502                 qdev_prop_set_uint32(hpet, HPET_INTCAP, hpet_irqs);
1503             }
1504             qdev_init_nofail(hpet);
1505             sysbus_mmio_map(SYS_BUS_DEVICE(hpet), 0, HPET_BASE);
1506 
1507             for (i = 0; i < GSI_NUM_PINS; i++) {
1508                 sysbus_connect_irq(SYS_BUS_DEVICE(hpet), i, gsi[i]);
1509             }
1510             pit_isa_irq = -1;
1511             pit_alt_irq = qdev_get_gpio_in(hpet, HPET_LEGACY_PIT_INT);
1512             rtc_irq = qdev_get_gpio_in(hpet, HPET_LEGACY_RTC_INT);
1513         }
1514     }
1515     *rtc_state = rtc_init(isa_bus, 2000, rtc_irq);
1516 
1517     qemu_register_boot_set(pc_boot_set, *rtc_state);
1518 
1519     if (!xen_enabled()) {
1520         if (kvm_irqchip_in_kernel()) {
1521             pit = kvm_pit_init(isa_bus, 0x40);
1522         } else {
1523             pit = pit_init(isa_bus, 0x40, pit_isa_irq, pit_alt_irq);
1524         }
1525         if (hpet) {
1526             /* connect PIT to output control line of the HPET */
1527             qdev_connect_gpio_out(hpet, 0, qdev_get_gpio_in(DEVICE(pit), 0));
1528         }
1529         pcspk_init(isa_bus, pit);
1530     }
1531 
1532     serial_hds_isa_init(isa_bus, MAX_SERIAL_PORTS);
1533     parallel_hds_isa_init(isa_bus, MAX_PARALLEL_PORTS);
1534 
1535     a20_line = qemu_allocate_irqs(handle_a20_line_change, first_cpu, 2);
1536     i8042 = isa_create_simple(isa_bus, "i8042");
1537     i8042_setup_a20_line(i8042, &a20_line[0]);
1538     if (!no_vmport) {
1539         vmport_init(isa_bus);
1540         vmmouse = isa_try_create(isa_bus, "vmmouse");
1541     } else {
1542         vmmouse = NULL;
1543     }
1544     if (vmmouse) {
1545         DeviceState *dev = DEVICE(vmmouse);
1546         qdev_prop_set_ptr(dev, "ps2_mouse", i8042);
1547         qdev_init_nofail(dev);
1548     }
1549     port92 = isa_create_simple(isa_bus, "port92");
1550     port92_init(port92, &a20_line[1]);
1551 
1552     DMA_init(0);
1553 
1554     for(i = 0; i < MAX_FD; i++) {
1555         fd[i] = drive_get(IF_FLOPPY, 0, i);
1556         create_fdctrl |= !!fd[i];
1557     }
1558     if (create_fdctrl) {
1559         fdctrl_init_isa(isa_bus, fd);
1560     }
1561 }
1562 
1563 void pc_nic_init(ISABus *isa_bus, PCIBus *pci_bus)
1564 {
1565     int i;
1566 
1567     for (i = 0; i < nb_nics; i++) {
1568         NICInfo *nd = &nd_table[i];
1569 
1570         if (!pci_bus || (nd->model && strcmp(nd->model, "ne2k_isa") == 0)) {
1571             pc_init_ne2k_isa(isa_bus, nd);
1572         } else {
1573             pci_nic_init_nofail(nd, pci_bus, "e1000", NULL);
1574         }
1575     }
1576 }
1577 
1578 void pc_pci_device_init(PCIBus *pci_bus)
1579 {
1580     int max_bus;
1581     int bus;
1582 
1583     max_bus = drive_get_max_bus(IF_SCSI);
1584     for (bus = 0; bus <= max_bus; bus++) {
1585         pci_create_simple(pci_bus, -1, "lsi53c895a");
1586     }
1587 }
1588 
1589 void ioapic_init_gsi(GSIState *gsi_state, const char *parent_name)
1590 {
1591     DeviceState *dev;
1592     SysBusDevice *d;
1593     unsigned int i;
1594 
1595     if (kvm_irqchip_in_kernel()) {
1596         dev = qdev_create(NULL, "kvm-ioapic");
1597     } else {
1598         dev = qdev_create(NULL, "ioapic");
1599     }
1600     if (parent_name) {
1601         object_property_add_child(object_resolve_path(parent_name, NULL),
1602                                   "ioapic", OBJECT(dev), NULL);
1603     }
1604     qdev_init_nofail(dev);
1605     d = SYS_BUS_DEVICE(dev);
1606     sysbus_mmio_map(d, 0, IO_APIC_DEFAULT_ADDRESS);
1607 
1608     for (i = 0; i < IOAPIC_NUM_PINS; i++) {
1609         gsi_state->ioapic_irq[i] = qdev_get_gpio_in(dev, i);
1610     }
1611 }
1612 
1613 static void pc_dimm_plug(HotplugHandler *hotplug_dev,
1614                          DeviceState *dev, Error **errp)
1615 {
1616     HotplugHandlerClass *hhc;
1617     Error *local_err = NULL;
1618     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1619     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
1620     PCDIMMDevice *dimm = PC_DIMM(dev);
1621     PCDIMMDeviceClass *ddc = PC_DIMM_GET_CLASS(dimm);
1622     MemoryRegion *mr = ddc->get_memory_region(dimm);
1623     uint64_t align = TARGET_PAGE_SIZE;
1624 
1625     if (memory_region_get_alignment(mr) && pcms->enforce_aligned_dimm) {
1626         align = memory_region_get_alignment(mr);
1627     }
1628 
1629     if (!pcms->acpi_dev) {
1630         error_setg(&local_err,
1631                    "memory hotplug is not enabled: missing acpi device");
1632         goto out;
1633     }
1634 
1635     pc_dimm_memory_plug(dev, &pcms->hotplug_memory, mr, align,
1636                         pcmc->inter_dimm_gap, &local_err);
1637     if (local_err) {
1638         goto out;
1639     }
1640 
1641     hhc = HOTPLUG_HANDLER_GET_CLASS(pcms->acpi_dev);
1642     hhc->plug(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &error_abort);
1643 out:
1644     error_propagate(errp, local_err);
1645 }
1646 
1647 static void pc_dimm_unplug_request(HotplugHandler *hotplug_dev,
1648                                    DeviceState *dev, Error **errp)
1649 {
1650     HotplugHandlerClass *hhc;
1651     Error *local_err = NULL;
1652     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1653 
1654     if (!pcms->acpi_dev) {
1655         error_setg(&local_err,
1656                    "memory hotplug is not enabled: missing acpi device");
1657         goto out;
1658     }
1659 
1660     hhc = HOTPLUG_HANDLER_GET_CLASS(pcms->acpi_dev);
1661     hhc->unplug_request(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &local_err);
1662 
1663 out:
1664     error_propagate(errp, local_err);
1665 }
1666 
1667 static void pc_dimm_unplug(HotplugHandler *hotplug_dev,
1668                            DeviceState *dev, Error **errp)
1669 {
1670     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1671     PCDIMMDevice *dimm = PC_DIMM(dev);
1672     PCDIMMDeviceClass *ddc = PC_DIMM_GET_CLASS(dimm);
1673     MemoryRegion *mr = ddc->get_memory_region(dimm);
1674     HotplugHandlerClass *hhc;
1675     Error *local_err = NULL;
1676 
1677     hhc = HOTPLUG_HANDLER_GET_CLASS(pcms->acpi_dev);
1678     hhc->unplug(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &local_err);
1679 
1680     if (local_err) {
1681         goto out;
1682     }
1683 
1684     pc_dimm_memory_unplug(dev, &pcms->hotplug_memory, mr);
1685     object_unparent(OBJECT(dev));
1686 
1687  out:
1688     error_propagate(errp, local_err);
1689 }
1690 
1691 static void pc_cpu_plug(HotplugHandler *hotplug_dev,
1692                         DeviceState *dev, Error **errp)
1693 {
1694     HotplugHandlerClass *hhc;
1695     Error *local_err = NULL;
1696     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1697 
1698     if (!dev->hotplugged) {
1699         goto out;
1700     }
1701 
1702     if (!pcms->acpi_dev) {
1703         error_setg(&local_err,
1704                    "cpu hotplug is not enabled: missing acpi device");
1705         goto out;
1706     }
1707 
1708     hhc = HOTPLUG_HANDLER_GET_CLASS(pcms->acpi_dev);
1709     hhc->plug(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &local_err);
1710     if (local_err) {
1711         goto out;
1712     }
1713 
1714     /* increment the number of CPUs */
1715     rtc_set_memory(pcms->rtc, 0x5f, rtc_get_memory(pcms->rtc, 0x5f) + 1);
1716 out:
1717     error_propagate(errp, local_err);
1718 }
1719 
1720 static void pc_machine_device_plug_cb(HotplugHandler *hotplug_dev,
1721                                       DeviceState *dev, Error **errp)
1722 {
1723     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
1724         pc_dimm_plug(hotplug_dev, dev, errp);
1725     } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
1726         pc_cpu_plug(hotplug_dev, dev, errp);
1727     }
1728 }
1729 
1730 static void pc_machine_device_unplug_request_cb(HotplugHandler *hotplug_dev,
1731                                                 DeviceState *dev, Error **errp)
1732 {
1733     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
1734         pc_dimm_unplug_request(hotplug_dev, dev, errp);
1735     } else {
1736         error_setg(errp, "acpi: device unplug request for not supported device"
1737                    " type: %s", object_get_typename(OBJECT(dev)));
1738     }
1739 }
1740 
1741 static void pc_machine_device_unplug_cb(HotplugHandler *hotplug_dev,
1742                                         DeviceState *dev, Error **errp)
1743 {
1744     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
1745         pc_dimm_unplug(hotplug_dev, dev, errp);
1746     } else {
1747         error_setg(errp, "acpi: device unplug for not supported device"
1748                    " type: %s", object_get_typename(OBJECT(dev)));
1749     }
1750 }
1751 
1752 static HotplugHandler *pc_get_hotpug_handler(MachineState *machine,
1753                                              DeviceState *dev)
1754 {
1755     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(machine);
1756 
1757     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM) ||
1758         object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
1759         return HOTPLUG_HANDLER(machine);
1760     }
1761 
1762     return pcmc->get_hotplug_handler ?
1763         pcmc->get_hotplug_handler(machine, dev) : NULL;
1764 }
1765 
1766 static void
1767 pc_machine_get_hotplug_memory_region_size(Object *obj, Visitor *v, void *opaque,
1768                                           const char *name, Error **errp)
1769 {
1770     PCMachineState *pcms = PC_MACHINE(obj);
1771     int64_t value = memory_region_size(&pcms->hotplug_memory.mr);
1772 
1773     visit_type_int(v, &value, name, errp);
1774 }
1775 
1776 static void pc_machine_get_max_ram_below_4g(Object *obj, Visitor *v,
1777                                          void *opaque, const char *name,
1778                                          Error **errp)
1779 {
1780     PCMachineState *pcms = PC_MACHINE(obj);
1781     uint64_t value = pcms->max_ram_below_4g;
1782 
1783     visit_type_size(v, &value, name, errp);
1784 }
1785 
1786 static void pc_machine_set_max_ram_below_4g(Object *obj, Visitor *v,
1787                                          void *opaque, const char *name,
1788                                          Error **errp)
1789 {
1790     PCMachineState *pcms = PC_MACHINE(obj);
1791     Error *error = NULL;
1792     uint64_t value;
1793 
1794     visit_type_size(v, &value, name, &error);
1795     if (error) {
1796         error_propagate(errp, error);
1797         return;
1798     }
1799     if (value > (1ULL << 32)) {
1800         error_set(&error, ERROR_CLASS_GENERIC_ERROR,
1801                   "Machine option 'max-ram-below-4g=%"PRIu64
1802                   "' expects size less than or equal to 4G", value);
1803         error_propagate(errp, error);
1804         return;
1805     }
1806 
1807     if (value < (1ULL << 20)) {
1808         error_report("Warning: small max_ram_below_4g(%"PRIu64
1809                      ") less than 1M.  BIOS may not work..",
1810                      value);
1811     }
1812 
1813     pcms->max_ram_below_4g = value;
1814 }
1815 
1816 static void pc_machine_get_vmport(Object *obj, Visitor *v, void *opaque,
1817                                   const char *name, Error **errp)
1818 {
1819     PCMachineState *pcms = PC_MACHINE(obj);
1820     OnOffAuto vmport = pcms->vmport;
1821 
1822     visit_type_OnOffAuto(v, &vmport, name, errp);
1823 }
1824 
1825 static void pc_machine_set_vmport(Object *obj, Visitor *v, void *opaque,
1826                                   const char *name, Error **errp)
1827 {
1828     PCMachineState *pcms = PC_MACHINE(obj);
1829 
1830     visit_type_OnOffAuto(v, &pcms->vmport, name, errp);
1831 }
1832 
1833 bool pc_machine_is_smm_enabled(PCMachineState *pcms)
1834 {
1835     bool smm_available = false;
1836 
1837     if (pcms->smm == ON_OFF_AUTO_OFF) {
1838         return false;
1839     }
1840 
1841     if (tcg_enabled() || qtest_enabled()) {
1842         smm_available = true;
1843     } else if (kvm_enabled()) {
1844         smm_available = kvm_has_smm();
1845     }
1846 
1847     if (smm_available) {
1848         return true;
1849     }
1850 
1851     if (pcms->smm == ON_OFF_AUTO_ON) {
1852         error_report("System Management Mode not supported by this hypervisor.");
1853         exit(1);
1854     }
1855     return false;
1856 }
1857 
1858 static void pc_machine_get_smm(Object *obj, Visitor *v, void *opaque,
1859                               const char *name, Error **errp)
1860 {
1861     PCMachineState *pcms = PC_MACHINE(obj);
1862     OnOffAuto smm = pcms->smm;
1863 
1864     visit_type_OnOffAuto(v, &smm, name, errp);
1865 }
1866 
1867 static void pc_machine_set_smm(Object *obj, Visitor *v, void *opaque,
1868                                   const char *name, Error **errp)
1869 {
1870     PCMachineState *pcms = PC_MACHINE(obj);
1871 
1872     visit_type_OnOffAuto(v, &pcms->smm, name, errp);
1873 }
1874 
1875 static bool pc_machine_get_aligned_dimm(Object *obj, Error **errp)
1876 {
1877     PCMachineState *pcms = PC_MACHINE(obj);
1878 
1879     return pcms->enforce_aligned_dimm;
1880 }
1881 
1882 static void pc_machine_initfn(Object *obj)
1883 {
1884     PCMachineState *pcms = PC_MACHINE(obj);
1885 
1886     object_property_add(obj, PC_MACHINE_MEMHP_REGION_SIZE, "int",
1887                         pc_machine_get_hotplug_memory_region_size,
1888                         NULL, NULL, NULL, &error_abort);
1889 
1890     pcms->max_ram_below_4g = 1ULL << 32; /* 4G */
1891     object_property_add(obj, PC_MACHINE_MAX_RAM_BELOW_4G, "size",
1892                         pc_machine_get_max_ram_below_4g,
1893                         pc_machine_set_max_ram_below_4g,
1894                         NULL, NULL, &error_abort);
1895     object_property_set_description(obj, PC_MACHINE_MAX_RAM_BELOW_4G,
1896                                     "Maximum ram below the 4G boundary (32bit boundary)",
1897                                     &error_abort);
1898 
1899     pcms->smm = ON_OFF_AUTO_AUTO;
1900     object_property_add(obj, PC_MACHINE_SMM, "OnOffAuto",
1901                         pc_machine_get_smm,
1902                         pc_machine_set_smm,
1903                         NULL, NULL, &error_abort);
1904     object_property_set_description(obj, PC_MACHINE_SMM,
1905                                     "Enable SMM (pc & q35)",
1906                                     &error_abort);
1907 
1908     pcms->vmport = ON_OFF_AUTO_AUTO;
1909     object_property_add(obj, PC_MACHINE_VMPORT, "OnOffAuto",
1910                         pc_machine_get_vmport,
1911                         pc_machine_set_vmport,
1912                         NULL, NULL, &error_abort);
1913     object_property_set_description(obj, PC_MACHINE_VMPORT,
1914                                     "Enable vmport (pc & q35)",
1915                                     &error_abort);
1916 
1917     pcms->enforce_aligned_dimm = true;
1918     object_property_add_bool(obj, PC_MACHINE_ENFORCE_ALIGNED_DIMM,
1919                              pc_machine_get_aligned_dimm,
1920                              NULL, &error_abort);
1921 }
1922 
1923 static void pc_machine_reset(void)
1924 {
1925     CPUState *cs;
1926     X86CPU *cpu;
1927 
1928     qemu_devices_reset();
1929 
1930     /* Reset APIC after devices have been reset to cancel
1931      * any changes that qemu_devices_reset() might have done.
1932      */
1933     CPU_FOREACH(cs) {
1934         cpu = X86_CPU(cs);
1935 
1936         if (cpu->apic_state) {
1937             device_reset(cpu->apic_state);
1938         }
1939     }
1940 }
1941 
1942 static unsigned pc_cpu_index_to_socket_id(unsigned cpu_index)
1943 {
1944     X86CPUTopoInfo topo;
1945     x86_topo_ids_from_idx(smp_cores, smp_threads, cpu_index,
1946                           &topo);
1947     return topo.pkg_id;
1948 }
1949 
1950 static void pc_machine_class_init(ObjectClass *oc, void *data)
1951 {
1952     MachineClass *mc = MACHINE_CLASS(oc);
1953     PCMachineClass *pcmc = PC_MACHINE_CLASS(oc);
1954     HotplugHandlerClass *hc = HOTPLUG_HANDLER_CLASS(oc);
1955 
1956     pcmc->inter_dimm_gap = true;
1957     pcmc->get_hotplug_handler = mc->get_hotplug_handler;
1958     mc->get_hotplug_handler = pc_get_hotpug_handler;
1959     mc->cpu_index_to_socket_id = pc_cpu_index_to_socket_id;
1960     mc->default_boot_order = "cad";
1961     mc->hot_add_cpu = pc_hot_add_cpu;
1962     mc->max_cpus = 255;
1963     mc->reset = pc_machine_reset;
1964     hc->plug = pc_machine_device_plug_cb;
1965     hc->unplug_request = pc_machine_device_unplug_request_cb;
1966     hc->unplug = pc_machine_device_unplug_cb;
1967 }
1968 
1969 static const TypeInfo pc_machine_info = {
1970     .name = TYPE_PC_MACHINE,
1971     .parent = TYPE_MACHINE,
1972     .abstract = true,
1973     .instance_size = sizeof(PCMachineState),
1974     .instance_init = pc_machine_initfn,
1975     .class_size = sizeof(PCMachineClass),
1976     .class_init = pc_machine_class_init,
1977     .interfaces = (InterfaceInfo[]) {
1978          { TYPE_HOTPLUG_HANDLER },
1979          { }
1980     },
1981 };
1982 
1983 static void pc_machine_register_types(void)
1984 {
1985     type_register_static(&pc_machine_info);
1986 }
1987 
1988 type_init(pc_machine_register_types)
1989