xref: /openbmc/qemu/hw/i386/pc.c (revision 9c4218e9)
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 "qemu/osdep.h"
25 #include "hw/hw.h"
26 #include "hw/i386/pc.h"
27 #include "hw/char/serial.h"
28 #include "hw/i386/apic.h"
29 #include "hw/i386/topology.h"
30 #include "sysemu/cpus.h"
31 #include "hw/block/fdc.h"
32 #include "hw/ide.h"
33 #include "hw/pci/pci.h"
34 #include "hw/pci/pci_bus.h"
35 #include "hw/nvram/fw_cfg.h"
36 #include "hw/timer/hpet.h"
37 #include "hw/smbios/smbios.h"
38 #include "hw/loader.h"
39 #include "elf.h"
40 #include "multiboot.h"
41 #include "hw/timer/mc146818rtc.h"
42 #include "hw/timer/i8254.h"
43 #include "hw/audio/pcspk.h"
44 #include "hw/pci/msi.h"
45 #include "hw/sysbus.h"
46 #include "sysemu/sysemu.h"
47 #include "sysemu/numa.h"
48 #include "sysemu/kvm.h"
49 #include "sysemu/qtest.h"
50 #include "kvm_i386.h"
51 #include "hw/xen/xen.h"
52 #include "sysemu/block-backend.h"
53 #include "hw/block/block.h"
54 #include "ui/qemu-spice.h"
55 #include "exec/memory.h"
56 #include "exec/address-spaces.h"
57 #include "sysemu/arch_init.h"
58 #include "qemu/bitmap.h"
59 #include "qemu/config-file.h"
60 #include "qemu/error-report.h"
61 #include "hw/acpi/acpi.h"
62 #include "hw/acpi/cpu_hotplug.h"
63 #include "hw/boards.h"
64 #include "hw/pci/pci_host.h"
65 #include "acpi-build.h"
66 #include "hw/mem/pc-dimm.h"
67 #include "qapi/visitor.h"
68 #include "qapi-visit.h"
69 #include "qom/cpu.h"
70 
71 /* debug PC/ISA interrupts */
72 //#define DEBUG_IRQ
73 
74 #ifdef DEBUG_IRQ
75 #define DPRINTF(fmt, ...)                                       \
76     do { printf("CPUIRQ: " fmt , ## __VA_ARGS__); } while (0)
77 #else
78 #define DPRINTF(fmt, ...)
79 #endif
80 
81 #define BIOS_CFG_IOPORT 0x510
82 #define FW_CFG_ACPI_TABLES (FW_CFG_ARCH_LOCAL + 0)
83 #define FW_CFG_SMBIOS_ENTRIES (FW_CFG_ARCH_LOCAL + 1)
84 #define FW_CFG_IRQ0_OVERRIDE (FW_CFG_ARCH_LOCAL + 2)
85 #define FW_CFG_E820_TABLE (FW_CFG_ARCH_LOCAL + 3)
86 #define FW_CFG_HPET (FW_CFG_ARCH_LOCAL + 4)
87 
88 #define E820_NR_ENTRIES		16
89 
90 struct e820_entry {
91     uint64_t address;
92     uint64_t length;
93     uint32_t type;
94 } QEMU_PACKED __attribute((__aligned__(4)));
95 
96 struct e820_table {
97     uint32_t count;
98     struct e820_entry entry[E820_NR_ENTRIES];
99 } QEMU_PACKED __attribute((__aligned__(4)));
100 
101 static struct e820_table e820_reserve;
102 static struct e820_entry *e820_table;
103 static unsigned e820_entries;
104 struct hpet_fw_config hpet_cfg = {.count = UINT8_MAX};
105 
106 void gsi_handler(void *opaque, int n, int level)
107 {
108     GSIState *s = opaque;
109 
110     DPRINTF("pc: %s GSI %d\n", level ? "raising" : "lowering", n);
111     if (n < ISA_NUM_IRQS) {
112         qemu_set_irq(s->i8259_irq[n], level);
113     }
114     qemu_set_irq(s->ioapic_irq[n], level);
115 }
116 
117 static void ioport80_write(void *opaque, hwaddr addr, uint64_t data,
118                            unsigned size)
119 {
120 }
121 
122 static uint64_t ioport80_read(void *opaque, hwaddr addr, unsigned size)
123 {
124     return 0xffffffffffffffffULL;
125 }
126 
127 /* MSDOS compatibility mode FPU exception support */
128 static qemu_irq ferr_irq;
129 
130 void pc_register_ferr_irq(qemu_irq irq)
131 {
132     ferr_irq = irq;
133 }
134 
135 /* XXX: add IGNNE support */
136 void cpu_set_ferr(CPUX86State *s)
137 {
138     qemu_irq_raise(ferr_irq);
139 }
140 
141 static void ioportF0_write(void *opaque, hwaddr addr, uint64_t data,
142                            unsigned size)
143 {
144     qemu_irq_lower(ferr_irq);
145 }
146 
147 static uint64_t ioportF0_read(void *opaque, hwaddr addr, unsigned size)
148 {
149     return 0xffffffffffffffffULL;
150 }
151 
152 /* TSC handling */
153 uint64_t cpu_get_tsc(CPUX86State *env)
154 {
155     return cpu_get_ticks();
156 }
157 
158 /* IRQ handling */
159 int cpu_get_pic_interrupt(CPUX86State *env)
160 {
161     X86CPU *cpu = x86_env_get_cpu(env);
162     int intno;
163 
164     intno = apic_get_interrupt(cpu->apic_state);
165     if (intno >= 0) {
166         return intno;
167     }
168     /* read the irq from the PIC */
169     if (!apic_accept_pic_intr(cpu->apic_state)) {
170         return -1;
171     }
172 
173     intno = pic_read_irq(isa_pic);
174     return intno;
175 }
176 
177 static void pic_irq_request(void *opaque, int irq, int level)
178 {
179     CPUState *cs = first_cpu;
180     X86CPU *cpu = X86_CPU(cs);
181 
182     DPRINTF("pic_irqs: %s irq %d\n", level? "raise" : "lower", irq);
183     if (cpu->apic_state) {
184         CPU_FOREACH(cs) {
185             cpu = X86_CPU(cs);
186             if (apic_accept_pic_intr(cpu->apic_state)) {
187                 apic_deliver_pic_intr(cpu->apic_state, level);
188             }
189         }
190     } else {
191         if (level) {
192             cpu_interrupt(cs, CPU_INTERRUPT_HARD);
193         } else {
194             cpu_reset_interrupt(cs, CPU_INTERRUPT_HARD);
195         }
196     }
197 }
198 
199 /* PC cmos mappings */
200 
201 #define REG_EQUIPMENT_BYTE          0x14
202 
203 static int cmos_get_fd_drive_type(FloppyDriveType fd0)
204 {
205     int val;
206 
207     switch (fd0) {
208     case FLOPPY_DRIVE_TYPE_144:
209         /* 1.44 Mb 3"5 drive */
210         val = 4;
211         break;
212     case FLOPPY_DRIVE_TYPE_288:
213         /* 2.88 Mb 3"5 drive */
214         val = 5;
215         break;
216     case FLOPPY_DRIVE_TYPE_120:
217         /* 1.2 Mb 5"5 drive */
218         val = 2;
219         break;
220     case FLOPPY_DRIVE_TYPE_NONE:
221     default:
222         val = 0;
223         break;
224     }
225     return val;
226 }
227 
228 static void cmos_init_hd(ISADevice *s, int type_ofs, int info_ofs,
229                          int16_t cylinders, int8_t heads, int8_t sectors)
230 {
231     rtc_set_memory(s, type_ofs, 47);
232     rtc_set_memory(s, info_ofs, cylinders);
233     rtc_set_memory(s, info_ofs + 1, cylinders >> 8);
234     rtc_set_memory(s, info_ofs + 2, heads);
235     rtc_set_memory(s, info_ofs + 3, 0xff);
236     rtc_set_memory(s, info_ofs + 4, 0xff);
237     rtc_set_memory(s, info_ofs + 5, 0xc0 | ((heads > 8) << 3));
238     rtc_set_memory(s, info_ofs + 6, cylinders);
239     rtc_set_memory(s, info_ofs + 7, cylinders >> 8);
240     rtc_set_memory(s, info_ofs + 8, sectors);
241 }
242 
243 /* convert boot_device letter to something recognizable by the bios */
244 static int boot_device2nibble(char boot_device)
245 {
246     switch(boot_device) {
247     case 'a':
248     case 'b':
249         return 0x01; /* floppy boot */
250     case 'c':
251         return 0x02; /* hard drive boot */
252     case 'd':
253         return 0x03; /* CD-ROM boot */
254     case 'n':
255         return 0x04; /* Network boot */
256     }
257     return 0;
258 }
259 
260 static void set_boot_dev(ISADevice *s, const char *boot_device, Error **errp)
261 {
262 #define PC_MAX_BOOT_DEVICES 3
263     int nbds, bds[3] = { 0, };
264     int i;
265 
266     nbds = strlen(boot_device);
267     if (nbds > PC_MAX_BOOT_DEVICES) {
268         error_setg(errp, "Too many boot devices for PC");
269         return;
270     }
271     for (i = 0; i < nbds; i++) {
272         bds[i] = boot_device2nibble(boot_device[i]);
273         if (bds[i] == 0) {
274             error_setg(errp, "Invalid boot device for PC: '%c'",
275                        boot_device[i]);
276             return;
277         }
278     }
279     rtc_set_memory(s, 0x3d, (bds[1] << 4) | bds[0]);
280     rtc_set_memory(s, 0x38, (bds[2] << 4) | (fd_bootchk ? 0x0 : 0x1));
281 }
282 
283 static void pc_boot_set(void *opaque, const char *boot_device, Error **errp)
284 {
285     set_boot_dev(opaque, boot_device, errp);
286 }
287 
288 static void pc_cmos_init_floppy(ISADevice *rtc_state, ISADevice *floppy)
289 {
290     int val, nb, i;
291     FloppyDriveType fd_type[2] = { FLOPPY_DRIVE_TYPE_NONE,
292                                    FLOPPY_DRIVE_TYPE_NONE };
293 
294     /* floppy type */
295     if (floppy) {
296         for (i = 0; i < 2; i++) {
297             fd_type[i] = isa_fdc_get_drive_type(floppy, i);
298         }
299     }
300     val = (cmos_get_fd_drive_type(fd_type[0]) << 4) |
301         cmos_get_fd_drive_type(fd_type[1]);
302     rtc_set_memory(rtc_state, 0x10, val);
303 
304     val = rtc_get_memory(rtc_state, REG_EQUIPMENT_BYTE);
305     nb = 0;
306     if (fd_type[0] != FLOPPY_DRIVE_TYPE_NONE) {
307         nb++;
308     }
309     if (fd_type[1] != FLOPPY_DRIVE_TYPE_NONE) {
310         nb++;
311     }
312     switch (nb) {
313     case 0:
314         break;
315     case 1:
316         val |= 0x01; /* 1 drive, ready for boot */
317         break;
318     case 2:
319         val |= 0x41; /* 2 drives, ready for boot */
320         break;
321     }
322     rtc_set_memory(rtc_state, REG_EQUIPMENT_BYTE, val);
323 }
324 
325 typedef struct pc_cmos_init_late_arg {
326     ISADevice *rtc_state;
327     BusState *idebus[2];
328 } pc_cmos_init_late_arg;
329 
330 typedef struct check_fdc_state {
331     ISADevice *floppy;
332     bool multiple;
333 } CheckFdcState;
334 
335 static int check_fdc(Object *obj, void *opaque)
336 {
337     CheckFdcState *state = opaque;
338     Object *fdc;
339     uint32_t iobase;
340     Error *local_err = NULL;
341 
342     fdc = object_dynamic_cast(obj, TYPE_ISA_FDC);
343     if (!fdc) {
344         return 0;
345     }
346 
347     iobase = object_property_get_int(obj, "iobase", &local_err);
348     if (local_err || iobase != 0x3f0) {
349         error_free(local_err);
350         return 0;
351     }
352 
353     if (state->floppy) {
354         state->multiple = true;
355     } else {
356         state->floppy = ISA_DEVICE(obj);
357     }
358     return 0;
359 }
360 
361 static const char * const fdc_container_path[] = {
362     "/unattached", "/peripheral", "/peripheral-anon"
363 };
364 
365 /*
366  * Locate the FDC at IO address 0x3f0, in order to configure the CMOS registers
367  * and ACPI objects.
368  */
369 ISADevice *pc_find_fdc0(void)
370 {
371     int i;
372     Object *container;
373     CheckFdcState state = { 0 };
374 
375     for (i = 0; i < ARRAY_SIZE(fdc_container_path); i++) {
376         container = container_get(qdev_get_machine(), fdc_container_path[i]);
377         object_child_foreach(container, check_fdc, &state);
378     }
379 
380     if (state.multiple) {
381         error_report("warning: multiple floppy disk controllers with "
382                      "iobase=0x3f0 have been found");
383         error_printf("the one being picked for CMOS setup might not reflect "
384                      "your intent");
385     }
386 
387     return state.floppy;
388 }
389 
390 static void pc_cmos_init_late(void *opaque)
391 {
392     pc_cmos_init_late_arg *arg = opaque;
393     ISADevice *s = arg->rtc_state;
394     int16_t cylinders;
395     int8_t heads, sectors;
396     int val;
397     int i, trans;
398 
399     val = 0;
400     if (ide_get_geometry(arg->idebus[0], 0,
401                          &cylinders, &heads, &sectors) >= 0) {
402         cmos_init_hd(s, 0x19, 0x1b, cylinders, heads, sectors);
403         val |= 0xf0;
404     }
405     if (ide_get_geometry(arg->idebus[0], 1,
406                          &cylinders, &heads, &sectors) >= 0) {
407         cmos_init_hd(s, 0x1a, 0x24, cylinders, heads, sectors);
408         val |= 0x0f;
409     }
410     rtc_set_memory(s, 0x12, val);
411 
412     val = 0;
413     for (i = 0; i < 4; i++) {
414         /* NOTE: ide_get_geometry() returns the physical
415            geometry.  It is always such that: 1 <= sects <= 63, 1
416            <= heads <= 16, 1 <= cylinders <= 16383. The BIOS
417            geometry can be different if a translation is done. */
418         if (ide_get_geometry(arg->idebus[i / 2], i % 2,
419                              &cylinders, &heads, &sectors) >= 0) {
420             trans = ide_get_bios_chs_trans(arg->idebus[i / 2], i % 2) - 1;
421             assert((trans & ~3) == 0);
422             val |= trans << (i * 2);
423         }
424     }
425     rtc_set_memory(s, 0x39, val);
426 
427     pc_cmos_init_floppy(s, pc_find_fdc0());
428 
429     qemu_unregister_reset(pc_cmos_init_late, opaque);
430 }
431 
432 void pc_cmos_init(PCMachineState *pcms,
433                   BusState *idebus0, BusState *idebus1,
434                   ISADevice *s)
435 {
436     int val;
437     static pc_cmos_init_late_arg arg;
438 
439     /* various important CMOS locations needed by PC/Bochs bios */
440 
441     /* memory size */
442     /* base memory (first MiB) */
443     val = MIN(pcms->below_4g_mem_size / 1024, 640);
444     rtc_set_memory(s, 0x15, val);
445     rtc_set_memory(s, 0x16, val >> 8);
446     /* extended memory (next 64MiB) */
447     if (pcms->below_4g_mem_size > 1024 * 1024) {
448         val = (pcms->below_4g_mem_size - 1024 * 1024) / 1024;
449     } else {
450         val = 0;
451     }
452     if (val > 65535)
453         val = 65535;
454     rtc_set_memory(s, 0x17, val);
455     rtc_set_memory(s, 0x18, val >> 8);
456     rtc_set_memory(s, 0x30, val);
457     rtc_set_memory(s, 0x31, val >> 8);
458     /* memory between 16MiB and 4GiB */
459     if (pcms->below_4g_mem_size > 16 * 1024 * 1024) {
460         val = (pcms->below_4g_mem_size - 16 * 1024 * 1024) / 65536;
461     } else {
462         val = 0;
463     }
464     if (val > 65535)
465         val = 65535;
466     rtc_set_memory(s, 0x34, val);
467     rtc_set_memory(s, 0x35, val >> 8);
468     /* memory above 4GiB */
469     val = pcms->above_4g_mem_size / 65536;
470     rtc_set_memory(s, 0x5b, val);
471     rtc_set_memory(s, 0x5c, val >> 8);
472     rtc_set_memory(s, 0x5d, val >> 16);
473 
474     /* set the number of CPU */
475     rtc_set_memory(s, 0x5f, smp_cpus - 1);
476 
477     object_property_add_link(OBJECT(pcms), "rtc_state",
478                              TYPE_ISA_DEVICE,
479                              (Object **)&pcms->rtc,
480                              object_property_allow_set_link,
481                              OBJ_PROP_LINK_UNREF_ON_RELEASE, &error_abort);
482     object_property_set_link(OBJECT(pcms), OBJECT(s),
483                              "rtc_state", &error_abort);
484 
485     set_boot_dev(s, MACHINE(pcms)->boot_order, &error_fatal);
486 
487     val = 0;
488     val |= 0x02; /* FPU is there */
489     val |= 0x04; /* PS/2 mouse installed */
490     rtc_set_memory(s, REG_EQUIPMENT_BYTE, val);
491 
492     /* hard drives and FDC */
493     arg.rtc_state = s;
494     arg.idebus[0] = idebus0;
495     arg.idebus[1] = idebus1;
496     qemu_register_reset(pc_cmos_init_late, &arg);
497 }
498 
499 #define TYPE_PORT92 "port92"
500 #define PORT92(obj) OBJECT_CHECK(Port92State, (obj), TYPE_PORT92)
501 
502 /* port 92 stuff: could be split off */
503 typedef struct Port92State {
504     ISADevice parent_obj;
505 
506     MemoryRegion io;
507     uint8_t outport;
508     qemu_irq *a20_out;
509 } Port92State;
510 
511 static void port92_write(void *opaque, hwaddr addr, uint64_t val,
512                          unsigned size)
513 {
514     Port92State *s = opaque;
515     int oldval = s->outport;
516 
517     DPRINTF("port92: write 0x%02" PRIx64 "\n", val);
518     s->outport = val;
519     qemu_set_irq(*s->a20_out, (val >> 1) & 1);
520     if ((val & 1) && !(oldval & 1)) {
521         qemu_system_reset_request();
522     }
523 }
524 
525 static uint64_t port92_read(void *opaque, hwaddr addr,
526                             unsigned size)
527 {
528     Port92State *s = opaque;
529     uint32_t ret;
530 
531     ret = s->outport;
532     DPRINTF("port92: read 0x%02x\n", ret);
533     return ret;
534 }
535 
536 static void port92_init(ISADevice *dev, qemu_irq *a20_out)
537 {
538     Port92State *s = PORT92(dev);
539 
540     s->a20_out = a20_out;
541 }
542 
543 static const VMStateDescription vmstate_port92_isa = {
544     .name = "port92",
545     .version_id = 1,
546     .minimum_version_id = 1,
547     .fields = (VMStateField[]) {
548         VMSTATE_UINT8(outport, Port92State),
549         VMSTATE_END_OF_LIST()
550     }
551 };
552 
553 static void port92_reset(DeviceState *d)
554 {
555     Port92State *s = PORT92(d);
556 
557     s->outport &= ~1;
558 }
559 
560 static const MemoryRegionOps port92_ops = {
561     .read = port92_read,
562     .write = port92_write,
563     .impl = {
564         .min_access_size = 1,
565         .max_access_size = 1,
566     },
567     .endianness = DEVICE_LITTLE_ENDIAN,
568 };
569 
570 static void port92_initfn(Object *obj)
571 {
572     Port92State *s = PORT92(obj);
573 
574     memory_region_init_io(&s->io, OBJECT(s), &port92_ops, s, "port92", 1);
575 
576     s->outport = 0;
577 }
578 
579 static void port92_realizefn(DeviceState *dev, Error **errp)
580 {
581     ISADevice *isadev = ISA_DEVICE(dev);
582     Port92State *s = PORT92(dev);
583 
584     isa_register_ioport(isadev, &s->io, 0x92);
585 }
586 
587 static void port92_class_initfn(ObjectClass *klass, void *data)
588 {
589     DeviceClass *dc = DEVICE_CLASS(klass);
590 
591     dc->realize = port92_realizefn;
592     dc->reset = port92_reset;
593     dc->vmsd = &vmstate_port92_isa;
594     /*
595      * Reason: unlike ordinary ISA devices, this one needs additional
596      * wiring: its A20 output line needs to be wired up by
597      * port92_init().
598      */
599     dc->cannot_instantiate_with_device_add_yet = true;
600 }
601 
602 static const TypeInfo port92_info = {
603     .name          = TYPE_PORT92,
604     .parent        = TYPE_ISA_DEVICE,
605     .instance_size = sizeof(Port92State),
606     .instance_init = port92_initfn,
607     .class_init    = port92_class_initfn,
608 };
609 
610 static void port92_register_types(void)
611 {
612     type_register_static(&port92_info);
613 }
614 
615 type_init(port92_register_types)
616 
617 static void handle_a20_line_change(void *opaque, int irq, int level)
618 {
619     X86CPU *cpu = opaque;
620 
621     /* XXX: send to all CPUs ? */
622     /* XXX: add logic to handle multiple A20 line sources */
623     x86_cpu_set_a20(cpu, level);
624 }
625 
626 int e820_add_entry(uint64_t address, uint64_t length, uint32_t type)
627 {
628     int index = le32_to_cpu(e820_reserve.count);
629     struct e820_entry *entry;
630 
631     if (type != E820_RAM) {
632         /* old FW_CFG_E820_TABLE entry -- reservations only */
633         if (index >= E820_NR_ENTRIES) {
634             return -EBUSY;
635         }
636         entry = &e820_reserve.entry[index++];
637 
638         entry->address = cpu_to_le64(address);
639         entry->length = cpu_to_le64(length);
640         entry->type = cpu_to_le32(type);
641 
642         e820_reserve.count = cpu_to_le32(index);
643     }
644 
645     /* new "etc/e820" file -- include ram too */
646     e820_table = g_renew(struct e820_entry, e820_table, e820_entries + 1);
647     e820_table[e820_entries].address = cpu_to_le64(address);
648     e820_table[e820_entries].length = cpu_to_le64(length);
649     e820_table[e820_entries].type = cpu_to_le32(type);
650     e820_entries++;
651 
652     return e820_entries;
653 }
654 
655 int e820_get_num_entries(void)
656 {
657     return e820_entries;
658 }
659 
660 bool e820_get_entry(int idx, uint32_t type, uint64_t *address, uint64_t *length)
661 {
662     if (idx < e820_entries && e820_table[idx].type == cpu_to_le32(type)) {
663         *address = le64_to_cpu(e820_table[idx].address);
664         *length = le64_to_cpu(e820_table[idx].length);
665         return true;
666     }
667     return false;
668 }
669 
670 /* Enables contiguous-apic-ID mode, for compatibility */
671 static bool compat_apic_id_mode;
672 
673 void enable_compat_apic_id_mode(void)
674 {
675     compat_apic_id_mode = true;
676 }
677 
678 /* Calculates initial APIC ID for a specific CPU index
679  *
680  * Currently we need to be able to calculate the APIC ID from the CPU index
681  * alone (without requiring a CPU object), as the QEMU<->Seabios interfaces have
682  * no concept of "CPU index", and the NUMA tables on fw_cfg need the APIC ID of
683  * all CPUs up to max_cpus.
684  */
685 static uint32_t x86_cpu_apic_id_from_index(unsigned int cpu_index)
686 {
687     uint32_t correct_id;
688     static bool warned;
689 
690     correct_id = x86_apicid_from_cpu_idx(smp_cores, smp_threads, cpu_index);
691     if (compat_apic_id_mode) {
692         if (cpu_index != correct_id && !warned && !qtest_enabled()) {
693             error_report("APIC IDs set in compatibility mode, "
694                          "CPU topology won't match the configuration");
695             warned = true;
696         }
697         return cpu_index;
698     } else {
699         return correct_id;
700     }
701 }
702 
703 /* Calculates the limit to CPU APIC ID values
704  *
705  * This function returns the limit for the APIC ID value, so that all
706  * CPU APIC IDs are < pc_apic_id_limit().
707  *
708  * This is used for FW_CFG_MAX_CPUS. See comments on bochs_bios_init().
709  */
710 static unsigned int pc_apic_id_limit(unsigned int max_cpus)
711 {
712     return x86_cpu_apic_id_from_index(max_cpus - 1) + 1;
713 }
714 
715 static void pc_build_smbios(FWCfgState *fw_cfg)
716 {
717     uint8_t *smbios_tables, *smbios_anchor;
718     size_t smbios_tables_len, smbios_anchor_len;
719     struct smbios_phys_mem_area *mem_array;
720     unsigned i, array_count;
721 
722     smbios_tables = smbios_get_table_legacy(&smbios_tables_len);
723     if (smbios_tables) {
724         fw_cfg_add_bytes(fw_cfg, FW_CFG_SMBIOS_ENTRIES,
725                          smbios_tables, smbios_tables_len);
726     }
727 
728     /* build the array of physical mem area from e820 table */
729     mem_array = g_malloc0(sizeof(*mem_array) * e820_get_num_entries());
730     for (i = 0, array_count = 0; i < e820_get_num_entries(); i++) {
731         uint64_t addr, len;
732 
733         if (e820_get_entry(i, E820_RAM, &addr, &len)) {
734             mem_array[array_count].address = addr;
735             mem_array[array_count].length = len;
736             array_count++;
737         }
738     }
739     smbios_get_tables(mem_array, array_count,
740                       &smbios_tables, &smbios_tables_len,
741                       &smbios_anchor, &smbios_anchor_len);
742     g_free(mem_array);
743 
744     if (smbios_anchor) {
745         fw_cfg_add_file(fw_cfg, "etc/smbios/smbios-tables",
746                         smbios_tables, smbios_tables_len);
747         fw_cfg_add_file(fw_cfg, "etc/smbios/smbios-anchor",
748                         smbios_anchor, smbios_anchor_len);
749     }
750 }
751 
752 static FWCfgState *bochs_bios_init(AddressSpace *as)
753 {
754     FWCfgState *fw_cfg;
755     uint64_t *numa_fw_cfg;
756     int i, j;
757     unsigned int apic_id_limit = pc_apic_id_limit(max_cpus);
758 
759     fw_cfg = fw_cfg_init_io_dma(BIOS_CFG_IOPORT, BIOS_CFG_IOPORT + 4, as);
760 
761     /* FW_CFG_MAX_CPUS is a bit confusing/problematic on x86:
762      *
763      * SeaBIOS needs FW_CFG_MAX_CPUS for CPU hotplug, but the CPU hotplug
764      * QEMU<->SeaBIOS interface is not based on the "CPU index", but on the APIC
765      * ID of hotplugged CPUs[1]. This means that FW_CFG_MAX_CPUS is not the
766      * "maximum number of CPUs", but the "limit to the APIC ID values SeaBIOS
767      * may see".
768      *
769      * So, this means we must not use max_cpus, here, but the maximum possible
770      * APIC ID value, plus one.
771      *
772      * [1] The only kind of "CPU identifier" used between SeaBIOS and QEMU is
773      *     the APIC ID, not the "CPU index"
774      */
775     fw_cfg_add_i16(fw_cfg, FW_CFG_MAX_CPUS, (uint16_t)apic_id_limit);
776     fw_cfg_add_i64(fw_cfg, FW_CFG_RAM_SIZE, (uint64_t)ram_size);
777     fw_cfg_add_bytes(fw_cfg, FW_CFG_ACPI_TABLES,
778                      acpi_tables, acpi_tables_len);
779     fw_cfg_add_i32(fw_cfg, FW_CFG_IRQ0_OVERRIDE, kvm_allows_irq0_override());
780 
781     pc_build_smbios(fw_cfg);
782 
783     fw_cfg_add_bytes(fw_cfg, FW_CFG_E820_TABLE,
784                      &e820_reserve, sizeof(e820_reserve));
785     fw_cfg_add_file(fw_cfg, "etc/e820", e820_table,
786                     sizeof(struct e820_entry) * e820_entries);
787 
788     fw_cfg_add_bytes(fw_cfg, FW_CFG_HPET, &hpet_cfg, sizeof(hpet_cfg));
789     /* allocate memory for the NUMA channel: one (64bit) word for the number
790      * of nodes, one word for each VCPU->node and one word for each node to
791      * hold the amount of memory.
792      */
793     numa_fw_cfg = g_new0(uint64_t, 1 + apic_id_limit + nb_numa_nodes);
794     numa_fw_cfg[0] = cpu_to_le64(nb_numa_nodes);
795     for (i = 0; i < max_cpus; i++) {
796         unsigned int apic_id = x86_cpu_apic_id_from_index(i);
797         assert(apic_id < apic_id_limit);
798         for (j = 0; j < nb_numa_nodes; j++) {
799             if (test_bit(i, numa_info[j].node_cpu)) {
800                 numa_fw_cfg[apic_id + 1] = cpu_to_le64(j);
801                 break;
802             }
803         }
804     }
805     for (i = 0; i < nb_numa_nodes; i++) {
806         numa_fw_cfg[apic_id_limit + 1 + i] = cpu_to_le64(numa_info[i].node_mem);
807     }
808     fw_cfg_add_bytes(fw_cfg, FW_CFG_NUMA, numa_fw_cfg,
809                      (1 + apic_id_limit + nb_numa_nodes) *
810                      sizeof(*numa_fw_cfg));
811 
812     return fw_cfg;
813 }
814 
815 static long get_file_size(FILE *f)
816 {
817     long where, size;
818 
819     /* XXX: on Unix systems, using fstat() probably makes more sense */
820 
821     where = ftell(f);
822     fseek(f, 0, SEEK_END);
823     size = ftell(f);
824     fseek(f, where, SEEK_SET);
825 
826     return size;
827 }
828 
829 static void load_linux(PCMachineState *pcms,
830                        FWCfgState *fw_cfg)
831 {
832     uint16_t protocol;
833     int setup_size, kernel_size, initrd_size = 0, cmdline_size;
834     uint32_t initrd_max;
835     uint8_t header[8192], *setup, *kernel, *initrd_data;
836     hwaddr real_addr, prot_addr, cmdline_addr, initrd_addr = 0;
837     FILE *f;
838     char *vmode;
839     MachineState *machine = MACHINE(pcms);
840     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
841     const char *kernel_filename = machine->kernel_filename;
842     const char *initrd_filename = machine->initrd_filename;
843     const char *kernel_cmdline = machine->kernel_cmdline;
844 
845     /* Align to 16 bytes as a paranoia measure */
846     cmdline_size = (strlen(kernel_cmdline)+16) & ~15;
847 
848     /* load the kernel header */
849     f = fopen(kernel_filename, "rb");
850     if (!f || !(kernel_size = get_file_size(f)) ||
851         fread(header, 1, MIN(ARRAY_SIZE(header), kernel_size), f) !=
852         MIN(ARRAY_SIZE(header), kernel_size)) {
853         fprintf(stderr, "qemu: could not load kernel '%s': %s\n",
854                 kernel_filename, strerror(errno));
855         exit(1);
856     }
857 
858     /* kernel protocol version */
859 #if 0
860     fprintf(stderr, "header magic: %#x\n", ldl_p(header+0x202));
861 #endif
862     if (ldl_p(header+0x202) == 0x53726448) {
863         protocol = lduw_p(header+0x206);
864     } else {
865         /* This looks like a multiboot kernel. If it is, let's stop
866            treating it like a Linux kernel. */
867         if (load_multiboot(fw_cfg, f, kernel_filename, initrd_filename,
868                            kernel_cmdline, kernel_size, header)) {
869             return;
870         }
871         protocol = 0;
872     }
873 
874     if (protocol < 0x200 || !(header[0x211] & 0x01)) {
875         /* Low kernel */
876         real_addr    = 0x90000;
877         cmdline_addr = 0x9a000 - cmdline_size;
878         prot_addr    = 0x10000;
879     } else if (protocol < 0x202) {
880         /* High but ancient kernel */
881         real_addr    = 0x90000;
882         cmdline_addr = 0x9a000 - cmdline_size;
883         prot_addr    = 0x100000;
884     } else {
885         /* High and recent kernel */
886         real_addr    = 0x10000;
887         cmdline_addr = 0x20000;
888         prot_addr    = 0x100000;
889     }
890 
891 #if 0
892     fprintf(stderr,
893             "qemu: real_addr     = 0x" TARGET_FMT_plx "\n"
894             "qemu: cmdline_addr  = 0x" TARGET_FMT_plx "\n"
895             "qemu: prot_addr     = 0x" TARGET_FMT_plx "\n",
896             real_addr,
897             cmdline_addr,
898             prot_addr);
899 #endif
900 
901     /* highest address for loading the initrd */
902     if (protocol >= 0x203) {
903         initrd_max = ldl_p(header+0x22c);
904     } else {
905         initrd_max = 0x37ffffff;
906     }
907 
908     if (initrd_max >= pcms->below_4g_mem_size - pcmc->acpi_data_size) {
909         initrd_max = pcms->below_4g_mem_size - pcmc->acpi_data_size - 1;
910     }
911 
912     fw_cfg_add_i32(fw_cfg, FW_CFG_CMDLINE_ADDR, cmdline_addr);
913     fw_cfg_add_i32(fw_cfg, FW_CFG_CMDLINE_SIZE, strlen(kernel_cmdline)+1);
914     fw_cfg_add_string(fw_cfg, FW_CFG_CMDLINE_DATA, kernel_cmdline);
915 
916     if (protocol >= 0x202) {
917         stl_p(header+0x228, cmdline_addr);
918     } else {
919         stw_p(header+0x20, 0xA33F);
920         stw_p(header+0x22, cmdline_addr-real_addr);
921     }
922 
923     /* handle vga= parameter */
924     vmode = strstr(kernel_cmdline, "vga=");
925     if (vmode) {
926         unsigned int video_mode;
927         /* skip "vga=" */
928         vmode += 4;
929         if (!strncmp(vmode, "normal", 6)) {
930             video_mode = 0xffff;
931         } else if (!strncmp(vmode, "ext", 3)) {
932             video_mode = 0xfffe;
933         } else if (!strncmp(vmode, "ask", 3)) {
934             video_mode = 0xfffd;
935         } else {
936             video_mode = strtol(vmode, NULL, 0);
937         }
938         stw_p(header+0x1fa, video_mode);
939     }
940 
941     /* loader type */
942     /* High nybble = B reserved for QEMU; low nybble is revision number.
943        If this code is substantially changed, you may want to consider
944        incrementing the revision. */
945     if (protocol >= 0x200) {
946         header[0x210] = 0xB0;
947     }
948     /* heap */
949     if (protocol >= 0x201) {
950         header[0x211] |= 0x80;	/* CAN_USE_HEAP */
951         stw_p(header+0x224, cmdline_addr-real_addr-0x200);
952     }
953 
954     /* load initrd */
955     if (initrd_filename) {
956         if (protocol < 0x200) {
957             fprintf(stderr, "qemu: linux kernel too old to load a ram disk\n");
958             exit(1);
959         }
960 
961         initrd_size = get_image_size(initrd_filename);
962         if (initrd_size < 0) {
963             fprintf(stderr, "qemu: error reading initrd %s: %s\n",
964                     initrd_filename, strerror(errno));
965             exit(1);
966         }
967 
968         initrd_addr = (initrd_max-initrd_size) & ~4095;
969 
970         initrd_data = g_malloc(initrd_size);
971         load_image(initrd_filename, initrd_data);
972 
973         fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_ADDR, initrd_addr);
974         fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_SIZE, initrd_size);
975         fw_cfg_add_bytes(fw_cfg, FW_CFG_INITRD_DATA, initrd_data, initrd_size);
976 
977         stl_p(header+0x218, initrd_addr);
978         stl_p(header+0x21c, initrd_size);
979     }
980 
981     /* load kernel and setup */
982     setup_size = header[0x1f1];
983     if (setup_size == 0) {
984         setup_size = 4;
985     }
986     setup_size = (setup_size+1)*512;
987     if (setup_size > kernel_size) {
988         fprintf(stderr, "qemu: invalid kernel header\n");
989         exit(1);
990     }
991     kernel_size -= setup_size;
992 
993     setup  = g_malloc(setup_size);
994     kernel = g_malloc(kernel_size);
995     fseek(f, 0, SEEK_SET);
996     if (fread(setup, 1, setup_size, f) != setup_size) {
997         fprintf(stderr, "fread() failed\n");
998         exit(1);
999     }
1000     if (fread(kernel, 1, kernel_size, f) != kernel_size) {
1001         fprintf(stderr, "fread() failed\n");
1002         exit(1);
1003     }
1004     fclose(f);
1005     memcpy(setup, header, MIN(sizeof(header), setup_size));
1006 
1007     fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_ADDR, prot_addr);
1008     fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_SIZE, kernel_size);
1009     fw_cfg_add_bytes(fw_cfg, FW_CFG_KERNEL_DATA, kernel, kernel_size);
1010 
1011     fw_cfg_add_i32(fw_cfg, FW_CFG_SETUP_ADDR, real_addr);
1012     fw_cfg_add_i32(fw_cfg, FW_CFG_SETUP_SIZE, setup_size);
1013     fw_cfg_add_bytes(fw_cfg, FW_CFG_SETUP_DATA, setup, setup_size);
1014 
1015     option_rom[nb_option_roms].name = "linuxboot.bin";
1016     option_rom[nb_option_roms].bootindex = 0;
1017     nb_option_roms++;
1018 }
1019 
1020 #define NE2000_NB_MAX 6
1021 
1022 static const int ne2000_io[NE2000_NB_MAX] = { 0x300, 0x320, 0x340, 0x360,
1023                                               0x280, 0x380 };
1024 static const int ne2000_irq[NE2000_NB_MAX] = { 9, 10, 11, 3, 4, 5 };
1025 
1026 void pc_init_ne2k_isa(ISABus *bus, NICInfo *nd)
1027 {
1028     static int nb_ne2k = 0;
1029 
1030     if (nb_ne2k == NE2000_NB_MAX)
1031         return;
1032     isa_ne2000_init(bus, ne2000_io[nb_ne2k],
1033                     ne2000_irq[nb_ne2k], nd);
1034     nb_ne2k++;
1035 }
1036 
1037 DeviceState *cpu_get_current_apic(void)
1038 {
1039     if (current_cpu) {
1040         X86CPU *cpu = X86_CPU(current_cpu);
1041         return cpu->apic_state;
1042     } else {
1043         return NULL;
1044     }
1045 }
1046 
1047 void pc_acpi_smi_interrupt(void *opaque, int irq, int level)
1048 {
1049     X86CPU *cpu = opaque;
1050 
1051     if (level) {
1052         cpu_interrupt(CPU(cpu), CPU_INTERRUPT_SMI);
1053     }
1054 }
1055 
1056 static X86CPU *pc_new_cpu(const char *cpu_model, int64_t apic_id,
1057                           Error **errp)
1058 {
1059     X86CPU *cpu = NULL;
1060     Error *local_err = NULL;
1061 
1062     cpu = cpu_x86_create(cpu_model, &local_err);
1063     if (local_err != NULL) {
1064         goto out;
1065     }
1066 
1067     object_property_set_int(OBJECT(cpu), apic_id, "apic-id", &local_err);
1068     object_property_set_bool(OBJECT(cpu), true, "realized", &local_err);
1069 
1070 out:
1071     if (local_err) {
1072         error_propagate(errp, local_err);
1073         object_unref(OBJECT(cpu));
1074         cpu = NULL;
1075     }
1076     return cpu;
1077 }
1078 
1079 void pc_hot_add_cpu(const int64_t id, Error **errp)
1080 {
1081     X86CPU *cpu;
1082     MachineState *machine = MACHINE(qdev_get_machine());
1083     int64_t apic_id = x86_cpu_apic_id_from_index(id);
1084     Error *local_err = NULL;
1085 
1086     if (id < 0) {
1087         error_setg(errp, "Invalid CPU id: %" PRIi64, id);
1088         return;
1089     }
1090 
1091     if (cpu_exists(apic_id)) {
1092         error_setg(errp, "Unable to add CPU: %" PRIi64
1093                    ", it already exists", id);
1094         return;
1095     }
1096 
1097     if (id >= max_cpus) {
1098         error_setg(errp, "Unable to add CPU: %" PRIi64
1099                    ", max allowed: %d", id, max_cpus - 1);
1100         return;
1101     }
1102 
1103     if (apic_id >= ACPI_CPU_HOTPLUG_ID_LIMIT) {
1104         error_setg(errp, "Unable to add CPU: %" PRIi64
1105                    ", resulting APIC ID (%" PRIi64 ") is too large",
1106                    id, apic_id);
1107         return;
1108     }
1109 
1110     cpu = pc_new_cpu(machine->cpu_model, apic_id, &local_err);
1111     if (local_err) {
1112         error_propagate(errp, local_err);
1113         return;
1114     }
1115     object_unref(OBJECT(cpu));
1116 }
1117 
1118 void pc_cpus_init(PCMachineState *pcms)
1119 {
1120     int i;
1121     X86CPU *cpu = NULL;
1122     MachineState *machine = MACHINE(pcms);
1123     unsigned long apic_id_limit;
1124 
1125     /* init CPUs */
1126     if (machine->cpu_model == NULL) {
1127 #ifdef TARGET_X86_64
1128         machine->cpu_model = "qemu64";
1129 #else
1130         machine->cpu_model = "qemu32";
1131 #endif
1132     }
1133 
1134     apic_id_limit = pc_apic_id_limit(max_cpus);
1135     if (apic_id_limit > ACPI_CPU_HOTPLUG_ID_LIMIT) {
1136         error_report("max_cpus is too large. APIC ID of last CPU is %lu",
1137                      apic_id_limit - 1);
1138         exit(1);
1139     }
1140 
1141     for (i = 0; i < smp_cpus; i++) {
1142         cpu = pc_new_cpu(machine->cpu_model, x86_cpu_apic_id_from_index(i),
1143                          &error_fatal);
1144         object_unref(OBJECT(cpu));
1145     }
1146 
1147     /* tell smbios about cpuid version and features */
1148     smbios_set_cpuid(cpu->env.cpuid_version, cpu->env.features[FEAT_1_EDX]);
1149 }
1150 
1151 /* pci-info ROM file. Little endian format */
1152 typedef struct PcRomPciInfo {
1153     uint64_t w32_min;
1154     uint64_t w32_max;
1155     uint64_t w64_min;
1156     uint64_t w64_max;
1157 } PcRomPciInfo;
1158 
1159 typedef struct PcGuestInfoState {
1160     PcGuestInfo info;
1161     Notifier machine_done;
1162 } PcGuestInfoState;
1163 
1164 static
1165 void pc_guest_info_machine_done(Notifier *notifier, void *data)
1166 {
1167     PcGuestInfoState *guest_info_state = container_of(notifier,
1168                                                       PcGuestInfoState,
1169                                                       machine_done);
1170     PCIBus *bus = PC_MACHINE(qdev_get_machine())->bus;
1171 
1172     if (bus) {
1173         int extra_hosts = 0;
1174 
1175         QLIST_FOREACH(bus, &bus->child, sibling) {
1176             /* look for expander root buses */
1177             if (pci_bus_is_root(bus)) {
1178                 extra_hosts++;
1179             }
1180         }
1181         if (extra_hosts && guest_info_state->info.fw_cfg) {
1182             uint64_t *val = g_malloc(sizeof(*val));
1183             *val = cpu_to_le64(extra_hosts);
1184             fw_cfg_add_file(guest_info_state->info.fw_cfg,
1185                     "etc/extra-pci-roots", val, sizeof(*val));
1186         }
1187     }
1188 
1189     acpi_setup(&guest_info_state->info);
1190 }
1191 
1192 PcGuestInfo *pc_guest_info_init(PCMachineState *pcms)
1193 {
1194     PcGuestInfoState *guest_info_state = g_malloc0(sizeof *guest_info_state);
1195     PcGuestInfo *guest_info = &guest_info_state->info;
1196     int i, j;
1197 
1198     guest_info->ram_size_below_4g = pcms->below_4g_mem_size;
1199     guest_info->ram_size = pcms->below_4g_mem_size + pcms->above_4g_mem_size;
1200     guest_info->apic_id_limit = pc_apic_id_limit(max_cpus);
1201     guest_info->apic_xrupt_override = kvm_allows_irq0_override();
1202     guest_info->numa_nodes = nb_numa_nodes;
1203     guest_info->node_mem = g_malloc0(guest_info->numa_nodes *
1204                                     sizeof *guest_info->node_mem);
1205     for (i = 0; i < nb_numa_nodes; i++) {
1206         guest_info->node_mem[i] = numa_info[i].node_mem;
1207     }
1208 
1209     guest_info->node_cpu = g_malloc0(guest_info->apic_id_limit *
1210                                      sizeof *guest_info->node_cpu);
1211 
1212     for (i = 0; i < max_cpus; i++) {
1213         unsigned int apic_id = x86_cpu_apic_id_from_index(i);
1214         assert(apic_id < guest_info->apic_id_limit);
1215         for (j = 0; j < nb_numa_nodes; j++) {
1216             if (test_bit(i, numa_info[j].node_cpu)) {
1217                 guest_info->node_cpu[apic_id] = j;
1218                 break;
1219             }
1220         }
1221     }
1222 
1223     guest_info_state->machine_done.notify = pc_guest_info_machine_done;
1224     qemu_add_machine_init_done_notifier(&guest_info_state->machine_done);
1225     return guest_info;
1226 }
1227 
1228 /* setup pci memory address space mapping into system address space */
1229 void pc_pci_as_mapping_init(Object *owner, MemoryRegion *system_memory,
1230                             MemoryRegion *pci_address_space)
1231 {
1232     /* Set to lower priority than RAM */
1233     memory_region_add_subregion_overlap(system_memory, 0x0,
1234                                         pci_address_space, -1);
1235 }
1236 
1237 void pc_acpi_init(const char *default_dsdt)
1238 {
1239     char *filename;
1240 
1241     if (acpi_tables != NULL) {
1242         /* manually set via -acpitable, leave it alone */
1243         return;
1244     }
1245 
1246     filename = qemu_find_file(QEMU_FILE_TYPE_BIOS, default_dsdt);
1247     if (filename == NULL) {
1248         fprintf(stderr, "WARNING: failed to find %s\n", default_dsdt);
1249     } else {
1250         QemuOpts *opts = qemu_opts_create(qemu_find_opts("acpi"), NULL, 0,
1251                                           &error_abort);
1252         Error *err = NULL;
1253 
1254         qemu_opt_set(opts, "file", filename, &error_abort);
1255 
1256         acpi_table_add_builtin(opts, &err);
1257         if (err) {
1258             error_reportf_err(err, "WARNING: failed to load %s: ",
1259                               filename);
1260         }
1261         g_free(filename);
1262     }
1263 }
1264 
1265 FWCfgState *xen_load_linux(PCMachineState *pcms,
1266                            PcGuestInfo *guest_info)
1267 {
1268     int i;
1269     FWCfgState *fw_cfg;
1270 
1271     assert(MACHINE(pcms)->kernel_filename != NULL);
1272 
1273     fw_cfg = fw_cfg_init_io(BIOS_CFG_IOPORT);
1274     rom_set_fw(fw_cfg);
1275 
1276     load_linux(pcms, fw_cfg);
1277     for (i = 0; i < nb_option_roms; i++) {
1278         assert(!strcmp(option_rom[i].name, "linuxboot.bin") ||
1279                !strcmp(option_rom[i].name, "multiboot.bin"));
1280         rom_add_option(option_rom[i].name, option_rom[i].bootindex);
1281     }
1282     guest_info->fw_cfg = fw_cfg;
1283     return fw_cfg;
1284 }
1285 
1286 FWCfgState *pc_memory_init(PCMachineState *pcms,
1287                            MemoryRegion *system_memory,
1288                            MemoryRegion *rom_memory,
1289                            MemoryRegion **ram_memory,
1290                            PcGuestInfo *guest_info)
1291 {
1292     int linux_boot, i;
1293     MemoryRegion *ram, *option_rom_mr;
1294     MemoryRegion *ram_below_4g, *ram_above_4g;
1295     FWCfgState *fw_cfg;
1296     MachineState *machine = MACHINE(pcms);
1297     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
1298 
1299     assert(machine->ram_size == pcms->below_4g_mem_size +
1300                                 pcms->above_4g_mem_size);
1301 
1302     linux_boot = (machine->kernel_filename != NULL);
1303 
1304     /* Allocate RAM.  We allocate it as a single memory region and use
1305      * aliases to address portions of it, mostly for backwards compatibility
1306      * with older qemus that used qemu_ram_alloc().
1307      */
1308     ram = g_malloc(sizeof(*ram));
1309     memory_region_allocate_system_memory(ram, NULL, "pc.ram",
1310                                          machine->ram_size);
1311     *ram_memory = ram;
1312     ram_below_4g = g_malloc(sizeof(*ram_below_4g));
1313     memory_region_init_alias(ram_below_4g, NULL, "ram-below-4g", ram,
1314                              0, pcms->below_4g_mem_size);
1315     memory_region_add_subregion(system_memory, 0, ram_below_4g);
1316     e820_add_entry(0, pcms->below_4g_mem_size, E820_RAM);
1317     if (pcms->above_4g_mem_size > 0) {
1318         ram_above_4g = g_malloc(sizeof(*ram_above_4g));
1319         memory_region_init_alias(ram_above_4g, NULL, "ram-above-4g", ram,
1320                                  pcms->below_4g_mem_size,
1321                                  pcms->above_4g_mem_size);
1322         memory_region_add_subregion(system_memory, 0x100000000ULL,
1323                                     ram_above_4g);
1324         e820_add_entry(0x100000000ULL, pcms->above_4g_mem_size, E820_RAM);
1325     }
1326 
1327     if (!guest_info->has_reserved_memory &&
1328         (machine->ram_slots ||
1329          (machine->maxram_size > machine->ram_size))) {
1330         MachineClass *mc = MACHINE_GET_CLASS(machine);
1331 
1332         error_report("\"-memory 'slots|maxmem'\" is not supported by: %s",
1333                      mc->name);
1334         exit(EXIT_FAILURE);
1335     }
1336 
1337     /* initialize hotplug memory address space */
1338     if (guest_info->has_reserved_memory &&
1339         (machine->ram_size < machine->maxram_size)) {
1340         ram_addr_t hotplug_mem_size =
1341             machine->maxram_size - machine->ram_size;
1342 
1343         if (machine->ram_slots > ACPI_MAX_RAM_SLOTS) {
1344             error_report("unsupported amount of memory slots: %"PRIu64,
1345                          machine->ram_slots);
1346             exit(EXIT_FAILURE);
1347         }
1348 
1349         if (QEMU_ALIGN_UP(machine->maxram_size,
1350                           TARGET_PAGE_SIZE) != machine->maxram_size) {
1351             error_report("maximum memory size must by aligned to multiple of "
1352                          "%d bytes", TARGET_PAGE_SIZE);
1353             exit(EXIT_FAILURE);
1354         }
1355 
1356         pcms->hotplug_memory.base =
1357             ROUND_UP(0x100000000ULL + pcms->above_4g_mem_size, 1ULL << 30);
1358 
1359         if (pcmc->enforce_aligned_dimm) {
1360             /* size hotplug region assuming 1G page max alignment per slot */
1361             hotplug_mem_size += (1ULL << 30) * machine->ram_slots;
1362         }
1363 
1364         if ((pcms->hotplug_memory.base + hotplug_mem_size) <
1365             hotplug_mem_size) {
1366             error_report("unsupported amount of maximum memory: " RAM_ADDR_FMT,
1367                          machine->maxram_size);
1368             exit(EXIT_FAILURE);
1369         }
1370 
1371         memory_region_init(&pcms->hotplug_memory.mr, OBJECT(pcms),
1372                            "hotplug-memory", hotplug_mem_size);
1373         memory_region_add_subregion(system_memory, pcms->hotplug_memory.base,
1374                                     &pcms->hotplug_memory.mr);
1375     }
1376 
1377     /* Initialize PC system firmware */
1378     pc_system_firmware_init(rom_memory, guest_info->isapc_ram_fw);
1379 
1380     option_rom_mr = g_malloc(sizeof(*option_rom_mr));
1381     memory_region_init_ram(option_rom_mr, NULL, "pc.rom", PC_ROM_SIZE,
1382                            &error_fatal);
1383     vmstate_register_ram_global(option_rom_mr);
1384     memory_region_add_subregion_overlap(rom_memory,
1385                                         PC_ROM_MIN_VGA,
1386                                         option_rom_mr,
1387                                         1);
1388 
1389     fw_cfg = bochs_bios_init(&address_space_memory);
1390 
1391     rom_set_fw(fw_cfg);
1392 
1393     if (guest_info->has_reserved_memory && pcms->hotplug_memory.base) {
1394         uint64_t *val = g_malloc(sizeof(*val));
1395         PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
1396         uint64_t res_mem_end = pcms->hotplug_memory.base;
1397 
1398         if (!pcmc->broken_reserved_end) {
1399             res_mem_end += memory_region_size(&pcms->hotplug_memory.mr);
1400         }
1401         *val = cpu_to_le64(ROUND_UP(res_mem_end, 0x1ULL << 30));
1402         fw_cfg_add_file(fw_cfg, "etc/reserved-memory-end", val, sizeof(*val));
1403     }
1404 
1405     if (linux_boot) {
1406         load_linux(pcms, fw_cfg);
1407     }
1408 
1409     for (i = 0; i < nb_option_roms; i++) {
1410         rom_add_option(option_rom[i].name, option_rom[i].bootindex);
1411     }
1412     guest_info->fw_cfg = fw_cfg;
1413     return fw_cfg;
1414 }
1415 
1416 qemu_irq pc_allocate_cpu_irq(void)
1417 {
1418     return qemu_allocate_irq(pic_irq_request, NULL, 0);
1419 }
1420 
1421 DeviceState *pc_vga_init(ISABus *isa_bus, PCIBus *pci_bus)
1422 {
1423     DeviceState *dev = NULL;
1424 
1425     if (pci_bus) {
1426         PCIDevice *pcidev = pci_vga_init(pci_bus);
1427         dev = pcidev ? &pcidev->qdev : NULL;
1428     } else if (isa_bus) {
1429         ISADevice *isadev = isa_vga_init(isa_bus);
1430         dev = isadev ? DEVICE(isadev) : NULL;
1431     }
1432     return dev;
1433 }
1434 
1435 static const MemoryRegionOps ioport80_io_ops = {
1436     .write = ioport80_write,
1437     .read = ioport80_read,
1438     .endianness = DEVICE_NATIVE_ENDIAN,
1439     .impl = {
1440         .min_access_size = 1,
1441         .max_access_size = 1,
1442     },
1443 };
1444 
1445 static const MemoryRegionOps ioportF0_io_ops = {
1446     .write = ioportF0_write,
1447     .read = ioportF0_read,
1448     .endianness = DEVICE_NATIVE_ENDIAN,
1449     .impl = {
1450         .min_access_size = 1,
1451         .max_access_size = 1,
1452     },
1453 };
1454 
1455 void pc_basic_device_init(ISABus *isa_bus, qemu_irq *gsi,
1456                           ISADevice **rtc_state,
1457                           bool create_fdctrl,
1458                           bool no_vmport,
1459                           uint32_t hpet_irqs)
1460 {
1461     int i;
1462     DriveInfo *fd[MAX_FD];
1463     DeviceState *hpet = NULL;
1464     int pit_isa_irq = 0;
1465     qemu_irq pit_alt_irq = NULL;
1466     qemu_irq rtc_irq = NULL;
1467     qemu_irq *a20_line;
1468     ISADevice *i8042, *port92, *vmmouse, *pit = NULL;
1469     MemoryRegion *ioport80_io = g_new(MemoryRegion, 1);
1470     MemoryRegion *ioportF0_io = g_new(MemoryRegion, 1);
1471 
1472     memory_region_init_io(ioport80_io, NULL, &ioport80_io_ops, NULL, "ioport80", 1);
1473     memory_region_add_subregion(isa_bus->address_space_io, 0x80, ioport80_io);
1474 
1475     memory_region_init_io(ioportF0_io, NULL, &ioportF0_io_ops, NULL, "ioportF0", 1);
1476     memory_region_add_subregion(isa_bus->address_space_io, 0xf0, ioportF0_io);
1477 
1478     /*
1479      * Check if an HPET shall be created.
1480      *
1481      * Without KVM_CAP_PIT_STATE2, we cannot switch off the in-kernel PIT
1482      * when the HPET wants to take over. Thus we have to disable the latter.
1483      */
1484     if (!no_hpet && (!kvm_irqchip_in_kernel() || kvm_has_pit_state2())) {
1485         /* In order to set property, here not using sysbus_try_create_simple */
1486         hpet = qdev_try_create(NULL, TYPE_HPET);
1487         if (hpet) {
1488             /* For pc-piix-*, hpet's intcap is always IRQ2. For pc-q35-1.7
1489              * and earlier, use IRQ2 for compat. Otherwise, use IRQ16~23,
1490              * IRQ8 and IRQ2.
1491              */
1492             uint8_t compat = object_property_get_int(OBJECT(hpet),
1493                     HPET_INTCAP, NULL);
1494             if (!compat) {
1495                 qdev_prop_set_uint32(hpet, HPET_INTCAP, hpet_irqs);
1496             }
1497             qdev_init_nofail(hpet);
1498             sysbus_mmio_map(SYS_BUS_DEVICE(hpet), 0, HPET_BASE);
1499 
1500             for (i = 0; i < GSI_NUM_PINS; i++) {
1501                 sysbus_connect_irq(SYS_BUS_DEVICE(hpet), i, gsi[i]);
1502             }
1503             pit_isa_irq = -1;
1504             pit_alt_irq = qdev_get_gpio_in(hpet, HPET_LEGACY_PIT_INT);
1505             rtc_irq = qdev_get_gpio_in(hpet, HPET_LEGACY_RTC_INT);
1506         }
1507     }
1508     *rtc_state = rtc_init(isa_bus, 2000, rtc_irq);
1509 
1510     qemu_register_boot_set(pc_boot_set, *rtc_state);
1511 
1512     if (!xen_enabled()) {
1513         if (kvm_pit_in_kernel()) {
1514             pit = kvm_pit_init(isa_bus, 0x40);
1515         } else {
1516             pit = pit_init(isa_bus, 0x40, pit_isa_irq, pit_alt_irq);
1517         }
1518         if (hpet) {
1519             /* connect PIT to output control line of the HPET */
1520             qdev_connect_gpio_out(hpet, 0, qdev_get_gpio_in(DEVICE(pit), 0));
1521         }
1522         pcspk_init(isa_bus, pit);
1523     }
1524 
1525     serial_hds_isa_init(isa_bus, MAX_SERIAL_PORTS);
1526     parallel_hds_isa_init(isa_bus, MAX_PARALLEL_PORTS);
1527 
1528     a20_line = qemu_allocate_irqs(handle_a20_line_change, first_cpu, 2);
1529     i8042 = isa_create_simple(isa_bus, "i8042");
1530     i8042_setup_a20_line(i8042, &a20_line[0]);
1531     if (!no_vmport) {
1532         vmport_init(isa_bus);
1533         vmmouse = isa_try_create(isa_bus, "vmmouse");
1534     } else {
1535         vmmouse = NULL;
1536     }
1537     if (vmmouse) {
1538         DeviceState *dev = DEVICE(vmmouse);
1539         qdev_prop_set_ptr(dev, "ps2_mouse", i8042);
1540         qdev_init_nofail(dev);
1541     }
1542     port92 = isa_create_simple(isa_bus, "port92");
1543     port92_init(port92, &a20_line[1]);
1544 
1545     DMA_init(0);
1546 
1547     for(i = 0; i < MAX_FD; i++) {
1548         fd[i] = drive_get(IF_FLOPPY, 0, i);
1549         create_fdctrl |= !!fd[i];
1550     }
1551     if (create_fdctrl) {
1552         fdctrl_init_isa(isa_bus, fd);
1553     }
1554 }
1555 
1556 void pc_nic_init(ISABus *isa_bus, PCIBus *pci_bus)
1557 {
1558     int i;
1559 
1560     for (i = 0; i < nb_nics; i++) {
1561         NICInfo *nd = &nd_table[i];
1562 
1563         if (!pci_bus || (nd->model && strcmp(nd->model, "ne2k_isa") == 0)) {
1564             pc_init_ne2k_isa(isa_bus, nd);
1565         } else {
1566             pci_nic_init_nofail(nd, pci_bus, "e1000", NULL);
1567         }
1568     }
1569 }
1570 
1571 void pc_pci_device_init(PCIBus *pci_bus)
1572 {
1573     int max_bus;
1574     int bus;
1575 
1576     max_bus = drive_get_max_bus(IF_SCSI);
1577     for (bus = 0; bus <= max_bus; bus++) {
1578         pci_create_simple(pci_bus, -1, "lsi53c895a");
1579     }
1580 }
1581 
1582 void ioapic_init_gsi(GSIState *gsi_state, const char *parent_name)
1583 {
1584     DeviceState *dev;
1585     SysBusDevice *d;
1586     unsigned int i;
1587 
1588     if (kvm_ioapic_in_kernel()) {
1589         dev = qdev_create(NULL, "kvm-ioapic");
1590     } else {
1591         dev = qdev_create(NULL, "ioapic");
1592     }
1593     if (parent_name) {
1594         object_property_add_child(object_resolve_path(parent_name, NULL),
1595                                   "ioapic", OBJECT(dev), NULL);
1596     }
1597     qdev_init_nofail(dev);
1598     d = SYS_BUS_DEVICE(dev);
1599     sysbus_mmio_map(d, 0, IO_APIC_DEFAULT_ADDRESS);
1600 
1601     for (i = 0; i < IOAPIC_NUM_PINS; i++) {
1602         gsi_state->ioapic_irq[i] = qdev_get_gpio_in(dev, i);
1603     }
1604 }
1605 
1606 static void pc_dimm_plug(HotplugHandler *hotplug_dev,
1607                          DeviceState *dev, Error **errp)
1608 {
1609     HotplugHandlerClass *hhc;
1610     Error *local_err = NULL;
1611     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1612     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
1613     PCDIMMDevice *dimm = PC_DIMM(dev);
1614     PCDIMMDeviceClass *ddc = PC_DIMM_GET_CLASS(dimm);
1615     MemoryRegion *mr = ddc->get_memory_region(dimm);
1616     uint64_t align = TARGET_PAGE_SIZE;
1617 
1618     if (memory_region_get_alignment(mr) && pcmc->enforce_aligned_dimm) {
1619         align = memory_region_get_alignment(mr);
1620     }
1621 
1622     if (!pcms->acpi_dev) {
1623         error_setg(&local_err,
1624                    "memory hotplug is not enabled: missing acpi device");
1625         goto out;
1626     }
1627 
1628     pc_dimm_memory_plug(dev, &pcms->hotplug_memory, mr, align, &local_err);
1629     if (local_err) {
1630         goto out;
1631     }
1632 
1633     hhc = HOTPLUG_HANDLER_GET_CLASS(pcms->acpi_dev);
1634     hhc->plug(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &error_abort);
1635 out:
1636     error_propagate(errp, local_err);
1637 }
1638 
1639 static void pc_dimm_unplug_request(HotplugHandler *hotplug_dev,
1640                                    DeviceState *dev, Error **errp)
1641 {
1642     HotplugHandlerClass *hhc;
1643     Error *local_err = NULL;
1644     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1645 
1646     if (!pcms->acpi_dev) {
1647         error_setg(&local_err,
1648                    "memory hotplug is not enabled: missing acpi device");
1649         goto out;
1650     }
1651 
1652     hhc = HOTPLUG_HANDLER_GET_CLASS(pcms->acpi_dev);
1653     hhc->unplug_request(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &local_err);
1654 
1655 out:
1656     error_propagate(errp, local_err);
1657 }
1658 
1659 static void pc_dimm_unplug(HotplugHandler *hotplug_dev,
1660                            DeviceState *dev, Error **errp)
1661 {
1662     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1663     PCDIMMDevice *dimm = PC_DIMM(dev);
1664     PCDIMMDeviceClass *ddc = PC_DIMM_GET_CLASS(dimm);
1665     MemoryRegion *mr = ddc->get_memory_region(dimm);
1666     HotplugHandlerClass *hhc;
1667     Error *local_err = NULL;
1668 
1669     hhc = HOTPLUG_HANDLER_GET_CLASS(pcms->acpi_dev);
1670     hhc->unplug(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &local_err);
1671 
1672     if (local_err) {
1673         goto out;
1674     }
1675 
1676     pc_dimm_memory_unplug(dev, &pcms->hotplug_memory, mr);
1677     object_unparent(OBJECT(dev));
1678 
1679  out:
1680     error_propagate(errp, local_err);
1681 }
1682 
1683 static void pc_cpu_plug(HotplugHandler *hotplug_dev,
1684                         DeviceState *dev, Error **errp)
1685 {
1686     HotplugHandlerClass *hhc;
1687     Error *local_err = NULL;
1688     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1689 
1690     if (!dev->hotplugged) {
1691         goto out;
1692     }
1693 
1694     if (!pcms->acpi_dev) {
1695         error_setg(&local_err,
1696                    "cpu hotplug is not enabled: missing acpi device");
1697         goto out;
1698     }
1699 
1700     hhc = HOTPLUG_HANDLER_GET_CLASS(pcms->acpi_dev);
1701     hhc->plug(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &local_err);
1702     if (local_err) {
1703         goto out;
1704     }
1705 
1706     /* increment the number of CPUs */
1707     rtc_set_memory(pcms->rtc, 0x5f, rtc_get_memory(pcms->rtc, 0x5f) + 1);
1708 out:
1709     error_propagate(errp, local_err);
1710 }
1711 
1712 static void pc_machine_device_plug_cb(HotplugHandler *hotplug_dev,
1713                                       DeviceState *dev, Error **errp)
1714 {
1715     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
1716         pc_dimm_plug(hotplug_dev, dev, errp);
1717     } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
1718         pc_cpu_plug(hotplug_dev, dev, errp);
1719     }
1720 }
1721 
1722 static void pc_machine_device_unplug_request_cb(HotplugHandler *hotplug_dev,
1723                                                 DeviceState *dev, Error **errp)
1724 {
1725     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
1726         pc_dimm_unplug_request(hotplug_dev, dev, errp);
1727     } else {
1728         error_setg(errp, "acpi: device unplug request for not supported device"
1729                    " type: %s", object_get_typename(OBJECT(dev)));
1730     }
1731 }
1732 
1733 static void pc_machine_device_unplug_cb(HotplugHandler *hotplug_dev,
1734                                         DeviceState *dev, Error **errp)
1735 {
1736     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
1737         pc_dimm_unplug(hotplug_dev, dev, errp);
1738     } else {
1739         error_setg(errp, "acpi: device unplug for not supported device"
1740                    " type: %s", object_get_typename(OBJECT(dev)));
1741     }
1742 }
1743 
1744 static HotplugHandler *pc_get_hotpug_handler(MachineState *machine,
1745                                              DeviceState *dev)
1746 {
1747     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(machine);
1748 
1749     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM) ||
1750         object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
1751         return HOTPLUG_HANDLER(machine);
1752     }
1753 
1754     return pcmc->get_hotplug_handler ?
1755         pcmc->get_hotplug_handler(machine, dev) : NULL;
1756 }
1757 
1758 static void
1759 pc_machine_get_hotplug_memory_region_size(Object *obj, Visitor *v, void *opaque,
1760                                           const char *name, Error **errp)
1761 {
1762     PCMachineState *pcms = PC_MACHINE(obj);
1763     int64_t value = memory_region_size(&pcms->hotplug_memory.mr);
1764 
1765     visit_type_int(v, &value, name, errp);
1766 }
1767 
1768 static void pc_machine_get_max_ram_below_4g(Object *obj, Visitor *v,
1769                                          void *opaque, const char *name,
1770                                          Error **errp)
1771 {
1772     PCMachineState *pcms = PC_MACHINE(obj);
1773     uint64_t value = pcms->max_ram_below_4g;
1774 
1775     visit_type_size(v, &value, name, errp);
1776 }
1777 
1778 static void pc_machine_set_max_ram_below_4g(Object *obj, Visitor *v,
1779                                          void *opaque, const char *name,
1780                                          Error **errp)
1781 {
1782     PCMachineState *pcms = PC_MACHINE(obj);
1783     Error *error = NULL;
1784     uint64_t value;
1785 
1786     visit_type_size(v, &value, name, &error);
1787     if (error) {
1788         error_propagate(errp, error);
1789         return;
1790     }
1791     if (value > (1ULL << 32)) {
1792         error_setg(&error,
1793                    "Machine option 'max-ram-below-4g=%"PRIu64
1794                    "' expects size less than or equal to 4G", value);
1795         error_propagate(errp, error);
1796         return;
1797     }
1798 
1799     if (value < (1ULL << 20)) {
1800         error_report("Warning: small max_ram_below_4g(%"PRIu64
1801                      ") less than 1M.  BIOS may not work..",
1802                      value);
1803     }
1804 
1805     pcms->max_ram_below_4g = value;
1806 }
1807 
1808 static void pc_machine_get_vmport(Object *obj, Visitor *v, void *opaque,
1809                                   const char *name, Error **errp)
1810 {
1811     PCMachineState *pcms = PC_MACHINE(obj);
1812     OnOffAuto vmport = pcms->vmport;
1813 
1814     visit_type_OnOffAuto(v, &vmport, name, errp);
1815 }
1816 
1817 static void pc_machine_set_vmport(Object *obj, Visitor *v, void *opaque,
1818                                   const char *name, Error **errp)
1819 {
1820     PCMachineState *pcms = PC_MACHINE(obj);
1821 
1822     visit_type_OnOffAuto(v, &pcms->vmport, name, errp);
1823 }
1824 
1825 bool pc_machine_is_smm_enabled(PCMachineState *pcms)
1826 {
1827     bool smm_available = false;
1828 
1829     if (pcms->smm == ON_OFF_AUTO_OFF) {
1830         return false;
1831     }
1832 
1833     if (tcg_enabled() || qtest_enabled()) {
1834         smm_available = true;
1835     } else if (kvm_enabled()) {
1836         smm_available = kvm_has_smm();
1837     }
1838 
1839     if (smm_available) {
1840         return true;
1841     }
1842 
1843     if (pcms->smm == ON_OFF_AUTO_ON) {
1844         error_report("System Management Mode not supported by this hypervisor.");
1845         exit(1);
1846     }
1847     return false;
1848 }
1849 
1850 static void pc_machine_get_smm(Object *obj, Visitor *v, void *opaque,
1851                               const char *name, Error **errp)
1852 {
1853     PCMachineState *pcms = PC_MACHINE(obj);
1854     OnOffAuto smm = pcms->smm;
1855 
1856     visit_type_OnOffAuto(v, &smm, name, errp);
1857 }
1858 
1859 static void pc_machine_set_smm(Object *obj, Visitor *v, void *opaque,
1860                                   const char *name, Error **errp)
1861 {
1862     PCMachineState *pcms = PC_MACHINE(obj);
1863 
1864     visit_type_OnOffAuto(v, &pcms->smm, name, errp);
1865 }
1866 
1867 static bool pc_machine_get_nvdimm(Object *obj, Error **errp)
1868 {
1869     PCMachineState *pcms = PC_MACHINE(obj);
1870 
1871     return pcms->nvdimm;
1872 }
1873 
1874 static void pc_machine_set_nvdimm(Object *obj, bool value, Error **errp)
1875 {
1876     PCMachineState *pcms = PC_MACHINE(obj);
1877 
1878     pcms->nvdimm = value;
1879 }
1880 
1881 static void pc_machine_initfn(Object *obj)
1882 {
1883     PCMachineState *pcms = PC_MACHINE(obj);
1884 
1885     object_property_add(obj, PC_MACHINE_MEMHP_REGION_SIZE, "int",
1886                         pc_machine_get_hotplug_memory_region_size,
1887                         NULL, NULL, NULL, &error_abort);
1888 
1889     pcms->max_ram_below_4g = 1ULL << 32; /* 4G */
1890     object_property_add(obj, PC_MACHINE_MAX_RAM_BELOW_4G, "size",
1891                         pc_machine_get_max_ram_below_4g,
1892                         pc_machine_set_max_ram_below_4g,
1893                         NULL, NULL, &error_abort);
1894     object_property_set_description(obj, PC_MACHINE_MAX_RAM_BELOW_4G,
1895                                     "Maximum ram below the 4G boundary (32bit boundary)",
1896                                     &error_abort);
1897 
1898     pcms->smm = ON_OFF_AUTO_AUTO;
1899     object_property_add(obj, PC_MACHINE_SMM, "OnOffAuto",
1900                         pc_machine_get_smm,
1901                         pc_machine_set_smm,
1902                         NULL, NULL, &error_abort);
1903     object_property_set_description(obj, PC_MACHINE_SMM,
1904                                     "Enable SMM (pc & q35)",
1905                                     &error_abort);
1906 
1907     pcms->vmport = ON_OFF_AUTO_AUTO;
1908     object_property_add(obj, PC_MACHINE_VMPORT, "OnOffAuto",
1909                         pc_machine_get_vmport,
1910                         pc_machine_set_vmport,
1911                         NULL, NULL, &error_abort);
1912     object_property_set_description(obj, PC_MACHINE_VMPORT,
1913                                     "Enable vmport (pc & q35)",
1914                                     &error_abort);
1915 
1916     /* nvdimm is disabled on default. */
1917     pcms->nvdimm = false;
1918     object_property_add_bool(obj, PC_MACHINE_NVDIMM, pc_machine_get_nvdimm,
1919                              pc_machine_set_nvdimm, &error_abort);
1920 }
1921 
1922 static void pc_machine_reset(void)
1923 {
1924     CPUState *cs;
1925     X86CPU *cpu;
1926 
1927     qemu_devices_reset();
1928 
1929     /* Reset APIC after devices have been reset to cancel
1930      * any changes that qemu_devices_reset() might have done.
1931      */
1932     CPU_FOREACH(cs) {
1933         cpu = X86_CPU(cs);
1934 
1935         if (cpu->apic_state) {
1936             device_reset(cpu->apic_state);
1937         }
1938     }
1939 }
1940 
1941 static unsigned pc_cpu_index_to_socket_id(unsigned cpu_index)
1942 {
1943     X86CPUTopoInfo topo;
1944     x86_topo_ids_from_idx(smp_cores, smp_threads, cpu_index,
1945                           &topo);
1946     return topo.pkg_id;
1947 }
1948 
1949 static void pc_machine_class_init(ObjectClass *oc, void *data)
1950 {
1951     MachineClass *mc = MACHINE_CLASS(oc);
1952     PCMachineClass *pcmc = PC_MACHINE_CLASS(oc);
1953     HotplugHandlerClass *hc = HOTPLUG_HANDLER_CLASS(oc);
1954 
1955     pcmc->get_hotplug_handler = mc->get_hotplug_handler;
1956     pcmc->pci_enabled = true;
1957     pcmc->has_acpi_build = true;
1958     pcmc->rsdp_in_ram = true;
1959     pcmc->smbios_defaults = true;
1960     pcmc->smbios_uuid_encoded = true;
1961     pcmc->gigabyte_align = true;
1962     pcmc->has_reserved_memory = true;
1963     pcmc->kvmclock_enabled = true;
1964     pcmc->enforce_aligned_dimm = true;
1965     /* BIOS ACPI tables: 128K. Other BIOS datastructures: less than 4K reported
1966      * to be used at the moment, 32K should be enough for a while.  */
1967     pcmc->acpi_data_size = 0x20000 + 0x8000;
1968     pcmc->save_tsc_khz = true;
1969     mc->get_hotplug_handler = pc_get_hotpug_handler;
1970     mc->cpu_index_to_socket_id = pc_cpu_index_to_socket_id;
1971     mc->default_boot_order = "cad";
1972     mc->hot_add_cpu = pc_hot_add_cpu;
1973     mc->max_cpus = 255;
1974     mc->reset = pc_machine_reset;
1975     hc->plug = pc_machine_device_plug_cb;
1976     hc->unplug_request = pc_machine_device_unplug_request_cb;
1977     hc->unplug = pc_machine_device_unplug_cb;
1978 }
1979 
1980 static const TypeInfo pc_machine_info = {
1981     .name = TYPE_PC_MACHINE,
1982     .parent = TYPE_MACHINE,
1983     .abstract = true,
1984     .instance_size = sizeof(PCMachineState),
1985     .instance_init = pc_machine_initfn,
1986     .class_size = sizeof(PCMachineClass),
1987     .class_init = pc_machine_class_init,
1988     .interfaces = (InterfaceInfo[]) {
1989          { TYPE_HOTPLUG_HANDLER },
1990          { }
1991     },
1992 };
1993 
1994 static void pc_machine_register_types(void)
1995 {
1996     type_register_static(&pc_machine_info);
1997 }
1998 
1999 type_init(pc_machine_register_types)
2000