xref: /openbmc/qemu/hw/i386/pc.c (revision 1602001195dca96aaea8b16f740ac860238555a5)
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     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 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     pc_build_smbios(fw_cfg);
769 
770     fw_cfg_add_bytes(fw_cfg, FW_CFG_E820_TABLE,
771                      &e820_reserve, sizeof(e820_reserve));
772     fw_cfg_add_file(fw_cfg, "etc/e820", e820_table,
773                     sizeof(struct e820_entry) * e820_entries);
774 
775     fw_cfg_add_bytes(fw_cfg, FW_CFG_HPET, &hpet_cfg, sizeof(hpet_cfg));
776     /* allocate memory for the NUMA channel: one (64bit) word for the number
777      * of nodes, one word for each VCPU->node and one word for each node to
778      * hold the amount of memory.
779      */
780     numa_fw_cfg = g_new0(uint64_t, 1 + pcms->apic_id_limit + nb_numa_nodes);
781     numa_fw_cfg[0] = cpu_to_le64(nb_numa_nodes);
782     for (i = 0; i < max_cpus; i++) {
783         unsigned int apic_id = x86_cpu_apic_id_from_index(i);
784         assert(apic_id < pcms->apic_id_limit);
785         for (j = 0; j < nb_numa_nodes; j++) {
786             if (test_bit(i, numa_info[j].node_cpu)) {
787                 numa_fw_cfg[apic_id + 1] = cpu_to_le64(j);
788                 break;
789             }
790         }
791     }
792     for (i = 0; i < nb_numa_nodes; i++) {
793         numa_fw_cfg[pcms->apic_id_limit + 1 + i] =
794             cpu_to_le64(numa_info[i].node_mem);
795     }
796     fw_cfg_add_bytes(fw_cfg, FW_CFG_NUMA, numa_fw_cfg,
797                      (1 + pcms->apic_id_limit + nb_numa_nodes) *
798                      sizeof(*numa_fw_cfg));
799 
800     return fw_cfg;
801 }
802 
803 static long get_file_size(FILE *f)
804 {
805     long where, size;
806 
807     /* XXX: on Unix systems, using fstat() probably makes more sense */
808 
809     where = ftell(f);
810     fseek(f, 0, SEEK_END);
811     size = ftell(f);
812     fseek(f, where, SEEK_SET);
813 
814     return size;
815 }
816 
817 static void load_linux(PCMachineState *pcms,
818                        FWCfgState *fw_cfg)
819 {
820     uint16_t protocol;
821     int setup_size, kernel_size, initrd_size = 0, cmdline_size;
822     uint32_t initrd_max;
823     uint8_t header[8192], *setup, *kernel, *initrd_data;
824     hwaddr real_addr, prot_addr, cmdline_addr, initrd_addr = 0;
825     FILE *f;
826     char *vmode;
827     MachineState *machine = MACHINE(pcms);
828     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
829     const char *kernel_filename = machine->kernel_filename;
830     const char *initrd_filename = machine->initrd_filename;
831     const char *kernel_cmdline = machine->kernel_cmdline;
832 
833     /* Align to 16 bytes as a paranoia measure */
834     cmdline_size = (strlen(kernel_cmdline)+16) & ~15;
835 
836     /* load the kernel header */
837     f = fopen(kernel_filename, "rb");
838     if (!f || !(kernel_size = get_file_size(f)) ||
839         fread(header, 1, MIN(ARRAY_SIZE(header), kernel_size), f) !=
840         MIN(ARRAY_SIZE(header), kernel_size)) {
841         fprintf(stderr, "qemu: could not load kernel '%s': %s\n",
842                 kernel_filename, strerror(errno));
843         exit(1);
844     }
845 
846     /* kernel protocol version */
847 #if 0
848     fprintf(stderr, "header magic: %#x\n", ldl_p(header+0x202));
849 #endif
850     if (ldl_p(header+0x202) == 0x53726448) {
851         protocol = lduw_p(header+0x206);
852     } else {
853         /* This looks like a multiboot kernel. If it is, let's stop
854            treating it like a Linux kernel. */
855         if (load_multiboot(fw_cfg, f, kernel_filename, initrd_filename,
856                            kernel_cmdline, kernel_size, header)) {
857             return;
858         }
859         protocol = 0;
860     }
861 
862     if (protocol < 0x200 || !(header[0x211] & 0x01)) {
863         /* Low kernel */
864         real_addr    = 0x90000;
865         cmdline_addr = 0x9a000 - cmdline_size;
866         prot_addr    = 0x10000;
867     } else if (protocol < 0x202) {
868         /* High but ancient kernel */
869         real_addr    = 0x90000;
870         cmdline_addr = 0x9a000 - cmdline_size;
871         prot_addr    = 0x100000;
872     } else {
873         /* High and recent kernel */
874         real_addr    = 0x10000;
875         cmdline_addr = 0x20000;
876         prot_addr    = 0x100000;
877     }
878 
879 #if 0
880     fprintf(stderr,
881             "qemu: real_addr     = 0x" TARGET_FMT_plx "\n"
882             "qemu: cmdline_addr  = 0x" TARGET_FMT_plx "\n"
883             "qemu: prot_addr     = 0x" TARGET_FMT_plx "\n",
884             real_addr,
885             cmdline_addr,
886             prot_addr);
887 #endif
888 
889     /* highest address for loading the initrd */
890     if (protocol >= 0x203) {
891         initrd_max = ldl_p(header+0x22c);
892     } else {
893         initrd_max = 0x37ffffff;
894     }
895 
896     if (initrd_max >= pcms->below_4g_mem_size - pcmc->acpi_data_size) {
897         initrd_max = pcms->below_4g_mem_size - pcmc->acpi_data_size - 1;
898     }
899 
900     fw_cfg_add_i32(fw_cfg, FW_CFG_CMDLINE_ADDR, cmdline_addr);
901     fw_cfg_add_i32(fw_cfg, FW_CFG_CMDLINE_SIZE, strlen(kernel_cmdline)+1);
902     fw_cfg_add_string(fw_cfg, FW_CFG_CMDLINE_DATA, kernel_cmdline);
903 
904     if (protocol >= 0x202) {
905         stl_p(header+0x228, cmdline_addr);
906     } else {
907         stw_p(header+0x20, 0xA33F);
908         stw_p(header+0x22, cmdline_addr-real_addr);
909     }
910 
911     /* handle vga= parameter */
912     vmode = strstr(kernel_cmdline, "vga=");
913     if (vmode) {
914         unsigned int video_mode;
915         /* skip "vga=" */
916         vmode += 4;
917         if (!strncmp(vmode, "normal", 6)) {
918             video_mode = 0xffff;
919         } else if (!strncmp(vmode, "ext", 3)) {
920             video_mode = 0xfffe;
921         } else if (!strncmp(vmode, "ask", 3)) {
922             video_mode = 0xfffd;
923         } else {
924             video_mode = strtol(vmode, NULL, 0);
925         }
926         stw_p(header+0x1fa, video_mode);
927     }
928 
929     /* loader type */
930     /* High nybble = B reserved for QEMU; low nybble is revision number.
931        If this code is substantially changed, you may want to consider
932        incrementing the revision. */
933     if (protocol >= 0x200) {
934         header[0x210] = 0xB0;
935     }
936     /* heap */
937     if (protocol >= 0x201) {
938         header[0x211] |= 0x80;	/* CAN_USE_HEAP */
939         stw_p(header+0x224, cmdline_addr-real_addr-0x200);
940     }
941 
942     /* load initrd */
943     if (initrd_filename) {
944         if (protocol < 0x200) {
945             fprintf(stderr, "qemu: linux kernel too old to load a ram disk\n");
946             exit(1);
947         }
948 
949         initrd_size = get_image_size(initrd_filename);
950         if (initrd_size < 0) {
951             fprintf(stderr, "qemu: error reading initrd %s: %s\n",
952                     initrd_filename, strerror(errno));
953             exit(1);
954         }
955 
956         initrd_addr = (initrd_max-initrd_size) & ~4095;
957 
958         initrd_data = g_malloc(initrd_size);
959         load_image(initrd_filename, initrd_data);
960 
961         fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_ADDR, initrd_addr);
962         fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_SIZE, initrd_size);
963         fw_cfg_add_bytes(fw_cfg, FW_CFG_INITRD_DATA, initrd_data, initrd_size);
964 
965         stl_p(header+0x218, initrd_addr);
966         stl_p(header+0x21c, initrd_size);
967     }
968 
969     /* load kernel and setup */
970     setup_size = header[0x1f1];
971     if (setup_size == 0) {
972         setup_size = 4;
973     }
974     setup_size = (setup_size+1)*512;
975     if (setup_size > kernel_size) {
976         fprintf(stderr, "qemu: invalid kernel header\n");
977         exit(1);
978     }
979     kernel_size -= setup_size;
980 
981     setup  = g_malloc(setup_size);
982     kernel = g_malloc(kernel_size);
983     fseek(f, 0, SEEK_SET);
984     if (fread(setup, 1, setup_size, f) != setup_size) {
985         fprintf(stderr, "fread() failed\n");
986         exit(1);
987     }
988     if (fread(kernel, 1, kernel_size, f) != kernel_size) {
989         fprintf(stderr, "fread() failed\n");
990         exit(1);
991     }
992     fclose(f);
993     memcpy(setup, header, MIN(sizeof(header), setup_size));
994 
995     fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_ADDR, prot_addr);
996     fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_SIZE, kernel_size);
997     fw_cfg_add_bytes(fw_cfg, FW_CFG_KERNEL_DATA, kernel, kernel_size);
998 
999     fw_cfg_add_i32(fw_cfg, FW_CFG_SETUP_ADDR, real_addr);
1000     fw_cfg_add_i32(fw_cfg, FW_CFG_SETUP_SIZE, setup_size);
1001     fw_cfg_add_bytes(fw_cfg, FW_CFG_SETUP_DATA, setup, setup_size);
1002 
1003     option_rom[nb_option_roms].name = "linuxboot.bin";
1004     option_rom[nb_option_roms].bootindex = 0;
1005     nb_option_roms++;
1006 }
1007 
1008 #define NE2000_NB_MAX 6
1009 
1010 static const int ne2000_io[NE2000_NB_MAX] = { 0x300, 0x320, 0x340, 0x360,
1011                                               0x280, 0x380 };
1012 static const int ne2000_irq[NE2000_NB_MAX] = { 9, 10, 11, 3, 4, 5 };
1013 
1014 void pc_init_ne2k_isa(ISABus *bus, NICInfo *nd)
1015 {
1016     static int nb_ne2k = 0;
1017 
1018     if (nb_ne2k == NE2000_NB_MAX)
1019         return;
1020     isa_ne2000_init(bus, ne2000_io[nb_ne2k],
1021                     ne2000_irq[nb_ne2k], nd);
1022     nb_ne2k++;
1023 }
1024 
1025 DeviceState *cpu_get_current_apic(void)
1026 {
1027     if (current_cpu) {
1028         X86CPU *cpu = X86_CPU(current_cpu);
1029         return cpu->apic_state;
1030     } else {
1031         return NULL;
1032     }
1033 }
1034 
1035 void pc_acpi_smi_interrupt(void *opaque, int irq, int level)
1036 {
1037     X86CPU *cpu = opaque;
1038 
1039     if (level) {
1040         cpu_interrupt(CPU(cpu), CPU_INTERRUPT_SMI);
1041     }
1042 }
1043 
1044 static X86CPU *pc_new_cpu(const char *cpu_model, int64_t apic_id,
1045                           Error **errp)
1046 {
1047     X86CPU *cpu = NULL;
1048     Error *local_err = NULL;
1049 
1050     cpu = cpu_x86_create(cpu_model, &local_err);
1051     if (local_err != NULL) {
1052         goto out;
1053     }
1054 
1055     object_property_set_int(OBJECT(cpu), apic_id, "apic-id", &local_err);
1056     object_property_set_bool(OBJECT(cpu), true, "realized", &local_err);
1057 
1058 out:
1059     if (local_err) {
1060         error_propagate(errp, local_err);
1061         object_unref(OBJECT(cpu));
1062         cpu = NULL;
1063     }
1064     return cpu;
1065 }
1066 
1067 void pc_hot_add_cpu(const int64_t id, Error **errp)
1068 {
1069     X86CPU *cpu;
1070     MachineState *machine = MACHINE(qdev_get_machine());
1071     int64_t apic_id = x86_cpu_apic_id_from_index(id);
1072     Error *local_err = NULL;
1073 
1074     if (id < 0) {
1075         error_setg(errp, "Invalid CPU id: %" PRIi64, id);
1076         return;
1077     }
1078 
1079     if (cpu_exists(apic_id)) {
1080         error_setg(errp, "Unable to add CPU: %" PRIi64
1081                    ", it already exists", id);
1082         return;
1083     }
1084 
1085     if (id >= max_cpus) {
1086         error_setg(errp, "Unable to add CPU: %" PRIi64
1087                    ", max allowed: %d", id, max_cpus - 1);
1088         return;
1089     }
1090 
1091     if (apic_id >= ACPI_CPU_HOTPLUG_ID_LIMIT) {
1092         error_setg(errp, "Unable to add CPU: %" PRIi64
1093                    ", resulting APIC ID (%" PRIi64 ") is too large",
1094                    id, apic_id);
1095         return;
1096     }
1097 
1098     cpu = pc_new_cpu(machine->cpu_model, apic_id, &local_err);
1099     if (local_err) {
1100         error_propagate(errp, local_err);
1101         return;
1102     }
1103     object_unref(OBJECT(cpu));
1104 }
1105 
1106 void pc_cpus_init(PCMachineState *pcms)
1107 {
1108     int i;
1109     X86CPU *cpu = NULL;
1110     MachineState *machine = MACHINE(pcms);
1111 
1112     /* init CPUs */
1113     if (machine->cpu_model == NULL) {
1114 #ifdef TARGET_X86_64
1115         machine->cpu_model = "qemu64";
1116 #else
1117         machine->cpu_model = "qemu32";
1118 #endif
1119     }
1120 
1121     /* Calculates the limit to CPU APIC ID values
1122      *
1123      * Limit for the APIC ID value, so that all
1124      * CPU APIC IDs are < pcms->apic_id_limit.
1125      *
1126      * This is used for FW_CFG_MAX_CPUS. See comments on bochs_bios_init().
1127      */
1128     pcms->apic_id_limit = x86_cpu_apic_id_from_index(max_cpus - 1) + 1;
1129     if (pcms->apic_id_limit > ACPI_CPU_HOTPLUG_ID_LIMIT) {
1130         error_report("max_cpus is too large. APIC ID of last CPU is %u",
1131                      pcms->apic_id_limit - 1);
1132         exit(1);
1133     }
1134 
1135     pcms->possible_cpus = g_malloc0(sizeof(CPUArchIdList) +
1136                                     sizeof(CPUArchId) * max_cpus);
1137     for (i = 0; i < max_cpus; i++) {
1138         pcms->possible_cpus->cpus[i].arch_id = x86_cpu_apic_id_from_index(i);
1139         pcms->possible_cpus->len++;
1140         if (i < smp_cpus) {
1141             cpu = pc_new_cpu(machine->cpu_model, x86_cpu_apic_id_from_index(i),
1142                              &error_fatal);
1143             pcms->possible_cpus->cpus[i].cpu = CPU(cpu);
1144             object_unref(OBJECT(cpu));
1145         }
1146     }
1147 
1148     /* tell smbios about cpuid version and features */
1149     smbios_set_cpuid(cpu->env.cpuid_version, cpu->env.features[FEAT_1_EDX]);
1150 }
1151 
1152 /* pci-info ROM file. Little endian format */
1153 typedef struct PcRomPciInfo {
1154     uint64_t w32_min;
1155     uint64_t w32_max;
1156     uint64_t w64_min;
1157     uint64_t w64_max;
1158 } PcRomPciInfo;
1159 
1160 static
1161 void pc_machine_done(Notifier *notifier, void *data)
1162 {
1163     PCMachineState *pcms = container_of(notifier,
1164                                         PCMachineState, machine_done);
1165     PCIBus *bus = pcms->bus;
1166 
1167     if (bus) {
1168         int extra_hosts = 0;
1169 
1170         QLIST_FOREACH(bus, &bus->child, sibling) {
1171             /* look for expander root buses */
1172             if (pci_bus_is_root(bus)) {
1173                 extra_hosts++;
1174             }
1175         }
1176         if (extra_hosts && pcms->fw_cfg) {
1177             uint64_t *val = g_malloc(sizeof(*val));
1178             *val = cpu_to_le64(extra_hosts);
1179             fw_cfg_add_file(pcms->fw_cfg,
1180                     "etc/extra-pci-roots", val, sizeof(*val));
1181         }
1182     }
1183 
1184     acpi_setup();
1185 }
1186 
1187 void pc_guest_info_init(PCMachineState *pcms)
1188 {
1189     int i, j;
1190 
1191     pcms->apic_xrupt_override = kvm_allows_irq0_override();
1192     pcms->numa_nodes = nb_numa_nodes;
1193     pcms->node_mem = g_malloc0(pcms->numa_nodes *
1194                                     sizeof *pcms->node_mem);
1195     for (i = 0; i < nb_numa_nodes; i++) {
1196         pcms->node_mem[i] = numa_info[i].node_mem;
1197     }
1198 
1199     pcms->node_cpu = g_malloc0(pcms->apic_id_limit *
1200                                      sizeof *pcms->node_cpu);
1201 
1202     for (i = 0; i < max_cpus; i++) {
1203         unsigned int apic_id = x86_cpu_apic_id_from_index(i);
1204         assert(apic_id < pcms->apic_id_limit);
1205         for (j = 0; j < nb_numa_nodes; j++) {
1206             if (test_bit(i, numa_info[j].node_cpu)) {
1207                 pcms->node_cpu[apic_id] = j;
1208                 break;
1209             }
1210         }
1211     }
1212 
1213     pcms->machine_done.notify = pc_machine_done;
1214     qemu_add_machine_init_done_notifier(&pcms->machine_done);
1215 }
1216 
1217 /* setup pci memory address space mapping into system address space */
1218 void pc_pci_as_mapping_init(Object *owner, MemoryRegion *system_memory,
1219                             MemoryRegion *pci_address_space)
1220 {
1221     /* Set to lower priority than RAM */
1222     memory_region_add_subregion_overlap(system_memory, 0x0,
1223                                         pci_address_space, -1);
1224 }
1225 
1226 void pc_acpi_init(const char *default_dsdt)
1227 {
1228     char *filename;
1229 
1230     if (acpi_tables != NULL) {
1231         /* manually set via -acpitable, leave it alone */
1232         return;
1233     }
1234 
1235     filename = qemu_find_file(QEMU_FILE_TYPE_BIOS, default_dsdt);
1236     if (filename == NULL) {
1237         fprintf(stderr, "WARNING: failed to find %s\n", default_dsdt);
1238     } else {
1239         QemuOpts *opts = qemu_opts_create(qemu_find_opts("acpi"), NULL, 0,
1240                                           &error_abort);
1241         Error *err = NULL;
1242 
1243         qemu_opt_set(opts, "file", filename, &error_abort);
1244 
1245         acpi_table_add_builtin(opts, &err);
1246         if (err) {
1247             error_reportf_err(err, "WARNING: failed to load %s: ",
1248                               filename);
1249         }
1250         g_free(filename);
1251     }
1252 }
1253 
1254 void xen_load_linux(PCMachineState *pcms)
1255 {
1256     int i;
1257     FWCfgState *fw_cfg;
1258 
1259     assert(MACHINE(pcms)->kernel_filename != NULL);
1260 
1261     fw_cfg = fw_cfg_init_io(FW_CFG_IO_BASE);
1262     rom_set_fw(fw_cfg);
1263 
1264     load_linux(pcms, fw_cfg);
1265     for (i = 0; i < nb_option_roms; i++) {
1266         assert(!strcmp(option_rom[i].name, "linuxboot.bin") ||
1267                !strcmp(option_rom[i].name, "multiboot.bin"));
1268         rom_add_option(option_rom[i].name, option_rom[i].bootindex);
1269     }
1270     pcms->fw_cfg = fw_cfg;
1271 }
1272 
1273 void pc_memory_init(PCMachineState *pcms,
1274                     MemoryRegion *system_memory,
1275                     MemoryRegion *rom_memory,
1276                     MemoryRegion **ram_memory)
1277 {
1278     int linux_boot, i;
1279     MemoryRegion *ram, *option_rom_mr;
1280     MemoryRegion *ram_below_4g, *ram_above_4g;
1281     FWCfgState *fw_cfg;
1282     MachineState *machine = MACHINE(pcms);
1283     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
1284 
1285     assert(machine->ram_size == pcms->below_4g_mem_size +
1286                                 pcms->above_4g_mem_size);
1287 
1288     linux_boot = (machine->kernel_filename != NULL);
1289 
1290     /* Allocate RAM.  We allocate it as a single memory region and use
1291      * aliases to address portions of it, mostly for backwards compatibility
1292      * with older qemus that used qemu_ram_alloc().
1293      */
1294     ram = g_malloc(sizeof(*ram));
1295     memory_region_allocate_system_memory(ram, NULL, "pc.ram",
1296                                          machine->ram_size);
1297     *ram_memory = ram;
1298     ram_below_4g = g_malloc(sizeof(*ram_below_4g));
1299     memory_region_init_alias(ram_below_4g, NULL, "ram-below-4g", ram,
1300                              0, pcms->below_4g_mem_size);
1301     memory_region_add_subregion(system_memory, 0, ram_below_4g);
1302     e820_add_entry(0, pcms->below_4g_mem_size, E820_RAM);
1303     if (pcms->above_4g_mem_size > 0) {
1304         ram_above_4g = g_malloc(sizeof(*ram_above_4g));
1305         memory_region_init_alias(ram_above_4g, NULL, "ram-above-4g", ram,
1306                                  pcms->below_4g_mem_size,
1307                                  pcms->above_4g_mem_size);
1308         memory_region_add_subregion(system_memory, 0x100000000ULL,
1309                                     ram_above_4g);
1310         e820_add_entry(0x100000000ULL, pcms->above_4g_mem_size, E820_RAM);
1311     }
1312 
1313     if (!pcmc->has_reserved_memory &&
1314         (machine->ram_slots ||
1315          (machine->maxram_size > machine->ram_size))) {
1316         MachineClass *mc = MACHINE_GET_CLASS(machine);
1317 
1318         error_report("\"-memory 'slots|maxmem'\" is not supported by: %s",
1319                      mc->name);
1320         exit(EXIT_FAILURE);
1321     }
1322 
1323     /* initialize hotplug memory address space */
1324     if (pcmc->has_reserved_memory &&
1325         (machine->ram_size < machine->maxram_size)) {
1326         ram_addr_t hotplug_mem_size =
1327             machine->maxram_size - machine->ram_size;
1328 
1329         if (machine->ram_slots > ACPI_MAX_RAM_SLOTS) {
1330             error_report("unsupported amount of memory slots: %"PRIu64,
1331                          machine->ram_slots);
1332             exit(EXIT_FAILURE);
1333         }
1334 
1335         if (QEMU_ALIGN_UP(machine->maxram_size,
1336                           TARGET_PAGE_SIZE) != machine->maxram_size) {
1337             error_report("maximum memory size must by aligned to multiple of "
1338                          "%d bytes", TARGET_PAGE_SIZE);
1339             exit(EXIT_FAILURE);
1340         }
1341 
1342         pcms->hotplug_memory.base =
1343             ROUND_UP(0x100000000ULL + pcms->above_4g_mem_size, 1ULL << 30);
1344 
1345         if (pcmc->enforce_aligned_dimm) {
1346             /* size hotplug region assuming 1G page max alignment per slot */
1347             hotplug_mem_size += (1ULL << 30) * machine->ram_slots;
1348         }
1349 
1350         if ((pcms->hotplug_memory.base + hotplug_mem_size) <
1351             hotplug_mem_size) {
1352             error_report("unsupported amount of maximum memory: " RAM_ADDR_FMT,
1353                          machine->maxram_size);
1354             exit(EXIT_FAILURE);
1355         }
1356 
1357         memory_region_init(&pcms->hotplug_memory.mr, OBJECT(pcms),
1358                            "hotplug-memory", hotplug_mem_size);
1359         memory_region_add_subregion(system_memory, pcms->hotplug_memory.base,
1360                                     &pcms->hotplug_memory.mr);
1361     }
1362 
1363     /* Initialize PC system firmware */
1364     pc_system_firmware_init(rom_memory, !pcmc->pci_enabled);
1365 
1366     option_rom_mr = g_malloc(sizeof(*option_rom_mr));
1367     memory_region_init_ram(option_rom_mr, NULL, "pc.rom", PC_ROM_SIZE,
1368                            &error_fatal);
1369     vmstate_register_ram_global(option_rom_mr);
1370     memory_region_add_subregion_overlap(rom_memory,
1371                                         PC_ROM_MIN_VGA,
1372                                         option_rom_mr,
1373                                         1);
1374 
1375     fw_cfg = bochs_bios_init(&address_space_memory, pcms);
1376 
1377     rom_set_fw(fw_cfg);
1378 
1379     if (pcmc->has_reserved_memory && pcms->hotplug_memory.base) {
1380         uint64_t *val = g_malloc(sizeof(*val));
1381         PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
1382         uint64_t res_mem_end = pcms->hotplug_memory.base;
1383 
1384         if (!pcmc->broken_reserved_end) {
1385             res_mem_end += memory_region_size(&pcms->hotplug_memory.mr);
1386         }
1387         *val = cpu_to_le64(ROUND_UP(res_mem_end, 0x1ULL << 30));
1388         fw_cfg_add_file(fw_cfg, "etc/reserved-memory-end", val, sizeof(*val));
1389     }
1390 
1391     if (linux_boot) {
1392         load_linux(pcms, fw_cfg);
1393     }
1394 
1395     for (i = 0; i < nb_option_roms; i++) {
1396         rom_add_option(option_rom[i].name, option_rom[i].bootindex);
1397     }
1398     pcms->fw_cfg = fw_cfg;
1399 }
1400 
1401 qemu_irq pc_allocate_cpu_irq(void)
1402 {
1403     return qemu_allocate_irq(pic_irq_request, NULL, 0);
1404 }
1405 
1406 DeviceState *pc_vga_init(ISABus *isa_bus, PCIBus *pci_bus)
1407 {
1408     DeviceState *dev = NULL;
1409 
1410     rom_set_order_override(FW_CFG_ORDER_OVERRIDE_VGA);
1411     if (pci_bus) {
1412         PCIDevice *pcidev = pci_vga_init(pci_bus);
1413         dev = pcidev ? &pcidev->qdev : NULL;
1414     } else if (isa_bus) {
1415         ISADevice *isadev = isa_vga_init(isa_bus);
1416         dev = isadev ? DEVICE(isadev) : NULL;
1417     }
1418     rom_reset_order_override();
1419     return dev;
1420 }
1421 
1422 static const MemoryRegionOps ioport80_io_ops = {
1423     .write = ioport80_write,
1424     .read = ioport80_read,
1425     .endianness = DEVICE_NATIVE_ENDIAN,
1426     .impl = {
1427         .min_access_size = 1,
1428         .max_access_size = 1,
1429     },
1430 };
1431 
1432 static const MemoryRegionOps ioportF0_io_ops = {
1433     .write = ioportF0_write,
1434     .read = ioportF0_read,
1435     .endianness = DEVICE_NATIVE_ENDIAN,
1436     .impl = {
1437         .min_access_size = 1,
1438         .max_access_size = 1,
1439     },
1440 };
1441 
1442 void pc_basic_device_init(ISABus *isa_bus, qemu_irq *gsi,
1443                           ISADevice **rtc_state,
1444                           bool create_fdctrl,
1445                           bool no_vmport,
1446                           uint32_t hpet_irqs)
1447 {
1448     int i;
1449     DriveInfo *fd[MAX_FD];
1450     DeviceState *hpet = NULL;
1451     int pit_isa_irq = 0;
1452     qemu_irq pit_alt_irq = NULL;
1453     qemu_irq rtc_irq = NULL;
1454     qemu_irq *a20_line;
1455     ISADevice *i8042, *port92, *vmmouse, *pit = NULL;
1456     MemoryRegion *ioport80_io = g_new(MemoryRegion, 1);
1457     MemoryRegion *ioportF0_io = g_new(MemoryRegion, 1);
1458 
1459     memory_region_init_io(ioport80_io, NULL, &ioport80_io_ops, NULL, "ioport80", 1);
1460     memory_region_add_subregion(isa_bus->address_space_io, 0x80, ioport80_io);
1461 
1462     memory_region_init_io(ioportF0_io, NULL, &ioportF0_io_ops, NULL, "ioportF0", 1);
1463     memory_region_add_subregion(isa_bus->address_space_io, 0xf0, ioportF0_io);
1464 
1465     /*
1466      * Check if an HPET shall be created.
1467      *
1468      * Without KVM_CAP_PIT_STATE2, we cannot switch off the in-kernel PIT
1469      * when the HPET wants to take over. Thus we have to disable the latter.
1470      */
1471     if (!no_hpet && (!kvm_irqchip_in_kernel() || kvm_has_pit_state2())) {
1472         /* In order to set property, here not using sysbus_try_create_simple */
1473         hpet = qdev_try_create(NULL, TYPE_HPET);
1474         if (hpet) {
1475             /* For pc-piix-*, hpet's intcap is always IRQ2. For pc-q35-1.7
1476              * and earlier, use IRQ2 for compat. Otherwise, use IRQ16~23,
1477              * IRQ8 and IRQ2.
1478              */
1479             uint8_t compat = object_property_get_int(OBJECT(hpet),
1480                     HPET_INTCAP, NULL);
1481             if (!compat) {
1482                 qdev_prop_set_uint32(hpet, HPET_INTCAP, hpet_irqs);
1483             }
1484             qdev_init_nofail(hpet);
1485             sysbus_mmio_map(SYS_BUS_DEVICE(hpet), 0, HPET_BASE);
1486 
1487             for (i = 0; i < GSI_NUM_PINS; i++) {
1488                 sysbus_connect_irq(SYS_BUS_DEVICE(hpet), i, gsi[i]);
1489             }
1490             pit_isa_irq = -1;
1491             pit_alt_irq = qdev_get_gpio_in(hpet, HPET_LEGACY_PIT_INT);
1492             rtc_irq = qdev_get_gpio_in(hpet, HPET_LEGACY_RTC_INT);
1493         }
1494     }
1495     *rtc_state = rtc_init(isa_bus, 2000, rtc_irq);
1496 
1497     qemu_register_boot_set(pc_boot_set, *rtc_state);
1498 
1499     if (!xen_enabled()) {
1500         if (kvm_pit_in_kernel()) {
1501             pit = kvm_pit_init(isa_bus, 0x40);
1502         } else {
1503             pit = pit_init(isa_bus, 0x40, pit_isa_irq, pit_alt_irq);
1504         }
1505         if (hpet) {
1506             /* connect PIT to output control line of the HPET */
1507             qdev_connect_gpio_out(hpet, 0, qdev_get_gpio_in(DEVICE(pit), 0));
1508         }
1509         pcspk_init(isa_bus, pit);
1510     }
1511 
1512     serial_hds_isa_init(isa_bus, MAX_SERIAL_PORTS);
1513     parallel_hds_isa_init(isa_bus, MAX_PARALLEL_PORTS);
1514 
1515     a20_line = qemu_allocate_irqs(handle_a20_line_change, first_cpu, 2);
1516     i8042 = isa_create_simple(isa_bus, "i8042");
1517     i8042_setup_a20_line(i8042, &a20_line[0]);
1518     if (!no_vmport) {
1519         vmport_init(isa_bus);
1520         vmmouse = isa_try_create(isa_bus, "vmmouse");
1521     } else {
1522         vmmouse = NULL;
1523     }
1524     if (vmmouse) {
1525         DeviceState *dev = DEVICE(vmmouse);
1526         qdev_prop_set_ptr(dev, "ps2_mouse", i8042);
1527         qdev_init_nofail(dev);
1528     }
1529     port92 = isa_create_simple(isa_bus, "port92");
1530     port92_init(port92, &a20_line[1]);
1531 
1532     DMA_init(isa_bus, 0);
1533 
1534     for(i = 0; i < MAX_FD; i++) {
1535         fd[i] = drive_get(IF_FLOPPY, 0, i);
1536         create_fdctrl |= !!fd[i];
1537     }
1538     if (create_fdctrl) {
1539         fdctrl_init_isa(isa_bus, fd);
1540     }
1541 }
1542 
1543 void pc_nic_init(ISABus *isa_bus, PCIBus *pci_bus)
1544 {
1545     int i;
1546 
1547     rom_set_order_override(FW_CFG_ORDER_OVERRIDE_NIC);
1548     for (i = 0; i < nb_nics; i++) {
1549         NICInfo *nd = &nd_table[i];
1550 
1551         if (!pci_bus || (nd->model && strcmp(nd->model, "ne2k_isa") == 0)) {
1552             pc_init_ne2k_isa(isa_bus, nd);
1553         } else {
1554             pci_nic_init_nofail(nd, pci_bus, "e1000", NULL);
1555         }
1556     }
1557     rom_reset_order_override();
1558 }
1559 
1560 void pc_pci_device_init(PCIBus *pci_bus)
1561 {
1562     int max_bus;
1563     int bus;
1564 
1565     max_bus = drive_get_max_bus(IF_SCSI);
1566     for (bus = 0; bus <= max_bus; bus++) {
1567         pci_create_simple(pci_bus, -1, "lsi53c895a");
1568     }
1569 }
1570 
1571 void ioapic_init_gsi(GSIState *gsi_state, const char *parent_name)
1572 {
1573     DeviceState *dev;
1574     SysBusDevice *d;
1575     unsigned int i;
1576 
1577     if (kvm_ioapic_in_kernel()) {
1578         dev = qdev_create(NULL, "kvm-ioapic");
1579     } else {
1580         dev = qdev_create(NULL, "ioapic");
1581     }
1582     if (parent_name) {
1583         object_property_add_child(object_resolve_path(parent_name, NULL),
1584                                   "ioapic", OBJECT(dev), NULL);
1585     }
1586     qdev_init_nofail(dev);
1587     d = SYS_BUS_DEVICE(dev);
1588     sysbus_mmio_map(d, 0, IO_APIC_DEFAULT_ADDRESS);
1589 
1590     for (i = 0; i < IOAPIC_NUM_PINS; i++) {
1591         gsi_state->ioapic_irq[i] = qdev_get_gpio_in(dev, i);
1592     }
1593 }
1594 
1595 static void pc_dimm_plug(HotplugHandler *hotplug_dev,
1596                          DeviceState *dev, Error **errp)
1597 {
1598     HotplugHandlerClass *hhc;
1599     Error *local_err = NULL;
1600     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1601     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
1602     PCDIMMDevice *dimm = PC_DIMM(dev);
1603     PCDIMMDeviceClass *ddc = PC_DIMM_GET_CLASS(dimm);
1604     MemoryRegion *mr = ddc->get_memory_region(dimm);
1605     uint64_t align = TARGET_PAGE_SIZE;
1606 
1607     if (memory_region_get_alignment(mr) && pcmc->enforce_aligned_dimm) {
1608         align = memory_region_get_alignment(mr);
1609     }
1610 
1611     if (!pcms->acpi_dev) {
1612         error_setg(&local_err,
1613                    "memory hotplug is not enabled: missing acpi device");
1614         goto out;
1615     }
1616 
1617     pc_dimm_memory_plug(dev, &pcms->hotplug_memory, mr, align, &local_err);
1618     if (local_err) {
1619         goto out;
1620     }
1621 
1622     hhc = HOTPLUG_HANDLER_GET_CLASS(pcms->acpi_dev);
1623     hhc->plug(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &error_abort);
1624 out:
1625     error_propagate(errp, local_err);
1626 }
1627 
1628 static void pc_dimm_unplug_request(HotplugHandler *hotplug_dev,
1629                                    DeviceState *dev, Error **errp)
1630 {
1631     HotplugHandlerClass *hhc;
1632     Error *local_err = NULL;
1633     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1634 
1635     if (!pcms->acpi_dev) {
1636         error_setg(&local_err,
1637                    "memory hotplug is not enabled: missing acpi device");
1638         goto out;
1639     }
1640 
1641     hhc = HOTPLUG_HANDLER_GET_CLASS(pcms->acpi_dev);
1642     hhc->unplug_request(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &local_err);
1643 
1644 out:
1645     error_propagate(errp, local_err);
1646 }
1647 
1648 static void pc_dimm_unplug(HotplugHandler *hotplug_dev,
1649                            DeviceState *dev, Error **errp)
1650 {
1651     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1652     PCDIMMDevice *dimm = PC_DIMM(dev);
1653     PCDIMMDeviceClass *ddc = PC_DIMM_GET_CLASS(dimm);
1654     MemoryRegion *mr = ddc->get_memory_region(dimm);
1655     HotplugHandlerClass *hhc;
1656     Error *local_err = NULL;
1657 
1658     hhc = HOTPLUG_HANDLER_GET_CLASS(pcms->acpi_dev);
1659     hhc->unplug(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &local_err);
1660 
1661     if (local_err) {
1662         goto out;
1663     }
1664 
1665     pc_dimm_memory_unplug(dev, &pcms->hotplug_memory, mr);
1666     object_unparent(OBJECT(dev));
1667 
1668  out:
1669     error_propagate(errp, local_err);
1670 }
1671 
1672 static int pc_apic_cmp(const void *a, const void *b)
1673 {
1674    CPUArchId *apic_a = (CPUArchId *)a;
1675    CPUArchId *apic_b = (CPUArchId *)b;
1676 
1677    return apic_a->arch_id - apic_b->arch_id;
1678 }
1679 
1680 static void pc_cpu_plug(HotplugHandler *hotplug_dev,
1681                         DeviceState *dev, Error **errp)
1682 {
1683     CPUClass *cc = CPU_GET_CLASS(dev);
1684     CPUArchId apic_id, *found_cpu;
1685     HotplugHandlerClass *hhc;
1686     Error *local_err = NULL;
1687     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1688 
1689     if (!dev->hotplugged) {
1690         goto out;
1691     }
1692 
1693     if (!pcms->acpi_dev) {
1694         error_setg(&local_err,
1695                    "cpu hotplug is not enabled: missing acpi device");
1696         goto out;
1697     }
1698 
1699     hhc = HOTPLUG_HANDLER_GET_CLASS(pcms->acpi_dev);
1700     hhc->plug(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &local_err);
1701     if (local_err) {
1702         goto out;
1703     }
1704 
1705     /* increment the number of CPUs */
1706     rtc_set_memory(pcms->rtc, 0x5f, rtc_get_memory(pcms->rtc, 0x5f) + 1);
1707 
1708     apic_id.arch_id = cc->get_arch_id(CPU(dev));
1709     found_cpu = bsearch(&apic_id, pcms->possible_cpus->cpus,
1710         pcms->possible_cpus->len, sizeof(*pcms->possible_cpus->cpus),
1711         pc_apic_cmp);
1712     assert(found_cpu);
1713     found_cpu->cpu = CPU(dev);
1714 out:
1715     error_propagate(errp, local_err);
1716 }
1717 
1718 static void pc_machine_device_plug_cb(HotplugHandler *hotplug_dev,
1719                                       DeviceState *dev, Error **errp)
1720 {
1721     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
1722         pc_dimm_plug(hotplug_dev, dev, errp);
1723     } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
1724         pc_cpu_plug(hotplug_dev, dev, errp);
1725     }
1726 }
1727 
1728 static void pc_machine_device_unplug_request_cb(HotplugHandler *hotplug_dev,
1729                                                 DeviceState *dev, Error **errp)
1730 {
1731     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
1732         pc_dimm_unplug_request(hotplug_dev, dev, errp);
1733     } else {
1734         error_setg(errp, "acpi: device unplug request for not supported device"
1735                    " type: %s", object_get_typename(OBJECT(dev)));
1736     }
1737 }
1738 
1739 static void pc_machine_device_unplug_cb(HotplugHandler *hotplug_dev,
1740                                         DeviceState *dev, Error **errp)
1741 {
1742     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
1743         pc_dimm_unplug(hotplug_dev, dev, errp);
1744     } else {
1745         error_setg(errp, "acpi: device unplug for not supported device"
1746                    " type: %s", object_get_typename(OBJECT(dev)));
1747     }
1748 }
1749 
1750 static HotplugHandler *pc_get_hotpug_handler(MachineState *machine,
1751                                              DeviceState *dev)
1752 {
1753     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(machine);
1754 
1755     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM) ||
1756         object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
1757         return HOTPLUG_HANDLER(machine);
1758     }
1759 
1760     return pcmc->get_hotplug_handler ?
1761         pcmc->get_hotplug_handler(machine, dev) : NULL;
1762 }
1763 
1764 static void
1765 pc_machine_get_hotplug_memory_region_size(Object *obj, Visitor *v,
1766                                           const char *name, void *opaque,
1767                                           Error **errp)
1768 {
1769     PCMachineState *pcms = PC_MACHINE(obj);
1770     int64_t value = memory_region_size(&pcms->hotplug_memory.mr);
1771 
1772     visit_type_int(v, name, &value, errp);
1773 }
1774 
1775 static void pc_machine_get_max_ram_below_4g(Object *obj, Visitor *v,
1776                                             const char *name, void *opaque,
1777                                             Error **errp)
1778 {
1779     PCMachineState *pcms = PC_MACHINE(obj);
1780     uint64_t value = pcms->max_ram_below_4g;
1781 
1782     visit_type_size(v, name, &value, errp);
1783 }
1784 
1785 static void pc_machine_set_max_ram_below_4g(Object *obj, Visitor *v,
1786                                             const char *name, void *opaque,
1787                                             Error **errp)
1788 {
1789     PCMachineState *pcms = PC_MACHINE(obj);
1790     Error *error = NULL;
1791     uint64_t value;
1792 
1793     visit_type_size(v, name, &value, &error);
1794     if (error) {
1795         error_propagate(errp, error);
1796         return;
1797     }
1798     if (value > (1ULL << 32)) {
1799         error_setg(&error,
1800                    "Machine option 'max-ram-below-4g=%"PRIu64
1801                    "' expects size less than or equal to 4G", value);
1802         error_propagate(errp, error);
1803         return;
1804     }
1805 
1806     if (value < (1ULL << 20)) {
1807         error_report("Warning: small max_ram_below_4g(%"PRIu64
1808                      ") less than 1M.  BIOS may not work..",
1809                      value);
1810     }
1811 
1812     pcms->max_ram_below_4g = value;
1813 }
1814 
1815 static void pc_machine_get_vmport(Object *obj, Visitor *v, const char *name,
1816                                   void *opaque, Error **errp)
1817 {
1818     PCMachineState *pcms = PC_MACHINE(obj);
1819     OnOffAuto vmport = pcms->vmport;
1820 
1821     visit_type_OnOffAuto(v, name, &vmport, errp);
1822 }
1823 
1824 static void pc_machine_set_vmport(Object *obj, Visitor *v, const char *name,
1825                                   void *opaque, Error **errp)
1826 {
1827     PCMachineState *pcms = PC_MACHINE(obj);
1828 
1829     visit_type_OnOffAuto(v, name, &pcms->vmport, errp);
1830 }
1831 
1832 bool pc_machine_is_smm_enabled(PCMachineState *pcms)
1833 {
1834     bool smm_available = false;
1835 
1836     if (pcms->smm == ON_OFF_AUTO_OFF) {
1837         return false;
1838     }
1839 
1840     if (tcg_enabled() || qtest_enabled()) {
1841         smm_available = true;
1842     } else if (kvm_enabled()) {
1843         smm_available = kvm_has_smm();
1844     }
1845 
1846     if (smm_available) {
1847         return true;
1848     }
1849 
1850     if (pcms->smm == ON_OFF_AUTO_ON) {
1851         error_report("System Management Mode not supported by this hypervisor.");
1852         exit(1);
1853     }
1854     return false;
1855 }
1856 
1857 static void pc_machine_get_smm(Object *obj, Visitor *v, const char *name,
1858                                void *opaque, Error **errp)
1859 {
1860     PCMachineState *pcms = PC_MACHINE(obj);
1861     OnOffAuto smm = pcms->smm;
1862 
1863     visit_type_OnOffAuto(v, name, &smm, errp);
1864 }
1865 
1866 static void pc_machine_set_smm(Object *obj, Visitor *v, const char *name,
1867                                void *opaque, Error **errp)
1868 {
1869     PCMachineState *pcms = PC_MACHINE(obj);
1870 
1871     visit_type_OnOffAuto(v, name, &pcms->smm, errp);
1872 }
1873 
1874 static bool pc_machine_get_nvdimm(Object *obj, Error **errp)
1875 {
1876     PCMachineState *pcms = PC_MACHINE(obj);
1877 
1878     return pcms->acpi_nvdimm_state.is_enabled;
1879 }
1880 
1881 static void pc_machine_set_nvdimm(Object *obj, bool value, Error **errp)
1882 {
1883     PCMachineState *pcms = PC_MACHINE(obj);
1884 
1885     pcms->acpi_nvdimm_state.is_enabled = value;
1886 }
1887 
1888 static void pc_machine_initfn(Object *obj)
1889 {
1890     PCMachineState *pcms = PC_MACHINE(obj);
1891 
1892     object_property_add(obj, PC_MACHINE_MEMHP_REGION_SIZE, "int",
1893                         pc_machine_get_hotplug_memory_region_size,
1894                         NULL, NULL, NULL, &error_abort);
1895 
1896     pcms->max_ram_below_4g = 1ULL << 32; /* 4G */
1897     object_property_add(obj, PC_MACHINE_MAX_RAM_BELOW_4G, "size",
1898                         pc_machine_get_max_ram_below_4g,
1899                         pc_machine_set_max_ram_below_4g,
1900                         NULL, NULL, &error_abort);
1901     object_property_set_description(obj, PC_MACHINE_MAX_RAM_BELOW_4G,
1902                                     "Maximum ram below the 4G boundary (32bit boundary)",
1903                                     &error_abort);
1904 
1905     pcms->smm = ON_OFF_AUTO_AUTO;
1906     object_property_add(obj, PC_MACHINE_SMM, "OnOffAuto",
1907                         pc_machine_get_smm,
1908                         pc_machine_set_smm,
1909                         NULL, NULL, &error_abort);
1910     object_property_set_description(obj, PC_MACHINE_SMM,
1911                                     "Enable SMM (pc & q35)",
1912                                     &error_abort);
1913 
1914     pcms->vmport = ON_OFF_AUTO_AUTO;
1915     object_property_add(obj, PC_MACHINE_VMPORT, "OnOffAuto",
1916                         pc_machine_get_vmport,
1917                         pc_machine_set_vmport,
1918                         NULL, NULL, &error_abort);
1919     object_property_set_description(obj, PC_MACHINE_VMPORT,
1920                                     "Enable vmport (pc & q35)",
1921                                     &error_abort);
1922 
1923     /* nvdimm is disabled on default. */
1924     pcms->acpi_nvdimm_state.is_enabled = false;
1925     object_property_add_bool(obj, PC_MACHINE_NVDIMM, pc_machine_get_nvdimm,
1926                              pc_machine_set_nvdimm, &error_abort);
1927 }
1928 
1929 static void pc_machine_reset(void)
1930 {
1931     CPUState *cs;
1932     X86CPU *cpu;
1933 
1934     qemu_devices_reset();
1935 
1936     /* Reset APIC after devices have been reset to cancel
1937      * any changes that qemu_devices_reset() might have done.
1938      */
1939     CPU_FOREACH(cs) {
1940         cpu = X86_CPU(cs);
1941 
1942         if (cpu->apic_state) {
1943             device_reset(cpu->apic_state);
1944         }
1945     }
1946 }
1947 
1948 static unsigned pc_cpu_index_to_socket_id(unsigned cpu_index)
1949 {
1950     X86CPUTopoInfo topo;
1951     x86_topo_ids_from_idx(smp_cores, smp_threads, cpu_index,
1952                           &topo);
1953     return topo.pkg_id;
1954 }
1955 
1956 static CPUArchIdList *pc_possible_cpu_arch_ids(MachineState *machine)
1957 {
1958     PCMachineState *pcms = PC_MACHINE(machine);
1959     int len = sizeof(CPUArchIdList) +
1960               sizeof(CPUArchId) * (pcms->possible_cpus->len);
1961     CPUArchIdList *list = g_malloc(len);
1962 
1963     memcpy(list, pcms->possible_cpus, len);
1964     return list;
1965 }
1966 
1967 static void x86_nmi(NMIState *n, int cpu_index, Error **errp)
1968 {
1969     /* cpu index isn't used */
1970     CPUState *cs;
1971 
1972     CPU_FOREACH(cs) {
1973         X86CPU *cpu = X86_CPU(cs);
1974 
1975         if (!cpu->apic_state) {
1976             cpu_interrupt(cs, CPU_INTERRUPT_NMI);
1977         } else {
1978             apic_deliver_nmi(cpu->apic_state);
1979         }
1980     }
1981 }
1982 
1983 static void pc_machine_class_init(ObjectClass *oc, void *data)
1984 {
1985     MachineClass *mc = MACHINE_CLASS(oc);
1986     PCMachineClass *pcmc = PC_MACHINE_CLASS(oc);
1987     HotplugHandlerClass *hc = HOTPLUG_HANDLER_CLASS(oc);
1988     NMIClass *nc = NMI_CLASS(oc);
1989 
1990     pcmc->get_hotplug_handler = mc->get_hotplug_handler;
1991     pcmc->pci_enabled = true;
1992     pcmc->has_acpi_build = true;
1993     pcmc->rsdp_in_ram = true;
1994     pcmc->smbios_defaults = true;
1995     pcmc->smbios_uuid_encoded = true;
1996     pcmc->gigabyte_align = true;
1997     pcmc->has_reserved_memory = true;
1998     pcmc->kvmclock_enabled = true;
1999     pcmc->enforce_aligned_dimm = true;
2000     /* BIOS ACPI tables: 128K. Other BIOS datastructures: less than 4K reported
2001      * to be used at the moment, 32K should be enough for a while.  */
2002     pcmc->acpi_data_size = 0x20000 + 0x8000;
2003     pcmc->save_tsc_khz = true;
2004     mc->get_hotplug_handler = pc_get_hotpug_handler;
2005     mc->cpu_index_to_socket_id = pc_cpu_index_to_socket_id;
2006     mc->possible_cpu_arch_ids = pc_possible_cpu_arch_ids;
2007     mc->default_boot_order = "cad";
2008     mc->hot_add_cpu = pc_hot_add_cpu;
2009     mc->max_cpus = 255;
2010     mc->reset = pc_machine_reset;
2011     hc->plug = pc_machine_device_plug_cb;
2012     hc->unplug_request = pc_machine_device_unplug_request_cb;
2013     hc->unplug = pc_machine_device_unplug_cb;
2014     nc->nmi_monitor_handler = x86_nmi;
2015 }
2016 
2017 static const TypeInfo pc_machine_info = {
2018     .name = TYPE_PC_MACHINE,
2019     .parent = TYPE_MACHINE,
2020     .abstract = true,
2021     .instance_size = sizeof(PCMachineState),
2022     .instance_init = pc_machine_initfn,
2023     .class_size = sizeof(PCMachineClass),
2024     .class_init = pc_machine_class_init,
2025     .interfaces = (InterfaceInfo[]) {
2026          { TYPE_HOTPLUG_HANDLER },
2027          { TYPE_NMI },
2028          { }
2029     },
2030 };
2031 
2032 static void pc_machine_register_types(void)
2033 {
2034     type_register_static(&pc_machine_info);
2035 }
2036 
2037 type_init(pc_machine_register_types)
2038