xref: /openbmc/qemu/hw/i386/pc.c (revision f186d64d)
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     if (!kvm_irqchip_in_kernel()) {
165         intno = apic_get_interrupt(cpu->apic_state);
166         if (intno >= 0) {
167             return intno;
168         }
169         /* read the irq from the PIC */
170         if (!apic_accept_pic_intr(cpu->apic_state)) {
171             return -1;
172         }
173     }
174 
175     intno = pic_read_irq(isa_pic);
176     return intno;
177 }
178 
179 static void pic_irq_request(void *opaque, int irq, int level)
180 {
181     CPUState *cs = first_cpu;
182     X86CPU *cpu = X86_CPU(cs);
183 
184     DPRINTF("pic_irqs: %s irq %d\n", level? "raise" : "lower", irq);
185     if (cpu->apic_state && !kvm_irqchip_in_kernel()) {
186         CPU_FOREACH(cs) {
187             cpu = X86_CPU(cs);
188             if (apic_accept_pic_intr(cpu->apic_state)) {
189                 apic_deliver_pic_intr(cpu->apic_state, level);
190             }
191         }
192     } else {
193         if (level) {
194             cpu_interrupt(cs, CPU_INTERRUPT_HARD);
195         } else {
196             cpu_reset_interrupt(cs, CPU_INTERRUPT_HARD);
197         }
198     }
199 }
200 
201 /* PC cmos mappings */
202 
203 #define REG_EQUIPMENT_BYTE          0x14
204 
205 int cmos_get_fd_drive_type(FloppyDriveType fd0)
206 {
207     int val;
208 
209     switch (fd0) {
210     case FLOPPY_DRIVE_TYPE_144:
211         /* 1.44 Mb 3"5 drive */
212         val = 4;
213         break;
214     case FLOPPY_DRIVE_TYPE_288:
215         /* 2.88 Mb 3"5 drive */
216         val = 5;
217         break;
218     case FLOPPY_DRIVE_TYPE_120:
219         /* 1.2 Mb 5"5 drive */
220         val = 2;
221         break;
222     case FLOPPY_DRIVE_TYPE_NONE:
223     default:
224         val = 0;
225         break;
226     }
227     return val;
228 }
229 
230 static void cmos_init_hd(ISADevice *s, int type_ofs, int info_ofs,
231                          int16_t cylinders, int8_t heads, int8_t sectors)
232 {
233     rtc_set_memory(s, type_ofs, 47);
234     rtc_set_memory(s, info_ofs, cylinders);
235     rtc_set_memory(s, info_ofs + 1, cylinders >> 8);
236     rtc_set_memory(s, info_ofs + 2, heads);
237     rtc_set_memory(s, info_ofs + 3, 0xff);
238     rtc_set_memory(s, info_ofs + 4, 0xff);
239     rtc_set_memory(s, info_ofs + 5, 0xc0 | ((heads > 8) << 3));
240     rtc_set_memory(s, info_ofs + 6, cylinders);
241     rtc_set_memory(s, info_ofs + 7, cylinders >> 8);
242     rtc_set_memory(s, info_ofs + 8, sectors);
243 }
244 
245 /* convert boot_device letter to something recognizable by the bios */
246 static int boot_device2nibble(char boot_device)
247 {
248     switch(boot_device) {
249     case 'a':
250     case 'b':
251         return 0x01; /* floppy boot */
252     case 'c':
253         return 0x02; /* hard drive boot */
254     case 'd':
255         return 0x03; /* CD-ROM boot */
256     case 'n':
257         return 0x04; /* Network boot */
258     }
259     return 0;
260 }
261 
262 static void set_boot_dev(ISADevice *s, const char *boot_device, Error **errp)
263 {
264 #define PC_MAX_BOOT_DEVICES 3
265     int nbds, bds[3] = { 0, };
266     int i;
267 
268     nbds = strlen(boot_device);
269     if (nbds > PC_MAX_BOOT_DEVICES) {
270         error_setg(errp, "Too many boot devices for PC");
271         return;
272     }
273     for (i = 0; i < nbds; i++) {
274         bds[i] = boot_device2nibble(boot_device[i]);
275         if (bds[i] == 0) {
276             error_setg(errp, "Invalid boot device for PC: '%c'",
277                        boot_device[i]);
278             return;
279         }
280     }
281     rtc_set_memory(s, 0x3d, (bds[1] << 4) | bds[0]);
282     rtc_set_memory(s, 0x38, (bds[2] << 4) | (fd_bootchk ? 0x0 : 0x1));
283 }
284 
285 static void pc_boot_set(void *opaque, const char *boot_device, Error **errp)
286 {
287     set_boot_dev(opaque, boot_device, errp);
288 }
289 
290 static void pc_cmos_init_floppy(ISADevice *rtc_state, ISADevice *floppy)
291 {
292     int val, nb, i;
293     FloppyDriveType fd_type[2] = { FLOPPY_DRIVE_TYPE_NONE,
294                                    FLOPPY_DRIVE_TYPE_NONE };
295 
296     /* floppy type */
297     if (floppy) {
298         for (i = 0; i < 2; i++) {
299             fd_type[i] = isa_fdc_get_drive_type(floppy, i);
300         }
301     }
302     val = (cmos_get_fd_drive_type(fd_type[0]) << 4) |
303         cmos_get_fd_drive_type(fd_type[1]);
304     rtc_set_memory(rtc_state, 0x10, val);
305 
306     val = rtc_get_memory(rtc_state, REG_EQUIPMENT_BYTE);
307     nb = 0;
308     if (fd_type[0] != FLOPPY_DRIVE_TYPE_NONE) {
309         nb++;
310     }
311     if (fd_type[1] != FLOPPY_DRIVE_TYPE_NONE) {
312         nb++;
313     }
314     switch (nb) {
315     case 0:
316         break;
317     case 1:
318         val |= 0x01; /* 1 drive, ready for boot */
319         break;
320     case 2:
321         val |= 0x41; /* 2 drives, ready for boot */
322         break;
323     }
324     rtc_set_memory(rtc_state, REG_EQUIPMENT_BYTE, val);
325 }
326 
327 typedef struct pc_cmos_init_late_arg {
328     ISADevice *rtc_state;
329     BusState *idebus[2];
330 } pc_cmos_init_late_arg;
331 
332 typedef struct check_fdc_state {
333     ISADevice *floppy;
334     bool multiple;
335 } CheckFdcState;
336 
337 static int check_fdc(Object *obj, void *opaque)
338 {
339     CheckFdcState *state = opaque;
340     Object *fdc;
341     uint32_t iobase;
342     Error *local_err = NULL;
343 
344     fdc = object_dynamic_cast(obj, TYPE_ISA_FDC);
345     if (!fdc) {
346         return 0;
347     }
348 
349     iobase = object_property_get_int(obj, "iobase", &local_err);
350     if (local_err || iobase != 0x3f0) {
351         error_free(local_err);
352         return 0;
353     }
354 
355     if (state->floppy) {
356         state->multiple = true;
357     } else {
358         state->floppy = ISA_DEVICE(obj);
359     }
360     return 0;
361 }
362 
363 static const char * const fdc_container_path[] = {
364     "/unattached", "/peripheral", "/peripheral-anon"
365 };
366 
367 /*
368  * Locate the FDC at IO address 0x3f0, in order to configure the CMOS registers
369  * and ACPI objects.
370  */
371 ISADevice *pc_find_fdc0(void)
372 {
373     int i;
374     Object *container;
375     CheckFdcState state = { 0 };
376 
377     for (i = 0; i < ARRAY_SIZE(fdc_container_path); i++) {
378         container = container_get(qdev_get_machine(), fdc_container_path[i]);
379         object_child_foreach(container, check_fdc, &state);
380     }
381 
382     if (state.multiple) {
383         error_report("warning: multiple floppy disk controllers with "
384                      "iobase=0x3f0 have been found");
385         error_printf("the one being picked for CMOS setup might not reflect "
386                      "your intent\n");
387     }
388 
389     return state.floppy;
390 }
391 
392 static void pc_cmos_init_late(void *opaque)
393 {
394     pc_cmos_init_late_arg *arg = opaque;
395     ISADevice *s = arg->rtc_state;
396     int16_t cylinders;
397     int8_t heads, sectors;
398     int val;
399     int i, trans;
400 
401     val = 0;
402     if (ide_get_geometry(arg->idebus[0], 0,
403                          &cylinders, &heads, &sectors) >= 0) {
404         cmos_init_hd(s, 0x19, 0x1b, cylinders, heads, sectors);
405         val |= 0xf0;
406     }
407     if (ide_get_geometry(arg->idebus[0], 1,
408                          &cylinders, &heads, &sectors) >= 0) {
409         cmos_init_hd(s, 0x1a, 0x24, cylinders, heads, sectors);
410         val |= 0x0f;
411     }
412     rtc_set_memory(s, 0x12, val);
413 
414     val = 0;
415     for (i = 0; i < 4; i++) {
416         /* NOTE: ide_get_geometry() returns the physical
417            geometry.  It is always such that: 1 <= sects <= 63, 1
418            <= heads <= 16, 1 <= cylinders <= 16383. The BIOS
419            geometry can be different if a translation is done. */
420         if (ide_get_geometry(arg->idebus[i / 2], i % 2,
421                              &cylinders, &heads, &sectors) >= 0) {
422             trans = ide_get_bios_chs_trans(arg->idebus[i / 2], i % 2) - 1;
423             assert((trans & ~3) == 0);
424             val |= trans << (i * 2);
425         }
426     }
427     rtc_set_memory(s, 0x39, val);
428 
429     pc_cmos_init_floppy(s, pc_find_fdc0());
430 
431     qemu_unregister_reset(pc_cmos_init_late, opaque);
432 }
433 
434 void pc_cmos_init(PCMachineState *pcms,
435                   BusState *idebus0, BusState *idebus1,
436                   ISADevice *s)
437 {
438     int val;
439     static pc_cmos_init_late_arg arg;
440 
441     /* various important CMOS locations needed by PC/Bochs bios */
442 
443     /* memory size */
444     /* base memory (first MiB) */
445     val = MIN(pcms->below_4g_mem_size / 1024, 640);
446     rtc_set_memory(s, 0x15, val);
447     rtc_set_memory(s, 0x16, val >> 8);
448     /* extended memory (next 64MiB) */
449     if (pcms->below_4g_mem_size > 1024 * 1024) {
450         val = (pcms->below_4g_mem_size - 1024 * 1024) / 1024;
451     } else {
452         val = 0;
453     }
454     if (val > 65535)
455         val = 65535;
456     rtc_set_memory(s, 0x17, val);
457     rtc_set_memory(s, 0x18, val >> 8);
458     rtc_set_memory(s, 0x30, val);
459     rtc_set_memory(s, 0x31, val >> 8);
460     /* memory between 16MiB and 4GiB */
461     if (pcms->below_4g_mem_size > 16 * 1024 * 1024) {
462         val = (pcms->below_4g_mem_size - 16 * 1024 * 1024) / 65536;
463     } else {
464         val = 0;
465     }
466     if (val > 65535)
467         val = 65535;
468     rtc_set_memory(s, 0x34, val);
469     rtc_set_memory(s, 0x35, val >> 8);
470     /* memory above 4GiB */
471     val = pcms->above_4g_mem_size / 65536;
472     rtc_set_memory(s, 0x5b, val);
473     rtc_set_memory(s, 0x5c, val >> 8);
474     rtc_set_memory(s, 0x5d, val >> 16);
475 
476     object_property_add_link(OBJECT(pcms), "rtc_state",
477                              TYPE_ISA_DEVICE,
478                              (Object **)&pcms->rtc,
479                              object_property_allow_set_link,
480                              OBJ_PROP_LINK_UNREF_ON_RELEASE, &error_abort);
481     object_property_set_link(OBJECT(pcms), OBJECT(s),
482                              "rtc_state", &error_abort);
483 
484     set_boot_dev(s, MACHINE(pcms)->boot_order, &error_fatal);
485 
486     val = 0;
487     val |= 0x02; /* FPU is there */
488     val |= 0x04; /* PS/2 mouse installed */
489     rtc_set_memory(s, REG_EQUIPMENT_BYTE, val);
490 
491     /* hard drives and FDC */
492     arg.rtc_state = s;
493     arg.idebus[0] = idebus0;
494     arg.idebus[1] = idebus1;
495     qemu_register_reset(pc_cmos_init_late, &arg);
496 }
497 
498 #define TYPE_PORT92 "port92"
499 #define PORT92(obj) OBJECT_CHECK(Port92State, (obj), TYPE_PORT92)
500 
501 /* port 92 stuff: could be split off */
502 typedef struct Port92State {
503     ISADevice parent_obj;
504 
505     MemoryRegion io;
506     uint8_t outport;
507     qemu_irq a20_out;
508 } Port92State;
509 
510 static void port92_write(void *opaque, hwaddr addr, uint64_t val,
511                          unsigned size)
512 {
513     Port92State *s = opaque;
514     int oldval = s->outport;
515 
516     DPRINTF("port92: write 0x%02" PRIx64 "\n", val);
517     s->outport = val;
518     qemu_set_irq(s->a20_out, (val >> 1) & 1);
519     if ((val & 1) && !(oldval & 1)) {
520         qemu_system_reset_request();
521     }
522 }
523 
524 static uint64_t port92_read(void *opaque, hwaddr addr,
525                             unsigned size)
526 {
527     Port92State *s = opaque;
528     uint32_t ret;
529 
530     ret = s->outport;
531     DPRINTF("port92: read 0x%02x\n", ret);
532     return ret;
533 }
534 
535 static void port92_init(ISADevice *dev, qemu_irq a20_out)
536 {
537     qdev_connect_gpio_out_named(DEVICE(dev), PORT92_A20_LINE, 0, a20_out);
538 }
539 
540 static const VMStateDescription vmstate_port92_isa = {
541     .name = "port92",
542     .version_id = 1,
543     .minimum_version_id = 1,
544     .fields = (VMStateField[]) {
545         VMSTATE_UINT8(outport, Port92State),
546         VMSTATE_END_OF_LIST()
547     }
548 };
549 
550 static void port92_reset(DeviceState *d)
551 {
552     Port92State *s = PORT92(d);
553 
554     s->outport &= ~1;
555 }
556 
557 static const MemoryRegionOps port92_ops = {
558     .read = port92_read,
559     .write = port92_write,
560     .impl = {
561         .min_access_size = 1,
562         .max_access_size = 1,
563     },
564     .endianness = DEVICE_LITTLE_ENDIAN,
565 };
566 
567 static void port92_initfn(Object *obj)
568 {
569     Port92State *s = PORT92(obj);
570 
571     memory_region_init_io(&s->io, OBJECT(s), &port92_ops, s, "port92", 1);
572 
573     s->outport = 0;
574 
575     qdev_init_gpio_out_named(DEVICE(obj), &s->a20_out, PORT92_A20_LINE, 1);
576 }
577 
578 static void port92_realizefn(DeviceState *dev, Error **errp)
579 {
580     ISADevice *isadev = ISA_DEVICE(dev);
581     Port92State *s = PORT92(dev);
582 
583     isa_register_ioport(isadev, &s->io, 0x92);
584 }
585 
586 static void port92_class_initfn(ObjectClass *klass, void *data)
587 {
588     DeviceClass *dc = DEVICE_CLASS(klass);
589 
590     dc->realize = port92_realizefn;
591     dc->reset = port92_reset;
592     dc->vmsd = &vmstate_port92_isa;
593     /*
594      * Reason: unlike ordinary ISA devices, this one needs additional
595      * wiring: its A20 output line needs to be wired up by
596      * port92_init().
597      */
598     dc->cannot_instantiate_with_device_add_yet = true;
599 }
600 
601 static const TypeInfo port92_info = {
602     .name          = TYPE_PORT92,
603     .parent        = TYPE_ISA_DEVICE,
604     .instance_size = sizeof(Port92State),
605     .instance_init = port92_initfn,
606     .class_init    = port92_class_initfn,
607 };
608 
609 static void port92_register_types(void)
610 {
611     type_register_static(&port92_info);
612 }
613 
614 type_init(port92_register_types)
615 
616 static void handle_a20_line_change(void *opaque, int irq, int level)
617 {
618     X86CPU *cpu = opaque;
619 
620     /* XXX: send to all CPUs ? */
621     /* XXX: add logic to handle multiple A20 line sources */
622     x86_cpu_set_a20(cpu, level);
623 }
624 
625 int e820_add_entry(uint64_t address, uint64_t length, uint32_t type)
626 {
627     int index = le32_to_cpu(e820_reserve.count);
628     struct e820_entry *entry;
629 
630     if (type != E820_RAM) {
631         /* old FW_CFG_E820_TABLE entry -- reservations only */
632         if (index >= E820_NR_ENTRIES) {
633             return -EBUSY;
634         }
635         entry = &e820_reserve.entry[index++];
636 
637         entry->address = cpu_to_le64(address);
638         entry->length = cpu_to_le64(length);
639         entry->type = cpu_to_le32(type);
640 
641         e820_reserve.count = cpu_to_le32(index);
642     }
643 
644     /* new "etc/e820" file -- include ram too */
645     e820_table = g_renew(struct e820_entry, e820_table, e820_entries + 1);
646     e820_table[e820_entries].address = cpu_to_le64(address);
647     e820_table[e820_entries].length = cpu_to_le64(length);
648     e820_table[e820_entries].type = cpu_to_le32(type);
649     e820_entries++;
650 
651     return e820_entries;
652 }
653 
654 int e820_get_num_entries(void)
655 {
656     return e820_entries;
657 }
658 
659 bool e820_get_entry(int idx, uint32_t type, uint64_t *address, uint64_t *length)
660 {
661     if (idx < e820_entries && e820_table[idx].type == cpu_to_le32(type)) {
662         *address = le64_to_cpu(e820_table[idx].address);
663         *length = le64_to_cpu(e820_table[idx].length);
664         return true;
665     }
666     return false;
667 }
668 
669 /* Enables contiguous-apic-ID mode, for compatibility */
670 static bool compat_apic_id_mode;
671 
672 void enable_compat_apic_id_mode(void)
673 {
674     compat_apic_id_mode = true;
675 }
676 
677 /* Calculates initial APIC ID for a specific CPU index
678  *
679  * Currently we need to be able to calculate the APIC ID from the CPU index
680  * alone (without requiring a CPU object), as the QEMU<->Seabios interfaces have
681  * no concept of "CPU index", and the NUMA tables on fw_cfg need the APIC ID of
682  * all CPUs up to max_cpus.
683  */
684 static uint32_t x86_cpu_apic_id_from_index(unsigned int cpu_index)
685 {
686     uint32_t correct_id;
687     static bool warned;
688 
689     correct_id = x86_apicid_from_cpu_idx(smp_cores, smp_threads, cpu_index);
690     if (compat_apic_id_mode) {
691         if (cpu_index != correct_id && !warned && !qtest_enabled()) {
692             error_report("APIC IDs set in compatibility mode, "
693                          "CPU topology won't match the configuration");
694             warned = true;
695         }
696         return cpu_index;
697     } else {
698         return correct_id;
699     }
700 }
701 
702 static void pc_build_smbios(FWCfgState *fw_cfg)
703 {
704     uint8_t *smbios_tables, *smbios_anchor;
705     size_t smbios_tables_len, smbios_anchor_len;
706     struct smbios_phys_mem_area *mem_array;
707     unsigned i, array_count;
708 
709     smbios_tables = smbios_get_table_legacy(&smbios_tables_len);
710     if (smbios_tables) {
711         fw_cfg_add_bytes(fw_cfg, FW_CFG_SMBIOS_ENTRIES,
712                          smbios_tables, smbios_tables_len);
713     }
714 
715     /* build the array of physical mem area from e820 table */
716     mem_array = g_malloc0(sizeof(*mem_array) * e820_get_num_entries());
717     for (i = 0, array_count = 0; i < e820_get_num_entries(); i++) {
718         uint64_t addr, len;
719 
720         if (e820_get_entry(i, E820_RAM, &addr, &len)) {
721             mem_array[array_count].address = addr;
722             mem_array[array_count].length = len;
723             array_count++;
724         }
725     }
726     smbios_get_tables(mem_array, array_count,
727                       &smbios_tables, &smbios_tables_len,
728                       &smbios_anchor, &smbios_anchor_len);
729     g_free(mem_array);
730 
731     if (smbios_anchor) {
732         fw_cfg_add_file(fw_cfg, "etc/smbios/smbios-tables",
733                         smbios_tables, smbios_tables_len);
734         fw_cfg_add_file(fw_cfg, "etc/smbios/smbios-anchor",
735                         smbios_anchor, smbios_anchor_len);
736     }
737 }
738 
739 static FWCfgState *bochs_bios_init(AddressSpace *as, PCMachineState *pcms)
740 {
741     FWCfgState *fw_cfg;
742     uint64_t *numa_fw_cfg;
743     int i, j;
744 
745     fw_cfg = fw_cfg_init_io_dma(FW_CFG_IO_BASE, FW_CFG_IO_BASE + 4, as);
746 
747     /* FW_CFG_MAX_CPUS is a bit confusing/problematic on x86:
748      *
749      * SeaBIOS needs FW_CFG_MAX_CPUS for CPU hotplug, but the CPU hotplug
750      * QEMU<->SeaBIOS interface is not based on the "CPU index", but on the APIC
751      * ID of hotplugged CPUs[1]. This means that FW_CFG_MAX_CPUS is not the
752      * "maximum number of CPUs", but the "limit to the APIC ID values SeaBIOS
753      * may see".
754      *
755      * So, this means we must not use max_cpus, here, but the maximum possible
756      * APIC ID value, plus one.
757      *
758      * [1] The only kind of "CPU identifier" used between SeaBIOS and QEMU is
759      *     the APIC ID, not the "CPU index"
760      */
761     fw_cfg_add_i16(fw_cfg, FW_CFG_MAX_CPUS, (uint16_t)pcms->apic_id_limit);
762     fw_cfg_add_i64(fw_cfg, FW_CFG_RAM_SIZE, (uint64_t)ram_size);
763     fw_cfg_add_bytes(fw_cfg, FW_CFG_ACPI_TABLES,
764                      acpi_tables, acpi_tables_len);
765     fw_cfg_add_i32(fw_cfg, FW_CFG_IRQ0_OVERRIDE, kvm_allows_irq0_override());
766 
767     fw_cfg_add_bytes(fw_cfg, FW_CFG_E820_TABLE,
768                      &e820_reserve, sizeof(e820_reserve));
769     fw_cfg_add_file(fw_cfg, "etc/e820", e820_table,
770                     sizeof(struct e820_entry) * e820_entries);
771 
772     fw_cfg_add_bytes(fw_cfg, FW_CFG_HPET, &hpet_cfg, sizeof(hpet_cfg));
773     /* allocate memory for the NUMA channel: one (64bit) word for the number
774      * of nodes, one word for each VCPU->node and one word for each node to
775      * hold the amount of memory.
776      */
777     numa_fw_cfg = g_new0(uint64_t, 1 + pcms->apic_id_limit + nb_numa_nodes);
778     numa_fw_cfg[0] = cpu_to_le64(nb_numa_nodes);
779     for (i = 0; i < max_cpus; i++) {
780         unsigned int apic_id = x86_cpu_apic_id_from_index(i);
781         assert(apic_id < pcms->apic_id_limit);
782         for (j = 0; j < nb_numa_nodes; j++) {
783             if (test_bit(i, numa_info[j].node_cpu)) {
784                 numa_fw_cfg[apic_id + 1] = cpu_to_le64(j);
785                 break;
786             }
787         }
788     }
789     for (i = 0; i < nb_numa_nodes; i++) {
790         numa_fw_cfg[pcms->apic_id_limit + 1 + i] =
791             cpu_to_le64(numa_info[i].node_mem);
792     }
793     fw_cfg_add_bytes(fw_cfg, FW_CFG_NUMA, numa_fw_cfg,
794                      (1 + pcms->apic_id_limit + nb_numa_nodes) *
795                      sizeof(*numa_fw_cfg));
796 
797     return fw_cfg;
798 }
799 
800 static long get_file_size(FILE *f)
801 {
802     long where, size;
803 
804     /* XXX: on Unix systems, using fstat() probably makes more sense */
805 
806     where = ftell(f);
807     fseek(f, 0, SEEK_END);
808     size = ftell(f);
809     fseek(f, where, SEEK_SET);
810 
811     return size;
812 }
813 
814 /* setup_data types */
815 #define SETUP_NONE     0
816 #define SETUP_E820_EXT 1
817 #define SETUP_DTB      2
818 #define SETUP_PCI      3
819 #define SETUP_EFI      4
820 
821 struct setup_data {
822     uint64_t next;
823     uint32_t type;
824     uint32_t len;
825     uint8_t data[0];
826 } __attribute__((packed));
827 
828 static void load_linux(PCMachineState *pcms,
829                        FWCfgState *fw_cfg)
830 {
831     uint16_t protocol;
832     int setup_size, kernel_size, initrd_size = 0, cmdline_size;
833     int dtb_size, setup_data_offset;
834     uint32_t initrd_max;
835     uint8_t header[8192], *setup, *kernel, *initrd_data;
836     hwaddr real_addr, prot_addr, cmdline_addr, initrd_addr = 0;
837     FILE *f;
838     char *vmode;
839     MachineState *machine = MACHINE(pcms);
840     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
841     struct setup_data *setup_data;
842     const char *kernel_filename = machine->kernel_filename;
843     const char *initrd_filename = machine->initrd_filename;
844     const char *dtb_filename = machine->dtb;
845     const char *kernel_cmdline = machine->kernel_cmdline;
846 
847     /* Align to 16 bytes as a paranoia measure */
848     cmdline_size = (strlen(kernel_cmdline)+16) & ~15;
849 
850     /* load the kernel header */
851     f = fopen(kernel_filename, "rb");
852     if (!f || !(kernel_size = get_file_size(f)) ||
853         fread(header, 1, MIN(ARRAY_SIZE(header), kernel_size), f) !=
854         MIN(ARRAY_SIZE(header), kernel_size)) {
855         fprintf(stderr, "qemu: could not load kernel '%s': %s\n",
856                 kernel_filename, strerror(errno));
857         exit(1);
858     }
859 
860     /* kernel protocol version */
861 #if 0
862     fprintf(stderr, "header magic: %#x\n", ldl_p(header+0x202));
863 #endif
864     if (ldl_p(header+0x202) == 0x53726448) {
865         protocol = lduw_p(header+0x206);
866     } else {
867         /* This looks like a multiboot kernel. If it is, let's stop
868            treating it like a Linux kernel. */
869         if (load_multiboot(fw_cfg, f, kernel_filename, initrd_filename,
870                            kernel_cmdline, kernel_size, header)) {
871             return;
872         }
873         protocol = 0;
874     }
875 
876     if (protocol < 0x200 || !(header[0x211] & 0x01)) {
877         /* Low kernel */
878         real_addr    = 0x90000;
879         cmdline_addr = 0x9a000 - cmdline_size;
880         prot_addr    = 0x10000;
881     } else if (protocol < 0x202) {
882         /* High but ancient kernel */
883         real_addr    = 0x90000;
884         cmdline_addr = 0x9a000 - cmdline_size;
885         prot_addr    = 0x100000;
886     } else {
887         /* High and recent kernel */
888         real_addr    = 0x10000;
889         cmdline_addr = 0x20000;
890         prot_addr    = 0x100000;
891     }
892 
893 #if 0
894     fprintf(stderr,
895             "qemu: real_addr     = 0x" TARGET_FMT_plx "\n"
896             "qemu: cmdline_addr  = 0x" TARGET_FMT_plx "\n"
897             "qemu: prot_addr     = 0x" TARGET_FMT_plx "\n",
898             real_addr,
899             cmdline_addr,
900             prot_addr);
901 #endif
902 
903     /* highest address for loading the initrd */
904     if (protocol >= 0x203) {
905         initrd_max = ldl_p(header+0x22c);
906     } else {
907         initrd_max = 0x37ffffff;
908     }
909 
910     if (initrd_max >= pcms->below_4g_mem_size - pcmc->acpi_data_size) {
911         initrd_max = pcms->below_4g_mem_size - pcmc->acpi_data_size - 1;
912     }
913 
914     fw_cfg_add_i32(fw_cfg, FW_CFG_CMDLINE_ADDR, cmdline_addr);
915     fw_cfg_add_i32(fw_cfg, FW_CFG_CMDLINE_SIZE, strlen(kernel_cmdline)+1);
916     fw_cfg_add_string(fw_cfg, FW_CFG_CMDLINE_DATA, kernel_cmdline);
917 
918     if (protocol >= 0x202) {
919         stl_p(header+0x228, cmdline_addr);
920     } else {
921         stw_p(header+0x20, 0xA33F);
922         stw_p(header+0x22, cmdline_addr-real_addr);
923     }
924 
925     /* handle vga= parameter */
926     vmode = strstr(kernel_cmdline, "vga=");
927     if (vmode) {
928         unsigned int video_mode;
929         /* skip "vga=" */
930         vmode += 4;
931         if (!strncmp(vmode, "normal", 6)) {
932             video_mode = 0xffff;
933         } else if (!strncmp(vmode, "ext", 3)) {
934             video_mode = 0xfffe;
935         } else if (!strncmp(vmode, "ask", 3)) {
936             video_mode = 0xfffd;
937         } else {
938             video_mode = strtol(vmode, NULL, 0);
939         }
940         stw_p(header+0x1fa, video_mode);
941     }
942 
943     /* loader type */
944     /* High nybble = B reserved for QEMU; low nybble is revision number.
945        If this code is substantially changed, you may want to consider
946        incrementing the revision. */
947     if (protocol >= 0x200) {
948         header[0x210] = 0xB0;
949     }
950     /* heap */
951     if (protocol >= 0x201) {
952         header[0x211] |= 0x80;	/* CAN_USE_HEAP */
953         stw_p(header+0x224, cmdline_addr-real_addr-0x200);
954     }
955 
956     /* load initrd */
957     if (initrd_filename) {
958         if (protocol < 0x200) {
959             fprintf(stderr, "qemu: linux kernel too old to load a ram disk\n");
960             exit(1);
961         }
962 
963         initrd_size = get_image_size(initrd_filename);
964         if (initrd_size < 0) {
965             fprintf(stderr, "qemu: error reading initrd %s: %s\n",
966                     initrd_filename, strerror(errno));
967             exit(1);
968         }
969 
970         initrd_addr = (initrd_max-initrd_size) & ~4095;
971 
972         initrd_data = g_malloc(initrd_size);
973         load_image(initrd_filename, initrd_data);
974 
975         fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_ADDR, initrd_addr);
976         fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_SIZE, initrd_size);
977         fw_cfg_add_bytes(fw_cfg, FW_CFG_INITRD_DATA, initrd_data, initrd_size);
978 
979         stl_p(header+0x218, initrd_addr);
980         stl_p(header+0x21c, initrd_size);
981     }
982 
983     /* load kernel and setup */
984     setup_size = header[0x1f1];
985     if (setup_size == 0) {
986         setup_size = 4;
987     }
988     setup_size = (setup_size+1)*512;
989     if (setup_size > kernel_size) {
990         fprintf(stderr, "qemu: invalid kernel header\n");
991         exit(1);
992     }
993     kernel_size -= setup_size;
994 
995     setup  = g_malloc(setup_size);
996     kernel = g_malloc(kernel_size);
997     fseek(f, 0, SEEK_SET);
998     if (fread(setup, 1, setup_size, f) != setup_size) {
999         fprintf(stderr, "fread() failed\n");
1000         exit(1);
1001     }
1002     if (fread(kernel, 1, kernel_size, f) != kernel_size) {
1003         fprintf(stderr, "fread() failed\n");
1004         exit(1);
1005     }
1006     fclose(f);
1007 
1008     /* append dtb to kernel */
1009     if (dtb_filename) {
1010         if (protocol < 0x209) {
1011             fprintf(stderr, "qemu: Linux kernel too old to load a dtb\n");
1012             exit(1);
1013         }
1014 
1015         dtb_size = get_image_size(dtb_filename);
1016         if (dtb_size <= 0) {
1017             fprintf(stderr, "qemu: error reading dtb %s: %s\n",
1018                     dtb_filename, strerror(errno));
1019             exit(1);
1020         }
1021 
1022         setup_data_offset = QEMU_ALIGN_UP(kernel_size, 16);
1023         kernel_size = setup_data_offset + sizeof(struct setup_data) + dtb_size;
1024         kernel = g_realloc(kernel, kernel_size);
1025 
1026         stq_p(header+0x250, prot_addr + setup_data_offset);
1027 
1028         setup_data = (struct setup_data *)(kernel + setup_data_offset);
1029         setup_data->next = 0;
1030         setup_data->type = cpu_to_le32(SETUP_DTB);
1031         setup_data->len = cpu_to_le32(dtb_size);
1032 
1033         load_image_size(dtb_filename, setup_data->data, dtb_size);
1034     }
1035 
1036     memcpy(setup, header, MIN(sizeof(header), setup_size));
1037 
1038     fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_ADDR, prot_addr);
1039     fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_SIZE, kernel_size);
1040     fw_cfg_add_bytes(fw_cfg, FW_CFG_KERNEL_DATA, kernel, kernel_size);
1041 
1042     fw_cfg_add_i32(fw_cfg, FW_CFG_SETUP_ADDR, real_addr);
1043     fw_cfg_add_i32(fw_cfg, FW_CFG_SETUP_SIZE, setup_size);
1044     fw_cfg_add_bytes(fw_cfg, FW_CFG_SETUP_DATA, setup, setup_size);
1045 
1046     if (fw_cfg_dma_enabled(fw_cfg)) {
1047         option_rom[nb_option_roms].name = "linuxboot_dma.bin";
1048         option_rom[nb_option_roms].bootindex = 0;
1049     } else {
1050         option_rom[nb_option_roms].name = "linuxboot.bin";
1051         option_rom[nb_option_roms].bootindex = 0;
1052     }
1053     nb_option_roms++;
1054 }
1055 
1056 #define NE2000_NB_MAX 6
1057 
1058 static const int ne2000_io[NE2000_NB_MAX] = { 0x300, 0x320, 0x340, 0x360,
1059                                               0x280, 0x380 };
1060 static const int ne2000_irq[NE2000_NB_MAX] = { 9, 10, 11, 3, 4, 5 };
1061 
1062 void pc_init_ne2k_isa(ISABus *bus, NICInfo *nd)
1063 {
1064     static int nb_ne2k = 0;
1065 
1066     if (nb_ne2k == NE2000_NB_MAX)
1067         return;
1068     isa_ne2000_init(bus, ne2000_io[nb_ne2k],
1069                     ne2000_irq[nb_ne2k], nd);
1070     nb_ne2k++;
1071 }
1072 
1073 DeviceState *cpu_get_current_apic(void)
1074 {
1075     if (current_cpu) {
1076         X86CPU *cpu = X86_CPU(current_cpu);
1077         return cpu->apic_state;
1078     } else {
1079         return NULL;
1080     }
1081 }
1082 
1083 void pc_acpi_smi_interrupt(void *opaque, int irq, int level)
1084 {
1085     X86CPU *cpu = opaque;
1086 
1087     if (level) {
1088         cpu_interrupt(CPU(cpu), CPU_INTERRUPT_SMI);
1089     }
1090 }
1091 
1092 static int pc_present_cpus_count(PCMachineState *pcms)
1093 {
1094     int i, boot_cpus = 0;
1095     for (i = 0; i < pcms->possible_cpus->len; i++) {
1096         if (pcms->possible_cpus->cpus[i].cpu) {
1097             boot_cpus++;
1098         }
1099     }
1100     return boot_cpus;
1101 }
1102 
1103 static X86CPU *pc_new_cpu(const char *typename, int64_t apic_id,
1104                           Error **errp)
1105 {
1106     X86CPU *cpu = NULL;
1107     Error *local_err = NULL;
1108 
1109     cpu = X86_CPU(object_new(typename));
1110 
1111     object_property_set_int(OBJECT(cpu), apic_id, "apic-id", &local_err);
1112     object_property_set_bool(OBJECT(cpu), true, "realized", &local_err);
1113 
1114     if (local_err) {
1115         error_propagate(errp, local_err);
1116         object_unref(OBJECT(cpu));
1117         cpu = NULL;
1118     }
1119     return cpu;
1120 }
1121 
1122 void pc_hot_add_cpu(const int64_t id, Error **errp)
1123 {
1124     X86CPU *cpu;
1125     ObjectClass *oc;
1126     PCMachineState *pcms = PC_MACHINE(qdev_get_machine());
1127     int64_t apic_id = x86_cpu_apic_id_from_index(id);
1128     Error *local_err = NULL;
1129 
1130     if (id < 0) {
1131         error_setg(errp, "Invalid CPU id: %" PRIi64, id);
1132         return;
1133     }
1134 
1135     if (apic_id >= ACPI_CPU_HOTPLUG_ID_LIMIT) {
1136         error_setg(errp, "Unable to add CPU: %" PRIi64
1137                    ", resulting APIC ID (%" PRIi64 ") is too large",
1138                    id, apic_id);
1139         return;
1140     }
1141 
1142     assert(pcms->possible_cpus->cpus[0].cpu); /* BSP is always present */
1143     oc = OBJECT_CLASS(CPU_GET_CLASS(pcms->possible_cpus->cpus[0].cpu));
1144     cpu = pc_new_cpu(object_class_get_name(oc), apic_id, &local_err);
1145     if (local_err) {
1146         error_propagate(errp, local_err);
1147         return;
1148     }
1149     object_unref(OBJECT(cpu));
1150 }
1151 
1152 void pc_cpus_init(PCMachineState *pcms)
1153 {
1154     int i;
1155     CPUClass *cc;
1156     ObjectClass *oc;
1157     const char *typename;
1158     gchar **model_pieces;
1159     X86CPU *cpu = NULL;
1160     MachineState *machine = MACHINE(pcms);
1161 
1162     /* init CPUs */
1163     if (machine->cpu_model == NULL) {
1164 #ifdef TARGET_X86_64
1165         machine->cpu_model = "qemu64";
1166 #else
1167         machine->cpu_model = "qemu32";
1168 #endif
1169     }
1170 
1171     model_pieces = g_strsplit(machine->cpu_model, ",", 2);
1172     if (!model_pieces[0]) {
1173         error_report("Invalid/empty CPU model name");
1174         exit(1);
1175     }
1176 
1177     oc = cpu_class_by_name(TYPE_X86_CPU, model_pieces[0]);
1178     if (oc == NULL) {
1179         error_report("Unable to find CPU definition: %s", model_pieces[0]);
1180         exit(1);
1181     }
1182     typename = object_class_get_name(oc);
1183     cc = CPU_CLASS(oc);
1184     cc->parse_features(typename, model_pieces[1], &error_fatal);
1185     g_strfreev(model_pieces);
1186 
1187     /* Calculates the limit to CPU APIC ID values
1188      *
1189      * Limit for the APIC ID value, so that all
1190      * CPU APIC IDs are < pcms->apic_id_limit.
1191      *
1192      * This is used for FW_CFG_MAX_CPUS. See comments on bochs_bios_init().
1193      */
1194     pcms->apic_id_limit = x86_cpu_apic_id_from_index(max_cpus - 1) + 1;
1195     if (pcms->apic_id_limit > ACPI_CPU_HOTPLUG_ID_LIMIT) {
1196         error_report("max_cpus is too large. APIC ID of last CPU is %u",
1197                      pcms->apic_id_limit - 1);
1198         exit(1);
1199     }
1200 
1201     pcms->possible_cpus = g_malloc0(sizeof(CPUArchIdList) +
1202                                     sizeof(CPUArchId) * max_cpus);
1203     for (i = 0; i < max_cpus; i++) {
1204         pcms->possible_cpus->cpus[i].arch_id = x86_cpu_apic_id_from_index(i);
1205         pcms->possible_cpus->len++;
1206         if (i < smp_cpus) {
1207             cpu = pc_new_cpu(typename, x86_cpu_apic_id_from_index(i),
1208                              &error_fatal);
1209             object_unref(OBJECT(cpu));
1210         }
1211     }
1212 
1213     /* tell smbios about cpuid version and features */
1214     smbios_set_cpuid(cpu->env.cpuid_version, cpu->env.features[FEAT_1_EDX]);
1215 }
1216 
1217 static void pc_build_feature_control_file(PCMachineState *pcms)
1218 {
1219     X86CPU *cpu = X86_CPU(pcms->possible_cpus->cpus[0].cpu);
1220     CPUX86State *env = &cpu->env;
1221     uint32_t unused, ecx, edx;
1222     uint64_t feature_control_bits = 0;
1223     uint64_t *val;
1224 
1225     cpu_x86_cpuid(env, 1, 0, &unused, &unused, &ecx, &edx);
1226     if (ecx & CPUID_EXT_VMX) {
1227         feature_control_bits |= FEATURE_CONTROL_VMXON_ENABLED_OUTSIDE_SMX;
1228     }
1229 
1230     if ((edx & (CPUID_EXT2_MCE | CPUID_EXT2_MCA)) ==
1231         (CPUID_EXT2_MCE | CPUID_EXT2_MCA) &&
1232         (env->mcg_cap & MCG_LMCE_P)) {
1233         feature_control_bits |= FEATURE_CONTROL_LMCE;
1234     }
1235 
1236     if (!feature_control_bits) {
1237         return;
1238     }
1239 
1240     val = g_malloc(sizeof(*val));
1241     *val = cpu_to_le64(feature_control_bits | FEATURE_CONTROL_LOCKED);
1242     fw_cfg_add_file(pcms->fw_cfg, "etc/msr_feature_control", val, sizeof(*val));
1243 }
1244 
1245 static
1246 void pc_machine_done(Notifier *notifier, void *data)
1247 {
1248     PCMachineState *pcms = container_of(notifier,
1249                                         PCMachineState, machine_done);
1250     PCIBus *bus = pcms->bus;
1251 
1252     /* set the number of CPUs */
1253     rtc_set_memory(pcms->rtc, 0x5f, pc_present_cpus_count(pcms) - 1);
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     g_free(a20_line);
1614 
1615     DMA_init(isa_bus, 0);
1616 
1617     for(i = 0; i < MAX_FD; i++) {
1618         fd[i] = drive_get(IF_FLOPPY, 0, i);
1619         create_fdctrl |= !!fd[i];
1620     }
1621     if (create_fdctrl) {
1622         fdctrl_init_isa(isa_bus, fd);
1623     }
1624 }
1625 
1626 void pc_nic_init(ISABus *isa_bus, PCIBus *pci_bus)
1627 {
1628     int i;
1629 
1630     rom_set_order_override(FW_CFG_ORDER_OVERRIDE_NIC);
1631     for (i = 0; i < nb_nics; i++) {
1632         NICInfo *nd = &nd_table[i];
1633 
1634         if (!pci_bus || (nd->model && strcmp(nd->model, "ne2k_isa") == 0)) {
1635             pc_init_ne2k_isa(isa_bus, nd);
1636         } else {
1637             pci_nic_init_nofail(nd, pci_bus, "e1000", NULL);
1638         }
1639     }
1640     rom_reset_order_override();
1641 }
1642 
1643 void pc_pci_device_init(PCIBus *pci_bus)
1644 {
1645     int max_bus;
1646     int bus;
1647 
1648     max_bus = drive_get_max_bus(IF_SCSI);
1649     for (bus = 0; bus <= max_bus; bus++) {
1650         pci_create_simple(pci_bus, -1, "lsi53c895a");
1651     }
1652 }
1653 
1654 void ioapic_init_gsi(GSIState *gsi_state, const char *parent_name)
1655 {
1656     DeviceState *dev;
1657     SysBusDevice *d;
1658     unsigned int i;
1659 
1660     if (kvm_ioapic_in_kernel()) {
1661         dev = qdev_create(NULL, "kvm-ioapic");
1662     } else {
1663         dev = qdev_create(NULL, "ioapic");
1664     }
1665     if (parent_name) {
1666         object_property_add_child(object_resolve_path(parent_name, NULL),
1667                                   "ioapic", OBJECT(dev), NULL);
1668     }
1669     qdev_init_nofail(dev);
1670     d = SYS_BUS_DEVICE(dev);
1671     sysbus_mmio_map(d, 0, IO_APIC_DEFAULT_ADDRESS);
1672 
1673     for (i = 0; i < IOAPIC_NUM_PINS; i++) {
1674         gsi_state->ioapic_irq[i] = qdev_get_gpio_in(dev, i);
1675     }
1676 }
1677 
1678 static void pc_dimm_plug(HotplugHandler *hotplug_dev,
1679                          DeviceState *dev, Error **errp)
1680 {
1681     HotplugHandlerClass *hhc;
1682     Error *local_err = NULL;
1683     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1684     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
1685     PCDIMMDevice *dimm = PC_DIMM(dev);
1686     PCDIMMDeviceClass *ddc = PC_DIMM_GET_CLASS(dimm);
1687     MemoryRegion *mr = ddc->get_memory_region(dimm);
1688     uint64_t align = TARGET_PAGE_SIZE;
1689 
1690     if (memory_region_get_alignment(mr) && pcmc->enforce_aligned_dimm) {
1691         align = memory_region_get_alignment(mr);
1692     }
1693 
1694     if (!pcms->acpi_dev) {
1695         error_setg(&local_err,
1696                    "memory hotplug is not enabled: missing acpi device");
1697         goto out;
1698     }
1699 
1700     pc_dimm_memory_plug(dev, &pcms->hotplug_memory, mr, align, &local_err);
1701     if (local_err) {
1702         goto out;
1703     }
1704 
1705     hhc = HOTPLUG_HANDLER_GET_CLASS(pcms->acpi_dev);
1706     hhc->plug(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &error_abort);
1707 out:
1708     error_propagate(errp, local_err);
1709 }
1710 
1711 static void pc_dimm_unplug_request(HotplugHandler *hotplug_dev,
1712                                    DeviceState *dev, Error **errp)
1713 {
1714     HotplugHandlerClass *hhc;
1715     Error *local_err = NULL;
1716     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1717 
1718     if (!pcms->acpi_dev) {
1719         error_setg(&local_err,
1720                    "memory hotplug is not enabled: missing acpi device");
1721         goto out;
1722     }
1723 
1724     hhc = HOTPLUG_HANDLER_GET_CLASS(pcms->acpi_dev);
1725     hhc->unplug_request(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &local_err);
1726 
1727 out:
1728     error_propagate(errp, local_err);
1729 }
1730 
1731 static void pc_dimm_unplug(HotplugHandler *hotplug_dev,
1732                            DeviceState *dev, Error **errp)
1733 {
1734     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1735     PCDIMMDevice *dimm = PC_DIMM(dev);
1736     PCDIMMDeviceClass *ddc = PC_DIMM_GET_CLASS(dimm);
1737     MemoryRegion *mr = ddc->get_memory_region(dimm);
1738     HotplugHandlerClass *hhc;
1739     Error *local_err = NULL;
1740 
1741     hhc = HOTPLUG_HANDLER_GET_CLASS(pcms->acpi_dev);
1742     hhc->unplug(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &local_err);
1743 
1744     if (local_err) {
1745         goto out;
1746     }
1747 
1748     pc_dimm_memory_unplug(dev, &pcms->hotplug_memory, mr);
1749     object_unparent(OBJECT(dev));
1750 
1751  out:
1752     error_propagate(errp, local_err);
1753 }
1754 
1755 static int pc_apic_cmp(const void *a, const void *b)
1756 {
1757    CPUArchId *apic_a = (CPUArchId *)a;
1758    CPUArchId *apic_b = (CPUArchId *)b;
1759 
1760    return apic_a->arch_id - apic_b->arch_id;
1761 }
1762 
1763 /* returns pointer to CPUArchId descriptor that matches CPU's apic_id
1764  * in pcms->possible_cpus->cpus, if pcms->possible_cpus->cpus has no
1765  * entry correponding to CPU's apic_id returns NULL.
1766  */
1767 static CPUArchId *pc_find_cpu_slot(PCMachineState *pcms, CPUState *cpu,
1768                                    int *idx)
1769 {
1770     CPUClass *cc = CPU_GET_CLASS(cpu);
1771     CPUArchId apic_id, *found_cpu;
1772 
1773     apic_id.arch_id = cc->get_arch_id(CPU(cpu));
1774     found_cpu = bsearch(&apic_id, pcms->possible_cpus->cpus,
1775         pcms->possible_cpus->len, sizeof(*pcms->possible_cpus->cpus),
1776         pc_apic_cmp);
1777     if (found_cpu && idx) {
1778         *idx = found_cpu - pcms->possible_cpus->cpus;
1779     }
1780     return found_cpu;
1781 }
1782 
1783 static void pc_cpu_plug(HotplugHandler *hotplug_dev,
1784                         DeviceState *dev, Error **errp)
1785 {
1786     CPUArchId *found_cpu;
1787     HotplugHandlerClass *hhc;
1788     Error *local_err = NULL;
1789     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1790 
1791     if (pcms->acpi_dev) {
1792         hhc = HOTPLUG_HANDLER_GET_CLASS(pcms->acpi_dev);
1793         hhc->plug(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &local_err);
1794         if (local_err) {
1795             goto out;
1796         }
1797     }
1798 
1799     if (dev->hotplugged) {
1800         /* increment the number of CPUs */
1801         rtc_set_memory(pcms->rtc, 0x5f, rtc_get_memory(pcms->rtc, 0x5f) + 1);
1802     }
1803 
1804     found_cpu = pc_find_cpu_slot(pcms, CPU(dev), NULL);
1805     found_cpu->cpu = CPU(dev);
1806 out:
1807     error_propagate(errp, local_err);
1808 }
1809 static void pc_cpu_unplug_request_cb(HotplugHandler *hotplug_dev,
1810                                      DeviceState *dev, Error **errp)
1811 {
1812     int idx = -1;
1813     HotplugHandlerClass *hhc;
1814     Error *local_err = NULL;
1815     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1816 
1817     pc_find_cpu_slot(pcms, CPU(dev), &idx);
1818     assert(idx != -1);
1819     if (idx == 0) {
1820         error_setg(&local_err, "Boot CPU is unpluggable");
1821         goto out;
1822     }
1823 
1824     hhc = HOTPLUG_HANDLER_GET_CLASS(pcms->acpi_dev);
1825     hhc->unplug_request(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &local_err);
1826 
1827     if (local_err) {
1828         goto out;
1829     }
1830 
1831  out:
1832     error_propagate(errp, local_err);
1833 
1834 }
1835 
1836 static void pc_cpu_unplug_cb(HotplugHandler *hotplug_dev,
1837                              DeviceState *dev, Error **errp)
1838 {
1839     CPUArchId *found_cpu;
1840     HotplugHandlerClass *hhc;
1841     Error *local_err = NULL;
1842     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1843 
1844     hhc = HOTPLUG_HANDLER_GET_CLASS(pcms->acpi_dev);
1845     hhc->unplug(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &local_err);
1846 
1847     if (local_err) {
1848         goto out;
1849     }
1850 
1851     found_cpu = pc_find_cpu_slot(pcms, CPU(dev), NULL);
1852     found_cpu->cpu = NULL;
1853     object_unparent(OBJECT(dev));
1854 
1855     rtc_set_memory(pcms->rtc, 0x5f, rtc_get_memory(pcms->rtc, 0x5f) - 1);
1856  out:
1857     error_propagate(errp, local_err);
1858 }
1859 
1860 static void pc_cpu_pre_plug(HotplugHandler *hotplug_dev,
1861                             DeviceState *dev, Error **errp)
1862 {
1863     int idx;
1864     CPUState *cs;
1865     CPUArchId *cpu_slot;
1866     X86CPUTopoInfo topo;
1867     X86CPU *cpu = X86_CPU(dev);
1868     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1869 
1870     /* if APIC ID is not set, set it based on socket/core/thread properties */
1871     if (cpu->apic_id == UNASSIGNED_APIC_ID) {
1872         int max_socket = (max_cpus - 1) / smp_threads / smp_cores;
1873 
1874         if (cpu->socket_id < 0) {
1875             error_setg(errp, "CPU socket-id is not set");
1876             return;
1877         } else if (cpu->socket_id > max_socket) {
1878             error_setg(errp, "Invalid CPU socket-id: %u must be in range 0:%u",
1879                        cpu->socket_id, max_socket);
1880             return;
1881         }
1882         if (cpu->core_id < 0) {
1883             error_setg(errp, "CPU core-id is not set");
1884             return;
1885         } else if (cpu->core_id > (smp_cores - 1)) {
1886             error_setg(errp, "Invalid CPU core-id: %u must be in range 0:%u",
1887                        cpu->core_id, smp_cores - 1);
1888             return;
1889         }
1890         if (cpu->thread_id < 0) {
1891             error_setg(errp, "CPU thread-id is not set");
1892             return;
1893         } else if (cpu->thread_id > (smp_threads - 1)) {
1894             error_setg(errp, "Invalid CPU thread-id: %u must be in range 0:%u",
1895                        cpu->thread_id, smp_threads - 1);
1896             return;
1897         }
1898 
1899         topo.pkg_id = cpu->socket_id;
1900         topo.core_id = cpu->core_id;
1901         topo.smt_id = cpu->thread_id;
1902         cpu->apic_id = apicid_from_topo_ids(smp_cores, smp_threads, &topo);
1903     }
1904 
1905     cpu_slot = pc_find_cpu_slot(pcms, CPU(dev), &idx);
1906     if (!cpu_slot) {
1907         x86_topo_ids_from_apicid(cpu->apic_id, smp_cores, smp_threads, &topo);
1908         error_setg(errp, "Invalid CPU [socket: %u, core: %u, thread: %u] with"
1909                   " APIC ID %" PRIu32 ", valid index range 0:%d",
1910                    topo.pkg_id, topo.core_id, topo.smt_id, cpu->apic_id,
1911                    pcms->possible_cpus->len - 1);
1912         return;
1913     }
1914 
1915     if (cpu_slot->cpu) {
1916         error_setg(errp, "CPU[%d] with APIC ID %" PRIu32 " exists",
1917                    idx, cpu->apic_id);
1918         return;
1919     }
1920 
1921     /* if 'address' properties socket-id/core-id/thread-id are not set, set them
1922      * so that query_hotpluggable_cpus would show correct values
1923      */
1924     /* TODO: move socket_id/core_id/thread_id checks into x86_cpu_realizefn()
1925      * once -smp refactoring is complete and there will be CPU private
1926      * CPUState::nr_cores and CPUState::nr_threads fields instead of globals */
1927     x86_topo_ids_from_apicid(cpu->apic_id, smp_cores, smp_threads, &topo);
1928     if (cpu->socket_id != -1 && cpu->socket_id != topo.pkg_id) {
1929         error_setg(errp, "property socket-id: %u doesn't match set apic-id:"
1930             " 0x%x (socket-id: %u)", cpu->socket_id, cpu->apic_id, topo.pkg_id);
1931         return;
1932     }
1933     cpu->socket_id = topo.pkg_id;
1934 
1935     if (cpu->core_id != -1 && cpu->core_id != topo.core_id) {
1936         error_setg(errp, "property core-id: %u doesn't match set apic-id:"
1937             " 0x%x (core-id: %u)", cpu->core_id, cpu->apic_id, topo.core_id);
1938         return;
1939     }
1940     cpu->core_id = topo.core_id;
1941 
1942     if (cpu->thread_id != -1 && cpu->thread_id != topo.smt_id) {
1943         error_setg(errp, "property thread-id: %u doesn't match set apic-id:"
1944             " 0x%x (thread-id: %u)", cpu->thread_id, cpu->apic_id, topo.smt_id);
1945         return;
1946     }
1947     cpu->thread_id = topo.smt_id;
1948 
1949     cs = CPU(cpu);
1950     cs->cpu_index = idx;
1951 }
1952 
1953 static void pc_machine_device_pre_plug_cb(HotplugHandler *hotplug_dev,
1954                                           DeviceState *dev, Error **errp)
1955 {
1956     if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
1957         pc_cpu_pre_plug(hotplug_dev, dev, errp);
1958     }
1959 }
1960 
1961 static void pc_machine_device_plug_cb(HotplugHandler *hotplug_dev,
1962                                       DeviceState *dev, Error **errp)
1963 {
1964     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
1965         pc_dimm_plug(hotplug_dev, dev, errp);
1966     } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
1967         pc_cpu_plug(hotplug_dev, dev, errp);
1968     }
1969 }
1970 
1971 static void pc_machine_device_unplug_request_cb(HotplugHandler *hotplug_dev,
1972                                                 DeviceState *dev, Error **errp)
1973 {
1974     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
1975         pc_dimm_unplug_request(hotplug_dev, dev, errp);
1976     } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
1977         pc_cpu_unplug_request_cb(hotplug_dev, dev, errp);
1978     } else {
1979         error_setg(errp, "acpi: device unplug request for not supported device"
1980                    " type: %s", object_get_typename(OBJECT(dev)));
1981     }
1982 }
1983 
1984 static void pc_machine_device_unplug_cb(HotplugHandler *hotplug_dev,
1985                                         DeviceState *dev, Error **errp)
1986 {
1987     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
1988         pc_dimm_unplug(hotplug_dev, dev, errp);
1989     } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
1990         pc_cpu_unplug_cb(hotplug_dev, dev, errp);
1991     } else {
1992         error_setg(errp, "acpi: device unplug for not supported device"
1993                    " type: %s", object_get_typename(OBJECT(dev)));
1994     }
1995 }
1996 
1997 static HotplugHandler *pc_get_hotpug_handler(MachineState *machine,
1998                                              DeviceState *dev)
1999 {
2000     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(machine);
2001 
2002     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM) ||
2003         object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
2004         return HOTPLUG_HANDLER(machine);
2005     }
2006 
2007     return pcmc->get_hotplug_handler ?
2008         pcmc->get_hotplug_handler(machine, dev) : NULL;
2009 }
2010 
2011 static void
2012 pc_machine_get_hotplug_memory_region_size(Object *obj, Visitor *v,
2013                                           const char *name, void *opaque,
2014                                           Error **errp)
2015 {
2016     PCMachineState *pcms = PC_MACHINE(obj);
2017     int64_t value = memory_region_size(&pcms->hotplug_memory.mr);
2018 
2019     visit_type_int(v, name, &value, errp);
2020 }
2021 
2022 static void pc_machine_get_max_ram_below_4g(Object *obj, Visitor *v,
2023                                             const char *name, void *opaque,
2024                                             Error **errp)
2025 {
2026     PCMachineState *pcms = PC_MACHINE(obj);
2027     uint64_t value = pcms->max_ram_below_4g;
2028 
2029     visit_type_size(v, name, &value, errp);
2030 }
2031 
2032 static void pc_machine_set_max_ram_below_4g(Object *obj, Visitor *v,
2033                                             const char *name, void *opaque,
2034                                             Error **errp)
2035 {
2036     PCMachineState *pcms = PC_MACHINE(obj);
2037     Error *error = NULL;
2038     uint64_t value;
2039 
2040     visit_type_size(v, name, &value, &error);
2041     if (error) {
2042         error_propagate(errp, error);
2043         return;
2044     }
2045     if (value > (1ULL << 32)) {
2046         error_setg(&error,
2047                    "Machine option 'max-ram-below-4g=%"PRIu64
2048                    "' expects size less than or equal to 4G", value);
2049         error_propagate(errp, error);
2050         return;
2051     }
2052 
2053     if (value < (1ULL << 20)) {
2054         error_report("Warning: small max_ram_below_4g(%"PRIu64
2055                      ") less than 1M.  BIOS may not work..",
2056                      value);
2057     }
2058 
2059     pcms->max_ram_below_4g = value;
2060 }
2061 
2062 static void pc_machine_get_vmport(Object *obj, Visitor *v, const char *name,
2063                                   void *opaque, Error **errp)
2064 {
2065     PCMachineState *pcms = PC_MACHINE(obj);
2066     OnOffAuto vmport = pcms->vmport;
2067 
2068     visit_type_OnOffAuto(v, name, &vmport, errp);
2069 }
2070 
2071 static void pc_machine_set_vmport(Object *obj, Visitor *v, const char *name,
2072                                   void *opaque, Error **errp)
2073 {
2074     PCMachineState *pcms = PC_MACHINE(obj);
2075 
2076     visit_type_OnOffAuto(v, name, &pcms->vmport, errp);
2077 }
2078 
2079 bool pc_machine_is_smm_enabled(PCMachineState *pcms)
2080 {
2081     bool smm_available = false;
2082 
2083     if (pcms->smm == ON_OFF_AUTO_OFF) {
2084         return false;
2085     }
2086 
2087     if (tcg_enabled() || qtest_enabled()) {
2088         smm_available = true;
2089     } else if (kvm_enabled()) {
2090         smm_available = kvm_has_smm();
2091     }
2092 
2093     if (smm_available) {
2094         return true;
2095     }
2096 
2097     if (pcms->smm == ON_OFF_AUTO_ON) {
2098         error_report("System Management Mode not supported by this hypervisor.");
2099         exit(1);
2100     }
2101     return false;
2102 }
2103 
2104 static void pc_machine_get_smm(Object *obj, Visitor *v, const char *name,
2105                                void *opaque, Error **errp)
2106 {
2107     PCMachineState *pcms = PC_MACHINE(obj);
2108     OnOffAuto smm = pcms->smm;
2109 
2110     visit_type_OnOffAuto(v, name, &smm, errp);
2111 }
2112 
2113 static void pc_machine_set_smm(Object *obj, Visitor *v, const char *name,
2114                                void *opaque, Error **errp)
2115 {
2116     PCMachineState *pcms = PC_MACHINE(obj);
2117 
2118     visit_type_OnOffAuto(v, name, &pcms->smm, errp);
2119 }
2120 
2121 static bool pc_machine_get_nvdimm(Object *obj, Error **errp)
2122 {
2123     PCMachineState *pcms = PC_MACHINE(obj);
2124 
2125     return pcms->acpi_nvdimm_state.is_enabled;
2126 }
2127 
2128 static void pc_machine_set_nvdimm(Object *obj, bool value, Error **errp)
2129 {
2130     PCMachineState *pcms = PC_MACHINE(obj);
2131 
2132     pcms->acpi_nvdimm_state.is_enabled = value;
2133 }
2134 
2135 static void pc_machine_initfn(Object *obj)
2136 {
2137     PCMachineState *pcms = PC_MACHINE(obj);
2138 
2139     object_property_add(obj, PC_MACHINE_MEMHP_REGION_SIZE, "int",
2140                         pc_machine_get_hotplug_memory_region_size,
2141                         NULL, NULL, NULL, &error_abort);
2142 
2143     pcms->max_ram_below_4g = 0; /* use default */
2144     object_property_add(obj, PC_MACHINE_MAX_RAM_BELOW_4G, "size",
2145                         pc_machine_get_max_ram_below_4g,
2146                         pc_machine_set_max_ram_below_4g,
2147                         NULL, NULL, &error_abort);
2148     object_property_set_description(obj, PC_MACHINE_MAX_RAM_BELOW_4G,
2149                                     "Maximum ram below the 4G boundary (32bit boundary)",
2150                                     &error_abort);
2151 
2152     pcms->smm = ON_OFF_AUTO_AUTO;
2153     object_property_add(obj, PC_MACHINE_SMM, "OnOffAuto",
2154                         pc_machine_get_smm,
2155                         pc_machine_set_smm,
2156                         NULL, NULL, &error_abort);
2157     object_property_set_description(obj, PC_MACHINE_SMM,
2158                                     "Enable SMM (pc & q35)",
2159                                     &error_abort);
2160 
2161     pcms->vmport = ON_OFF_AUTO_AUTO;
2162     object_property_add(obj, PC_MACHINE_VMPORT, "OnOffAuto",
2163                         pc_machine_get_vmport,
2164                         pc_machine_set_vmport,
2165                         NULL, NULL, &error_abort);
2166     object_property_set_description(obj, PC_MACHINE_VMPORT,
2167                                     "Enable vmport (pc & q35)",
2168                                     &error_abort);
2169 
2170     /* nvdimm is disabled on default. */
2171     pcms->acpi_nvdimm_state.is_enabled = false;
2172     object_property_add_bool(obj, PC_MACHINE_NVDIMM, pc_machine_get_nvdimm,
2173                              pc_machine_set_nvdimm, &error_abort);
2174 }
2175 
2176 static void pc_machine_reset(void)
2177 {
2178     CPUState *cs;
2179     X86CPU *cpu;
2180 
2181     qemu_devices_reset();
2182 
2183     /* Reset APIC after devices have been reset to cancel
2184      * any changes that qemu_devices_reset() might have done.
2185      */
2186     CPU_FOREACH(cs) {
2187         cpu = X86_CPU(cs);
2188 
2189         if (cpu->apic_state) {
2190             device_reset(cpu->apic_state);
2191         }
2192     }
2193 }
2194 
2195 static unsigned pc_cpu_index_to_socket_id(unsigned cpu_index)
2196 {
2197     X86CPUTopoInfo topo;
2198     x86_topo_ids_from_idx(smp_cores, smp_threads, cpu_index,
2199                           &topo);
2200     return topo.pkg_id;
2201 }
2202 
2203 static CPUArchIdList *pc_possible_cpu_arch_ids(MachineState *machine)
2204 {
2205     PCMachineState *pcms = PC_MACHINE(machine);
2206     int len = sizeof(CPUArchIdList) +
2207               sizeof(CPUArchId) * (pcms->possible_cpus->len);
2208     CPUArchIdList *list = g_malloc(len);
2209 
2210     memcpy(list, pcms->possible_cpus, len);
2211     return list;
2212 }
2213 
2214 static HotpluggableCPUList *pc_query_hotpluggable_cpus(MachineState *machine)
2215 {
2216     int i;
2217     CPUState *cpu;
2218     HotpluggableCPUList *head = NULL;
2219     PCMachineState *pcms = PC_MACHINE(machine);
2220     const char *cpu_type;
2221 
2222     cpu = pcms->possible_cpus->cpus[0].cpu;
2223     assert(cpu); /* BSP is always present */
2224     cpu_type = object_class_get_name(OBJECT_CLASS(CPU_GET_CLASS(cpu)));
2225 
2226     for (i = 0; i < pcms->possible_cpus->len; i++) {
2227         X86CPUTopoInfo topo;
2228         HotpluggableCPUList *list_item = g_new0(typeof(*list_item), 1);
2229         HotpluggableCPU *cpu_item = g_new0(typeof(*cpu_item), 1);
2230         CpuInstanceProperties *cpu_props = g_new0(typeof(*cpu_props), 1);
2231         const uint32_t apic_id = pcms->possible_cpus->cpus[i].arch_id;
2232 
2233         x86_topo_ids_from_apicid(apic_id, smp_cores, smp_threads, &topo);
2234 
2235         cpu_item->type = g_strdup(cpu_type);
2236         cpu_item->vcpus_count = 1;
2237         cpu_props->has_socket_id = true;
2238         cpu_props->socket_id = topo.pkg_id;
2239         cpu_props->has_core_id = true;
2240         cpu_props->core_id = topo.core_id;
2241         cpu_props->has_thread_id = true;
2242         cpu_props->thread_id = topo.smt_id;
2243         cpu_item->props = cpu_props;
2244 
2245         cpu = pcms->possible_cpus->cpus[i].cpu;
2246         if (cpu) {
2247             cpu_item->has_qom_path = true;
2248             cpu_item->qom_path = object_get_canonical_path(OBJECT(cpu));
2249         }
2250 
2251         list_item->value = cpu_item;
2252         list_item->next = head;
2253         head = list_item;
2254     }
2255     return head;
2256 }
2257 
2258 static void x86_nmi(NMIState *n, int cpu_index, Error **errp)
2259 {
2260     /* cpu index isn't used */
2261     CPUState *cs;
2262 
2263     CPU_FOREACH(cs) {
2264         X86CPU *cpu = X86_CPU(cs);
2265 
2266         if (!cpu->apic_state) {
2267             cpu_interrupt(cs, CPU_INTERRUPT_NMI);
2268         } else {
2269             apic_deliver_nmi(cpu->apic_state);
2270         }
2271     }
2272 }
2273 
2274 static void pc_machine_class_init(ObjectClass *oc, void *data)
2275 {
2276     MachineClass *mc = MACHINE_CLASS(oc);
2277     PCMachineClass *pcmc = PC_MACHINE_CLASS(oc);
2278     HotplugHandlerClass *hc = HOTPLUG_HANDLER_CLASS(oc);
2279     NMIClass *nc = NMI_CLASS(oc);
2280 
2281     pcmc->get_hotplug_handler = mc->get_hotplug_handler;
2282     pcmc->pci_enabled = true;
2283     pcmc->has_acpi_build = true;
2284     pcmc->rsdp_in_ram = true;
2285     pcmc->smbios_defaults = true;
2286     pcmc->smbios_uuid_encoded = true;
2287     pcmc->gigabyte_align = true;
2288     pcmc->has_reserved_memory = true;
2289     pcmc->kvmclock_enabled = true;
2290     pcmc->enforce_aligned_dimm = true;
2291     /* BIOS ACPI tables: 128K. Other BIOS datastructures: less than 4K reported
2292      * to be used at the moment, 32K should be enough for a while.  */
2293     pcmc->acpi_data_size = 0x20000 + 0x8000;
2294     pcmc->save_tsc_khz = true;
2295     mc->get_hotplug_handler = pc_get_hotpug_handler;
2296     mc->cpu_index_to_socket_id = pc_cpu_index_to_socket_id;
2297     mc->possible_cpu_arch_ids = pc_possible_cpu_arch_ids;
2298     mc->query_hotpluggable_cpus = pc_query_hotpluggable_cpus;
2299     mc->default_boot_order = "cad";
2300     mc->hot_add_cpu = pc_hot_add_cpu;
2301     mc->max_cpus = 255;
2302     mc->reset = pc_machine_reset;
2303     hc->pre_plug = pc_machine_device_pre_plug_cb;
2304     hc->plug = pc_machine_device_plug_cb;
2305     hc->unplug_request = pc_machine_device_unplug_request_cb;
2306     hc->unplug = pc_machine_device_unplug_cb;
2307     nc->nmi_monitor_handler = x86_nmi;
2308 }
2309 
2310 static const TypeInfo pc_machine_info = {
2311     .name = TYPE_PC_MACHINE,
2312     .parent = TYPE_MACHINE,
2313     .abstract = true,
2314     .instance_size = sizeof(PCMachineState),
2315     .instance_init = pc_machine_initfn,
2316     .class_size = sizeof(PCMachineClass),
2317     .class_init = pc_machine_class_init,
2318     .interfaces = (InterfaceInfo[]) {
2319          { TYPE_HOTPLUG_HANDLER },
2320          { TYPE_NMI },
2321          { }
2322     },
2323 };
2324 
2325 static void pc_machine_register_types(void)
2326 {
2327     type_register_static(&pc_machine_info);
2328 }
2329 
2330 type_init(pc_machine_register_types)
2331