xref: /openbmc/qemu/hw/i386/pc.c (revision 42f7a448)
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 #include "qom/cpu.h"
69 
70 /* debug PC/ISA interrupts */
71 //#define DEBUG_IRQ
72 
73 #ifdef DEBUG_IRQ
74 #define DPRINTF(fmt, ...)                                       \
75     do { printf("CPUIRQ: " fmt , ## __VA_ARGS__); } while (0)
76 #else
77 #define DPRINTF(fmt, ...)
78 #endif
79 
80 #define BIOS_CFG_IOPORT 0x510
81 #define FW_CFG_ACPI_TABLES (FW_CFG_ARCH_LOCAL + 0)
82 #define FW_CFG_SMBIOS_ENTRIES (FW_CFG_ARCH_LOCAL + 1)
83 #define FW_CFG_IRQ0_OVERRIDE (FW_CFG_ARCH_LOCAL + 2)
84 #define FW_CFG_E820_TABLE (FW_CFG_ARCH_LOCAL + 3)
85 #define FW_CFG_HPET (FW_CFG_ARCH_LOCAL + 4)
86 
87 #define E820_NR_ENTRIES		16
88 
89 struct e820_entry {
90     uint64_t address;
91     uint64_t length;
92     uint32_t type;
93 } QEMU_PACKED __attribute((__aligned__(4)));
94 
95 struct e820_table {
96     uint32_t count;
97     struct e820_entry entry[E820_NR_ENTRIES];
98 } QEMU_PACKED __attribute((__aligned__(4)));
99 
100 static struct e820_table e820_reserve;
101 static struct e820_entry *e820_table;
102 static unsigned e820_entries;
103 struct hpet_fw_config hpet_cfg = {.count = UINT8_MAX};
104 
105 void gsi_handler(void *opaque, int n, int level)
106 {
107     GSIState *s = opaque;
108 
109     DPRINTF("pc: %s GSI %d\n", level ? "raising" : "lowering", n);
110     if (n < ISA_NUM_IRQS) {
111         qemu_set_irq(s->i8259_irq[n], level);
112     }
113     qemu_set_irq(s->ioapic_irq[n], level);
114 }
115 
116 static void ioport80_write(void *opaque, hwaddr addr, uint64_t data,
117                            unsigned size)
118 {
119 }
120 
121 static uint64_t ioport80_read(void *opaque, hwaddr addr, unsigned size)
122 {
123     return 0xffffffffffffffffULL;
124 }
125 
126 /* MSDOS compatibility mode FPU exception support */
127 static qemu_irq ferr_irq;
128 
129 void pc_register_ferr_irq(qemu_irq irq)
130 {
131     ferr_irq = irq;
132 }
133 
134 /* XXX: add IGNNE support */
135 void cpu_set_ferr(CPUX86State *s)
136 {
137     qemu_irq_raise(ferr_irq);
138 }
139 
140 static void ioportF0_write(void *opaque, hwaddr addr, uint64_t data,
141                            unsigned size)
142 {
143     qemu_irq_lower(ferr_irq);
144 }
145 
146 static uint64_t ioportF0_read(void *opaque, hwaddr addr, unsigned size)
147 {
148     return 0xffffffffffffffffULL;
149 }
150 
151 /* TSC handling */
152 uint64_t cpu_get_tsc(CPUX86State *env)
153 {
154     return cpu_get_ticks();
155 }
156 
157 /* IRQ handling */
158 int cpu_get_pic_interrupt(CPUX86State *env)
159 {
160     X86CPU *cpu = x86_env_get_cpu(env);
161     int intno;
162 
163     intno = apic_get_interrupt(cpu->apic_state);
164     if (intno >= 0) {
165         return intno;
166     }
167     /* read the irq from the PIC */
168     if (!apic_accept_pic_intr(cpu->apic_state)) {
169         return -1;
170     }
171 
172     intno = pic_read_irq(isa_pic);
173     return intno;
174 }
175 
176 static void pic_irq_request(void *opaque, int irq, int level)
177 {
178     CPUState *cs = first_cpu;
179     X86CPU *cpu = X86_CPU(cs);
180 
181     DPRINTF("pic_irqs: %s irq %d\n", level? "raise" : "lower", irq);
182     if (cpu->apic_state) {
183         CPU_FOREACH(cs) {
184             cpu = X86_CPU(cs);
185             if (apic_accept_pic_intr(cpu->apic_state)) {
186                 apic_deliver_pic_intr(cpu->apic_state, level);
187             }
188         }
189     } else {
190         if (level) {
191             cpu_interrupt(cs, CPU_INTERRUPT_HARD);
192         } else {
193             cpu_reset_interrupt(cs, CPU_INTERRUPT_HARD);
194         }
195     }
196 }
197 
198 /* PC cmos mappings */
199 
200 #define REG_EQUIPMENT_BYTE          0x14
201 
202 static int cmos_get_fd_drive_type(FloppyDriveType fd0)
203 {
204     int val;
205 
206     switch (fd0) {
207     case FLOPPY_DRIVE_TYPE_144:
208         /* 1.44 Mb 3"5 drive */
209         val = 4;
210         break;
211     case FLOPPY_DRIVE_TYPE_288:
212         /* 2.88 Mb 3"5 drive */
213         val = 5;
214         break;
215     case FLOPPY_DRIVE_TYPE_120:
216         /* 1.2 Mb 5"5 drive */
217         val = 2;
218         break;
219     case FLOPPY_DRIVE_TYPE_NONE:
220     default:
221         val = 0;
222         break;
223     }
224     return val;
225 }
226 
227 static void cmos_init_hd(ISADevice *s, int type_ofs, int info_ofs,
228                          int16_t cylinders, int8_t heads, int8_t sectors)
229 {
230     rtc_set_memory(s, type_ofs, 47);
231     rtc_set_memory(s, info_ofs, cylinders);
232     rtc_set_memory(s, info_ofs + 1, cylinders >> 8);
233     rtc_set_memory(s, info_ofs + 2, heads);
234     rtc_set_memory(s, info_ofs + 3, 0xff);
235     rtc_set_memory(s, info_ofs + 4, 0xff);
236     rtc_set_memory(s, info_ofs + 5, 0xc0 | ((heads > 8) << 3));
237     rtc_set_memory(s, info_ofs + 6, cylinders);
238     rtc_set_memory(s, info_ofs + 7, cylinders >> 8);
239     rtc_set_memory(s, info_ofs + 8, sectors);
240 }
241 
242 /* convert boot_device letter to something recognizable by the bios */
243 static int boot_device2nibble(char boot_device)
244 {
245     switch(boot_device) {
246     case 'a':
247     case 'b':
248         return 0x01; /* floppy boot */
249     case 'c':
250         return 0x02; /* hard drive boot */
251     case 'd':
252         return 0x03; /* CD-ROM boot */
253     case 'n':
254         return 0x04; /* Network boot */
255     }
256     return 0;
257 }
258 
259 static void set_boot_dev(ISADevice *s, const char *boot_device, Error **errp)
260 {
261 #define PC_MAX_BOOT_DEVICES 3
262     int nbds, bds[3] = { 0, };
263     int i;
264 
265     nbds = strlen(boot_device);
266     if (nbds > PC_MAX_BOOT_DEVICES) {
267         error_setg(errp, "Too many boot devices for PC");
268         return;
269     }
270     for (i = 0; i < nbds; i++) {
271         bds[i] = boot_device2nibble(boot_device[i]);
272         if (bds[i] == 0) {
273             error_setg(errp, "Invalid boot device for PC: '%c'",
274                        boot_device[i]);
275             return;
276         }
277     }
278     rtc_set_memory(s, 0x3d, (bds[1] << 4) | bds[0]);
279     rtc_set_memory(s, 0x38, (bds[2] << 4) | (fd_bootchk ? 0x0 : 0x1));
280 }
281 
282 static void pc_boot_set(void *opaque, const char *boot_device, Error **errp)
283 {
284     set_boot_dev(opaque, boot_device, errp);
285 }
286 
287 static void pc_cmos_init_floppy(ISADevice *rtc_state, ISADevice *floppy)
288 {
289     int val, nb, i;
290     FloppyDriveType fd_type[2] = { FLOPPY_DRIVE_TYPE_NONE,
291                                    FLOPPY_DRIVE_TYPE_NONE };
292 
293     /* floppy type */
294     if (floppy) {
295         for (i = 0; i < 2; i++) {
296             fd_type[i] = isa_fdc_get_drive_type(floppy, i);
297         }
298     }
299     val = (cmos_get_fd_drive_type(fd_type[0]) << 4) |
300         cmos_get_fd_drive_type(fd_type[1]);
301     rtc_set_memory(rtc_state, 0x10, val);
302 
303     val = rtc_get_memory(rtc_state, REG_EQUIPMENT_BYTE);
304     nb = 0;
305     if (fd_type[0] != FLOPPY_DRIVE_TYPE_NONE) {
306         nb++;
307     }
308     if (fd_type[1] != FLOPPY_DRIVE_TYPE_NONE) {
309         nb++;
310     }
311     switch (nb) {
312     case 0:
313         break;
314     case 1:
315         val |= 0x01; /* 1 drive, ready for boot */
316         break;
317     case 2:
318         val |= 0x41; /* 2 drives, ready for boot */
319         break;
320     }
321     rtc_set_memory(rtc_state, REG_EQUIPMENT_BYTE, val);
322 }
323 
324 typedef struct pc_cmos_init_late_arg {
325     ISADevice *rtc_state;
326     BusState *idebus[2];
327 } pc_cmos_init_late_arg;
328 
329 typedef struct check_fdc_state {
330     ISADevice *floppy;
331     bool multiple;
332 } CheckFdcState;
333 
334 static int check_fdc(Object *obj, void *opaque)
335 {
336     CheckFdcState *state = opaque;
337     Object *fdc;
338     uint32_t iobase;
339     Error *local_err = NULL;
340 
341     fdc = object_dynamic_cast(obj, TYPE_ISA_FDC);
342     if (!fdc) {
343         return 0;
344     }
345 
346     iobase = object_property_get_int(obj, "iobase", &local_err);
347     if (local_err || iobase != 0x3f0) {
348         error_free(local_err);
349         return 0;
350     }
351 
352     if (state->floppy) {
353         state->multiple = true;
354     } else {
355         state->floppy = ISA_DEVICE(obj);
356     }
357     return 0;
358 }
359 
360 static const char * const fdc_container_path[] = {
361     "/unattached", "/peripheral", "/peripheral-anon"
362 };
363 
364 /*
365  * Locate the FDC at IO address 0x3f0, in order to configure the CMOS registers
366  * and ACPI objects.
367  */
368 ISADevice *pc_find_fdc0(void)
369 {
370     int i;
371     Object *container;
372     CheckFdcState state = { 0 };
373 
374     for (i = 0; i < ARRAY_SIZE(fdc_container_path); i++) {
375         container = container_get(qdev_get_machine(), fdc_container_path[i]);
376         object_child_foreach(container, check_fdc, &state);
377     }
378 
379     if (state.multiple) {
380         error_report("warning: multiple floppy disk controllers with "
381                      "iobase=0x3f0 have been found");
382         error_printf("the one being picked for CMOS setup might not reflect "
383                      "your intent");
384     }
385 
386     return state.floppy;
387 }
388 
389 static void pc_cmos_init_late(void *opaque)
390 {
391     pc_cmos_init_late_arg *arg = opaque;
392     ISADevice *s = arg->rtc_state;
393     int16_t cylinders;
394     int8_t heads, sectors;
395     int val;
396     int i, trans;
397 
398     val = 0;
399     if (ide_get_geometry(arg->idebus[0], 0,
400                          &cylinders, &heads, &sectors) >= 0) {
401         cmos_init_hd(s, 0x19, 0x1b, cylinders, heads, sectors);
402         val |= 0xf0;
403     }
404     if (ide_get_geometry(arg->idebus[0], 1,
405                          &cylinders, &heads, &sectors) >= 0) {
406         cmos_init_hd(s, 0x1a, 0x24, cylinders, heads, sectors);
407         val |= 0x0f;
408     }
409     rtc_set_memory(s, 0x12, val);
410 
411     val = 0;
412     for (i = 0; i < 4; i++) {
413         /* NOTE: ide_get_geometry() returns the physical
414            geometry.  It is always such that: 1 <= sects <= 63, 1
415            <= heads <= 16, 1 <= cylinders <= 16383. The BIOS
416            geometry can be different if a translation is done. */
417         if (ide_get_geometry(arg->idebus[i / 2], i % 2,
418                              &cylinders, &heads, &sectors) >= 0) {
419             trans = ide_get_bios_chs_trans(arg->idebus[i / 2], i % 2) - 1;
420             assert((trans & ~3) == 0);
421             val |= trans << (i * 2);
422         }
423     }
424     rtc_set_memory(s, 0x39, val);
425 
426     pc_cmos_init_floppy(s, pc_find_fdc0());
427 
428     qemu_unregister_reset(pc_cmos_init_late, opaque);
429 }
430 
431 void pc_cmos_init(PCMachineState *pcms,
432                   BusState *idebus0, BusState *idebus1,
433                   ISADevice *s)
434 {
435     int val;
436     static pc_cmos_init_late_arg arg;
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, &error_fatal);
485 
486     val = 0;
487     val |= 0x02; /* FPU is there */
488     val |= 0x04; /* PS/2 mouse installed */
489     rtc_set_memory(s, REG_EQUIPMENT_BYTE, val);
490 
491     /* hard drives and FDC */
492     arg.rtc_state = s;
493     arg.idebus[0] = idebus0;
494     arg.idebus[1] = idebus1;
495     qemu_register_reset(pc_cmos_init_late, &arg);
496 }
497 
498 #define TYPE_PORT92 "port92"
499 #define PORT92(obj) OBJECT_CHECK(Port92State, (obj), TYPE_PORT92)
500 
501 /* port 92 stuff: could be split off */
502 typedef struct Port92State {
503     ISADevice parent_obj;
504 
505     MemoryRegion io;
506     uint8_t outport;
507     qemu_irq *a20_out;
508 } Port92State;
509 
510 static void port92_write(void *opaque, hwaddr addr, uint64_t val,
511                          unsigned size)
512 {
513     Port92State *s = opaque;
514     int oldval = s->outport;
515 
516     DPRINTF("port92: write 0x%02" PRIx64 "\n", val);
517     s->outport = val;
518     qemu_set_irq(*s->a20_out, (val >> 1) & 1);
519     if ((val & 1) && !(oldval & 1)) {
520         qemu_system_reset_request();
521     }
522 }
523 
524 static uint64_t port92_read(void *opaque, hwaddr addr,
525                             unsigned size)
526 {
527     Port92State *s = opaque;
528     uint32_t ret;
529 
530     ret = s->outport;
531     DPRINTF("port92: read 0x%02x\n", ret);
532     return ret;
533 }
534 
535 static void port92_init(ISADevice *dev, qemu_irq *a20_out)
536 {
537     Port92State *s = PORT92(dev);
538 
539     s->a20_out = a20_out;
540 }
541 
542 static const VMStateDescription vmstate_port92_isa = {
543     .name = "port92",
544     .version_id = 1,
545     .minimum_version_id = 1,
546     .fields = (VMStateField[]) {
547         VMSTATE_UINT8(outport, Port92State),
548         VMSTATE_END_OF_LIST()
549     }
550 };
551 
552 static void port92_reset(DeviceState *d)
553 {
554     Port92State *s = PORT92(d);
555 
556     s->outport &= ~1;
557 }
558 
559 static const MemoryRegionOps port92_ops = {
560     .read = port92_read,
561     .write = port92_write,
562     .impl = {
563         .min_access_size = 1,
564         .max_access_size = 1,
565     },
566     .endianness = DEVICE_LITTLE_ENDIAN,
567 };
568 
569 static void port92_initfn(Object *obj)
570 {
571     Port92State *s = PORT92(obj);
572 
573     memory_region_init_io(&s->io, OBJECT(s), &port92_ops, s, "port92", 1);
574 
575     s->outport = 0;
576 }
577 
578 static void port92_realizefn(DeviceState *dev, Error **errp)
579 {
580     ISADevice *isadev = ISA_DEVICE(dev);
581     Port92State *s = PORT92(dev);
582 
583     isa_register_ioport(isadev, &s->io, 0x92);
584 }
585 
586 static void port92_class_initfn(ObjectClass *klass, void *data)
587 {
588     DeviceClass *dc = DEVICE_CLASS(klass);
589 
590     dc->realize = port92_realizefn;
591     dc->reset = port92_reset;
592     dc->vmsd = &vmstate_port92_isa;
593     /*
594      * Reason: unlike ordinary ISA devices, this one needs additional
595      * wiring: its A20 output line needs to be wired up by
596      * port92_init().
597      */
598     dc->cannot_instantiate_with_device_add_yet = true;
599 }
600 
601 static const TypeInfo port92_info = {
602     .name          = TYPE_PORT92,
603     .parent        = TYPE_ISA_DEVICE,
604     .instance_size = sizeof(Port92State),
605     .instance_init = port92_initfn,
606     .class_init    = port92_class_initfn,
607 };
608 
609 static void port92_register_types(void)
610 {
611     type_register_static(&port92_info);
612 }
613 
614 type_init(port92_register_types)
615 
616 static void handle_a20_line_change(void *opaque, int irq, int level)
617 {
618     X86CPU *cpu = opaque;
619 
620     /* XXX: send to all CPUs ? */
621     /* XXX: add logic to handle multiple A20 line sources */
622     x86_cpu_set_a20(cpu, level);
623 }
624 
625 int e820_add_entry(uint64_t address, uint64_t length, uint32_t type)
626 {
627     int index = le32_to_cpu(e820_reserve.count);
628     struct e820_entry *entry;
629 
630     if (type != E820_RAM) {
631         /* old FW_CFG_E820_TABLE entry -- reservations only */
632         if (index >= E820_NR_ENTRIES) {
633             return -EBUSY;
634         }
635         entry = &e820_reserve.entry[index++];
636 
637         entry->address = cpu_to_le64(address);
638         entry->length = cpu_to_le64(length);
639         entry->type = cpu_to_le32(type);
640 
641         e820_reserve.count = cpu_to_le32(index);
642     }
643 
644     /* new "etc/e820" file -- include ram too */
645     e820_table = g_renew(struct e820_entry, e820_table, e820_entries + 1);
646     e820_table[e820_entries].address = cpu_to_le64(address);
647     e820_table[e820_entries].length = cpu_to_le64(length);
648     e820_table[e820_entries].type = cpu_to_le32(type);
649     e820_entries++;
650 
651     return e820_entries;
652 }
653 
654 int e820_get_num_entries(void)
655 {
656     return e820_entries;
657 }
658 
659 bool e820_get_entry(int idx, uint32_t type, uint64_t *address, uint64_t *length)
660 {
661     if (idx < e820_entries && e820_table[idx].type == cpu_to_le32(type)) {
662         *address = le64_to_cpu(e820_table[idx].address);
663         *length = le64_to_cpu(e820_table[idx].length);
664         return true;
665     }
666     return false;
667 }
668 
669 /* Enables contiguous-apic-ID mode, for compatibility */
670 static bool compat_apic_id_mode;
671 
672 void enable_compat_apic_id_mode(void)
673 {
674     compat_apic_id_mode = true;
675 }
676 
677 /* Calculates initial APIC ID for a specific CPU index
678  *
679  * Currently we need to be able to calculate the APIC ID from the CPU index
680  * alone (without requiring a CPU object), as the QEMU<->Seabios interfaces have
681  * no concept of "CPU index", and the NUMA tables on fw_cfg need the APIC ID of
682  * all CPUs up to max_cpus.
683  */
684 static uint32_t x86_cpu_apic_id_from_index(unsigned int cpu_index)
685 {
686     uint32_t correct_id;
687     static bool warned;
688 
689     correct_id = x86_apicid_from_cpu_idx(smp_cores, smp_threads, cpu_index);
690     if (compat_apic_id_mode) {
691         if (cpu_index != correct_id && !warned && !qtest_enabled()) {
692             error_report("APIC IDs set in compatibility mode, "
693                          "CPU topology won't match the configuration");
694             warned = true;
695         }
696         return cpu_index;
697     } else {
698         return correct_id;
699     }
700 }
701 
702 /* Calculates the limit to CPU APIC ID values
703  *
704  * This function returns the limit for the APIC ID value, so that all
705  * CPU APIC IDs are < pc_apic_id_limit().
706  *
707  * This is used for FW_CFG_MAX_CPUS. See comments on bochs_bios_init().
708  */
709 static unsigned int pc_apic_id_limit(unsigned int max_cpus)
710 {
711     return x86_cpu_apic_id_from_index(max_cpus - 1) + 1;
712 }
713 
714 static void pc_build_smbios(FWCfgState *fw_cfg)
715 {
716     uint8_t *smbios_tables, *smbios_anchor;
717     size_t smbios_tables_len, smbios_anchor_len;
718     struct smbios_phys_mem_area *mem_array;
719     unsigned i, array_count;
720 
721     smbios_tables = smbios_get_table_legacy(&smbios_tables_len);
722     if (smbios_tables) {
723         fw_cfg_add_bytes(fw_cfg, FW_CFG_SMBIOS_ENTRIES,
724                          smbios_tables, smbios_tables_len);
725     }
726 
727     /* build the array of physical mem area from e820 table */
728     mem_array = g_malloc0(sizeof(*mem_array) * e820_get_num_entries());
729     for (i = 0, array_count = 0; i < e820_get_num_entries(); i++) {
730         uint64_t addr, len;
731 
732         if (e820_get_entry(i, E820_RAM, &addr, &len)) {
733             mem_array[array_count].address = addr;
734             mem_array[array_count].length = len;
735             array_count++;
736         }
737     }
738     smbios_get_tables(mem_array, array_count,
739                       &smbios_tables, &smbios_tables_len,
740                       &smbios_anchor, &smbios_anchor_len);
741     g_free(mem_array);
742 
743     if (smbios_anchor) {
744         fw_cfg_add_file(fw_cfg, "etc/smbios/smbios-tables",
745                         smbios_tables, smbios_tables_len);
746         fw_cfg_add_file(fw_cfg, "etc/smbios/smbios-anchor",
747                         smbios_anchor, smbios_anchor_len);
748     }
749 }
750 
751 static FWCfgState *bochs_bios_init(AddressSpace *as)
752 {
753     FWCfgState *fw_cfg;
754     uint64_t *numa_fw_cfg;
755     int i, j;
756     unsigned int apic_id_limit = pc_apic_id_limit(max_cpus);
757 
758     fw_cfg = fw_cfg_init_io_dma(BIOS_CFG_IOPORT, BIOS_CFG_IOPORT + 4, as);
759 
760     /* FW_CFG_MAX_CPUS is a bit confusing/problematic on x86:
761      *
762      * SeaBIOS needs FW_CFG_MAX_CPUS for CPU hotplug, but the CPU hotplug
763      * QEMU<->SeaBIOS interface is not based on the "CPU index", but on the APIC
764      * ID of hotplugged CPUs[1]. This means that FW_CFG_MAX_CPUS is not the
765      * "maximum number of CPUs", but the "limit to the APIC ID values SeaBIOS
766      * may see".
767      *
768      * So, this means we must not use max_cpus, here, but the maximum possible
769      * APIC ID value, plus one.
770      *
771      * [1] The only kind of "CPU identifier" used between SeaBIOS and QEMU is
772      *     the APIC ID, not the "CPU index"
773      */
774     fw_cfg_add_i16(fw_cfg, FW_CFG_MAX_CPUS, (uint16_t)apic_id_limit);
775     fw_cfg_add_i64(fw_cfg, FW_CFG_RAM_SIZE, (uint64_t)ram_size);
776     fw_cfg_add_bytes(fw_cfg, FW_CFG_ACPI_TABLES,
777                      acpi_tables, acpi_tables_len);
778     fw_cfg_add_i32(fw_cfg, FW_CFG_IRQ0_OVERRIDE, kvm_allows_irq0_override());
779 
780     pc_build_smbios(fw_cfg);
781 
782     fw_cfg_add_bytes(fw_cfg, FW_CFG_E820_TABLE,
783                      &e820_reserve, sizeof(e820_reserve));
784     fw_cfg_add_file(fw_cfg, "etc/e820", e820_table,
785                     sizeof(struct e820_entry) * e820_entries);
786 
787     fw_cfg_add_bytes(fw_cfg, FW_CFG_HPET, &hpet_cfg, sizeof(hpet_cfg));
788     /* allocate memory for the NUMA channel: one (64bit) word for the number
789      * of nodes, one word for each VCPU->node and one word for each node to
790      * hold the amount of memory.
791      */
792     numa_fw_cfg = g_new0(uint64_t, 1 + apic_id_limit + nb_numa_nodes);
793     numa_fw_cfg[0] = cpu_to_le64(nb_numa_nodes);
794     for (i = 0; i < max_cpus; i++) {
795         unsigned int apic_id = x86_cpu_apic_id_from_index(i);
796         assert(apic_id < apic_id_limit);
797         for (j = 0; j < nb_numa_nodes; j++) {
798             if (test_bit(i, numa_info[j].node_cpu)) {
799                 numa_fw_cfg[apic_id + 1] = cpu_to_le64(j);
800                 break;
801             }
802         }
803     }
804     for (i = 0; i < nb_numa_nodes; i++) {
805         numa_fw_cfg[apic_id_limit + 1 + i] = cpu_to_le64(numa_info[i].node_mem);
806     }
807     fw_cfg_add_bytes(fw_cfg, FW_CFG_NUMA, numa_fw_cfg,
808                      (1 + apic_id_limit + nb_numa_nodes) *
809                      sizeof(*numa_fw_cfg));
810 
811     return fw_cfg;
812 }
813 
814 static long get_file_size(FILE *f)
815 {
816     long where, size;
817 
818     /* XXX: on Unix systems, using fstat() probably makes more sense */
819 
820     where = ftell(f);
821     fseek(f, 0, SEEK_END);
822     size = ftell(f);
823     fseek(f, where, SEEK_SET);
824 
825     return size;
826 }
827 
828 static void load_linux(PCMachineState *pcms,
829                        FWCfgState *fw_cfg)
830 {
831     uint16_t protocol;
832     int setup_size, kernel_size, initrd_size = 0, cmdline_size;
833     uint32_t initrd_max;
834     uint8_t header[8192], *setup, *kernel, *initrd_data;
835     hwaddr real_addr, prot_addr, cmdline_addr, initrd_addr = 0;
836     FILE *f;
837     char *vmode;
838     MachineState *machine = MACHINE(pcms);
839     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
840     const char *kernel_filename = machine->kernel_filename;
841     const char *initrd_filename = machine->initrd_filename;
842     const char *kernel_cmdline = machine->kernel_cmdline;
843 
844     /* Align to 16 bytes as a paranoia measure */
845     cmdline_size = (strlen(kernel_cmdline)+16) & ~15;
846 
847     /* load the kernel header */
848     f = fopen(kernel_filename, "rb");
849     if (!f || !(kernel_size = get_file_size(f)) ||
850         fread(header, 1, MIN(ARRAY_SIZE(header), kernel_size), f) !=
851         MIN(ARRAY_SIZE(header), kernel_size)) {
852         fprintf(stderr, "qemu: could not load kernel '%s': %s\n",
853                 kernel_filename, strerror(errno));
854         exit(1);
855     }
856 
857     /* kernel protocol version */
858 #if 0
859     fprintf(stderr, "header magic: %#x\n", ldl_p(header+0x202));
860 #endif
861     if (ldl_p(header+0x202) == 0x53726448) {
862         protocol = lduw_p(header+0x206);
863     } else {
864         /* This looks like a multiboot kernel. If it is, let's stop
865            treating it like a Linux kernel. */
866         if (load_multiboot(fw_cfg, f, kernel_filename, initrd_filename,
867                            kernel_cmdline, kernel_size, header)) {
868             return;
869         }
870         protocol = 0;
871     }
872 
873     if (protocol < 0x200 || !(header[0x211] & 0x01)) {
874         /* Low kernel */
875         real_addr    = 0x90000;
876         cmdline_addr = 0x9a000 - cmdline_size;
877         prot_addr    = 0x10000;
878     } else if (protocol < 0x202) {
879         /* High but ancient kernel */
880         real_addr    = 0x90000;
881         cmdline_addr = 0x9a000 - cmdline_size;
882         prot_addr    = 0x100000;
883     } else {
884         /* High and recent kernel */
885         real_addr    = 0x10000;
886         cmdline_addr = 0x20000;
887         prot_addr    = 0x100000;
888     }
889 
890 #if 0
891     fprintf(stderr,
892             "qemu: real_addr     = 0x" TARGET_FMT_plx "\n"
893             "qemu: cmdline_addr  = 0x" TARGET_FMT_plx "\n"
894             "qemu: prot_addr     = 0x" TARGET_FMT_plx "\n",
895             real_addr,
896             cmdline_addr,
897             prot_addr);
898 #endif
899 
900     /* highest address for loading the initrd */
901     if (protocol >= 0x203) {
902         initrd_max = ldl_p(header+0x22c);
903     } else {
904         initrd_max = 0x37ffffff;
905     }
906 
907     if (initrd_max >= pcms->below_4g_mem_size - pcmc->acpi_data_size) {
908         initrd_max = pcms->below_4g_mem_size - pcmc->acpi_data_size - 1;
909     }
910 
911     fw_cfg_add_i32(fw_cfg, FW_CFG_CMDLINE_ADDR, cmdline_addr);
912     fw_cfg_add_i32(fw_cfg, FW_CFG_CMDLINE_SIZE, strlen(kernel_cmdline)+1);
913     fw_cfg_add_string(fw_cfg, FW_CFG_CMDLINE_DATA, kernel_cmdline);
914 
915     if (protocol >= 0x202) {
916         stl_p(header+0x228, cmdline_addr);
917     } else {
918         stw_p(header+0x20, 0xA33F);
919         stw_p(header+0x22, cmdline_addr-real_addr);
920     }
921 
922     /* handle vga= parameter */
923     vmode = strstr(kernel_cmdline, "vga=");
924     if (vmode) {
925         unsigned int video_mode;
926         /* skip "vga=" */
927         vmode += 4;
928         if (!strncmp(vmode, "normal", 6)) {
929             video_mode = 0xffff;
930         } else if (!strncmp(vmode, "ext", 3)) {
931             video_mode = 0xfffe;
932         } else if (!strncmp(vmode, "ask", 3)) {
933             video_mode = 0xfffd;
934         } else {
935             video_mode = strtol(vmode, NULL, 0);
936         }
937         stw_p(header+0x1fa, video_mode);
938     }
939 
940     /* loader type */
941     /* High nybble = B reserved for QEMU; low nybble is revision number.
942        If this code is substantially changed, you may want to consider
943        incrementing the revision. */
944     if (protocol >= 0x200) {
945         header[0x210] = 0xB0;
946     }
947     /* heap */
948     if (protocol >= 0x201) {
949         header[0x211] |= 0x80;	/* CAN_USE_HEAP */
950         stw_p(header+0x224, cmdline_addr-real_addr-0x200);
951     }
952 
953     /* load initrd */
954     if (initrd_filename) {
955         if (protocol < 0x200) {
956             fprintf(stderr, "qemu: linux kernel too old to load a ram disk\n");
957             exit(1);
958         }
959 
960         initrd_size = get_image_size(initrd_filename);
961         if (initrd_size < 0) {
962             fprintf(stderr, "qemu: error reading initrd %s: %s\n",
963                     initrd_filename, strerror(errno));
964             exit(1);
965         }
966 
967         initrd_addr = (initrd_max-initrd_size) & ~4095;
968 
969         initrd_data = g_malloc(initrd_size);
970         load_image(initrd_filename, initrd_data);
971 
972         fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_ADDR, initrd_addr);
973         fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_SIZE, initrd_size);
974         fw_cfg_add_bytes(fw_cfg, FW_CFG_INITRD_DATA, initrd_data, initrd_size);
975 
976         stl_p(header+0x218, initrd_addr);
977         stl_p(header+0x21c, initrd_size);
978     }
979 
980     /* load kernel and setup */
981     setup_size = header[0x1f1];
982     if (setup_size == 0) {
983         setup_size = 4;
984     }
985     setup_size = (setup_size+1)*512;
986     if (setup_size > kernel_size) {
987         fprintf(stderr, "qemu: invalid kernel header\n");
988         exit(1);
989     }
990     kernel_size -= setup_size;
991 
992     setup  = g_malloc(setup_size);
993     kernel = g_malloc(kernel_size);
994     fseek(f, 0, SEEK_SET);
995     if (fread(setup, 1, setup_size, f) != setup_size) {
996         fprintf(stderr, "fread() failed\n");
997         exit(1);
998     }
999     if (fread(kernel, 1, kernel_size, f) != kernel_size) {
1000         fprintf(stderr, "fread() failed\n");
1001         exit(1);
1002     }
1003     fclose(f);
1004     memcpy(setup, header, MIN(sizeof(header), setup_size));
1005 
1006     fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_ADDR, prot_addr);
1007     fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_SIZE, kernel_size);
1008     fw_cfg_add_bytes(fw_cfg, FW_CFG_KERNEL_DATA, kernel, kernel_size);
1009 
1010     fw_cfg_add_i32(fw_cfg, FW_CFG_SETUP_ADDR, real_addr);
1011     fw_cfg_add_i32(fw_cfg, FW_CFG_SETUP_SIZE, setup_size);
1012     fw_cfg_add_bytes(fw_cfg, FW_CFG_SETUP_DATA, setup, setup_size);
1013 
1014     option_rom[nb_option_roms].name = "linuxboot.bin";
1015     option_rom[nb_option_roms].bootindex = 0;
1016     nb_option_roms++;
1017 }
1018 
1019 #define NE2000_NB_MAX 6
1020 
1021 static const int ne2000_io[NE2000_NB_MAX] = { 0x300, 0x320, 0x340, 0x360,
1022                                               0x280, 0x380 };
1023 static const int ne2000_irq[NE2000_NB_MAX] = { 9, 10, 11, 3, 4, 5 };
1024 
1025 void pc_init_ne2k_isa(ISABus *bus, NICInfo *nd)
1026 {
1027     static int nb_ne2k = 0;
1028 
1029     if (nb_ne2k == NE2000_NB_MAX)
1030         return;
1031     isa_ne2000_init(bus, ne2000_io[nb_ne2k],
1032                     ne2000_irq[nb_ne2k], nd);
1033     nb_ne2k++;
1034 }
1035 
1036 DeviceState *cpu_get_current_apic(void)
1037 {
1038     if (current_cpu) {
1039         X86CPU *cpu = X86_CPU(current_cpu);
1040         return cpu->apic_state;
1041     } else {
1042         return NULL;
1043     }
1044 }
1045 
1046 void pc_acpi_smi_interrupt(void *opaque, int irq, int level)
1047 {
1048     X86CPU *cpu = opaque;
1049 
1050     if (level) {
1051         cpu_interrupt(CPU(cpu), CPU_INTERRUPT_SMI);
1052     }
1053 }
1054 
1055 static X86CPU *pc_new_cpu(const char *cpu_model, int64_t apic_id,
1056                           Error **errp)
1057 {
1058     X86CPU *cpu = NULL;
1059     Error *local_err = NULL;
1060 
1061     cpu = cpu_x86_create(cpu_model, &local_err);
1062     if (local_err != NULL) {
1063         goto out;
1064     }
1065 
1066     object_property_set_int(OBJECT(cpu), apic_id, "apic-id", &local_err);
1067     object_property_set_bool(OBJECT(cpu), true, "realized", &local_err);
1068 
1069 out:
1070     if (local_err) {
1071         error_propagate(errp, local_err);
1072         object_unref(OBJECT(cpu));
1073         cpu = NULL;
1074     }
1075     return cpu;
1076 }
1077 
1078 void pc_hot_add_cpu(const int64_t id, Error **errp)
1079 {
1080     X86CPU *cpu;
1081     MachineState *machine = MACHINE(qdev_get_machine());
1082     int64_t apic_id = x86_cpu_apic_id_from_index(id);
1083     Error *local_err = NULL;
1084 
1085     if (id < 0) {
1086         error_setg(errp, "Invalid CPU id: %" PRIi64, id);
1087         return;
1088     }
1089 
1090     if (cpu_exists(apic_id)) {
1091         error_setg(errp, "Unable to add CPU: %" PRIi64
1092                    ", it already exists", id);
1093         return;
1094     }
1095 
1096     if (id >= max_cpus) {
1097         error_setg(errp, "Unable to add CPU: %" PRIi64
1098                    ", max allowed: %d", id, max_cpus - 1);
1099         return;
1100     }
1101 
1102     if (apic_id >= ACPI_CPU_HOTPLUG_ID_LIMIT) {
1103         error_setg(errp, "Unable to add CPU: %" PRIi64
1104                    ", resulting APIC ID (%" PRIi64 ") is too large",
1105                    id, apic_id);
1106         return;
1107     }
1108 
1109     cpu = pc_new_cpu(machine->cpu_model, apic_id, &local_err);
1110     if (local_err) {
1111         error_propagate(errp, local_err);
1112         return;
1113     }
1114     object_unref(OBJECT(cpu));
1115 }
1116 
1117 void pc_cpus_init(PCMachineState *pcms)
1118 {
1119     int i;
1120     X86CPU *cpu = NULL;
1121     MachineState *machine = MACHINE(pcms);
1122     unsigned long apic_id_limit;
1123 
1124     /* init CPUs */
1125     if (machine->cpu_model == NULL) {
1126 #ifdef TARGET_X86_64
1127         machine->cpu_model = "qemu64";
1128 #else
1129         machine->cpu_model = "qemu32";
1130 #endif
1131     }
1132 
1133     apic_id_limit = pc_apic_id_limit(max_cpus);
1134     if (apic_id_limit > ACPI_CPU_HOTPLUG_ID_LIMIT) {
1135         error_report("max_cpus is too large. APIC ID of last CPU is %lu",
1136                      apic_id_limit - 1);
1137         exit(1);
1138     }
1139 
1140     for (i = 0; i < smp_cpus; i++) {
1141         cpu = pc_new_cpu(machine->cpu_model, x86_cpu_apic_id_from_index(i),
1142                          &error_fatal);
1143         object_unref(OBJECT(cpu));
1144     }
1145 
1146     /* tell smbios about cpuid version and features */
1147     smbios_set_cpuid(cpu->env.cpuid_version, cpu->env.features[FEAT_1_EDX]);
1148 }
1149 
1150 /* pci-info ROM file. Little endian format */
1151 typedef struct PcRomPciInfo {
1152     uint64_t w32_min;
1153     uint64_t w32_max;
1154     uint64_t w64_min;
1155     uint64_t w64_max;
1156 } PcRomPciInfo;
1157 
1158 typedef struct PcGuestInfoState {
1159     PcGuestInfo info;
1160     Notifier machine_done;
1161 } PcGuestInfoState;
1162 
1163 static
1164 void pc_guest_info_machine_done(Notifier *notifier, void *data)
1165 {
1166     PcGuestInfoState *guest_info_state = container_of(notifier,
1167                                                       PcGuestInfoState,
1168                                                       machine_done);
1169     PCIBus *bus = PC_MACHINE(qdev_get_machine())->bus;
1170 
1171     if (bus) {
1172         int extra_hosts = 0;
1173 
1174         QLIST_FOREACH(bus, &bus->child, sibling) {
1175             /* look for expander root buses */
1176             if (pci_bus_is_root(bus)) {
1177                 extra_hosts++;
1178             }
1179         }
1180         if (extra_hosts && guest_info_state->info.fw_cfg) {
1181             uint64_t *val = g_malloc(sizeof(*val));
1182             *val = cpu_to_le64(extra_hosts);
1183             fw_cfg_add_file(guest_info_state->info.fw_cfg,
1184                     "etc/extra-pci-roots", val, sizeof(*val));
1185         }
1186     }
1187 
1188     acpi_setup(&guest_info_state->info);
1189 }
1190 
1191 PcGuestInfo *pc_guest_info_init(PCMachineState *pcms)
1192 {
1193     PcGuestInfoState *guest_info_state = g_malloc0(sizeof *guest_info_state);
1194     PcGuestInfo *guest_info = &guest_info_state->info;
1195     int i, j;
1196 
1197     guest_info->ram_size_below_4g = pcms->below_4g_mem_size;
1198     guest_info->ram_size = pcms->below_4g_mem_size + pcms->above_4g_mem_size;
1199     guest_info->apic_id_limit = pc_apic_id_limit(max_cpus);
1200     guest_info->apic_xrupt_override = kvm_allows_irq0_override();
1201     guest_info->numa_nodes = nb_numa_nodes;
1202     guest_info->node_mem = g_malloc0(guest_info->numa_nodes *
1203                                     sizeof *guest_info->node_mem);
1204     for (i = 0; i < nb_numa_nodes; i++) {
1205         guest_info->node_mem[i] = numa_info[i].node_mem;
1206     }
1207 
1208     guest_info->node_cpu = g_malloc0(guest_info->apic_id_limit *
1209                                      sizeof *guest_info->node_cpu);
1210 
1211     for (i = 0; i < max_cpus; i++) {
1212         unsigned int apic_id = x86_cpu_apic_id_from_index(i);
1213         assert(apic_id < guest_info->apic_id_limit);
1214         for (j = 0; j < nb_numa_nodes; j++) {
1215             if (test_bit(i, numa_info[j].node_cpu)) {
1216                 guest_info->node_cpu[apic_id] = j;
1217                 break;
1218             }
1219         }
1220     }
1221 
1222     guest_info_state->machine_done.notify = pc_guest_info_machine_done;
1223     qemu_add_machine_init_done_notifier(&guest_info_state->machine_done);
1224     return guest_info;
1225 }
1226 
1227 /* setup pci memory address space mapping into system address space */
1228 void pc_pci_as_mapping_init(Object *owner, MemoryRegion *system_memory,
1229                             MemoryRegion *pci_address_space)
1230 {
1231     /* Set to lower priority than RAM */
1232     memory_region_add_subregion_overlap(system_memory, 0x0,
1233                                         pci_address_space, -1);
1234 }
1235 
1236 void pc_acpi_init(const char *default_dsdt)
1237 {
1238     char *filename;
1239 
1240     if (acpi_tables != NULL) {
1241         /* manually set via -acpitable, leave it alone */
1242         return;
1243     }
1244 
1245     filename = qemu_find_file(QEMU_FILE_TYPE_BIOS, default_dsdt);
1246     if (filename == NULL) {
1247         fprintf(stderr, "WARNING: failed to find %s\n", default_dsdt);
1248     } else {
1249         QemuOpts *opts = qemu_opts_create(qemu_find_opts("acpi"), NULL, 0,
1250                                           &error_abort);
1251         Error *err = NULL;
1252 
1253         qemu_opt_set(opts, "file", filename, &error_abort);
1254 
1255         acpi_table_add_builtin(opts, &err);
1256         if (err) {
1257             error_reportf_err(err, "WARNING: failed to load %s: ",
1258                               filename);
1259         }
1260         g_free(filename);
1261     }
1262 }
1263 
1264 FWCfgState *xen_load_linux(PCMachineState *pcms,
1265                            PcGuestInfo *guest_info)
1266 {
1267     int i;
1268     FWCfgState *fw_cfg;
1269 
1270     assert(MACHINE(pcms)->kernel_filename != NULL);
1271 
1272     fw_cfg = fw_cfg_init_io(BIOS_CFG_IOPORT);
1273     rom_set_fw(fw_cfg);
1274 
1275     load_linux(pcms, fw_cfg);
1276     for (i = 0; i < nb_option_roms; i++) {
1277         assert(!strcmp(option_rom[i].name, "linuxboot.bin") ||
1278                !strcmp(option_rom[i].name, "multiboot.bin"));
1279         rom_add_option(option_rom[i].name, option_rom[i].bootindex);
1280     }
1281     guest_info->fw_cfg = fw_cfg;
1282     return fw_cfg;
1283 }
1284 
1285 FWCfgState *pc_memory_init(PCMachineState *pcms,
1286                            MemoryRegion *system_memory,
1287                            MemoryRegion *rom_memory,
1288                            MemoryRegion **ram_memory,
1289                            PcGuestInfo *guest_info)
1290 {
1291     int linux_boot, i;
1292     MemoryRegion *ram, *option_rom_mr;
1293     MemoryRegion *ram_below_4g, *ram_above_4g;
1294     FWCfgState *fw_cfg;
1295     MachineState *machine = MACHINE(pcms);
1296     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
1297 
1298     assert(machine->ram_size == pcms->below_4g_mem_size +
1299                                 pcms->above_4g_mem_size);
1300 
1301     linux_boot = (machine->kernel_filename != NULL);
1302 
1303     /* Allocate RAM.  We allocate it as a single memory region and use
1304      * aliases to address portions of it, mostly for backwards compatibility
1305      * with older qemus that used qemu_ram_alloc().
1306      */
1307     ram = g_malloc(sizeof(*ram));
1308     memory_region_allocate_system_memory(ram, NULL, "pc.ram",
1309                                          machine->ram_size);
1310     *ram_memory = ram;
1311     ram_below_4g = g_malloc(sizeof(*ram_below_4g));
1312     memory_region_init_alias(ram_below_4g, NULL, "ram-below-4g", ram,
1313                              0, pcms->below_4g_mem_size);
1314     memory_region_add_subregion(system_memory, 0, ram_below_4g);
1315     e820_add_entry(0, pcms->below_4g_mem_size, E820_RAM);
1316     if (pcms->above_4g_mem_size > 0) {
1317         ram_above_4g = g_malloc(sizeof(*ram_above_4g));
1318         memory_region_init_alias(ram_above_4g, NULL, "ram-above-4g", ram,
1319                                  pcms->below_4g_mem_size,
1320                                  pcms->above_4g_mem_size);
1321         memory_region_add_subregion(system_memory, 0x100000000ULL,
1322                                     ram_above_4g);
1323         e820_add_entry(0x100000000ULL, pcms->above_4g_mem_size, E820_RAM);
1324     }
1325 
1326     if (!guest_info->has_reserved_memory &&
1327         (machine->ram_slots ||
1328          (machine->maxram_size > machine->ram_size))) {
1329         MachineClass *mc = MACHINE_GET_CLASS(machine);
1330 
1331         error_report("\"-memory 'slots|maxmem'\" is not supported by: %s",
1332                      mc->name);
1333         exit(EXIT_FAILURE);
1334     }
1335 
1336     /* initialize hotplug memory address space */
1337     if (guest_info->has_reserved_memory &&
1338         (machine->ram_size < machine->maxram_size)) {
1339         ram_addr_t hotplug_mem_size =
1340             machine->maxram_size - machine->ram_size;
1341 
1342         if (machine->ram_slots > ACPI_MAX_RAM_SLOTS) {
1343             error_report("unsupported amount of memory slots: %"PRIu64,
1344                          machine->ram_slots);
1345             exit(EXIT_FAILURE);
1346         }
1347 
1348         if (QEMU_ALIGN_UP(machine->maxram_size,
1349                           TARGET_PAGE_SIZE) != machine->maxram_size) {
1350             error_report("maximum memory size must by aligned to multiple of "
1351                          "%d bytes", TARGET_PAGE_SIZE);
1352             exit(EXIT_FAILURE);
1353         }
1354 
1355         pcms->hotplug_memory.base =
1356             ROUND_UP(0x100000000ULL + pcms->above_4g_mem_size, 1ULL << 30);
1357 
1358         if (pcmc->enforce_aligned_dimm) {
1359             /* size hotplug region assuming 1G page max alignment per slot */
1360             hotplug_mem_size += (1ULL << 30) * machine->ram_slots;
1361         }
1362 
1363         if ((pcms->hotplug_memory.base + hotplug_mem_size) <
1364             hotplug_mem_size) {
1365             error_report("unsupported amount of maximum memory: " RAM_ADDR_FMT,
1366                          machine->maxram_size);
1367             exit(EXIT_FAILURE);
1368         }
1369 
1370         memory_region_init(&pcms->hotplug_memory.mr, OBJECT(pcms),
1371                            "hotplug-memory", hotplug_mem_size);
1372         memory_region_add_subregion(system_memory, pcms->hotplug_memory.base,
1373                                     &pcms->hotplug_memory.mr);
1374     }
1375 
1376     /* Initialize PC system firmware */
1377     pc_system_firmware_init(rom_memory, guest_info->isapc_ram_fw);
1378 
1379     option_rom_mr = g_malloc(sizeof(*option_rom_mr));
1380     memory_region_init_ram(option_rom_mr, NULL, "pc.rom", PC_ROM_SIZE,
1381                            &error_fatal);
1382     vmstate_register_ram_global(option_rom_mr);
1383     memory_region_add_subregion_overlap(rom_memory,
1384                                         PC_ROM_MIN_VGA,
1385                                         option_rom_mr,
1386                                         1);
1387 
1388     fw_cfg = bochs_bios_init(&address_space_memory);
1389 
1390     rom_set_fw(fw_cfg);
1391 
1392     if (guest_info->has_reserved_memory && pcms->hotplug_memory.base) {
1393         uint64_t *val = g_malloc(sizeof(*val));
1394         PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
1395         uint64_t res_mem_end = pcms->hotplug_memory.base;
1396 
1397         if (!pcmc->broken_reserved_end) {
1398             res_mem_end += memory_region_size(&pcms->hotplug_memory.mr);
1399         }
1400         *val = cpu_to_le64(ROUND_UP(res_mem_end, 0x1ULL << 30));
1401         fw_cfg_add_file(fw_cfg, "etc/reserved-memory-end", val, sizeof(*val));
1402     }
1403 
1404     if (linux_boot) {
1405         load_linux(pcms, fw_cfg);
1406     }
1407 
1408     for (i = 0; i < nb_option_roms; i++) {
1409         rom_add_option(option_rom[i].name, option_rom[i].bootindex);
1410     }
1411     guest_info->fw_cfg = fw_cfg;
1412     return fw_cfg;
1413 }
1414 
1415 qemu_irq pc_allocate_cpu_irq(void)
1416 {
1417     return qemu_allocate_irq(pic_irq_request, NULL, 0);
1418 }
1419 
1420 DeviceState *pc_vga_init(ISABus *isa_bus, PCIBus *pci_bus)
1421 {
1422     DeviceState *dev = NULL;
1423 
1424     if (pci_bus) {
1425         PCIDevice *pcidev = pci_vga_init(pci_bus);
1426         dev = pcidev ? &pcidev->qdev : NULL;
1427     } else if (isa_bus) {
1428         ISADevice *isadev = isa_vga_init(isa_bus);
1429         dev = isadev ? DEVICE(isadev) : NULL;
1430     }
1431     return dev;
1432 }
1433 
1434 static const MemoryRegionOps ioport80_io_ops = {
1435     .write = ioport80_write,
1436     .read = ioport80_read,
1437     .endianness = DEVICE_NATIVE_ENDIAN,
1438     .impl = {
1439         .min_access_size = 1,
1440         .max_access_size = 1,
1441     },
1442 };
1443 
1444 static const MemoryRegionOps ioportF0_io_ops = {
1445     .write = ioportF0_write,
1446     .read = ioportF0_read,
1447     .endianness = DEVICE_NATIVE_ENDIAN,
1448     .impl = {
1449         .min_access_size = 1,
1450         .max_access_size = 1,
1451     },
1452 };
1453 
1454 void pc_basic_device_init(ISABus *isa_bus, qemu_irq *gsi,
1455                           ISADevice **rtc_state,
1456                           bool create_fdctrl,
1457                           bool no_vmport,
1458                           uint32_t hpet_irqs)
1459 {
1460     int i;
1461     DriveInfo *fd[MAX_FD];
1462     DeviceState *hpet = NULL;
1463     int pit_isa_irq = 0;
1464     qemu_irq pit_alt_irq = NULL;
1465     qemu_irq rtc_irq = NULL;
1466     qemu_irq *a20_line;
1467     ISADevice *i8042, *port92, *vmmouse, *pit = NULL;
1468     MemoryRegion *ioport80_io = g_new(MemoryRegion, 1);
1469     MemoryRegion *ioportF0_io = g_new(MemoryRegion, 1);
1470 
1471     memory_region_init_io(ioport80_io, NULL, &ioport80_io_ops, NULL, "ioport80", 1);
1472     memory_region_add_subregion(isa_bus->address_space_io, 0x80, ioport80_io);
1473 
1474     memory_region_init_io(ioportF0_io, NULL, &ioportF0_io_ops, NULL, "ioportF0", 1);
1475     memory_region_add_subregion(isa_bus->address_space_io, 0xf0, ioportF0_io);
1476 
1477     /*
1478      * Check if an HPET shall be created.
1479      *
1480      * Without KVM_CAP_PIT_STATE2, we cannot switch off the in-kernel PIT
1481      * when the HPET wants to take over. Thus we have to disable the latter.
1482      */
1483     if (!no_hpet && (!kvm_irqchip_in_kernel() || kvm_has_pit_state2())) {
1484         /* In order to set property, here not using sysbus_try_create_simple */
1485         hpet = qdev_try_create(NULL, TYPE_HPET);
1486         if (hpet) {
1487             /* For pc-piix-*, hpet's intcap is always IRQ2. For pc-q35-1.7
1488              * and earlier, use IRQ2 for compat. Otherwise, use IRQ16~23,
1489              * IRQ8 and IRQ2.
1490              */
1491             uint8_t compat = object_property_get_int(OBJECT(hpet),
1492                     HPET_INTCAP, NULL);
1493             if (!compat) {
1494                 qdev_prop_set_uint32(hpet, HPET_INTCAP, hpet_irqs);
1495             }
1496             qdev_init_nofail(hpet);
1497             sysbus_mmio_map(SYS_BUS_DEVICE(hpet), 0, HPET_BASE);
1498 
1499             for (i = 0; i < GSI_NUM_PINS; i++) {
1500                 sysbus_connect_irq(SYS_BUS_DEVICE(hpet), i, gsi[i]);
1501             }
1502             pit_isa_irq = -1;
1503             pit_alt_irq = qdev_get_gpio_in(hpet, HPET_LEGACY_PIT_INT);
1504             rtc_irq = qdev_get_gpio_in(hpet, HPET_LEGACY_RTC_INT);
1505         }
1506     }
1507     *rtc_state = rtc_init(isa_bus, 2000, rtc_irq);
1508 
1509     qemu_register_boot_set(pc_boot_set, *rtc_state);
1510 
1511     if (!xen_enabled()) {
1512         if (kvm_pit_in_kernel()) {
1513             pit = kvm_pit_init(isa_bus, 0x40);
1514         } else {
1515             pit = pit_init(isa_bus, 0x40, pit_isa_irq, pit_alt_irq);
1516         }
1517         if (hpet) {
1518             /* connect PIT to output control line of the HPET */
1519             qdev_connect_gpio_out(hpet, 0, qdev_get_gpio_in(DEVICE(pit), 0));
1520         }
1521         pcspk_init(isa_bus, pit);
1522     }
1523 
1524     serial_hds_isa_init(isa_bus, MAX_SERIAL_PORTS);
1525     parallel_hds_isa_init(isa_bus, MAX_PARALLEL_PORTS);
1526 
1527     a20_line = qemu_allocate_irqs(handle_a20_line_change, first_cpu, 2);
1528     i8042 = isa_create_simple(isa_bus, "i8042");
1529     i8042_setup_a20_line(i8042, &a20_line[0]);
1530     if (!no_vmport) {
1531         vmport_init(isa_bus);
1532         vmmouse = isa_try_create(isa_bus, "vmmouse");
1533     } else {
1534         vmmouse = NULL;
1535     }
1536     if (vmmouse) {
1537         DeviceState *dev = DEVICE(vmmouse);
1538         qdev_prop_set_ptr(dev, "ps2_mouse", i8042);
1539         qdev_init_nofail(dev);
1540     }
1541     port92 = isa_create_simple(isa_bus, "port92");
1542     port92_init(port92, &a20_line[1]);
1543 
1544     DMA_init(0);
1545 
1546     for(i = 0; i < MAX_FD; i++) {
1547         fd[i] = drive_get(IF_FLOPPY, 0, i);
1548         create_fdctrl |= !!fd[i];
1549     }
1550     if (create_fdctrl) {
1551         fdctrl_init_isa(isa_bus, fd);
1552     }
1553 }
1554 
1555 void pc_nic_init(ISABus *isa_bus, PCIBus *pci_bus)
1556 {
1557     int i;
1558 
1559     for (i = 0; i < nb_nics; i++) {
1560         NICInfo *nd = &nd_table[i];
1561 
1562         if (!pci_bus || (nd->model && strcmp(nd->model, "ne2k_isa") == 0)) {
1563             pc_init_ne2k_isa(isa_bus, nd);
1564         } else {
1565             pci_nic_init_nofail(nd, pci_bus, "e1000", NULL);
1566         }
1567     }
1568 }
1569 
1570 void pc_pci_device_init(PCIBus *pci_bus)
1571 {
1572     int max_bus;
1573     int bus;
1574 
1575     max_bus = drive_get_max_bus(IF_SCSI);
1576     for (bus = 0; bus <= max_bus; bus++) {
1577         pci_create_simple(pci_bus, -1, "lsi53c895a");
1578     }
1579 }
1580 
1581 void ioapic_init_gsi(GSIState *gsi_state, const char *parent_name)
1582 {
1583     DeviceState *dev;
1584     SysBusDevice *d;
1585     unsigned int i;
1586 
1587     if (kvm_ioapic_in_kernel()) {
1588         dev = qdev_create(NULL, "kvm-ioapic");
1589     } else {
1590         dev = qdev_create(NULL, "ioapic");
1591     }
1592     if (parent_name) {
1593         object_property_add_child(object_resolve_path(parent_name, NULL),
1594                                   "ioapic", OBJECT(dev), NULL);
1595     }
1596     qdev_init_nofail(dev);
1597     d = SYS_BUS_DEVICE(dev);
1598     sysbus_mmio_map(d, 0, IO_APIC_DEFAULT_ADDRESS);
1599 
1600     for (i = 0; i < IOAPIC_NUM_PINS; i++) {
1601         gsi_state->ioapic_irq[i] = qdev_get_gpio_in(dev, i);
1602     }
1603 }
1604 
1605 static void pc_dimm_plug(HotplugHandler *hotplug_dev,
1606                          DeviceState *dev, Error **errp)
1607 {
1608     HotplugHandlerClass *hhc;
1609     Error *local_err = NULL;
1610     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1611     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
1612     PCDIMMDevice *dimm = PC_DIMM(dev);
1613     PCDIMMDeviceClass *ddc = PC_DIMM_GET_CLASS(dimm);
1614     MemoryRegion *mr = ddc->get_memory_region(dimm);
1615     uint64_t align = TARGET_PAGE_SIZE;
1616 
1617     if (memory_region_get_alignment(mr) && pcmc->enforce_aligned_dimm) {
1618         align = memory_region_get_alignment(mr);
1619     }
1620 
1621     if (!pcms->acpi_dev) {
1622         error_setg(&local_err,
1623                    "memory hotplug is not enabled: missing acpi device");
1624         goto out;
1625     }
1626 
1627     pc_dimm_memory_plug(dev, &pcms->hotplug_memory, mr, align, &local_err);
1628     if (local_err) {
1629         goto out;
1630     }
1631 
1632     hhc = HOTPLUG_HANDLER_GET_CLASS(pcms->acpi_dev);
1633     hhc->plug(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &error_abort);
1634 out:
1635     error_propagate(errp, local_err);
1636 }
1637 
1638 static void pc_dimm_unplug_request(HotplugHandler *hotplug_dev,
1639                                    DeviceState *dev, Error **errp)
1640 {
1641     HotplugHandlerClass *hhc;
1642     Error *local_err = NULL;
1643     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1644 
1645     if (!pcms->acpi_dev) {
1646         error_setg(&local_err,
1647                    "memory hotplug is not enabled: missing acpi device");
1648         goto out;
1649     }
1650 
1651     hhc = HOTPLUG_HANDLER_GET_CLASS(pcms->acpi_dev);
1652     hhc->unplug_request(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &local_err);
1653 
1654 out:
1655     error_propagate(errp, local_err);
1656 }
1657 
1658 static void pc_dimm_unplug(HotplugHandler *hotplug_dev,
1659                            DeviceState *dev, Error **errp)
1660 {
1661     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1662     PCDIMMDevice *dimm = PC_DIMM(dev);
1663     PCDIMMDeviceClass *ddc = PC_DIMM_GET_CLASS(dimm);
1664     MemoryRegion *mr = ddc->get_memory_region(dimm);
1665     HotplugHandlerClass *hhc;
1666     Error *local_err = NULL;
1667 
1668     hhc = HOTPLUG_HANDLER_GET_CLASS(pcms->acpi_dev);
1669     hhc->unplug(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &local_err);
1670 
1671     if (local_err) {
1672         goto out;
1673     }
1674 
1675     pc_dimm_memory_unplug(dev, &pcms->hotplug_memory, mr);
1676     object_unparent(OBJECT(dev));
1677 
1678  out:
1679     error_propagate(errp, local_err);
1680 }
1681 
1682 static void pc_cpu_plug(HotplugHandler *hotplug_dev,
1683                         DeviceState *dev, Error **errp)
1684 {
1685     HotplugHandlerClass *hhc;
1686     Error *local_err = NULL;
1687     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1688 
1689     if (!dev->hotplugged) {
1690         goto out;
1691     }
1692 
1693     if (!pcms->acpi_dev) {
1694         error_setg(&local_err,
1695                    "cpu hotplug is not enabled: missing acpi device");
1696         goto out;
1697     }
1698 
1699     hhc = HOTPLUG_HANDLER_GET_CLASS(pcms->acpi_dev);
1700     hhc->plug(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &local_err);
1701     if (local_err) {
1702         goto out;
1703     }
1704 
1705     /* increment the number of CPUs */
1706     rtc_set_memory(pcms->rtc, 0x5f, rtc_get_memory(pcms->rtc, 0x5f) + 1);
1707 out:
1708     error_propagate(errp, local_err);
1709 }
1710 
1711 static void pc_machine_device_plug_cb(HotplugHandler *hotplug_dev,
1712                                       DeviceState *dev, Error **errp)
1713 {
1714     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
1715         pc_dimm_plug(hotplug_dev, dev, errp);
1716     } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
1717         pc_cpu_plug(hotplug_dev, dev, errp);
1718     }
1719 }
1720 
1721 static void pc_machine_device_unplug_request_cb(HotplugHandler *hotplug_dev,
1722                                                 DeviceState *dev, Error **errp)
1723 {
1724     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
1725         pc_dimm_unplug_request(hotplug_dev, dev, errp);
1726     } else {
1727         error_setg(errp, "acpi: device unplug request for not supported device"
1728                    " type: %s", object_get_typename(OBJECT(dev)));
1729     }
1730 }
1731 
1732 static void pc_machine_device_unplug_cb(HotplugHandler *hotplug_dev,
1733                                         DeviceState *dev, Error **errp)
1734 {
1735     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
1736         pc_dimm_unplug(hotplug_dev, dev, errp);
1737     } else {
1738         error_setg(errp, "acpi: device unplug for not supported device"
1739                    " type: %s", object_get_typename(OBJECT(dev)));
1740     }
1741 }
1742 
1743 static HotplugHandler *pc_get_hotpug_handler(MachineState *machine,
1744                                              DeviceState *dev)
1745 {
1746     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(machine);
1747 
1748     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM) ||
1749         object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
1750         return HOTPLUG_HANDLER(machine);
1751     }
1752 
1753     return pcmc->get_hotplug_handler ?
1754         pcmc->get_hotplug_handler(machine, dev) : NULL;
1755 }
1756 
1757 static void
1758 pc_machine_get_hotplug_memory_region_size(Object *obj, Visitor *v, void *opaque,
1759                                           const char *name, Error **errp)
1760 {
1761     PCMachineState *pcms = PC_MACHINE(obj);
1762     int64_t value = memory_region_size(&pcms->hotplug_memory.mr);
1763 
1764     visit_type_int(v, &value, name, errp);
1765 }
1766 
1767 static void pc_machine_get_max_ram_below_4g(Object *obj, Visitor *v,
1768                                          void *opaque, const char *name,
1769                                          Error **errp)
1770 {
1771     PCMachineState *pcms = PC_MACHINE(obj);
1772     uint64_t value = pcms->max_ram_below_4g;
1773 
1774     visit_type_size(v, &value, name, errp);
1775 }
1776 
1777 static void pc_machine_set_max_ram_below_4g(Object *obj, Visitor *v,
1778                                          void *opaque, const char *name,
1779                                          Error **errp)
1780 {
1781     PCMachineState *pcms = PC_MACHINE(obj);
1782     Error *error = NULL;
1783     uint64_t value;
1784 
1785     visit_type_size(v, &value, name, &error);
1786     if (error) {
1787         error_propagate(errp, error);
1788         return;
1789     }
1790     if (value > (1ULL << 32)) {
1791         error_setg(&error,
1792                    "Machine option 'max-ram-below-4g=%"PRIu64
1793                    "' expects size less than or equal to 4G", value);
1794         error_propagate(errp, error);
1795         return;
1796     }
1797 
1798     if (value < (1ULL << 20)) {
1799         error_report("Warning: small max_ram_below_4g(%"PRIu64
1800                      ") less than 1M.  BIOS may not work..",
1801                      value);
1802     }
1803 
1804     pcms->max_ram_below_4g = value;
1805 }
1806 
1807 static void pc_machine_get_vmport(Object *obj, Visitor *v, void *opaque,
1808                                   const char *name, Error **errp)
1809 {
1810     PCMachineState *pcms = PC_MACHINE(obj);
1811     OnOffAuto vmport = pcms->vmport;
1812 
1813     visit_type_OnOffAuto(v, &vmport, name, errp);
1814 }
1815 
1816 static void pc_machine_set_vmport(Object *obj, Visitor *v, void *opaque,
1817                                   const char *name, Error **errp)
1818 {
1819     PCMachineState *pcms = PC_MACHINE(obj);
1820 
1821     visit_type_OnOffAuto(v, &pcms->vmport, name, errp);
1822 }
1823 
1824 bool pc_machine_is_smm_enabled(PCMachineState *pcms)
1825 {
1826     bool smm_available = false;
1827 
1828     if (pcms->smm == ON_OFF_AUTO_OFF) {
1829         return false;
1830     }
1831 
1832     if (tcg_enabled() || qtest_enabled()) {
1833         smm_available = true;
1834     } else if (kvm_enabled()) {
1835         smm_available = kvm_has_smm();
1836     }
1837 
1838     if (smm_available) {
1839         return true;
1840     }
1841 
1842     if (pcms->smm == ON_OFF_AUTO_ON) {
1843         error_report("System Management Mode not supported by this hypervisor.");
1844         exit(1);
1845     }
1846     return false;
1847 }
1848 
1849 static void pc_machine_get_smm(Object *obj, Visitor *v, void *opaque,
1850                               const char *name, Error **errp)
1851 {
1852     PCMachineState *pcms = PC_MACHINE(obj);
1853     OnOffAuto smm = pcms->smm;
1854 
1855     visit_type_OnOffAuto(v, &smm, name, errp);
1856 }
1857 
1858 static void pc_machine_set_smm(Object *obj, Visitor *v, void *opaque,
1859                                   const char *name, Error **errp)
1860 {
1861     PCMachineState *pcms = PC_MACHINE(obj);
1862 
1863     visit_type_OnOffAuto(v, &pcms->smm, name, errp);
1864 }
1865 
1866 static bool pc_machine_get_nvdimm(Object *obj, Error **errp)
1867 {
1868     PCMachineState *pcms = PC_MACHINE(obj);
1869 
1870     return pcms->nvdimm;
1871 }
1872 
1873 static void pc_machine_set_nvdimm(Object *obj, bool value, Error **errp)
1874 {
1875     PCMachineState *pcms = PC_MACHINE(obj);
1876 
1877     pcms->nvdimm = value;
1878 }
1879 
1880 static void pc_machine_initfn(Object *obj)
1881 {
1882     PCMachineState *pcms = PC_MACHINE(obj);
1883 
1884     object_property_add(obj, PC_MACHINE_MEMHP_REGION_SIZE, "int",
1885                         pc_machine_get_hotplug_memory_region_size,
1886                         NULL, NULL, NULL, &error_abort);
1887 
1888     pcms->max_ram_below_4g = 1ULL << 32; /* 4G */
1889     object_property_add(obj, PC_MACHINE_MAX_RAM_BELOW_4G, "size",
1890                         pc_machine_get_max_ram_below_4g,
1891                         pc_machine_set_max_ram_below_4g,
1892                         NULL, NULL, &error_abort);
1893     object_property_set_description(obj, PC_MACHINE_MAX_RAM_BELOW_4G,
1894                                     "Maximum ram below the 4G boundary (32bit boundary)",
1895                                     &error_abort);
1896 
1897     pcms->smm = ON_OFF_AUTO_AUTO;
1898     object_property_add(obj, PC_MACHINE_SMM, "OnOffAuto",
1899                         pc_machine_get_smm,
1900                         pc_machine_set_smm,
1901                         NULL, NULL, &error_abort);
1902     object_property_set_description(obj, PC_MACHINE_SMM,
1903                                     "Enable SMM (pc & q35)",
1904                                     &error_abort);
1905 
1906     pcms->vmport = ON_OFF_AUTO_AUTO;
1907     object_property_add(obj, PC_MACHINE_VMPORT, "OnOffAuto",
1908                         pc_machine_get_vmport,
1909                         pc_machine_set_vmport,
1910                         NULL, NULL, &error_abort);
1911     object_property_set_description(obj, PC_MACHINE_VMPORT,
1912                                     "Enable vmport (pc & q35)",
1913                                     &error_abort);
1914 
1915     /* nvdimm is disabled on default. */
1916     pcms->nvdimm = false;
1917     object_property_add_bool(obj, PC_MACHINE_NVDIMM, pc_machine_get_nvdimm,
1918                              pc_machine_set_nvdimm, &error_abort);
1919 }
1920 
1921 static void pc_machine_reset(void)
1922 {
1923     CPUState *cs;
1924     X86CPU *cpu;
1925 
1926     qemu_devices_reset();
1927 
1928     /* Reset APIC after devices have been reset to cancel
1929      * any changes that qemu_devices_reset() might have done.
1930      */
1931     CPU_FOREACH(cs) {
1932         cpu = X86_CPU(cs);
1933 
1934         if (cpu->apic_state) {
1935             device_reset(cpu->apic_state);
1936         }
1937     }
1938 }
1939 
1940 static unsigned pc_cpu_index_to_socket_id(unsigned cpu_index)
1941 {
1942     X86CPUTopoInfo topo;
1943     x86_topo_ids_from_idx(smp_cores, smp_threads, cpu_index,
1944                           &topo);
1945     return topo.pkg_id;
1946 }
1947 
1948 static void pc_machine_class_init(ObjectClass *oc, void *data)
1949 {
1950     MachineClass *mc = MACHINE_CLASS(oc);
1951     PCMachineClass *pcmc = PC_MACHINE_CLASS(oc);
1952     HotplugHandlerClass *hc = HOTPLUG_HANDLER_CLASS(oc);
1953 
1954     pcmc->get_hotplug_handler = mc->get_hotplug_handler;
1955     pcmc->pci_enabled = true;
1956     pcmc->has_acpi_build = true;
1957     pcmc->rsdp_in_ram = true;
1958     pcmc->smbios_defaults = true;
1959     pcmc->smbios_uuid_encoded = true;
1960     pcmc->gigabyte_align = true;
1961     pcmc->has_reserved_memory = true;
1962     pcmc->kvmclock_enabled = true;
1963     pcmc->enforce_aligned_dimm = true;
1964     /* BIOS ACPI tables: 128K. Other BIOS datastructures: less than 4K reported
1965      * to be used at the moment, 32K should be enough for a while.  */
1966     pcmc->acpi_data_size = 0x20000 + 0x8000;
1967     pcmc->save_tsc_khz = true;
1968     mc->get_hotplug_handler = pc_get_hotpug_handler;
1969     mc->cpu_index_to_socket_id = pc_cpu_index_to_socket_id;
1970     mc->default_boot_order = "cad";
1971     mc->hot_add_cpu = pc_hot_add_cpu;
1972     mc->max_cpus = 255;
1973     mc->reset = pc_machine_reset;
1974     hc->plug = pc_machine_device_plug_cb;
1975     hc->unplug_request = pc_machine_device_unplug_request_cb;
1976     hc->unplug = pc_machine_device_unplug_cb;
1977 }
1978 
1979 static const TypeInfo pc_machine_info = {
1980     .name = TYPE_PC_MACHINE,
1981     .parent = TYPE_MACHINE,
1982     .abstract = true,
1983     .instance_size = sizeof(PCMachineState),
1984     .instance_init = pc_machine_initfn,
1985     .class_size = sizeof(PCMachineClass),
1986     .class_init = pc_machine_class_init,
1987     .interfaces = (InterfaceInfo[]) {
1988          { TYPE_HOTPLUG_HANDLER },
1989          { }
1990     },
1991 };
1992 
1993 static void pc_machine_register_types(void)
1994 {
1995     type_register_static(&pc_machine_info);
1996 }
1997 
1998 type_init(pc_machine_register_types)
1999