xref: /openbmc/qemu/hw/i386/pc.c (revision cb135f59b8059c3a372652377ef92fa4a49ad550)
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 #include "hw/nmi.h"
71 
72 /* debug PC/ISA interrupts */
73 //#define DEBUG_IRQ
74 
75 #ifdef DEBUG_IRQ
76 #define DPRINTF(fmt, ...)                                       \
77     do { printf("CPUIRQ: " fmt , ## __VA_ARGS__); } while (0)
78 #else
79 #define DPRINTF(fmt, ...)
80 #endif
81 
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 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     qdev_connect_gpio_out_named(DEVICE(dev), PORT92_A20_LINE, 0, *a20_out);
539 }
540 
541 static const VMStateDescription vmstate_port92_isa = {
542     .name = "port92",
543     .version_id = 1,
544     .minimum_version_id = 1,
545     .fields = (VMStateField[]) {
546         VMSTATE_UINT8(outport, Port92State),
547         VMSTATE_END_OF_LIST()
548     }
549 };
550 
551 static void port92_reset(DeviceState *d)
552 {
553     Port92State *s = PORT92(d);
554 
555     s->outport &= ~1;
556 }
557 
558 static const MemoryRegionOps port92_ops = {
559     .read = port92_read,
560     .write = port92_write,
561     .impl = {
562         .min_access_size = 1,
563         .max_access_size = 1,
564     },
565     .endianness = DEVICE_LITTLE_ENDIAN,
566 };
567 
568 static void port92_initfn(Object *obj)
569 {
570     Port92State *s = PORT92(obj);
571 
572     memory_region_init_io(&s->io, OBJECT(s), &port92_ops, s, "port92", 1);
573 
574     s->outport = 0;
575 
576     qdev_init_gpio_out_named(DEVICE(obj), &s->a20_out, PORT92_A20_LINE, 1);
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 static void pc_build_smbios(FWCfgState *fw_cfg)
704 {
705     uint8_t *smbios_tables, *smbios_anchor;
706     size_t smbios_tables_len, smbios_anchor_len;
707     struct smbios_phys_mem_area *mem_array;
708     unsigned i, array_count;
709 
710     smbios_tables = smbios_get_table_legacy(&smbios_tables_len);
711     if (smbios_tables) {
712         fw_cfg_add_bytes(fw_cfg, FW_CFG_SMBIOS_ENTRIES,
713                          smbios_tables, smbios_tables_len);
714     }
715 
716     /* build the array of physical mem area from e820 table */
717     mem_array = g_malloc0(sizeof(*mem_array) * e820_get_num_entries());
718     for (i = 0, array_count = 0; i < e820_get_num_entries(); i++) {
719         uint64_t addr, len;
720 
721         if (e820_get_entry(i, E820_RAM, &addr, &len)) {
722             mem_array[array_count].address = addr;
723             mem_array[array_count].length = len;
724             array_count++;
725         }
726     }
727     smbios_get_tables(mem_array, array_count,
728                       &smbios_tables, &smbios_tables_len,
729                       &smbios_anchor, &smbios_anchor_len);
730     g_free(mem_array);
731 
732     if (smbios_anchor) {
733         fw_cfg_add_file(fw_cfg, "etc/smbios/smbios-tables",
734                         smbios_tables, smbios_tables_len);
735         fw_cfg_add_file(fw_cfg, "etc/smbios/smbios-anchor",
736                         smbios_anchor, smbios_anchor_len);
737     }
738 }
739 
740 static FWCfgState *bochs_bios_init(AddressSpace *as, PCMachineState *pcms)
741 {
742     FWCfgState *fw_cfg;
743     uint64_t *numa_fw_cfg;
744     int i, j;
745 
746     fw_cfg = fw_cfg_init_io_dma(FW_CFG_IO_BASE, FW_CFG_IO_BASE + 4, as);
747 
748     /* FW_CFG_MAX_CPUS is a bit confusing/problematic on x86:
749      *
750      * SeaBIOS needs FW_CFG_MAX_CPUS for CPU hotplug, but the CPU hotplug
751      * QEMU<->SeaBIOS interface is not based on the "CPU index", but on the APIC
752      * ID of hotplugged CPUs[1]. This means that FW_CFG_MAX_CPUS is not the
753      * "maximum number of CPUs", but the "limit to the APIC ID values SeaBIOS
754      * may see".
755      *
756      * So, this means we must not use max_cpus, here, but the maximum possible
757      * APIC ID value, plus one.
758      *
759      * [1] The only kind of "CPU identifier" used between SeaBIOS and QEMU is
760      *     the APIC ID, not the "CPU index"
761      */
762     fw_cfg_add_i16(fw_cfg, FW_CFG_MAX_CPUS, (uint16_t)pcms->apic_id_limit);
763     fw_cfg_add_i64(fw_cfg, FW_CFG_RAM_SIZE, (uint64_t)ram_size);
764     fw_cfg_add_bytes(fw_cfg, FW_CFG_ACPI_TABLES,
765                      acpi_tables, acpi_tables_len);
766     fw_cfg_add_i32(fw_cfg, FW_CFG_IRQ0_OVERRIDE, kvm_allows_irq0_override());
767 
768     fw_cfg_add_bytes(fw_cfg, FW_CFG_E820_TABLE,
769                      &e820_reserve, sizeof(e820_reserve));
770     fw_cfg_add_file(fw_cfg, "etc/e820", e820_table,
771                     sizeof(struct e820_entry) * e820_entries);
772 
773     fw_cfg_add_bytes(fw_cfg, FW_CFG_HPET, &hpet_cfg, sizeof(hpet_cfg));
774     /* allocate memory for the NUMA channel: one (64bit) word for the number
775      * of nodes, one word for each VCPU->node and one word for each node to
776      * hold the amount of memory.
777      */
778     numa_fw_cfg = g_new0(uint64_t, 1 + pcms->apic_id_limit + nb_numa_nodes);
779     numa_fw_cfg[0] = cpu_to_le64(nb_numa_nodes);
780     for (i = 0; i < max_cpus; i++) {
781         unsigned int apic_id = x86_cpu_apic_id_from_index(i);
782         assert(apic_id < pcms->apic_id_limit);
783         for (j = 0; j < nb_numa_nodes; j++) {
784             if (test_bit(i, numa_info[j].node_cpu)) {
785                 numa_fw_cfg[apic_id + 1] = cpu_to_le64(j);
786                 break;
787             }
788         }
789     }
790     for (i = 0; i < nb_numa_nodes; i++) {
791         numa_fw_cfg[pcms->apic_id_limit + 1 + i] =
792             cpu_to_le64(numa_info[i].node_mem);
793     }
794     fw_cfg_add_bytes(fw_cfg, FW_CFG_NUMA, numa_fw_cfg,
795                      (1 + pcms->apic_id_limit + nb_numa_nodes) *
796                      sizeof(*numa_fw_cfg));
797 
798     return fw_cfg;
799 }
800 
801 static long get_file_size(FILE *f)
802 {
803     long where, size;
804 
805     /* XXX: on Unix systems, using fstat() probably makes more sense */
806 
807     where = ftell(f);
808     fseek(f, 0, SEEK_END);
809     size = ftell(f);
810     fseek(f, where, SEEK_SET);
811 
812     return size;
813 }
814 
815 /* setup_data types */
816 #define SETUP_NONE     0
817 #define SETUP_E820_EXT 1
818 #define SETUP_DTB      2
819 #define SETUP_PCI      3
820 #define SETUP_EFI      4
821 
822 struct setup_data {
823     uint64_t next;
824     uint32_t type;
825     uint32_t len;
826     uint8_t data[0];
827 } __attribute__((packed));
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     int dtb_size, setup_data_offset;
835     uint32_t initrd_max;
836     uint8_t header[8192], *setup, *kernel, *initrd_data;
837     hwaddr real_addr, prot_addr, cmdline_addr, initrd_addr = 0;
838     FILE *f;
839     char *vmode;
840     MachineState *machine = MACHINE(pcms);
841     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
842     struct setup_data *setup_data;
843     const char *kernel_filename = machine->kernel_filename;
844     const char *initrd_filename = machine->initrd_filename;
845     const char *dtb_filename = machine->dtb;
846     const char *kernel_cmdline = machine->kernel_cmdline;
847 
848     /* Align to 16 bytes as a paranoia measure */
849     cmdline_size = (strlen(kernel_cmdline)+16) & ~15;
850 
851     /* load the kernel header */
852     f = fopen(kernel_filename, "rb");
853     if (!f || !(kernel_size = get_file_size(f)) ||
854         fread(header, 1, MIN(ARRAY_SIZE(header), kernel_size), f) !=
855         MIN(ARRAY_SIZE(header), kernel_size)) {
856         fprintf(stderr, "qemu: could not load kernel '%s': %s\n",
857                 kernel_filename, strerror(errno));
858         exit(1);
859     }
860 
861     /* kernel protocol version */
862 #if 0
863     fprintf(stderr, "header magic: %#x\n", ldl_p(header+0x202));
864 #endif
865     if (ldl_p(header+0x202) == 0x53726448) {
866         protocol = lduw_p(header+0x206);
867     } else {
868         /* This looks like a multiboot kernel. If it is, let's stop
869            treating it like a Linux kernel. */
870         if (load_multiboot(fw_cfg, f, kernel_filename, initrd_filename,
871                            kernel_cmdline, kernel_size, header)) {
872             return;
873         }
874         protocol = 0;
875     }
876 
877     if (protocol < 0x200 || !(header[0x211] & 0x01)) {
878         /* Low kernel */
879         real_addr    = 0x90000;
880         cmdline_addr = 0x9a000 - cmdline_size;
881         prot_addr    = 0x10000;
882     } else if (protocol < 0x202) {
883         /* High but ancient kernel */
884         real_addr    = 0x90000;
885         cmdline_addr = 0x9a000 - cmdline_size;
886         prot_addr    = 0x100000;
887     } else {
888         /* High and recent kernel */
889         real_addr    = 0x10000;
890         cmdline_addr = 0x20000;
891         prot_addr    = 0x100000;
892     }
893 
894 #if 0
895     fprintf(stderr,
896             "qemu: real_addr     = 0x" TARGET_FMT_plx "\n"
897             "qemu: cmdline_addr  = 0x" TARGET_FMT_plx "\n"
898             "qemu: prot_addr     = 0x" TARGET_FMT_plx "\n",
899             real_addr,
900             cmdline_addr,
901             prot_addr);
902 #endif
903 
904     /* highest address for loading the initrd */
905     if (protocol >= 0x203) {
906         initrd_max = ldl_p(header+0x22c);
907     } else {
908         initrd_max = 0x37ffffff;
909     }
910 
911     if (initrd_max >= pcms->below_4g_mem_size - pcmc->acpi_data_size) {
912         initrd_max = pcms->below_4g_mem_size - pcmc->acpi_data_size - 1;
913     }
914 
915     fw_cfg_add_i32(fw_cfg, FW_CFG_CMDLINE_ADDR, cmdline_addr);
916     fw_cfg_add_i32(fw_cfg, FW_CFG_CMDLINE_SIZE, strlen(kernel_cmdline)+1);
917     fw_cfg_add_string(fw_cfg, FW_CFG_CMDLINE_DATA, kernel_cmdline);
918 
919     if (protocol >= 0x202) {
920         stl_p(header+0x228, cmdline_addr);
921     } else {
922         stw_p(header+0x20, 0xA33F);
923         stw_p(header+0x22, cmdline_addr-real_addr);
924     }
925 
926     /* handle vga= parameter */
927     vmode = strstr(kernel_cmdline, "vga=");
928     if (vmode) {
929         unsigned int video_mode;
930         /* skip "vga=" */
931         vmode += 4;
932         if (!strncmp(vmode, "normal", 6)) {
933             video_mode = 0xffff;
934         } else if (!strncmp(vmode, "ext", 3)) {
935             video_mode = 0xfffe;
936         } else if (!strncmp(vmode, "ask", 3)) {
937             video_mode = 0xfffd;
938         } else {
939             video_mode = strtol(vmode, NULL, 0);
940         }
941         stw_p(header+0x1fa, video_mode);
942     }
943 
944     /* loader type */
945     /* High nybble = B reserved for QEMU; low nybble is revision number.
946        If this code is substantially changed, you may want to consider
947        incrementing the revision. */
948     if (protocol >= 0x200) {
949         header[0x210] = 0xB0;
950     }
951     /* heap */
952     if (protocol >= 0x201) {
953         header[0x211] |= 0x80;	/* CAN_USE_HEAP */
954         stw_p(header+0x224, cmdline_addr-real_addr-0x200);
955     }
956 
957     /* load initrd */
958     if (initrd_filename) {
959         if (protocol < 0x200) {
960             fprintf(stderr, "qemu: linux kernel too old to load a ram disk\n");
961             exit(1);
962         }
963 
964         initrd_size = get_image_size(initrd_filename);
965         if (initrd_size < 0) {
966             fprintf(stderr, "qemu: error reading initrd %s: %s\n",
967                     initrd_filename, strerror(errno));
968             exit(1);
969         }
970 
971         initrd_addr = (initrd_max-initrd_size) & ~4095;
972 
973         initrd_data = g_malloc(initrd_size);
974         load_image(initrd_filename, initrd_data);
975 
976         fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_ADDR, initrd_addr);
977         fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_SIZE, initrd_size);
978         fw_cfg_add_bytes(fw_cfg, FW_CFG_INITRD_DATA, initrd_data, initrd_size);
979 
980         stl_p(header+0x218, initrd_addr);
981         stl_p(header+0x21c, initrd_size);
982     }
983 
984     /* load kernel and setup */
985     setup_size = header[0x1f1];
986     if (setup_size == 0) {
987         setup_size = 4;
988     }
989     setup_size = (setup_size+1)*512;
990     if (setup_size > kernel_size) {
991         fprintf(stderr, "qemu: invalid kernel header\n");
992         exit(1);
993     }
994     kernel_size -= setup_size;
995 
996     setup  = g_malloc(setup_size);
997     kernel = g_malloc(kernel_size);
998     fseek(f, 0, SEEK_SET);
999     if (fread(setup, 1, setup_size, f) != setup_size) {
1000         fprintf(stderr, "fread() failed\n");
1001         exit(1);
1002     }
1003     if (fread(kernel, 1, kernel_size, f) != kernel_size) {
1004         fprintf(stderr, "fread() failed\n");
1005         exit(1);
1006     }
1007     fclose(f);
1008 
1009     /* append dtb to kernel */
1010     if (dtb_filename) {
1011         if (protocol < 0x209) {
1012             fprintf(stderr, "qemu: Linux kernel too old to load a dtb\n");
1013             exit(1);
1014         }
1015 
1016         dtb_size = get_image_size(dtb_filename);
1017         if (dtb_size <= 0) {
1018             fprintf(stderr, "qemu: error reading dtb %s: %s\n",
1019                     dtb_filename, strerror(errno));
1020             exit(1);
1021         }
1022 
1023         setup_data_offset = QEMU_ALIGN_UP(kernel_size, 16);
1024         kernel_size = setup_data_offset + sizeof(struct setup_data) + dtb_size;
1025         kernel = g_realloc(kernel, kernel_size);
1026 
1027         stq_p(header+0x250, prot_addr + setup_data_offset);
1028 
1029         setup_data = (struct setup_data *)(kernel + setup_data_offset);
1030         setup_data->next = 0;
1031         setup_data->type = cpu_to_le32(SETUP_DTB);
1032         setup_data->len = cpu_to_le32(dtb_size);
1033 
1034         load_image_size(dtb_filename, setup_data->data, dtb_size);
1035     }
1036 
1037     memcpy(setup, header, MIN(sizeof(header), setup_size));
1038 
1039     fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_ADDR, prot_addr);
1040     fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_SIZE, kernel_size);
1041     fw_cfg_add_bytes(fw_cfg, FW_CFG_KERNEL_DATA, kernel, kernel_size);
1042 
1043     fw_cfg_add_i32(fw_cfg, FW_CFG_SETUP_ADDR, real_addr);
1044     fw_cfg_add_i32(fw_cfg, FW_CFG_SETUP_SIZE, setup_size);
1045     fw_cfg_add_bytes(fw_cfg, FW_CFG_SETUP_DATA, setup, setup_size);
1046 
1047     if (fw_cfg_dma_enabled(fw_cfg)) {
1048         option_rom[nb_option_roms].name = "linuxboot_dma.bin";
1049         option_rom[nb_option_roms].bootindex = 0;
1050     } else {
1051         option_rom[nb_option_roms].name = "linuxboot.bin";
1052         option_rom[nb_option_roms].bootindex = 0;
1053     }
1054     nb_option_roms++;
1055 }
1056 
1057 #define NE2000_NB_MAX 6
1058 
1059 static const int ne2000_io[NE2000_NB_MAX] = { 0x300, 0x320, 0x340, 0x360,
1060                                               0x280, 0x380 };
1061 static const int ne2000_irq[NE2000_NB_MAX] = { 9, 10, 11, 3, 4, 5 };
1062 
1063 void pc_init_ne2k_isa(ISABus *bus, NICInfo *nd)
1064 {
1065     static int nb_ne2k = 0;
1066 
1067     if (nb_ne2k == NE2000_NB_MAX)
1068         return;
1069     isa_ne2000_init(bus, ne2000_io[nb_ne2k],
1070                     ne2000_irq[nb_ne2k], nd);
1071     nb_ne2k++;
1072 }
1073 
1074 DeviceState *cpu_get_current_apic(void)
1075 {
1076     if (current_cpu) {
1077         X86CPU *cpu = X86_CPU(current_cpu);
1078         return cpu->apic_state;
1079     } else {
1080         return NULL;
1081     }
1082 }
1083 
1084 void pc_acpi_smi_interrupt(void *opaque, int irq, int level)
1085 {
1086     X86CPU *cpu = opaque;
1087 
1088     if (level) {
1089         cpu_interrupt(CPU(cpu), CPU_INTERRUPT_SMI);
1090     }
1091 }
1092 
1093 static X86CPU *pc_new_cpu(const char *typename, int64_t apic_id,
1094                           Error **errp)
1095 {
1096     X86CPU *cpu = NULL;
1097     Error *local_err = NULL;
1098 
1099     cpu = X86_CPU(object_new(typename));
1100 
1101     object_property_set_int(OBJECT(cpu), apic_id, "apic-id", &local_err);
1102     object_property_set_bool(OBJECT(cpu), true, "realized", &local_err);
1103 
1104     if (local_err) {
1105         error_propagate(errp, local_err);
1106         object_unref(OBJECT(cpu));
1107         cpu = NULL;
1108     }
1109     return cpu;
1110 }
1111 
1112 void pc_hot_add_cpu(const int64_t id, Error **errp)
1113 {
1114     X86CPU *cpu;
1115     ObjectClass *oc;
1116     PCMachineState *pcms = PC_MACHINE(qdev_get_machine());
1117     int64_t apic_id = x86_cpu_apic_id_from_index(id);
1118     Error *local_err = NULL;
1119 
1120     if (id < 0) {
1121         error_setg(errp, "Invalid CPU id: %" PRIi64, id);
1122         return;
1123     }
1124 
1125     if (cpu_exists(apic_id)) {
1126         error_setg(errp, "Unable to add CPU: %" PRIi64
1127                    ", it already exists", id);
1128         return;
1129     }
1130 
1131     if (id >= max_cpus) {
1132         error_setg(errp, "Unable to add CPU: %" PRIi64
1133                    ", max allowed: %d", id, max_cpus - 1);
1134         return;
1135     }
1136 
1137     if (apic_id >= ACPI_CPU_HOTPLUG_ID_LIMIT) {
1138         error_setg(errp, "Unable to add CPU: %" PRIi64
1139                    ", resulting APIC ID (%" PRIi64 ") is too large",
1140                    id, apic_id);
1141         return;
1142     }
1143 
1144     assert(pcms->possible_cpus->cpus[0].cpu); /* BSP is always present */
1145     oc = OBJECT_CLASS(CPU_GET_CLASS(pcms->possible_cpus->cpus[0].cpu));
1146     cpu = pc_new_cpu(object_class_get_name(oc), apic_id, &local_err);
1147     if (local_err) {
1148         error_propagate(errp, local_err);
1149         return;
1150     }
1151     object_unref(OBJECT(cpu));
1152 }
1153 
1154 void pc_cpus_init(PCMachineState *pcms)
1155 {
1156     int i;
1157     CPUClass *cc;
1158     ObjectClass *oc;
1159     const char *typename;
1160     gchar **model_pieces;
1161     X86CPU *cpu = NULL;
1162     MachineState *machine = MACHINE(pcms);
1163 
1164     /* init CPUs */
1165     if (machine->cpu_model == NULL) {
1166 #ifdef TARGET_X86_64
1167         machine->cpu_model = "qemu64";
1168 #else
1169         machine->cpu_model = "qemu32";
1170 #endif
1171     }
1172 
1173     model_pieces = g_strsplit(machine->cpu_model, ",", 2);
1174     if (!model_pieces[0]) {
1175         error_report("Invalid/empty CPU model name");
1176         exit(1);
1177     }
1178 
1179     oc = cpu_class_by_name(TYPE_X86_CPU, model_pieces[0]);
1180     if (oc == NULL) {
1181         error_report("Unable to find CPU definition: %s", model_pieces[0]);
1182         exit(1);
1183     }
1184     typename = object_class_get_name(oc);
1185     cc = CPU_CLASS(oc);
1186     cc->parse_features(typename, model_pieces[1], &error_fatal);
1187     g_strfreev(model_pieces);
1188 
1189     /* Calculates the limit to CPU APIC ID values
1190      *
1191      * Limit for the APIC ID value, so that all
1192      * CPU APIC IDs are < pcms->apic_id_limit.
1193      *
1194      * This is used for FW_CFG_MAX_CPUS. See comments on bochs_bios_init().
1195      */
1196     pcms->apic_id_limit = x86_cpu_apic_id_from_index(max_cpus - 1) + 1;
1197     if (pcms->apic_id_limit > ACPI_CPU_HOTPLUG_ID_LIMIT) {
1198         error_report("max_cpus is too large. APIC ID of last CPU is %u",
1199                      pcms->apic_id_limit - 1);
1200         exit(1);
1201     }
1202 
1203     pcms->possible_cpus = g_malloc0(sizeof(CPUArchIdList) +
1204                                     sizeof(CPUArchId) * max_cpus);
1205     for (i = 0; i < max_cpus; i++) {
1206         pcms->possible_cpus->cpus[i].arch_id = x86_cpu_apic_id_from_index(i);
1207         pcms->possible_cpus->len++;
1208         if (i < smp_cpus) {
1209             cpu = pc_new_cpu(typename, x86_cpu_apic_id_from_index(i),
1210                              &error_fatal);
1211             pcms->possible_cpus->cpus[i].cpu = CPU(cpu);
1212             object_unref(OBJECT(cpu));
1213         }
1214     }
1215 
1216     /* tell smbios about cpuid version and features */
1217     smbios_set_cpuid(cpu->env.cpuid_version, cpu->env.features[FEAT_1_EDX]);
1218 }
1219 
1220 static void pc_build_feature_control_file(PCMachineState *pcms)
1221 {
1222     X86CPU *cpu = X86_CPU(pcms->possible_cpus->cpus[0].cpu);
1223     CPUX86State *env = &cpu->env;
1224     uint32_t unused, ecx, edx;
1225     uint64_t feature_control_bits = 0;
1226     uint64_t *val;
1227 
1228     cpu_x86_cpuid(env, 1, 0, &unused, &unused, &ecx, &edx);
1229     if (ecx & CPUID_EXT_VMX) {
1230         feature_control_bits |= FEATURE_CONTROL_VMXON_ENABLED_OUTSIDE_SMX;
1231     }
1232 
1233     if ((edx & (CPUID_EXT2_MCE | CPUID_EXT2_MCA)) ==
1234         (CPUID_EXT2_MCE | CPUID_EXT2_MCA) &&
1235         (env->mcg_cap & MCG_LMCE_P)) {
1236         feature_control_bits |= FEATURE_CONTROL_LMCE;
1237     }
1238 
1239     if (!feature_control_bits) {
1240         return;
1241     }
1242 
1243     val = g_malloc(sizeof(*val));
1244     *val = cpu_to_le64(feature_control_bits | FEATURE_CONTROL_LOCKED);
1245     fw_cfg_add_file(pcms->fw_cfg, "etc/msr_feature_control", val, sizeof(*val));
1246 }
1247 
1248 static
1249 void pc_machine_done(Notifier *notifier, void *data)
1250 {
1251     PCMachineState *pcms = container_of(notifier,
1252                                         PCMachineState, machine_done);
1253     PCIBus *bus = pcms->bus;
1254 
1255     if (bus) {
1256         int extra_hosts = 0;
1257 
1258         QLIST_FOREACH(bus, &bus->child, sibling) {
1259             /* look for expander root buses */
1260             if (pci_bus_is_root(bus)) {
1261                 extra_hosts++;
1262             }
1263         }
1264         if (extra_hosts && pcms->fw_cfg) {
1265             uint64_t *val = g_malloc(sizeof(*val));
1266             *val = cpu_to_le64(extra_hosts);
1267             fw_cfg_add_file(pcms->fw_cfg,
1268                     "etc/extra-pci-roots", val, sizeof(*val));
1269         }
1270     }
1271 
1272     acpi_setup();
1273     if (pcms->fw_cfg) {
1274         pc_build_smbios(pcms->fw_cfg);
1275         pc_build_feature_control_file(pcms);
1276     }
1277 }
1278 
1279 void pc_guest_info_init(PCMachineState *pcms)
1280 {
1281     int i;
1282 
1283     pcms->apic_xrupt_override = kvm_allows_irq0_override();
1284     pcms->numa_nodes = nb_numa_nodes;
1285     pcms->node_mem = g_malloc0(pcms->numa_nodes *
1286                                     sizeof *pcms->node_mem);
1287     for (i = 0; i < nb_numa_nodes; i++) {
1288         pcms->node_mem[i] = numa_info[i].node_mem;
1289     }
1290 
1291     pcms->machine_done.notify = pc_machine_done;
1292     qemu_add_machine_init_done_notifier(&pcms->machine_done);
1293 }
1294 
1295 /* setup pci memory address space mapping into system address space */
1296 void pc_pci_as_mapping_init(Object *owner, MemoryRegion *system_memory,
1297                             MemoryRegion *pci_address_space)
1298 {
1299     /* Set to lower priority than RAM */
1300     memory_region_add_subregion_overlap(system_memory, 0x0,
1301                                         pci_address_space, -1);
1302 }
1303 
1304 void pc_acpi_init(const char *default_dsdt)
1305 {
1306     char *filename;
1307 
1308     if (acpi_tables != NULL) {
1309         /* manually set via -acpitable, leave it alone */
1310         return;
1311     }
1312 
1313     filename = qemu_find_file(QEMU_FILE_TYPE_BIOS, default_dsdt);
1314     if (filename == NULL) {
1315         fprintf(stderr, "WARNING: failed to find %s\n", default_dsdt);
1316     } else {
1317         QemuOpts *opts = qemu_opts_create(qemu_find_opts("acpi"), NULL, 0,
1318                                           &error_abort);
1319         Error *err = NULL;
1320 
1321         qemu_opt_set(opts, "file", filename, &error_abort);
1322 
1323         acpi_table_add_builtin(opts, &err);
1324         if (err) {
1325             error_reportf_err(err, "WARNING: failed to load %s: ",
1326                               filename);
1327         }
1328         g_free(filename);
1329     }
1330 }
1331 
1332 void xen_load_linux(PCMachineState *pcms)
1333 {
1334     int i;
1335     FWCfgState *fw_cfg;
1336 
1337     assert(MACHINE(pcms)->kernel_filename != NULL);
1338 
1339     fw_cfg = fw_cfg_init_io(FW_CFG_IO_BASE);
1340     rom_set_fw(fw_cfg);
1341 
1342     load_linux(pcms, fw_cfg);
1343     for (i = 0; i < nb_option_roms; i++) {
1344         assert(!strcmp(option_rom[i].name, "linuxboot.bin") ||
1345                !strcmp(option_rom[i].name, "linuxboot_dma.bin") ||
1346                !strcmp(option_rom[i].name, "multiboot.bin"));
1347         rom_add_option(option_rom[i].name, option_rom[i].bootindex);
1348     }
1349     pcms->fw_cfg = fw_cfg;
1350 }
1351 
1352 void pc_memory_init(PCMachineState *pcms,
1353                     MemoryRegion *system_memory,
1354                     MemoryRegion *rom_memory,
1355                     MemoryRegion **ram_memory)
1356 {
1357     int linux_boot, i;
1358     MemoryRegion *ram, *option_rom_mr;
1359     MemoryRegion *ram_below_4g, *ram_above_4g;
1360     FWCfgState *fw_cfg;
1361     MachineState *machine = MACHINE(pcms);
1362     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
1363 
1364     assert(machine->ram_size == pcms->below_4g_mem_size +
1365                                 pcms->above_4g_mem_size);
1366 
1367     linux_boot = (machine->kernel_filename != NULL);
1368 
1369     /* Allocate RAM.  We allocate it as a single memory region and use
1370      * aliases to address portions of it, mostly for backwards compatibility
1371      * with older qemus that used qemu_ram_alloc().
1372      */
1373     ram = g_malloc(sizeof(*ram));
1374     memory_region_allocate_system_memory(ram, NULL, "pc.ram",
1375                                          machine->ram_size);
1376     *ram_memory = ram;
1377     ram_below_4g = g_malloc(sizeof(*ram_below_4g));
1378     memory_region_init_alias(ram_below_4g, NULL, "ram-below-4g", ram,
1379                              0, pcms->below_4g_mem_size);
1380     memory_region_add_subregion(system_memory, 0, ram_below_4g);
1381     e820_add_entry(0, pcms->below_4g_mem_size, E820_RAM);
1382     if (pcms->above_4g_mem_size > 0) {
1383         ram_above_4g = g_malloc(sizeof(*ram_above_4g));
1384         memory_region_init_alias(ram_above_4g, NULL, "ram-above-4g", ram,
1385                                  pcms->below_4g_mem_size,
1386                                  pcms->above_4g_mem_size);
1387         memory_region_add_subregion(system_memory, 0x100000000ULL,
1388                                     ram_above_4g);
1389         e820_add_entry(0x100000000ULL, pcms->above_4g_mem_size, E820_RAM);
1390     }
1391 
1392     if (!pcmc->has_reserved_memory &&
1393         (machine->ram_slots ||
1394          (machine->maxram_size > machine->ram_size))) {
1395         MachineClass *mc = MACHINE_GET_CLASS(machine);
1396 
1397         error_report("\"-memory 'slots|maxmem'\" is not supported by: %s",
1398                      mc->name);
1399         exit(EXIT_FAILURE);
1400     }
1401 
1402     /* initialize hotplug memory address space */
1403     if (pcmc->has_reserved_memory &&
1404         (machine->ram_size < machine->maxram_size)) {
1405         ram_addr_t hotplug_mem_size =
1406             machine->maxram_size - machine->ram_size;
1407 
1408         if (machine->ram_slots > ACPI_MAX_RAM_SLOTS) {
1409             error_report("unsupported amount of memory slots: %"PRIu64,
1410                          machine->ram_slots);
1411             exit(EXIT_FAILURE);
1412         }
1413 
1414         if (QEMU_ALIGN_UP(machine->maxram_size,
1415                           TARGET_PAGE_SIZE) != machine->maxram_size) {
1416             error_report("maximum memory size must by aligned to multiple of "
1417                          "%d bytes", TARGET_PAGE_SIZE);
1418             exit(EXIT_FAILURE);
1419         }
1420 
1421         pcms->hotplug_memory.base =
1422             ROUND_UP(0x100000000ULL + pcms->above_4g_mem_size, 1ULL << 30);
1423 
1424         if (pcmc->enforce_aligned_dimm) {
1425             /* size hotplug region assuming 1G page max alignment per slot */
1426             hotplug_mem_size += (1ULL << 30) * machine->ram_slots;
1427         }
1428 
1429         if ((pcms->hotplug_memory.base + hotplug_mem_size) <
1430             hotplug_mem_size) {
1431             error_report("unsupported amount of maximum memory: " RAM_ADDR_FMT,
1432                          machine->maxram_size);
1433             exit(EXIT_FAILURE);
1434         }
1435 
1436         memory_region_init(&pcms->hotplug_memory.mr, OBJECT(pcms),
1437                            "hotplug-memory", hotplug_mem_size);
1438         memory_region_add_subregion(system_memory, pcms->hotplug_memory.base,
1439                                     &pcms->hotplug_memory.mr);
1440     }
1441 
1442     /* Initialize PC system firmware */
1443     pc_system_firmware_init(rom_memory, !pcmc->pci_enabled);
1444 
1445     option_rom_mr = g_malloc(sizeof(*option_rom_mr));
1446     memory_region_init_ram(option_rom_mr, NULL, "pc.rom", PC_ROM_SIZE,
1447                            &error_fatal);
1448     vmstate_register_ram_global(option_rom_mr);
1449     memory_region_add_subregion_overlap(rom_memory,
1450                                         PC_ROM_MIN_VGA,
1451                                         option_rom_mr,
1452                                         1);
1453 
1454     fw_cfg = bochs_bios_init(&address_space_memory, pcms);
1455 
1456     rom_set_fw(fw_cfg);
1457 
1458     if (pcmc->has_reserved_memory && pcms->hotplug_memory.base) {
1459         uint64_t *val = g_malloc(sizeof(*val));
1460         PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
1461         uint64_t res_mem_end = pcms->hotplug_memory.base;
1462 
1463         if (!pcmc->broken_reserved_end) {
1464             res_mem_end += memory_region_size(&pcms->hotplug_memory.mr);
1465         }
1466         *val = cpu_to_le64(ROUND_UP(res_mem_end, 0x1ULL << 30));
1467         fw_cfg_add_file(fw_cfg, "etc/reserved-memory-end", val, sizeof(*val));
1468     }
1469 
1470     if (linux_boot) {
1471         load_linux(pcms, fw_cfg);
1472     }
1473 
1474     for (i = 0; i < nb_option_roms; i++) {
1475         rom_add_option(option_rom[i].name, option_rom[i].bootindex);
1476     }
1477     pcms->fw_cfg = fw_cfg;
1478 
1479     /* Init default IOAPIC address space */
1480     pcms->ioapic_as = &address_space_memory;
1481 }
1482 
1483 qemu_irq pc_allocate_cpu_irq(void)
1484 {
1485     return qemu_allocate_irq(pic_irq_request, NULL, 0);
1486 }
1487 
1488 DeviceState *pc_vga_init(ISABus *isa_bus, PCIBus *pci_bus)
1489 {
1490     DeviceState *dev = NULL;
1491 
1492     rom_set_order_override(FW_CFG_ORDER_OVERRIDE_VGA);
1493     if (pci_bus) {
1494         PCIDevice *pcidev = pci_vga_init(pci_bus);
1495         dev = pcidev ? &pcidev->qdev : NULL;
1496     } else if (isa_bus) {
1497         ISADevice *isadev = isa_vga_init(isa_bus);
1498         dev = isadev ? DEVICE(isadev) : NULL;
1499     }
1500     rom_reset_order_override();
1501     return dev;
1502 }
1503 
1504 static const MemoryRegionOps ioport80_io_ops = {
1505     .write = ioport80_write,
1506     .read = ioport80_read,
1507     .endianness = DEVICE_NATIVE_ENDIAN,
1508     .impl = {
1509         .min_access_size = 1,
1510         .max_access_size = 1,
1511     },
1512 };
1513 
1514 static const MemoryRegionOps ioportF0_io_ops = {
1515     .write = ioportF0_write,
1516     .read = ioportF0_read,
1517     .endianness = DEVICE_NATIVE_ENDIAN,
1518     .impl = {
1519         .min_access_size = 1,
1520         .max_access_size = 1,
1521     },
1522 };
1523 
1524 void pc_basic_device_init(ISABus *isa_bus, qemu_irq *gsi,
1525                           ISADevice **rtc_state,
1526                           bool create_fdctrl,
1527                           bool no_vmport,
1528                           uint32_t hpet_irqs)
1529 {
1530     int i;
1531     DriveInfo *fd[MAX_FD];
1532     DeviceState *hpet = NULL;
1533     int pit_isa_irq = 0;
1534     qemu_irq pit_alt_irq = NULL;
1535     qemu_irq rtc_irq = NULL;
1536     qemu_irq *a20_line;
1537     ISADevice *i8042, *port92, *vmmouse, *pit = NULL;
1538     MemoryRegion *ioport80_io = g_new(MemoryRegion, 1);
1539     MemoryRegion *ioportF0_io = g_new(MemoryRegion, 1);
1540 
1541     memory_region_init_io(ioport80_io, NULL, &ioport80_io_ops, NULL, "ioport80", 1);
1542     memory_region_add_subregion(isa_bus->address_space_io, 0x80, ioport80_io);
1543 
1544     memory_region_init_io(ioportF0_io, NULL, &ioportF0_io_ops, NULL, "ioportF0", 1);
1545     memory_region_add_subregion(isa_bus->address_space_io, 0xf0, ioportF0_io);
1546 
1547     /*
1548      * Check if an HPET shall be created.
1549      *
1550      * Without KVM_CAP_PIT_STATE2, we cannot switch off the in-kernel PIT
1551      * when the HPET wants to take over. Thus we have to disable the latter.
1552      */
1553     if (!no_hpet && (!kvm_irqchip_in_kernel() || kvm_has_pit_state2())) {
1554         /* In order to set property, here not using sysbus_try_create_simple */
1555         hpet = qdev_try_create(NULL, TYPE_HPET);
1556         if (hpet) {
1557             /* For pc-piix-*, hpet's intcap is always IRQ2. For pc-q35-1.7
1558              * and earlier, use IRQ2 for compat. Otherwise, use IRQ16~23,
1559              * IRQ8 and IRQ2.
1560              */
1561             uint8_t compat = object_property_get_int(OBJECT(hpet),
1562                     HPET_INTCAP, NULL);
1563             if (!compat) {
1564                 qdev_prop_set_uint32(hpet, HPET_INTCAP, hpet_irqs);
1565             }
1566             qdev_init_nofail(hpet);
1567             sysbus_mmio_map(SYS_BUS_DEVICE(hpet), 0, HPET_BASE);
1568 
1569             for (i = 0; i < GSI_NUM_PINS; i++) {
1570                 sysbus_connect_irq(SYS_BUS_DEVICE(hpet), i, gsi[i]);
1571             }
1572             pit_isa_irq = -1;
1573             pit_alt_irq = qdev_get_gpio_in(hpet, HPET_LEGACY_PIT_INT);
1574             rtc_irq = qdev_get_gpio_in(hpet, HPET_LEGACY_RTC_INT);
1575         }
1576     }
1577     *rtc_state = rtc_init(isa_bus, 2000, rtc_irq);
1578 
1579     qemu_register_boot_set(pc_boot_set, *rtc_state);
1580 
1581     if (!xen_enabled()) {
1582         if (kvm_pit_in_kernel()) {
1583             pit = kvm_pit_init(isa_bus, 0x40);
1584         } else {
1585             pit = pit_init(isa_bus, 0x40, pit_isa_irq, pit_alt_irq);
1586         }
1587         if (hpet) {
1588             /* connect PIT to output control line of the HPET */
1589             qdev_connect_gpio_out(hpet, 0, qdev_get_gpio_in(DEVICE(pit), 0));
1590         }
1591         pcspk_init(isa_bus, pit);
1592     }
1593 
1594     serial_hds_isa_init(isa_bus, MAX_SERIAL_PORTS);
1595     parallel_hds_isa_init(isa_bus, MAX_PARALLEL_PORTS);
1596 
1597     a20_line = qemu_allocate_irqs(handle_a20_line_change, first_cpu, 2);
1598     i8042 = isa_create_simple(isa_bus, "i8042");
1599     i8042_setup_a20_line(i8042, &a20_line[0]);
1600     if (!no_vmport) {
1601         vmport_init(isa_bus);
1602         vmmouse = isa_try_create(isa_bus, "vmmouse");
1603     } else {
1604         vmmouse = NULL;
1605     }
1606     if (vmmouse) {
1607         DeviceState *dev = DEVICE(vmmouse);
1608         qdev_prop_set_ptr(dev, "ps2_mouse", i8042);
1609         qdev_init_nofail(dev);
1610     }
1611     port92 = isa_create_simple(isa_bus, "port92");
1612     port92_init(port92, &a20_line[1]);
1613 
1614     DMA_init(isa_bus, 0);
1615 
1616     for(i = 0; i < MAX_FD; i++) {
1617         fd[i] = drive_get(IF_FLOPPY, 0, i);
1618         create_fdctrl |= !!fd[i];
1619     }
1620     if (create_fdctrl) {
1621         fdctrl_init_isa(isa_bus, fd);
1622     }
1623 }
1624 
1625 void pc_nic_init(ISABus *isa_bus, PCIBus *pci_bus)
1626 {
1627     int i;
1628 
1629     rom_set_order_override(FW_CFG_ORDER_OVERRIDE_NIC);
1630     for (i = 0; i < nb_nics; i++) {
1631         NICInfo *nd = &nd_table[i];
1632 
1633         if (!pci_bus || (nd->model && strcmp(nd->model, "ne2k_isa") == 0)) {
1634             pc_init_ne2k_isa(isa_bus, nd);
1635         } else {
1636             pci_nic_init_nofail(nd, pci_bus, "e1000", NULL);
1637         }
1638     }
1639     rom_reset_order_override();
1640 }
1641 
1642 void pc_pci_device_init(PCIBus *pci_bus)
1643 {
1644     int max_bus;
1645     int bus;
1646 
1647     max_bus = drive_get_max_bus(IF_SCSI);
1648     for (bus = 0; bus <= max_bus; bus++) {
1649         pci_create_simple(pci_bus, -1, "lsi53c895a");
1650     }
1651 }
1652 
1653 void ioapic_init_gsi(GSIState *gsi_state, const char *parent_name)
1654 {
1655     DeviceState *dev;
1656     SysBusDevice *d;
1657     unsigned int i;
1658 
1659     if (kvm_ioapic_in_kernel()) {
1660         dev = qdev_create(NULL, "kvm-ioapic");
1661     } else {
1662         dev = qdev_create(NULL, "ioapic");
1663     }
1664     if (parent_name) {
1665         object_property_add_child(object_resolve_path(parent_name, NULL),
1666                                   "ioapic", OBJECT(dev), NULL);
1667     }
1668     qdev_init_nofail(dev);
1669     d = SYS_BUS_DEVICE(dev);
1670     sysbus_mmio_map(d, 0, IO_APIC_DEFAULT_ADDRESS);
1671 
1672     for (i = 0; i < IOAPIC_NUM_PINS; i++) {
1673         gsi_state->ioapic_irq[i] = qdev_get_gpio_in(dev, i);
1674     }
1675 }
1676 
1677 static void pc_dimm_plug(HotplugHandler *hotplug_dev,
1678                          DeviceState *dev, Error **errp)
1679 {
1680     HotplugHandlerClass *hhc;
1681     Error *local_err = NULL;
1682     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1683     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
1684     PCDIMMDevice *dimm = PC_DIMM(dev);
1685     PCDIMMDeviceClass *ddc = PC_DIMM_GET_CLASS(dimm);
1686     MemoryRegion *mr = ddc->get_memory_region(dimm);
1687     uint64_t align = TARGET_PAGE_SIZE;
1688 
1689     if (memory_region_get_alignment(mr) && pcmc->enforce_aligned_dimm) {
1690         align = memory_region_get_alignment(mr);
1691     }
1692 
1693     if (!pcms->acpi_dev) {
1694         error_setg(&local_err,
1695                    "memory hotplug is not enabled: missing acpi device");
1696         goto out;
1697     }
1698 
1699     pc_dimm_memory_plug(dev, &pcms->hotplug_memory, mr, align, &local_err);
1700     if (local_err) {
1701         goto out;
1702     }
1703 
1704     hhc = HOTPLUG_HANDLER_GET_CLASS(pcms->acpi_dev);
1705     hhc->plug(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &error_abort);
1706 out:
1707     error_propagate(errp, local_err);
1708 }
1709 
1710 static void pc_dimm_unplug_request(HotplugHandler *hotplug_dev,
1711                                    DeviceState *dev, Error **errp)
1712 {
1713     HotplugHandlerClass *hhc;
1714     Error *local_err = NULL;
1715     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1716 
1717     if (!pcms->acpi_dev) {
1718         error_setg(&local_err,
1719                    "memory hotplug is not enabled: missing acpi device");
1720         goto out;
1721     }
1722 
1723     hhc = HOTPLUG_HANDLER_GET_CLASS(pcms->acpi_dev);
1724     hhc->unplug_request(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &local_err);
1725 
1726 out:
1727     error_propagate(errp, local_err);
1728 }
1729 
1730 static void pc_dimm_unplug(HotplugHandler *hotplug_dev,
1731                            DeviceState *dev, Error **errp)
1732 {
1733     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1734     PCDIMMDevice *dimm = PC_DIMM(dev);
1735     PCDIMMDeviceClass *ddc = PC_DIMM_GET_CLASS(dimm);
1736     MemoryRegion *mr = ddc->get_memory_region(dimm);
1737     HotplugHandlerClass *hhc;
1738     Error *local_err = NULL;
1739 
1740     hhc = HOTPLUG_HANDLER_GET_CLASS(pcms->acpi_dev);
1741     hhc->unplug(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &local_err);
1742 
1743     if (local_err) {
1744         goto out;
1745     }
1746 
1747     pc_dimm_memory_unplug(dev, &pcms->hotplug_memory, mr);
1748     object_unparent(OBJECT(dev));
1749 
1750  out:
1751     error_propagate(errp, local_err);
1752 }
1753 
1754 static int pc_apic_cmp(const void *a, const void *b)
1755 {
1756    CPUArchId *apic_a = (CPUArchId *)a;
1757    CPUArchId *apic_b = (CPUArchId *)b;
1758 
1759    return apic_a->arch_id - apic_b->arch_id;
1760 }
1761 
1762 static void pc_cpu_plug(HotplugHandler *hotplug_dev,
1763                         DeviceState *dev, Error **errp)
1764 {
1765     CPUClass *cc = CPU_GET_CLASS(dev);
1766     CPUArchId apic_id, *found_cpu;
1767     HotplugHandlerClass *hhc;
1768     Error *local_err = NULL;
1769     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1770 
1771     if (!dev->hotplugged) {
1772         goto out;
1773     }
1774 
1775     if (!pcms->acpi_dev) {
1776         error_setg(&local_err,
1777                    "cpu hotplug is not enabled: missing acpi device");
1778         goto out;
1779     }
1780 
1781     hhc = HOTPLUG_HANDLER_GET_CLASS(pcms->acpi_dev);
1782     hhc->plug(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &local_err);
1783     if (local_err) {
1784         goto out;
1785     }
1786 
1787     /* increment the number of CPUs */
1788     rtc_set_memory(pcms->rtc, 0x5f, rtc_get_memory(pcms->rtc, 0x5f) + 1);
1789 
1790     apic_id.arch_id = cc->get_arch_id(CPU(dev));
1791     found_cpu = bsearch(&apic_id, pcms->possible_cpus->cpus,
1792         pcms->possible_cpus->len, sizeof(*pcms->possible_cpus->cpus),
1793         pc_apic_cmp);
1794     assert(found_cpu);
1795     found_cpu->cpu = CPU(dev);
1796 out:
1797     error_propagate(errp, local_err);
1798 }
1799 static void pc_cpu_unplug_request_cb(HotplugHandler *hotplug_dev,
1800                                      DeviceState *dev, Error **errp)
1801 {
1802     HotplugHandlerClass *hhc;
1803     Error *local_err = NULL;
1804     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1805 
1806     hhc = HOTPLUG_HANDLER_GET_CLASS(pcms->acpi_dev);
1807     hhc->unplug_request(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &local_err);
1808 
1809     if (local_err) {
1810         goto out;
1811     }
1812 
1813  out:
1814     error_propagate(errp, local_err);
1815 
1816 }
1817 
1818 static void pc_cpu_unplug_cb(HotplugHandler *hotplug_dev,
1819                              DeviceState *dev, Error **errp)
1820 {
1821     HotplugHandlerClass *hhc;
1822     Error *local_err = NULL;
1823     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1824 
1825     hhc = HOTPLUG_HANDLER_GET_CLASS(pcms->acpi_dev);
1826     hhc->unplug(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &local_err);
1827 
1828     if (local_err) {
1829         goto out;
1830     }
1831 
1832     /*
1833      * TODO: enable unplug once generic CPU remove bits land
1834      * for now guest will be able to eject CPU ACPI wise but
1835      * it will come back again on machine reset.
1836      */
1837     /*  object_unparent(OBJECT(dev)); */
1838 
1839  out:
1840     error_propagate(errp, local_err);
1841 }
1842 
1843 static void pc_machine_device_plug_cb(HotplugHandler *hotplug_dev,
1844                                       DeviceState *dev, Error **errp)
1845 {
1846     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
1847         pc_dimm_plug(hotplug_dev, dev, errp);
1848     } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
1849         pc_cpu_plug(hotplug_dev, dev, errp);
1850     }
1851 }
1852 
1853 static void pc_machine_device_unplug_request_cb(HotplugHandler *hotplug_dev,
1854                                                 DeviceState *dev, Error **errp)
1855 {
1856     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
1857         pc_dimm_unplug_request(hotplug_dev, dev, errp);
1858     } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
1859         pc_cpu_unplug_request_cb(hotplug_dev, dev, errp);
1860     } else {
1861         error_setg(errp, "acpi: device unplug request for not supported device"
1862                    " type: %s", object_get_typename(OBJECT(dev)));
1863     }
1864 }
1865 
1866 static void pc_machine_device_unplug_cb(HotplugHandler *hotplug_dev,
1867                                         DeviceState *dev, Error **errp)
1868 {
1869     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
1870         pc_dimm_unplug(hotplug_dev, dev, errp);
1871     } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
1872         pc_cpu_unplug_cb(hotplug_dev, dev, errp);
1873     } else {
1874         error_setg(errp, "acpi: device unplug for not supported device"
1875                    " type: %s", object_get_typename(OBJECT(dev)));
1876     }
1877 }
1878 
1879 static HotplugHandler *pc_get_hotpug_handler(MachineState *machine,
1880                                              DeviceState *dev)
1881 {
1882     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(machine);
1883 
1884     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM) ||
1885         object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
1886         return HOTPLUG_HANDLER(machine);
1887     }
1888 
1889     return pcmc->get_hotplug_handler ?
1890         pcmc->get_hotplug_handler(machine, dev) : NULL;
1891 }
1892 
1893 static void
1894 pc_machine_get_hotplug_memory_region_size(Object *obj, Visitor *v,
1895                                           const char *name, void *opaque,
1896                                           Error **errp)
1897 {
1898     PCMachineState *pcms = PC_MACHINE(obj);
1899     int64_t value = memory_region_size(&pcms->hotplug_memory.mr);
1900 
1901     visit_type_int(v, name, &value, errp);
1902 }
1903 
1904 static void pc_machine_get_max_ram_below_4g(Object *obj, Visitor *v,
1905                                             const char *name, void *opaque,
1906                                             Error **errp)
1907 {
1908     PCMachineState *pcms = PC_MACHINE(obj);
1909     uint64_t value = pcms->max_ram_below_4g;
1910 
1911     visit_type_size(v, name, &value, errp);
1912 }
1913 
1914 static void pc_machine_set_max_ram_below_4g(Object *obj, Visitor *v,
1915                                             const char *name, void *opaque,
1916                                             Error **errp)
1917 {
1918     PCMachineState *pcms = PC_MACHINE(obj);
1919     Error *error = NULL;
1920     uint64_t value;
1921 
1922     visit_type_size(v, name, &value, &error);
1923     if (error) {
1924         error_propagate(errp, error);
1925         return;
1926     }
1927     if (value > (1ULL << 32)) {
1928         error_setg(&error,
1929                    "Machine option 'max-ram-below-4g=%"PRIu64
1930                    "' expects size less than or equal to 4G", value);
1931         error_propagate(errp, error);
1932         return;
1933     }
1934 
1935     if (value < (1ULL << 20)) {
1936         error_report("Warning: small max_ram_below_4g(%"PRIu64
1937                      ") less than 1M.  BIOS may not work..",
1938                      value);
1939     }
1940 
1941     pcms->max_ram_below_4g = value;
1942 }
1943 
1944 static void pc_machine_get_vmport(Object *obj, Visitor *v, const char *name,
1945                                   void *opaque, Error **errp)
1946 {
1947     PCMachineState *pcms = PC_MACHINE(obj);
1948     OnOffAuto vmport = pcms->vmport;
1949 
1950     visit_type_OnOffAuto(v, name, &vmport, errp);
1951 }
1952 
1953 static void pc_machine_set_vmport(Object *obj, Visitor *v, const char *name,
1954                                   void *opaque, Error **errp)
1955 {
1956     PCMachineState *pcms = PC_MACHINE(obj);
1957 
1958     visit_type_OnOffAuto(v, name, &pcms->vmport, errp);
1959 }
1960 
1961 bool pc_machine_is_smm_enabled(PCMachineState *pcms)
1962 {
1963     bool smm_available = false;
1964 
1965     if (pcms->smm == ON_OFF_AUTO_OFF) {
1966         return false;
1967     }
1968 
1969     if (tcg_enabled() || qtest_enabled()) {
1970         smm_available = true;
1971     } else if (kvm_enabled()) {
1972         smm_available = kvm_has_smm();
1973     }
1974 
1975     if (smm_available) {
1976         return true;
1977     }
1978 
1979     if (pcms->smm == ON_OFF_AUTO_ON) {
1980         error_report("System Management Mode not supported by this hypervisor.");
1981         exit(1);
1982     }
1983     return false;
1984 }
1985 
1986 static void pc_machine_get_smm(Object *obj, Visitor *v, const char *name,
1987                                void *opaque, Error **errp)
1988 {
1989     PCMachineState *pcms = PC_MACHINE(obj);
1990     OnOffAuto smm = pcms->smm;
1991 
1992     visit_type_OnOffAuto(v, name, &smm, errp);
1993 }
1994 
1995 static void pc_machine_set_smm(Object *obj, Visitor *v, const char *name,
1996                                void *opaque, Error **errp)
1997 {
1998     PCMachineState *pcms = PC_MACHINE(obj);
1999 
2000     visit_type_OnOffAuto(v, name, &pcms->smm, errp);
2001 }
2002 
2003 static bool pc_machine_get_nvdimm(Object *obj, Error **errp)
2004 {
2005     PCMachineState *pcms = PC_MACHINE(obj);
2006 
2007     return pcms->acpi_nvdimm_state.is_enabled;
2008 }
2009 
2010 static void pc_machine_set_nvdimm(Object *obj, bool value, Error **errp)
2011 {
2012     PCMachineState *pcms = PC_MACHINE(obj);
2013 
2014     pcms->acpi_nvdimm_state.is_enabled = value;
2015 }
2016 
2017 static void pc_machine_initfn(Object *obj)
2018 {
2019     PCMachineState *pcms = PC_MACHINE(obj);
2020 
2021     object_property_add(obj, PC_MACHINE_MEMHP_REGION_SIZE, "int",
2022                         pc_machine_get_hotplug_memory_region_size,
2023                         NULL, NULL, NULL, &error_abort);
2024 
2025     pcms->max_ram_below_4g = 0; /* use default */
2026     object_property_add(obj, PC_MACHINE_MAX_RAM_BELOW_4G, "size",
2027                         pc_machine_get_max_ram_below_4g,
2028                         pc_machine_set_max_ram_below_4g,
2029                         NULL, NULL, &error_abort);
2030     object_property_set_description(obj, PC_MACHINE_MAX_RAM_BELOW_4G,
2031                                     "Maximum ram below the 4G boundary (32bit boundary)",
2032                                     &error_abort);
2033 
2034     pcms->smm = ON_OFF_AUTO_AUTO;
2035     object_property_add(obj, PC_MACHINE_SMM, "OnOffAuto",
2036                         pc_machine_get_smm,
2037                         pc_machine_set_smm,
2038                         NULL, NULL, &error_abort);
2039     object_property_set_description(obj, PC_MACHINE_SMM,
2040                                     "Enable SMM (pc & q35)",
2041                                     &error_abort);
2042 
2043     pcms->vmport = ON_OFF_AUTO_AUTO;
2044     object_property_add(obj, PC_MACHINE_VMPORT, "OnOffAuto",
2045                         pc_machine_get_vmport,
2046                         pc_machine_set_vmport,
2047                         NULL, NULL, &error_abort);
2048     object_property_set_description(obj, PC_MACHINE_VMPORT,
2049                                     "Enable vmport (pc & q35)",
2050                                     &error_abort);
2051 
2052     /* nvdimm is disabled on default. */
2053     pcms->acpi_nvdimm_state.is_enabled = false;
2054     object_property_add_bool(obj, PC_MACHINE_NVDIMM, pc_machine_get_nvdimm,
2055                              pc_machine_set_nvdimm, &error_abort);
2056 }
2057 
2058 static void pc_machine_reset(void)
2059 {
2060     CPUState *cs;
2061     X86CPU *cpu;
2062 
2063     qemu_devices_reset();
2064 
2065     /* Reset APIC after devices have been reset to cancel
2066      * any changes that qemu_devices_reset() might have done.
2067      */
2068     CPU_FOREACH(cs) {
2069         cpu = X86_CPU(cs);
2070 
2071         if (cpu->apic_state) {
2072             device_reset(cpu->apic_state);
2073         }
2074     }
2075 }
2076 
2077 static unsigned pc_cpu_index_to_socket_id(unsigned cpu_index)
2078 {
2079     X86CPUTopoInfo topo;
2080     x86_topo_ids_from_idx(smp_cores, smp_threads, cpu_index,
2081                           &topo);
2082     return topo.pkg_id;
2083 }
2084 
2085 static CPUArchIdList *pc_possible_cpu_arch_ids(MachineState *machine)
2086 {
2087     PCMachineState *pcms = PC_MACHINE(machine);
2088     int len = sizeof(CPUArchIdList) +
2089               sizeof(CPUArchId) * (pcms->possible_cpus->len);
2090     CPUArchIdList *list = g_malloc(len);
2091 
2092     memcpy(list, pcms->possible_cpus, len);
2093     return list;
2094 }
2095 
2096 static void x86_nmi(NMIState *n, int cpu_index, Error **errp)
2097 {
2098     /* cpu index isn't used */
2099     CPUState *cs;
2100 
2101     CPU_FOREACH(cs) {
2102         X86CPU *cpu = X86_CPU(cs);
2103 
2104         if (!cpu->apic_state) {
2105             cpu_interrupt(cs, CPU_INTERRUPT_NMI);
2106         } else {
2107             apic_deliver_nmi(cpu->apic_state);
2108         }
2109     }
2110 }
2111 
2112 static void pc_machine_class_init(ObjectClass *oc, void *data)
2113 {
2114     MachineClass *mc = MACHINE_CLASS(oc);
2115     PCMachineClass *pcmc = PC_MACHINE_CLASS(oc);
2116     HotplugHandlerClass *hc = HOTPLUG_HANDLER_CLASS(oc);
2117     NMIClass *nc = NMI_CLASS(oc);
2118 
2119     pcmc->get_hotplug_handler = mc->get_hotplug_handler;
2120     pcmc->pci_enabled = true;
2121     pcmc->has_acpi_build = true;
2122     pcmc->rsdp_in_ram = true;
2123     pcmc->smbios_defaults = true;
2124     pcmc->smbios_uuid_encoded = true;
2125     pcmc->gigabyte_align = true;
2126     pcmc->has_reserved_memory = true;
2127     pcmc->kvmclock_enabled = true;
2128     pcmc->enforce_aligned_dimm = true;
2129     /* BIOS ACPI tables: 128K. Other BIOS datastructures: less than 4K reported
2130      * to be used at the moment, 32K should be enough for a while.  */
2131     pcmc->acpi_data_size = 0x20000 + 0x8000;
2132     pcmc->save_tsc_khz = true;
2133     mc->get_hotplug_handler = pc_get_hotpug_handler;
2134     mc->cpu_index_to_socket_id = pc_cpu_index_to_socket_id;
2135     mc->possible_cpu_arch_ids = pc_possible_cpu_arch_ids;
2136     mc->default_boot_order = "cad";
2137     mc->hot_add_cpu = pc_hot_add_cpu;
2138     mc->max_cpus = 255;
2139     mc->reset = pc_machine_reset;
2140     hc->plug = pc_machine_device_plug_cb;
2141     hc->unplug_request = pc_machine_device_unplug_request_cb;
2142     hc->unplug = pc_machine_device_unplug_cb;
2143     nc->nmi_monitor_handler = x86_nmi;
2144 }
2145 
2146 static const TypeInfo pc_machine_info = {
2147     .name = TYPE_PC_MACHINE,
2148     .parent = TYPE_MACHINE,
2149     .abstract = true,
2150     .instance_size = sizeof(PCMachineState),
2151     .instance_init = pc_machine_initfn,
2152     .class_size = sizeof(PCMachineClass),
2153     .class_init = pc_machine_class_init,
2154     .interfaces = (InterfaceInfo[]) {
2155          { TYPE_HOTPLUG_HANDLER },
2156          { TYPE_NMI },
2157          { }
2158     },
2159 };
2160 
2161 static void pc_machine_register_types(void)
2162 {
2163     type_register_static(&pc_machine_info);
2164 }
2165 
2166 type_init(pc_machine_register_types)
2167