xref: /qemu/hw/i386/pc.c (revision 67cc32eb)
1 /*
2  * QEMU PC System Emulator
3  *
4  * Copyright (c) 2003-2004 Fabrice Bellard
5  *
6  * Permission is hereby granted, free of charge, to any person obtaining a copy
7  * of this software and associated documentation files (the "Software"), to deal
8  * in the Software without restriction, including without limitation the rights
9  * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
10  * copies of the Software, and to permit persons to whom the Software is
11  * furnished to do so, subject to the following conditions:
12  *
13  * The above copyright notice and this permission notice shall be included in
14  * all copies or substantial portions of the Software.
15  *
16  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
17  * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
18  * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
19  * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
20  * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
21  * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
22  * THE SOFTWARE.
23  */
24 #include "hw/hw.h"
25 #include "hw/i386/pc.h"
26 #include "hw/char/serial.h"
27 #include "hw/i386/apic.h"
28 #include "hw/i386/topology.h"
29 #include "sysemu/cpus.h"
30 #include "hw/block/fdc.h"
31 #include "hw/ide.h"
32 #include "hw/pci/pci.h"
33 #include "hw/pci/pci_bus.h"
34 #include "hw/nvram/fw_cfg.h"
35 #include "hw/timer/hpet.h"
36 #include "hw/smbios/smbios.h"
37 #include "hw/loader.h"
38 #include "elf.h"
39 #include "multiboot.h"
40 #include "hw/timer/mc146818rtc.h"
41 #include "hw/timer/i8254.h"
42 #include "hw/audio/pcspk.h"
43 #include "hw/pci/msi.h"
44 #include "hw/sysbus.h"
45 #include "sysemu/sysemu.h"
46 #include "sysemu/numa.h"
47 #include "sysemu/kvm.h"
48 #include "sysemu/qtest.h"
49 #include "kvm_i386.h"
50 #include "hw/xen/xen.h"
51 #include "sysemu/block-backend.h"
52 #include "hw/block/block.h"
53 #include "ui/qemu-spice.h"
54 #include "exec/memory.h"
55 #include "exec/address-spaces.h"
56 #include "sysemu/arch_init.h"
57 #include "qemu/bitmap.h"
58 #include "qemu/config-file.h"
59 #include "qemu/error-report.h"
60 #include "hw/acpi/acpi.h"
61 #include "hw/acpi/cpu_hotplug.h"
62 #include "hw/cpu/icc_bus.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 
70 /* debug PC/ISA interrupts */
71 //#define DEBUG_IRQ
72 
73 #ifdef DEBUG_IRQ
74 #define DPRINTF(fmt, ...)                                       \
75     do { printf("CPUIRQ: " fmt , ## __VA_ARGS__); } while (0)
76 #else
77 #define DPRINTF(fmt, ...)
78 #endif
79 
80 /* Leave a chunk of memory at the top of RAM for the BIOS ACPI tables
81  * (128K) and other BIOS datastructures (less than 4K reported to be used at
82  * the moment, 32K should be enough for a while).  */
83 static unsigned acpi_data_size = 0x20000 + 0x8000;
84 void pc_set_legacy_acpi_data_size(void)
85 {
86     acpi_data_size = 0x10000;
87 }
88 
89 #define BIOS_CFG_IOPORT 0x510
90 #define FW_CFG_ACPI_TABLES (FW_CFG_ARCH_LOCAL + 0)
91 #define FW_CFG_SMBIOS_ENTRIES (FW_CFG_ARCH_LOCAL + 1)
92 #define FW_CFG_IRQ0_OVERRIDE (FW_CFG_ARCH_LOCAL + 2)
93 #define FW_CFG_E820_TABLE (FW_CFG_ARCH_LOCAL + 3)
94 #define FW_CFG_HPET (FW_CFG_ARCH_LOCAL + 4)
95 
96 #define E820_NR_ENTRIES		16
97 
98 struct e820_entry {
99     uint64_t address;
100     uint64_t length;
101     uint32_t type;
102 } QEMU_PACKED __attribute((__aligned__(4)));
103 
104 struct e820_table {
105     uint32_t count;
106     struct e820_entry entry[E820_NR_ENTRIES];
107 } QEMU_PACKED __attribute((__aligned__(4)));
108 
109 static struct e820_table e820_reserve;
110 static struct e820_entry *e820_table;
111 static unsigned e820_entries;
112 struct hpet_fw_config hpet_cfg = {.count = UINT8_MAX};
113 
114 void gsi_handler(void *opaque, int n, int level)
115 {
116     GSIState *s = opaque;
117 
118     DPRINTF("pc: %s GSI %d\n", level ? "raising" : "lowering", n);
119     if (n < ISA_NUM_IRQS) {
120         qemu_set_irq(s->i8259_irq[n], level);
121     }
122     qemu_set_irq(s->ioapic_irq[n], level);
123 }
124 
125 static void ioport80_write(void *opaque, hwaddr addr, uint64_t data,
126                            unsigned size)
127 {
128 }
129 
130 static uint64_t ioport80_read(void *opaque, hwaddr addr, unsigned size)
131 {
132     return 0xffffffffffffffffULL;
133 }
134 
135 /* MSDOS compatibility mode FPU exception support */
136 static qemu_irq ferr_irq;
137 
138 void pc_register_ferr_irq(qemu_irq irq)
139 {
140     ferr_irq = irq;
141 }
142 
143 /* XXX: add IGNNE support */
144 void cpu_set_ferr(CPUX86State *s)
145 {
146     qemu_irq_raise(ferr_irq);
147 }
148 
149 static void ioportF0_write(void *opaque, hwaddr addr, uint64_t data,
150                            unsigned size)
151 {
152     qemu_irq_lower(ferr_irq);
153 }
154 
155 static uint64_t ioportF0_read(void *opaque, hwaddr addr, unsigned size)
156 {
157     return 0xffffffffffffffffULL;
158 }
159 
160 /* TSC handling */
161 uint64_t cpu_get_tsc(CPUX86State *env)
162 {
163     return cpu_get_ticks();
164 }
165 
166 /* IRQ handling */
167 int cpu_get_pic_interrupt(CPUX86State *env)
168 {
169     X86CPU *cpu = x86_env_get_cpu(env);
170     int intno;
171 
172     intno = apic_get_interrupt(cpu->apic_state);
173     if (intno >= 0) {
174         return intno;
175     }
176     /* read the irq from the PIC */
177     if (!apic_accept_pic_intr(cpu->apic_state)) {
178         return -1;
179     }
180 
181     intno = pic_read_irq(isa_pic);
182     return intno;
183 }
184 
185 static void pic_irq_request(void *opaque, int irq, int level)
186 {
187     CPUState *cs = first_cpu;
188     X86CPU *cpu = X86_CPU(cs);
189 
190     DPRINTF("pic_irqs: %s irq %d\n", level? "raise" : "lower", irq);
191     if (cpu->apic_state) {
192         CPU_FOREACH(cs) {
193             cpu = X86_CPU(cs);
194             if (apic_accept_pic_intr(cpu->apic_state)) {
195                 apic_deliver_pic_intr(cpu->apic_state, level);
196             }
197         }
198     } else {
199         if (level) {
200             cpu_interrupt(cs, CPU_INTERRUPT_HARD);
201         } else {
202             cpu_reset_interrupt(cs, CPU_INTERRUPT_HARD);
203         }
204     }
205 }
206 
207 /* PC cmos mappings */
208 
209 #define REG_EQUIPMENT_BYTE          0x14
210 
211 static int cmos_get_fd_drive_type(FDriveType fd0)
212 {
213     int val;
214 
215     switch (fd0) {
216     case FDRIVE_DRV_144:
217         /* 1.44 Mb 3"5 drive */
218         val = 4;
219         break;
220     case FDRIVE_DRV_288:
221         /* 2.88 Mb 3"5 drive */
222         val = 5;
223         break;
224     case FDRIVE_DRV_120:
225         /* 1.2 Mb 5"5 drive */
226         val = 2;
227         break;
228     case FDRIVE_DRV_NONE:
229     default:
230         val = 0;
231         break;
232     }
233     return val;
234 }
235 
236 static void cmos_init_hd(ISADevice *s, int type_ofs, int info_ofs,
237                          int16_t cylinders, int8_t heads, int8_t sectors)
238 {
239     rtc_set_memory(s, type_ofs, 47);
240     rtc_set_memory(s, info_ofs, cylinders);
241     rtc_set_memory(s, info_ofs + 1, cylinders >> 8);
242     rtc_set_memory(s, info_ofs + 2, heads);
243     rtc_set_memory(s, info_ofs + 3, 0xff);
244     rtc_set_memory(s, info_ofs + 4, 0xff);
245     rtc_set_memory(s, info_ofs + 5, 0xc0 | ((heads > 8) << 3));
246     rtc_set_memory(s, info_ofs + 6, cylinders);
247     rtc_set_memory(s, info_ofs + 7, cylinders >> 8);
248     rtc_set_memory(s, info_ofs + 8, sectors);
249 }
250 
251 /* convert boot_device letter to something recognizable by the bios */
252 static int boot_device2nibble(char boot_device)
253 {
254     switch(boot_device) {
255     case 'a':
256     case 'b':
257         return 0x01; /* floppy boot */
258     case 'c':
259         return 0x02; /* hard drive boot */
260     case 'd':
261         return 0x03; /* CD-ROM boot */
262     case 'n':
263         return 0x04; /* Network boot */
264     }
265     return 0;
266 }
267 
268 static void set_boot_dev(ISADevice *s, const char *boot_device, Error **errp)
269 {
270 #define PC_MAX_BOOT_DEVICES 3
271     int nbds, bds[3] = { 0, };
272     int i;
273 
274     nbds = strlen(boot_device);
275     if (nbds > PC_MAX_BOOT_DEVICES) {
276         error_setg(errp, "Too many boot devices for PC");
277         return;
278     }
279     for (i = 0; i < nbds; i++) {
280         bds[i] = boot_device2nibble(boot_device[i]);
281         if (bds[i] == 0) {
282             error_setg(errp, "Invalid boot device for PC: '%c'",
283                        boot_device[i]);
284             return;
285         }
286     }
287     rtc_set_memory(s, 0x3d, (bds[1] << 4) | bds[0]);
288     rtc_set_memory(s, 0x38, (bds[2] << 4) | (fd_bootchk ? 0x0 : 0x1));
289 }
290 
291 static void pc_boot_set(void *opaque, const char *boot_device, Error **errp)
292 {
293     set_boot_dev(opaque, boot_device, errp);
294 }
295 
296 static void pc_cmos_init_floppy(ISADevice *rtc_state, ISADevice *floppy)
297 {
298     int val, nb, i;
299     FDriveType fd_type[2] = { FDRIVE_DRV_NONE, FDRIVE_DRV_NONE };
300 
301     /* floppy type */
302     if (floppy) {
303         for (i = 0; i < 2; i++) {
304             fd_type[i] = isa_fdc_get_drive_type(floppy, i);
305         }
306     }
307     val = (cmos_get_fd_drive_type(fd_type[0]) << 4) |
308         cmos_get_fd_drive_type(fd_type[1]);
309     rtc_set_memory(rtc_state, 0x10, val);
310 
311     val = rtc_get_memory(rtc_state, REG_EQUIPMENT_BYTE);
312     nb = 0;
313     if (fd_type[0] < FDRIVE_DRV_NONE) {
314         nb++;
315     }
316     if (fd_type[1] < FDRIVE_DRV_NONE) {
317         nb++;
318     }
319     switch (nb) {
320     case 0:
321         break;
322     case 1:
323         val |= 0x01; /* 1 drive, ready for boot */
324         break;
325     case 2:
326         val |= 0x41; /* 2 drives, ready for boot */
327         break;
328     }
329     rtc_set_memory(rtc_state, REG_EQUIPMENT_BYTE, val);
330 }
331 
332 typedef struct pc_cmos_init_late_arg {
333     ISADevice *rtc_state;
334     BusState *idebus[2];
335 } pc_cmos_init_late_arg;
336 
337 typedef struct check_fdc_state {
338     ISADevice *floppy;
339     bool multiple;
340 } CheckFdcState;
341 
342 static int check_fdc(Object *obj, void *opaque)
343 {
344     CheckFdcState *state = opaque;
345     Object *fdc;
346     uint32_t iobase;
347     Error *local_err = NULL;
348 
349     fdc = object_dynamic_cast(obj, TYPE_ISA_FDC);
350     if (!fdc) {
351         return 0;
352     }
353 
354     iobase = object_property_get_int(obj, "iobase", &local_err);
355     if (local_err || iobase != 0x3f0) {
356         error_free(local_err);
357         return 0;
358     }
359 
360     if (state->floppy) {
361         state->multiple = true;
362     } else {
363         state->floppy = ISA_DEVICE(obj);
364     }
365     return 0;
366 }
367 
368 static const char * const fdc_container_path[] = {
369     "/unattached", "/peripheral", "/peripheral-anon"
370 };
371 
372 static void pc_cmos_init_late(void *opaque)
373 {
374     pc_cmos_init_late_arg *arg = opaque;
375     ISADevice *s = arg->rtc_state;
376     int16_t cylinders;
377     int8_t heads, sectors;
378     int val;
379     int i, trans;
380     Object *container;
381     CheckFdcState state = { 0 };
382 
383     val = 0;
384     if (ide_get_geometry(arg->idebus[0], 0,
385                          &cylinders, &heads, &sectors) >= 0) {
386         cmos_init_hd(s, 0x19, 0x1b, cylinders, heads, sectors);
387         val |= 0xf0;
388     }
389     if (ide_get_geometry(arg->idebus[0], 1,
390                          &cylinders, &heads, &sectors) >= 0) {
391         cmos_init_hd(s, 0x1a, 0x24, cylinders, heads, sectors);
392         val |= 0x0f;
393     }
394     rtc_set_memory(s, 0x12, val);
395 
396     val = 0;
397     for (i = 0; i < 4; i++) {
398         /* NOTE: ide_get_geometry() returns the physical
399            geometry.  It is always such that: 1 <= sects <= 63, 1
400            <= heads <= 16, 1 <= cylinders <= 16383. The BIOS
401            geometry can be different if a translation is done. */
402         if (ide_get_geometry(arg->idebus[i / 2], i % 2,
403                              &cylinders, &heads, &sectors) >= 0) {
404             trans = ide_get_bios_chs_trans(arg->idebus[i / 2], i % 2) - 1;
405             assert((trans & ~3) == 0);
406             val |= trans << (i * 2);
407         }
408     }
409     rtc_set_memory(s, 0x39, val);
410 
411     /*
412      * Locate the FDC at IO address 0x3f0, and configure the CMOS registers
413      * accordingly.
414      */
415     for (i = 0; i < ARRAY_SIZE(fdc_container_path); i++) {
416         container = container_get(qdev_get_machine(), fdc_container_path[i]);
417         object_child_foreach(container, check_fdc, &state);
418     }
419 
420     if (state.multiple) {
421         error_report("warning: multiple floppy disk controllers with "
422                      "iobase=0x3f0 have been found;\n"
423                      "the one being picked for CMOS setup might not reflect "
424                      "your intent");
425     }
426     pc_cmos_init_floppy(s, state.floppy);
427 
428     qemu_unregister_reset(pc_cmos_init_late, opaque);
429 }
430 
431 void pc_cmos_init(PCMachineState *pcms,
432                   BusState *idebus0, BusState *idebus1,
433                   ISADevice *s)
434 {
435     int val;
436     static pc_cmos_init_late_arg arg;
437     Error *local_err = NULL;
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, &local_err);
486     if (local_err) {
487         error_report_err(local_err);
488         exit(1);
489     }
490 
491     val = 0;
492     val |= 0x02; /* FPU is there */
493     val |= 0x04; /* PS/2 mouse installed */
494     rtc_set_memory(s, REG_EQUIPMENT_BYTE, val);
495 
496     /* hard drives and FDC */
497     arg.rtc_state = s;
498     arg.idebus[0] = idebus0;
499     arg.idebus[1] = idebus1;
500     qemu_register_reset(pc_cmos_init_late, &arg);
501 }
502 
503 #define TYPE_PORT92 "port92"
504 #define PORT92(obj) OBJECT_CHECK(Port92State, (obj), TYPE_PORT92)
505 
506 /* port 92 stuff: could be split off */
507 typedef struct Port92State {
508     ISADevice parent_obj;
509 
510     MemoryRegion io;
511     uint8_t outport;
512     qemu_irq *a20_out;
513 } Port92State;
514 
515 static void port92_write(void *opaque, hwaddr addr, uint64_t val,
516                          unsigned size)
517 {
518     Port92State *s = opaque;
519     int oldval = s->outport;
520 
521     DPRINTF("port92: write 0x%02" PRIx64 "\n", val);
522     s->outport = val;
523     qemu_set_irq(*s->a20_out, (val >> 1) & 1);
524     if ((val & 1) && !(oldval & 1)) {
525         qemu_system_reset_request();
526     }
527 }
528 
529 static uint64_t port92_read(void *opaque, hwaddr addr,
530                             unsigned size)
531 {
532     Port92State *s = opaque;
533     uint32_t ret;
534 
535     ret = s->outport;
536     DPRINTF("port92: read 0x%02x\n", ret);
537     return ret;
538 }
539 
540 static void port92_init(ISADevice *dev, qemu_irq *a20_out)
541 {
542     Port92State *s = PORT92(dev);
543 
544     s->a20_out = a20_out;
545 }
546 
547 static const VMStateDescription vmstate_port92_isa = {
548     .name = "port92",
549     .version_id = 1,
550     .minimum_version_id = 1,
551     .fields = (VMStateField[]) {
552         VMSTATE_UINT8(outport, Port92State),
553         VMSTATE_END_OF_LIST()
554     }
555 };
556 
557 static void port92_reset(DeviceState *d)
558 {
559     Port92State *s = PORT92(d);
560 
561     s->outport &= ~1;
562 }
563 
564 static const MemoryRegionOps port92_ops = {
565     .read = port92_read,
566     .write = port92_write,
567     .impl = {
568         .min_access_size = 1,
569         .max_access_size = 1,
570     },
571     .endianness = DEVICE_LITTLE_ENDIAN,
572 };
573 
574 static void port92_initfn(Object *obj)
575 {
576     Port92State *s = PORT92(obj);
577 
578     memory_region_init_io(&s->io, OBJECT(s), &port92_ops, s, "port92", 1);
579 
580     s->outport = 0;
581 }
582 
583 static void port92_realizefn(DeviceState *dev, Error **errp)
584 {
585     ISADevice *isadev = ISA_DEVICE(dev);
586     Port92State *s = PORT92(dev);
587 
588     isa_register_ioport(isadev, &s->io, 0x92);
589 }
590 
591 static void port92_class_initfn(ObjectClass *klass, void *data)
592 {
593     DeviceClass *dc = DEVICE_CLASS(klass);
594 
595     dc->realize = port92_realizefn;
596     dc->reset = port92_reset;
597     dc->vmsd = &vmstate_port92_isa;
598     /*
599      * Reason: unlike ordinary ISA devices, this one needs additional
600      * wiring: its A20 output line needs to be wired up by
601      * port92_init().
602      */
603     dc->cannot_instantiate_with_device_add_yet = true;
604 }
605 
606 static const TypeInfo port92_info = {
607     .name          = TYPE_PORT92,
608     .parent        = TYPE_ISA_DEVICE,
609     .instance_size = sizeof(Port92State),
610     .instance_init = port92_initfn,
611     .class_init    = port92_class_initfn,
612 };
613 
614 static void port92_register_types(void)
615 {
616     type_register_static(&port92_info);
617 }
618 
619 type_init(port92_register_types)
620 
621 static void handle_a20_line_change(void *opaque, int irq, int level)
622 {
623     X86CPU *cpu = opaque;
624 
625     /* XXX: send to all CPUs ? */
626     /* XXX: add logic to handle multiple A20 line sources */
627     x86_cpu_set_a20(cpu, level);
628 }
629 
630 int e820_add_entry(uint64_t address, uint64_t length, uint32_t type)
631 {
632     int index = le32_to_cpu(e820_reserve.count);
633     struct e820_entry *entry;
634 
635     if (type != E820_RAM) {
636         /* old FW_CFG_E820_TABLE entry -- reservations only */
637         if (index >= E820_NR_ENTRIES) {
638             return -EBUSY;
639         }
640         entry = &e820_reserve.entry[index++];
641 
642         entry->address = cpu_to_le64(address);
643         entry->length = cpu_to_le64(length);
644         entry->type = cpu_to_le32(type);
645 
646         e820_reserve.count = cpu_to_le32(index);
647     }
648 
649     /* new "etc/e820" file -- include ram too */
650     e820_table = g_renew(struct e820_entry, e820_table, e820_entries + 1);
651     e820_table[e820_entries].address = cpu_to_le64(address);
652     e820_table[e820_entries].length = cpu_to_le64(length);
653     e820_table[e820_entries].type = cpu_to_le32(type);
654     e820_entries++;
655 
656     return e820_entries;
657 }
658 
659 int e820_get_num_entries(void)
660 {
661     return e820_entries;
662 }
663 
664 bool e820_get_entry(int idx, uint32_t type, uint64_t *address, uint64_t *length)
665 {
666     if (idx < e820_entries && e820_table[idx].type == cpu_to_le32(type)) {
667         *address = le64_to_cpu(e820_table[idx].address);
668         *length = le64_to_cpu(e820_table[idx].length);
669         return true;
670     }
671     return false;
672 }
673 
674 /* Enables contiguous-apic-ID mode, for compatibility */
675 static bool compat_apic_id_mode;
676 
677 void enable_compat_apic_id_mode(void)
678 {
679     compat_apic_id_mode = true;
680 }
681 
682 /* Calculates initial APIC ID for a specific CPU index
683  *
684  * Currently we need to be able to calculate the APIC ID from the CPU index
685  * alone (without requiring a CPU object), as the QEMU<->Seabios interfaces have
686  * no concept of "CPU index", and the NUMA tables on fw_cfg need the APIC ID of
687  * all CPUs up to max_cpus.
688  */
689 static uint32_t x86_cpu_apic_id_from_index(unsigned int cpu_index)
690 {
691     uint32_t correct_id;
692     static bool warned;
693 
694     correct_id = x86_apicid_from_cpu_idx(smp_cores, smp_threads, cpu_index);
695     if (compat_apic_id_mode) {
696         if (cpu_index != correct_id && !warned && !qtest_enabled()) {
697             error_report("APIC IDs set in compatibility mode, "
698                          "CPU topology won't match the configuration");
699             warned = true;
700         }
701         return cpu_index;
702     } else {
703         return correct_id;
704     }
705 }
706 
707 /* Calculates the limit to CPU APIC ID values
708  *
709  * This function returns the limit for the APIC ID value, so that all
710  * CPU APIC IDs are < pc_apic_id_limit().
711  *
712  * This is used for FW_CFG_MAX_CPUS. See comments on bochs_bios_init().
713  */
714 static unsigned int pc_apic_id_limit(unsigned int max_cpus)
715 {
716     return x86_cpu_apic_id_from_index(max_cpus - 1) + 1;
717 }
718 
719 static void pc_build_smbios(FWCfgState *fw_cfg)
720 {
721     uint8_t *smbios_tables, *smbios_anchor;
722     size_t smbios_tables_len, smbios_anchor_len;
723     struct smbios_phys_mem_area *mem_array;
724     unsigned i, array_count;
725 
726     smbios_tables = smbios_get_table_legacy(&smbios_tables_len);
727     if (smbios_tables) {
728         fw_cfg_add_bytes(fw_cfg, FW_CFG_SMBIOS_ENTRIES,
729                          smbios_tables, smbios_tables_len);
730     }
731 
732     /* build the array of physical mem area from e820 table */
733     mem_array = g_malloc0(sizeof(*mem_array) * e820_get_num_entries());
734     for (i = 0, array_count = 0; i < e820_get_num_entries(); i++) {
735         uint64_t addr, len;
736 
737         if (e820_get_entry(i, E820_RAM, &addr, &len)) {
738             mem_array[array_count].address = addr;
739             mem_array[array_count].length = len;
740             array_count++;
741         }
742     }
743     smbios_get_tables(mem_array, array_count,
744                       &smbios_tables, &smbios_tables_len,
745                       &smbios_anchor, &smbios_anchor_len);
746     g_free(mem_array);
747 
748     if (smbios_anchor) {
749         fw_cfg_add_file(fw_cfg, "etc/smbios/smbios-tables",
750                         smbios_tables, smbios_tables_len);
751         fw_cfg_add_file(fw_cfg, "etc/smbios/smbios-anchor",
752                         smbios_anchor, smbios_anchor_len);
753     }
754 }
755 
756 static FWCfgState *bochs_bios_init(void)
757 {
758     FWCfgState *fw_cfg;
759     uint64_t *numa_fw_cfg;
760     int i, j;
761     unsigned int apic_id_limit = pc_apic_id_limit(max_cpus);
762 
763     fw_cfg = fw_cfg_init_io(BIOS_CFG_IOPORT);
764     /* FW_CFG_MAX_CPUS is a bit confusing/problematic on x86:
765      *
766      * SeaBIOS needs FW_CFG_MAX_CPUS for CPU hotplug, but the CPU hotplug
767      * QEMU<->SeaBIOS interface is not based on the "CPU index", but on the APIC
768      * ID of hotplugged CPUs[1]. This means that FW_CFG_MAX_CPUS is not the
769      * "maximum number of CPUs", but the "limit to the APIC ID values SeaBIOS
770      * may see".
771      *
772      * So, this means we must not use max_cpus, here, but the maximum possible
773      * APIC ID value, plus one.
774      *
775      * [1] The only kind of "CPU identifier" used between SeaBIOS and QEMU is
776      *     the APIC ID, not the "CPU index"
777      */
778     fw_cfg_add_i16(fw_cfg, FW_CFG_MAX_CPUS, (uint16_t)apic_id_limit);
779     fw_cfg_add_i64(fw_cfg, FW_CFG_RAM_SIZE, (uint64_t)ram_size);
780     fw_cfg_add_bytes(fw_cfg, FW_CFG_ACPI_TABLES,
781                      acpi_tables, acpi_tables_len);
782     fw_cfg_add_i32(fw_cfg, FW_CFG_IRQ0_OVERRIDE, kvm_allows_irq0_override());
783 
784     pc_build_smbios(fw_cfg);
785 
786     fw_cfg_add_bytes(fw_cfg, FW_CFG_E820_TABLE,
787                      &e820_reserve, sizeof(e820_reserve));
788     fw_cfg_add_file(fw_cfg, "etc/e820", e820_table,
789                     sizeof(struct e820_entry) * e820_entries);
790 
791     fw_cfg_add_bytes(fw_cfg, FW_CFG_HPET, &hpet_cfg, sizeof(hpet_cfg));
792     /* allocate memory for the NUMA channel: one (64bit) word for the number
793      * of nodes, one word for each VCPU->node and one word for each node to
794      * hold the amount of memory.
795      */
796     numa_fw_cfg = g_new0(uint64_t, 1 + apic_id_limit + nb_numa_nodes);
797     numa_fw_cfg[0] = cpu_to_le64(nb_numa_nodes);
798     for (i = 0; i < max_cpus; i++) {
799         unsigned int apic_id = x86_cpu_apic_id_from_index(i);
800         assert(apic_id < apic_id_limit);
801         for (j = 0; j < nb_numa_nodes; j++) {
802             if (test_bit(i, numa_info[j].node_cpu)) {
803                 numa_fw_cfg[apic_id + 1] = cpu_to_le64(j);
804                 break;
805             }
806         }
807     }
808     for (i = 0; i < nb_numa_nodes; i++) {
809         numa_fw_cfg[apic_id_limit + 1 + i] = cpu_to_le64(numa_info[i].node_mem);
810     }
811     fw_cfg_add_bytes(fw_cfg, FW_CFG_NUMA, numa_fw_cfg,
812                      (1 + apic_id_limit + nb_numa_nodes) *
813                      sizeof(*numa_fw_cfg));
814 
815     return fw_cfg;
816 }
817 
818 static long get_file_size(FILE *f)
819 {
820     long where, size;
821 
822     /* XXX: on Unix systems, using fstat() probably makes more sense */
823 
824     where = ftell(f);
825     fseek(f, 0, SEEK_END);
826     size = ftell(f);
827     fseek(f, where, SEEK_SET);
828 
829     return size;
830 }
831 
832 static void load_linux(PCMachineState *pcms,
833                        FWCfgState *fw_cfg)
834 {
835     uint16_t protocol;
836     int setup_size, kernel_size, initrd_size = 0, cmdline_size;
837     uint32_t initrd_max;
838     uint8_t header[8192], *setup, *kernel, *initrd_data;
839     hwaddr real_addr, prot_addr, cmdline_addr, initrd_addr = 0;
840     FILE *f;
841     char *vmode;
842     MachineState *machine = MACHINE(pcms);
843     const char *kernel_filename = machine->kernel_filename;
844     const char *initrd_filename = machine->initrd_filename;
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 - acpi_data_size) {
911         initrd_max = pcms->below_4g_mem_size - 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     kernel_size -= setup_size;
990 
991     setup  = g_malloc(setup_size);
992     kernel = g_malloc(kernel_size);
993     fseek(f, 0, SEEK_SET);
994     if (fread(setup, 1, setup_size, f) != setup_size) {
995         fprintf(stderr, "fread() failed\n");
996         exit(1);
997     }
998     if (fread(kernel, 1, kernel_size, f) != kernel_size) {
999         fprintf(stderr, "fread() failed\n");
1000         exit(1);
1001     }
1002     fclose(f);
1003     memcpy(setup, header, MIN(sizeof(header), setup_size));
1004 
1005     fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_ADDR, prot_addr);
1006     fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_SIZE, kernel_size);
1007     fw_cfg_add_bytes(fw_cfg, FW_CFG_KERNEL_DATA, kernel, kernel_size);
1008 
1009     fw_cfg_add_i32(fw_cfg, FW_CFG_SETUP_ADDR, real_addr);
1010     fw_cfg_add_i32(fw_cfg, FW_CFG_SETUP_SIZE, setup_size);
1011     fw_cfg_add_bytes(fw_cfg, FW_CFG_SETUP_DATA, setup, setup_size);
1012 
1013     option_rom[nb_option_roms].name = "linuxboot.bin";
1014     option_rom[nb_option_roms].bootindex = 0;
1015     nb_option_roms++;
1016 }
1017 
1018 #define NE2000_NB_MAX 6
1019 
1020 static const int ne2000_io[NE2000_NB_MAX] = { 0x300, 0x320, 0x340, 0x360,
1021                                               0x280, 0x380 };
1022 static const int ne2000_irq[NE2000_NB_MAX] = { 9, 10, 11, 3, 4, 5 };
1023 
1024 void pc_init_ne2k_isa(ISABus *bus, NICInfo *nd)
1025 {
1026     static int nb_ne2k = 0;
1027 
1028     if (nb_ne2k == NE2000_NB_MAX)
1029         return;
1030     isa_ne2000_init(bus, ne2000_io[nb_ne2k],
1031                     ne2000_irq[nb_ne2k], nd);
1032     nb_ne2k++;
1033 }
1034 
1035 DeviceState *cpu_get_current_apic(void)
1036 {
1037     if (current_cpu) {
1038         X86CPU *cpu = X86_CPU(current_cpu);
1039         return cpu->apic_state;
1040     } else {
1041         return NULL;
1042     }
1043 }
1044 
1045 void pc_acpi_smi_interrupt(void *opaque, int irq, int level)
1046 {
1047     X86CPU *cpu = opaque;
1048 
1049     if (level) {
1050         cpu_interrupt(CPU(cpu), CPU_INTERRUPT_SMI);
1051     }
1052 }
1053 
1054 static X86CPU *pc_new_cpu(const char *cpu_model, int64_t apic_id,
1055                           DeviceState *icc_bridge, Error **errp)
1056 {
1057     X86CPU *cpu = NULL;
1058     Error *local_err = NULL;
1059 
1060     if (icc_bridge == NULL) {
1061         error_setg(&local_err, "Invalid icc-bridge value");
1062         goto out;
1063     }
1064 
1065     cpu = cpu_x86_create(cpu_model, &local_err);
1066     if (local_err != NULL) {
1067         goto out;
1068     }
1069 
1070     qdev_set_parent_bus(DEVICE(cpu), qdev_get_child_bus(icc_bridge, "icc"));
1071 
1072     object_property_set_int(OBJECT(cpu), apic_id, "apic-id", &local_err);
1073     object_property_set_bool(OBJECT(cpu), true, "realized", &local_err);
1074 
1075 out:
1076     if (local_err) {
1077         error_propagate(errp, local_err);
1078         object_unref(OBJECT(cpu));
1079         cpu = NULL;
1080     }
1081     return cpu;
1082 }
1083 
1084 static const char *current_cpu_model;
1085 
1086 void pc_hot_add_cpu(const int64_t id, Error **errp)
1087 {
1088     DeviceState *icc_bridge;
1089     X86CPU *cpu;
1090     int64_t apic_id = x86_cpu_apic_id_from_index(id);
1091     Error *local_err = NULL;
1092 
1093     if (id < 0) {
1094         error_setg(errp, "Invalid CPU id: %" PRIi64, id);
1095         return;
1096     }
1097 
1098     if (cpu_exists(apic_id)) {
1099         error_setg(errp, "Unable to add CPU: %" PRIi64
1100                    ", it already exists", id);
1101         return;
1102     }
1103 
1104     if (id >= max_cpus) {
1105         error_setg(errp, "Unable to add CPU: %" PRIi64
1106                    ", max allowed: %d", id, max_cpus - 1);
1107         return;
1108     }
1109 
1110     if (apic_id >= ACPI_CPU_HOTPLUG_ID_LIMIT) {
1111         error_setg(errp, "Unable to add CPU: %" PRIi64
1112                    ", resulting APIC ID (%" PRIi64 ") is too large",
1113                    id, apic_id);
1114         return;
1115     }
1116 
1117     icc_bridge = DEVICE(object_resolve_path_type("icc-bridge",
1118                                                  TYPE_ICC_BRIDGE, NULL));
1119     cpu = pc_new_cpu(current_cpu_model, apic_id, icc_bridge, &local_err);
1120     if (local_err) {
1121         error_propagate(errp, local_err);
1122         return;
1123     }
1124     object_unref(OBJECT(cpu));
1125 }
1126 
1127 void pc_cpus_init(const char *cpu_model, DeviceState *icc_bridge)
1128 {
1129     int i;
1130     X86CPU *cpu = NULL;
1131     Error *error = NULL;
1132     unsigned long apic_id_limit;
1133 
1134     /* init CPUs */
1135     if (cpu_model == NULL) {
1136 #ifdef TARGET_X86_64
1137         cpu_model = "qemu64";
1138 #else
1139         cpu_model = "qemu32";
1140 #endif
1141     }
1142     current_cpu_model = cpu_model;
1143 
1144     apic_id_limit = pc_apic_id_limit(max_cpus);
1145     if (apic_id_limit > ACPI_CPU_HOTPLUG_ID_LIMIT) {
1146         error_report("max_cpus is too large. APIC ID of last CPU is %lu",
1147                      apic_id_limit - 1);
1148         exit(1);
1149     }
1150 
1151     for (i = 0; i < smp_cpus; i++) {
1152         cpu = pc_new_cpu(cpu_model, x86_cpu_apic_id_from_index(i),
1153                          icc_bridge, &error);
1154         if (error) {
1155             error_report_err(error);
1156             exit(1);
1157         }
1158         object_unref(OBJECT(cpu));
1159     }
1160 
1161     /* map APIC MMIO area if CPU has APIC */
1162     if (cpu && cpu->apic_state) {
1163         /* XXX: what if the base changes? */
1164         sysbus_mmio_map_overlap(SYS_BUS_DEVICE(icc_bridge), 0,
1165                                 APIC_DEFAULT_ADDRESS, 0x1000);
1166     }
1167 
1168     /* tell smbios about cpuid version and features */
1169     smbios_set_cpuid(cpu->env.cpuid_version, cpu->env.features[FEAT_1_EDX]);
1170 }
1171 
1172 /* pci-info ROM file. Little endian format */
1173 typedef struct PcRomPciInfo {
1174     uint64_t w32_min;
1175     uint64_t w32_max;
1176     uint64_t w64_min;
1177     uint64_t w64_max;
1178 } PcRomPciInfo;
1179 
1180 typedef struct PcGuestInfoState {
1181     PcGuestInfo info;
1182     Notifier machine_done;
1183 } PcGuestInfoState;
1184 
1185 static
1186 void pc_guest_info_machine_done(Notifier *notifier, void *data)
1187 {
1188     PcGuestInfoState *guest_info_state = container_of(notifier,
1189                                                       PcGuestInfoState,
1190                                                       machine_done);
1191     PCIBus *bus = find_i440fx();
1192 
1193     if (bus) {
1194         int extra_hosts = 0;
1195 
1196         QLIST_FOREACH(bus, &bus->child, sibling) {
1197             /* look for expander root buses */
1198             if (pci_bus_is_root(bus)) {
1199                 extra_hosts++;
1200             }
1201         }
1202         if (extra_hosts && guest_info_state->info.fw_cfg) {
1203             uint64_t *val = g_malloc(sizeof(*val));
1204             *val = cpu_to_le64(extra_hosts);
1205             fw_cfg_add_file(guest_info_state->info.fw_cfg,
1206                     "etc/extra-pci-roots", val, sizeof(*val));
1207         }
1208     }
1209 
1210     acpi_setup(&guest_info_state->info);
1211 }
1212 
1213 PcGuestInfo *pc_guest_info_init(PCMachineState *pcms)
1214 {
1215     PcGuestInfoState *guest_info_state = g_malloc0(sizeof *guest_info_state);
1216     PcGuestInfo *guest_info = &guest_info_state->info;
1217     int i, j;
1218 
1219     guest_info->ram_size_below_4g = pcms->below_4g_mem_size;
1220     guest_info->ram_size = pcms->below_4g_mem_size + pcms->above_4g_mem_size;
1221     guest_info->apic_id_limit = pc_apic_id_limit(max_cpus);
1222     guest_info->apic_xrupt_override = kvm_allows_irq0_override();
1223     guest_info->numa_nodes = nb_numa_nodes;
1224     guest_info->node_mem = g_malloc0(guest_info->numa_nodes *
1225                                     sizeof *guest_info->node_mem);
1226     for (i = 0; i < nb_numa_nodes; i++) {
1227         guest_info->node_mem[i] = numa_info[i].node_mem;
1228     }
1229 
1230     guest_info->node_cpu = g_malloc0(guest_info->apic_id_limit *
1231                                      sizeof *guest_info->node_cpu);
1232 
1233     for (i = 0; i < max_cpus; i++) {
1234         unsigned int apic_id = x86_cpu_apic_id_from_index(i);
1235         assert(apic_id < guest_info->apic_id_limit);
1236         for (j = 0; j < nb_numa_nodes; j++) {
1237             if (test_bit(i, numa_info[j].node_cpu)) {
1238                 guest_info->node_cpu[apic_id] = j;
1239                 break;
1240             }
1241         }
1242     }
1243 
1244     guest_info_state->machine_done.notify = pc_guest_info_machine_done;
1245     qemu_add_machine_init_done_notifier(&guest_info_state->machine_done);
1246     return guest_info;
1247 }
1248 
1249 /* setup pci memory address space mapping into system address space */
1250 void pc_pci_as_mapping_init(Object *owner, MemoryRegion *system_memory,
1251                             MemoryRegion *pci_address_space)
1252 {
1253     /* Set to lower priority than RAM */
1254     memory_region_add_subregion_overlap(system_memory, 0x0,
1255                                         pci_address_space, -1);
1256 }
1257 
1258 void pc_acpi_init(const char *default_dsdt)
1259 {
1260     char *filename;
1261 
1262     if (acpi_tables != NULL) {
1263         /* manually set via -acpitable, leave it alone */
1264         return;
1265     }
1266 
1267     filename = qemu_find_file(QEMU_FILE_TYPE_BIOS, default_dsdt);
1268     if (filename == NULL) {
1269         fprintf(stderr, "WARNING: failed to find %s\n", default_dsdt);
1270     } else {
1271         QemuOpts *opts = qemu_opts_create(qemu_find_opts("acpi"), NULL, 0,
1272                                           &error_abort);
1273         Error *err = NULL;
1274 
1275         qemu_opt_set(opts, "file", filename, &error_abort);
1276 
1277         acpi_table_add_builtin(opts, &err);
1278         if (err) {
1279             error_report("WARNING: failed to load %s: %s", filename,
1280                          error_get_pretty(err));
1281             error_free(err);
1282         }
1283         g_free(filename);
1284     }
1285 }
1286 
1287 FWCfgState *xen_load_linux(PCMachineState *pcms,
1288                            PcGuestInfo *guest_info)
1289 {
1290     int i;
1291     FWCfgState *fw_cfg;
1292 
1293     assert(MACHINE(pcms)->kernel_filename != NULL);
1294 
1295     fw_cfg = fw_cfg_init_io(BIOS_CFG_IOPORT);
1296     rom_set_fw(fw_cfg);
1297 
1298     load_linux(pcms, fw_cfg);
1299     for (i = 0; i < nb_option_roms; i++) {
1300         assert(!strcmp(option_rom[i].name, "linuxboot.bin") ||
1301                !strcmp(option_rom[i].name, "multiboot.bin"));
1302         rom_add_option(option_rom[i].name, option_rom[i].bootindex);
1303     }
1304     guest_info->fw_cfg = fw_cfg;
1305     return fw_cfg;
1306 }
1307 
1308 FWCfgState *pc_memory_init(PCMachineState *pcms,
1309                            MemoryRegion *system_memory,
1310                            MemoryRegion *rom_memory,
1311                            MemoryRegion **ram_memory,
1312                            PcGuestInfo *guest_info)
1313 {
1314     int linux_boot, i;
1315     MemoryRegion *ram, *option_rom_mr;
1316     MemoryRegion *ram_below_4g, *ram_above_4g;
1317     FWCfgState *fw_cfg;
1318     MachineState *machine = MACHINE(pcms);
1319 
1320     assert(machine->ram_size == pcms->below_4g_mem_size +
1321                                 pcms->above_4g_mem_size);
1322 
1323     linux_boot = (machine->kernel_filename != NULL);
1324 
1325     /* Allocate RAM.  We allocate it as a single memory region and use
1326      * aliases to address portions of it, mostly for backwards compatibility
1327      * with older qemus that used qemu_ram_alloc().
1328      */
1329     ram = g_malloc(sizeof(*ram));
1330     memory_region_allocate_system_memory(ram, NULL, "pc.ram",
1331                                          machine->ram_size);
1332     *ram_memory = ram;
1333     ram_below_4g = g_malloc(sizeof(*ram_below_4g));
1334     memory_region_init_alias(ram_below_4g, NULL, "ram-below-4g", ram,
1335                              0, pcms->below_4g_mem_size);
1336     memory_region_add_subregion(system_memory, 0, ram_below_4g);
1337     e820_add_entry(0, pcms->below_4g_mem_size, E820_RAM);
1338     if (pcms->above_4g_mem_size > 0) {
1339         ram_above_4g = g_malloc(sizeof(*ram_above_4g));
1340         memory_region_init_alias(ram_above_4g, NULL, "ram-above-4g", ram,
1341                                  pcms->below_4g_mem_size,
1342                                  pcms->above_4g_mem_size);
1343         memory_region_add_subregion(system_memory, 0x100000000ULL,
1344                                     ram_above_4g);
1345         e820_add_entry(0x100000000ULL, pcms->above_4g_mem_size, E820_RAM);
1346     }
1347 
1348     if (!guest_info->has_reserved_memory &&
1349         (machine->ram_slots ||
1350          (machine->maxram_size > machine->ram_size))) {
1351         MachineClass *mc = MACHINE_GET_CLASS(machine);
1352 
1353         error_report("\"-memory 'slots|maxmem'\" is not supported by: %s",
1354                      mc->name);
1355         exit(EXIT_FAILURE);
1356     }
1357 
1358     /* initialize hotplug memory address space */
1359     if (guest_info->has_reserved_memory &&
1360         (machine->ram_size < machine->maxram_size)) {
1361         ram_addr_t hotplug_mem_size =
1362             machine->maxram_size - machine->ram_size;
1363 
1364         if (machine->ram_slots > ACPI_MAX_RAM_SLOTS) {
1365             error_report("unsupported amount of memory slots: %"PRIu64,
1366                          machine->ram_slots);
1367             exit(EXIT_FAILURE);
1368         }
1369 
1370         if (QEMU_ALIGN_UP(machine->maxram_size,
1371                           TARGET_PAGE_SIZE) != machine->maxram_size) {
1372             error_report("maximum memory size must by aligned to multiple of "
1373                          "%d bytes", TARGET_PAGE_SIZE);
1374             exit(EXIT_FAILURE);
1375         }
1376 
1377         pcms->hotplug_memory.base =
1378             ROUND_UP(0x100000000ULL + pcms->above_4g_mem_size, 1ULL << 30);
1379 
1380         if (pcms->enforce_aligned_dimm) {
1381             /* size hotplug region assuming 1G page max alignment per slot */
1382             hotplug_mem_size += (1ULL << 30) * machine->ram_slots;
1383         }
1384 
1385         if ((pcms->hotplug_memory.base + hotplug_mem_size) <
1386             hotplug_mem_size) {
1387             error_report("unsupported amount of maximum memory: " RAM_ADDR_FMT,
1388                          machine->maxram_size);
1389             exit(EXIT_FAILURE);
1390         }
1391 
1392         memory_region_init(&pcms->hotplug_memory.mr, OBJECT(pcms),
1393                            "hotplug-memory", hotplug_mem_size);
1394         memory_region_add_subregion(system_memory, pcms->hotplug_memory.base,
1395                                     &pcms->hotplug_memory.mr);
1396     }
1397 
1398     /* Initialize PC system firmware */
1399     pc_system_firmware_init(rom_memory, guest_info->isapc_ram_fw);
1400 
1401     option_rom_mr = g_malloc(sizeof(*option_rom_mr));
1402     memory_region_init_ram(option_rom_mr, NULL, "pc.rom", PC_ROM_SIZE,
1403                            &error_abort);
1404     vmstate_register_ram_global(option_rom_mr);
1405     memory_region_add_subregion_overlap(rom_memory,
1406                                         PC_ROM_MIN_VGA,
1407                                         option_rom_mr,
1408                                         1);
1409 
1410     fw_cfg = bochs_bios_init();
1411     rom_set_fw(fw_cfg);
1412 
1413     if (guest_info->has_reserved_memory && pcms->hotplug_memory.base) {
1414         uint64_t *val = g_malloc(sizeof(*val));
1415         PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
1416         uint64_t res_mem_end = pcms->hotplug_memory.base;
1417 
1418         if (!pcmc->broken_reserved_end) {
1419             res_mem_end += memory_region_size(&pcms->hotplug_memory.mr);
1420         }
1421         *val = cpu_to_le64(ROUND_UP(res_mem_end, 0x1ULL << 30));
1422         fw_cfg_add_file(fw_cfg, "etc/reserved-memory-end", val, sizeof(*val));
1423     }
1424 
1425     if (linux_boot) {
1426         load_linux(pcms, fw_cfg);
1427     }
1428 
1429     for (i = 0; i < nb_option_roms; i++) {
1430         rom_add_option(option_rom[i].name, option_rom[i].bootindex);
1431     }
1432     guest_info->fw_cfg = fw_cfg;
1433     return fw_cfg;
1434 }
1435 
1436 qemu_irq pc_allocate_cpu_irq(void)
1437 {
1438     return qemu_allocate_irq(pic_irq_request, NULL, 0);
1439 }
1440 
1441 DeviceState *pc_vga_init(ISABus *isa_bus, PCIBus *pci_bus)
1442 {
1443     DeviceState *dev = NULL;
1444 
1445     if (pci_bus) {
1446         PCIDevice *pcidev = pci_vga_init(pci_bus);
1447         dev = pcidev ? &pcidev->qdev : NULL;
1448     } else if (isa_bus) {
1449         ISADevice *isadev = isa_vga_init(isa_bus);
1450         dev = isadev ? DEVICE(isadev) : NULL;
1451     }
1452     return dev;
1453 }
1454 
1455 static void cpu_request_exit(void *opaque, int irq, int level)
1456 {
1457     CPUState *cpu = current_cpu;
1458 
1459     if (cpu && level) {
1460         cpu_exit(cpu);
1461     }
1462 }
1463 
1464 static const MemoryRegionOps ioport80_io_ops = {
1465     .write = ioport80_write,
1466     .read = ioport80_read,
1467     .endianness = DEVICE_NATIVE_ENDIAN,
1468     .impl = {
1469         .min_access_size = 1,
1470         .max_access_size = 1,
1471     },
1472 };
1473 
1474 static const MemoryRegionOps ioportF0_io_ops = {
1475     .write = ioportF0_write,
1476     .read = ioportF0_read,
1477     .endianness = DEVICE_NATIVE_ENDIAN,
1478     .impl = {
1479         .min_access_size = 1,
1480         .max_access_size = 1,
1481     },
1482 };
1483 
1484 void pc_basic_device_init(ISABus *isa_bus, qemu_irq *gsi,
1485                           ISADevice **rtc_state,
1486                           bool create_fdctrl,
1487                           bool no_vmport,
1488                           uint32 hpet_irqs)
1489 {
1490     int i;
1491     DriveInfo *fd[MAX_FD];
1492     DeviceState *hpet = NULL;
1493     int pit_isa_irq = 0;
1494     qemu_irq pit_alt_irq = NULL;
1495     qemu_irq rtc_irq = NULL;
1496     qemu_irq *a20_line;
1497     ISADevice *i8042, *port92, *vmmouse, *pit = NULL;
1498     qemu_irq *cpu_exit_irq;
1499     MemoryRegion *ioport80_io = g_new(MemoryRegion, 1);
1500     MemoryRegion *ioportF0_io = g_new(MemoryRegion, 1);
1501 
1502     memory_region_init_io(ioport80_io, NULL, &ioport80_io_ops, NULL, "ioport80", 1);
1503     memory_region_add_subregion(isa_bus->address_space_io, 0x80, ioport80_io);
1504 
1505     memory_region_init_io(ioportF0_io, NULL, &ioportF0_io_ops, NULL, "ioportF0", 1);
1506     memory_region_add_subregion(isa_bus->address_space_io, 0xf0, ioportF0_io);
1507 
1508     /*
1509      * Check if an HPET shall be created.
1510      *
1511      * Without KVM_CAP_PIT_STATE2, we cannot switch off the in-kernel PIT
1512      * when the HPET wants to take over. Thus we have to disable the latter.
1513      */
1514     if (!no_hpet && (!kvm_irqchip_in_kernel() || kvm_has_pit_state2())) {
1515         /* In order to set property, here not using sysbus_try_create_simple */
1516         hpet = qdev_try_create(NULL, TYPE_HPET);
1517         if (hpet) {
1518             /* For pc-piix-*, hpet's intcap is always IRQ2. For pc-q35-1.7
1519              * and earlier, use IRQ2 for compat. Otherwise, use IRQ16~23,
1520              * IRQ8 and IRQ2.
1521              */
1522             uint8_t compat = object_property_get_int(OBJECT(hpet),
1523                     HPET_INTCAP, NULL);
1524             if (!compat) {
1525                 qdev_prop_set_uint32(hpet, HPET_INTCAP, hpet_irqs);
1526             }
1527             qdev_init_nofail(hpet);
1528             sysbus_mmio_map(SYS_BUS_DEVICE(hpet), 0, HPET_BASE);
1529 
1530             for (i = 0; i < GSI_NUM_PINS; i++) {
1531                 sysbus_connect_irq(SYS_BUS_DEVICE(hpet), i, gsi[i]);
1532             }
1533             pit_isa_irq = -1;
1534             pit_alt_irq = qdev_get_gpio_in(hpet, HPET_LEGACY_PIT_INT);
1535             rtc_irq = qdev_get_gpio_in(hpet, HPET_LEGACY_RTC_INT);
1536         }
1537     }
1538     *rtc_state = rtc_init(isa_bus, 2000, rtc_irq);
1539 
1540     qemu_register_boot_set(pc_boot_set, *rtc_state);
1541 
1542     if (!xen_enabled()) {
1543         if (kvm_irqchip_in_kernel()) {
1544             pit = kvm_pit_init(isa_bus, 0x40);
1545         } else {
1546             pit = pit_init(isa_bus, 0x40, pit_isa_irq, pit_alt_irq);
1547         }
1548         if (hpet) {
1549             /* connect PIT to output control line of the HPET */
1550             qdev_connect_gpio_out(hpet, 0, qdev_get_gpio_in(DEVICE(pit), 0));
1551         }
1552         pcspk_init(isa_bus, pit);
1553     }
1554 
1555     serial_hds_isa_init(isa_bus, MAX_SERIAL_PORTS);
1556     parallel_hds_isa_init(isa_bus, MAX_PARALLEL_PORTS);
1557 
1558     a20_line = qemu_allocate_irqs(handle_a20_line_change, first_cpu, 2);
1559     i8042 = isa_create_simple(isa_bus, "i8042");
1560     i8042_setup_a20_line(i8042, &a20_line[0]);
1561     if (!no_vmport) {
1562         vmport_init(isa_bus);
1563         vmmouse = isa_try_create(isa_bus, "vmmouse");
1564     } else {
1565         vmmouse = NULL;
1566     }
1567     if (vmmouse) {
1568         DeviceState *dev = DEVICE(vmmouse);
1569         qdev_prop_set_ptr(dev, "ps2_mouse", i8042);
1570         qdev_init_nofail(dev);
1571     }
1572     port92 = isa_create_simple(isa_bus, "port92");
1573     port92_init(port92, &a20_line[1]);
1574 
1575     cpu_exit_irq = qemu_allocate_irqs(cpu_request_exit, NULL, 1);
1576     DMA_init(0, cpu_exit_irq);
1577 
1578     for(i = 0; i < MAX_FD; i++) {
1579         fd[i] = drive_get(IF_FLOPPY, 0, i);
1580         create_fdctrl |= !!fd[i];
1581     }
1582     if (create_fdctrl) {
1583         fdctrl_init_isa(isa_bus, fd);
1584     }
1585 }
1586 
1587 void pc_nic_init(ISABus *isa_bus, PCIBus *pci_bus)
1588 {
1589     int i;
1590 
1591     for (i = 0; i < nb_nics; i++) {
1592         NICInfo *nd = &nd_table[i];
1593 
1594         if (!pci_bus || (nd->model && strcmp(nd->model, "ne2k_isa") == 0)) {
1595             pc_init_ne2k_isa(isa_bus, nd);
1596         } else {
1597             pci_nic_init_nofail(nd, pci_bus, "e1000", NULL);
1598         }
1599     }
1600 }
1601 
1602 void pc_pci_device_init(PCIBus *pci_bus)
1603 {
1604     int max_bus;
1605     int bus;
1606 
1607     max_bus = drive_get_max_bus(IF_SCSI);
1608     for (bus = 0; bus <= max_bus; bus++) {
1609         pci_create_simple(pci_bus, -1, "lsi53c895a");
1610     }
1611 }
1612 
1613 void ioapic_init_gsi(GSIState *gsi_state, const char *parent_name)
1614 {
1615     DeviceState *dev;
1616     SysBusDevice *d;
1617     unsigned int i;
1618 
1619     if (kvm_irqchip_in_kernel()) {
1620         dev = qdev_create(NULL, "kvm-ioapic");
1621     } else {
1622         dev = qdev_create(NULL, "ioapic");
1623     }
1624     if (parent_name) {
1625         object_property_add_child(object_resolve_path(parent_name, NULL),
1626                                   "ioapic", OBJECT(dev), NULL);
1627     }
1628     qdev_init_nofail(dev);
1629     d = SYS_BUS_DEVICE(dev);
1630     sysbus_mmio_map(d, 0, IO_APIC_DEFAULT_ADDRESS);
1631 
1632     for (i = 0; i < IOAPIC_NUM_PINS; i++) {
1633         gsi_state->ioapic_irq[i] = qdev_get_gpio_in(dev, i);
1634     }
1635 }
1636 
1637 static void pc_dimm_plug(HotplugHandler *hotplug_dev,
1638                          DeviceState *dev, Error **errp)
1639 {
1640     HotplugHandlerClass *hhc;
1641     Error *local_err = NULL;
1642     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1643     PCDIMMDevice *dimm = PC_DIMM(dev);
1644     PCDIMMDeviceClass *ddc = PC_DIMM_GET_CLASS(dimm);
1645     MemoryRegion *mr = ddc->get_memory_region(dimm);
1646     uint64_t align = TARGET_PAGE_SIZE;
1647 
1648     if (memory_region_get_alignment(mr) && pcms->enforce_aligned_dimm) {
1649         align = memory_region_get_alignment(mr);
1650     }
1651 
1652     if (!pcms->acpi_dev) {
1653         error_setg(&local_err,
1654                    "memory hotplug is not enabled: missing acpi device");
1655         goto out;
1656     }
1657 
1658     pc_dimm_memory_plug(dev, &pcms->hotplug_memory, mr, align, &local_err);
1659     if (local_err) {
1660         goto out;
1661     }
1662 
1663     hhc = HOTPLUG_HANDLER_GET_CLASS(pcms->acpi_dev);
1664     hhc->plug(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &error_abort);
1665 out:
1666     error_propagate(errp, local_err);
1667 }
1668 
1669 static void pc_dimm_unplug_request(HotplugHandler *hotplug_dev,
1670                                    DeviceState *dev, Error **errp)
1671 {
1672     HotplugHandlerClass *hhc;
1673     Error *local_err = NULL;
1674     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1675 
1676     if (!pcms->acpi_dev) {
1677         error_setg(&local_err,
1678                    "memory hotplug is not enabled: missing acpi device");
1679         goto out;
1680     }
1681 
1682     hhc = HOTPLUG_HANDLER_GET_CLASS(pcms->acpi_dev);
1683     hhc->unplug_request(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &local_err);
1684 
1685 out:
1686     error_propagate(errp, local_err);
1687 }
1688 
1689 static void pc_dimm_unplug(HotplugHandler *hotplug_dev,
1690                            DeviceState *dev, Error **errp)
1691 {
1692     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1693     PCDIMMDevice *dimm = PC_DIMM(dev);
1694     PCDIMMDeviceClass *ddc = PC_DIMM_GET_CLASS(dimm);
1695     MemoryRegion *mr = ddc->get_memory_region(dimm);
1696     HotplugHandlerClass *hhc;
1697     Error *local_err = NULL;
1698 
1699     hhc = HOTPLUG_HANDLER_GET_CLASS(pcms->acpi_dev);
1700     hhc->unplug(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &local_err);
1701 
1702     if (local_err) {
1703         goto out;
1704     }
1705 
1706     pc_dimm_memory_unplug(dev, &pcms->hotplug_memory, mr);
1707     object_unparent(OBJECT(dev));
1708 
1709  out:
1710     error_propagate(errp, local_err);
1711 }
1712 
1713 static void pc_cpu_plug(HotplugHandler *hotplug_dev,
1714                         DeviceState *dev, Error **errp)
1715 {
1716     HotplugHandlerClass *hhc;
1717     Error *local_err = NULL;
1718     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1719 
1720     if (!dev->hotplugged) {
1721         goto out;
1722     }
1723 
1724     if (!pcms->acpi_dev) {
1725         error_setg(&local_err,
1726                    "cpu hotplug is not enabled: missing acpi device");
1727         goto out;
1728     }
1729 
1730     hhc = HOTPLUG_HANDLER_GET_CLASS(pcms->acpi_dev);
1731     hhc->plug(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &local_err);
1732     if (local_err) {
1733         goto out;
1734     }
1735 
1736     /* increment the number of CPUs */
1737     rtc_set_memory(pcms->rtc, 0x5f, rtc_get_memory(pcms->rtc, 0x5f) + 1);
1738 out:
1739     error_propagate(errp, local_err);
1740 }
1741 
1742 static void pc_machine_device_plug_cb(HotplugHandler *hotplug_dev,
1743                                       DeviceState *dev, Error **errp)
1744 {
1745     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
1746         pc_dimm_plug(hotplug_dev, dev, errp);
1747     } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
1748         pc_cpu_plug(hotplug_dev, dev, errp);
1749     }
1750 }
1751 
1752 static void pc_machine_device_unplug_request_cb(HotplugHandler *hotplug_dev,
1753                                                 DeviceState *dev, Error **errp)
1754 {
1755     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
1756         pc_dimm_unplug_request(hotplug_dev, dev, errp);
1757     } else {
1758         error_setg(errp, "acpi: device unplug request for not supported device"
1759                    " type: %s", object_get_typename(OBJECT(dev)));
1760     }
1761 }
1762 
1763 static void pc_machine_device_unplug_cb(HotplugHandler *hotplug_dev,
1764                                         DeviceState *dev, Error **errp)
1765 {
1766     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
1767         pc_dimm_unplug(hotplug_dev, dev, errp);
1768     } else {
1769         error_setg(errp, "acpi: device unplug for not supported device"
1770                    " type: %s", object_get_typename(OBJECT(dev)));
1771     }
1772 }
1773 
1774 static HotplugHandler *pc_get_hotpug_handler(MachineState *machine,
1775                                              DeviceState *dev)
1776 {
1777     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(machine);
1778 
1779     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM) ||
1780         object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
1781         return HOTPLUG_HANDLER(machine);
1782     }
1783 
1784     return pcmc->get_hotplug_handler ?
1785         pcmc->get_hotplug_handler(machine, dev) : NULL;
1786 }
1787 
1788 static void
1789 pc_machine_get_hotplug_memory_region_size(Object *obj, Visitor *v, void *opaque,
1790                                           const char *name, Error **errp)
1791 {
1792     PCMachineState *pcms = PC_MACHINE(obj);
1793     int64_t value = memory_region_size(&pcms->hotplug_memory.mr);
1794 
1795     visit_type_int(v, &value, name, errp);
1796 }
1797 
1798 static void pc_machine_get_max_ram_below_4g(Object *obj, Visitor *v,
1799                                          void *opaque, const char *name,
1800                                          Error **errp)
1801 {
1802     PCMachineState *pcms = PC_MACHINE(obj);
1803     uint64_t value = pcms->max_ram_below_4g;
1804 
1805     visit_type_size(v, &value, name, errp);
1806 }
1807 
1808 static void pc_machine_set_max_ram_below_4g(Object *obj, Visitor *v,
1809                                          void *opaque, const char *name,
1810                                          Error **errp)
1811 {
1812     PCMachineState *pcms = PC_MACHINE(obj);
1813     Error *error = NULL;
1814     uint64_t value;
1815 
1816     visit_type_size(v, &value, name, &error);
1817     if (error) {
1818         error_propagate(errp, error);
1819         return;
1820     }
1821     if (value > (1ULL << 32)) {
1822         error_set(&error, ERROR_CLASS_GENERIC_ERROR,
1823                   "Machine option 'max-ram-below-4g=%"PRIu64
1824                   "' expects size less than or equal to 4G", value);
1825         error_propagate(errp, error);
1826         return;
1827     }
1828 
1829     if (value < (1ULL << 20)) {
1830         error_report("Warning: small max_ram_below_4g(%"PRIu64
1831                      ") less than 1M.  BIOS may not work..",
1832                      value);
1833     }
1834 
1835     pcms->max_ram_below_4g = value;
1836 }
1837 
1838 static void pc_machine_get_vmport(Object *obj, Visitor *v, void *opaque,
1839                                   const char *name, Error **errp)
1840 {
1841     PCMachineState *pcms = PC_MACHINE(obj);
1842     OnOffAuto vmport = pcms->vmport;
1843 
1844     visit_type_OnOffAuto(v, &vmport, name, errp);
1845 }
1846 
1847 static void pc_machine_set_vmport(Object *obj, Visitor *v, void *opaque,
1848                                   const char *name, Error **errp)
1849 {
1850     PCMachineState *pcms = PC_MACHINE(obj);
1851 
1852     visit_type_OnOffAuto(v, &pcms->vmport, name, errp);
1853 }
1854 
1855 bool pc_machine_is_smm_enabled(PCMachineState *pcms)
1856 {
1857     bool smm_available = false;
1858 
1859     if (pcms->smm == ON_OFF_AUTO_OFF) {
1860         return false;
1861     }
1862 
1863     if (tcg_enabled() || qtest_enabled()) {
1864         smm_available = true;
1865     } else if (kvm_enabled()) {
1866         smm_available = kvm_has_smm();
1867     }
1868 
1869     if (smm_available) {
1870         return true;
1871     }
1872 
1873     if (pcms->smm == ON_OFF_AUTO_ON) {
1874         error_report("System Management Mode not supported by this hypervisor.");
1875         exit(1);
1876     }
1877     return false;
1878 }
1879 
1880 static void pc_machine_get_smm(Object *obj, Visitor *v, void *opaque,
1881                               const char *name, Error **errp)
1882 {
1883     PCMachineState *pcms = PC_MACHINE(obj);
1884     OnOffAuto smm = pcms->smm;
1885 
1886     visit_type_OnOffAuto(v, &smm, name, errp);
1887 }
1888 
1889 static void pc_machine_set_smm(Object *obj, Visitor *v, void *opaque,
1890                                   const char *name, Error **errp)
1891 {
1892     PCMachineState *pcms = PC_MACHINE(obj);
1893 
1894     visit_type_OnOffAuto(v, &pcms->smm, name, errp);
1895 }
1896 
1897 static bool pc_machine_get_aligned_dimm(Object *obj, Error **errp)
1898 {
1899     PCMachineState *pcms = PC_MACHINE(obj);
1900 
1901     return pcms->enforce_aligned_dimm;
1902 }
1903 
1904 static void pc_machine_initfn(Object *obj)
1905 {
1906     PCMachineState *pcms = PC_MACHINE(obj);
1907 
1908     object_property_add(obj, PC_MACHINE_MEMHP_REGION_SIZE, "int",
1909                         pc_machine_get_hotplug_memory_region_size,
1910                         NULL, NULL, NULL, &error_abort);
1911 
1912     pcms->max_ram_below_4g = 1ULL << 32; /* 4G */
1913     object_property_add(obj, PC_MACHINE_MAX_RAM_BELOW_4G, "size",
1914                         pc_machine_get_max_ram_below_4g,
1915                         pc_machine_set_max_ram_below_4g,
1916                         NULL, NULL, &error_abort);
1917     object_property_set_description(obj, PC_MACHINE_MAX_RAM_BELOW_4G,
1918                                     "Maximum ram below the 4G boundary (32bit boundary)",
1919                                     &error_abort);
1920 
1921     pcms->smm = ON_OFF_AUTO_AUTO;
1922     object_property_add(obj, PC_MACHINE_SMM, "OnOffAuto",
1923                         pc_machine_get_smm,
1924                         pc_machine_set_smm,
1925                         NULL, NULL, &error_abort);
1926     object_property_set_description(obj, PC_MACHINE_SMM,
1927                                     "Enable SMM (pc & q35)",
1928                                     &error_abort);
1929 
1930     pcms->vmport = ON_OFF_AUTO_AUTO;
1931     object_property_add(obj, PC_MACHINE_VMPORT, "OnOffAuto",
1932                         pc_machine_get_vmport,
1933                         pc_machine_set_vmport,
1934                         NULL, NULL, &error_abort);
1935     object_property_set_description(obj, PC_MACHINE_VMPORT,
1936                                     "Enable vmport (pc & q35)",
1937                                     &error_abort);
1938 
1939     pcms->enforce_aligned_dimm = true;
1940     object_property_add_bool(obj, PC_MACHINE_ENFORCE_ALIGNED_DIMM,
1941                              pc_machine_get_aligned_dimm,
1942                              NULL, &error_abort);
1943 }
1944 
1945 static unsigned pc_cpu_index_to_socket_id(unsigned cpu_index)
1946 {
1947     unsigned pkg_id, core_id, smt_id;
1948     x86_topo_ids_from_idx(smp_cores, smp_threads, cpu_index,
1949                           &pkg_id, &core_id, &smt_id);
1950     return pkg_id;
1951 }
1952 
1953 static void pc_machine_class_init(ObjectClass *oc, void *data)
1954 {
1955     MachineClass *mc = MACHINE_CLASS(oc);
1956     PCMachineClass *pcmc = PC_MACHINE_CLASS(oc);
1957     HotplugHandlerClass *hc = HOTPLUG_HANDLER_CLASS(oc);
1958 
1959     pcmc->get_hotplug_handler = mc->get_hotplug_handler;
1960     mc->get_hotplug_handler = pc_get_hotpug_handler;
1961     mc->cpu_index_to_socket_id = pc_cpu_index_to_socket_id;
1962     mc->default_boot_order = "cad";
1963     mc->hot_add_cpu = pc_hot_add_cpu;
1964     mc->max_cpus = 255;
1965     hc->plug = pc_machine_device_plug_cb;
1966     hc->unplug_request = pc_machine_device_unplug_request_cb;
1967     hc->unplug = pc_machine_device_unplug_cb;
1968 }
1969 
1970 static const TypeInfo pc_machine_info = {
1971     .name = TYPE_PC_MACHINE,
1972     .parent = TYPE_MACHINE,
1973     .abstract = true,
1974     .instance_size = sizeof(PCMachineState),
1975     .instance_init = pc_machine_initfn,
1976     .class_size = sizeof(PCMachineClass),
1977     .class_init = pc_machine_class_init,
1978     .interfaces = (InterfaceInfo[]) {
1979          { TYPE_HOTPLUG_HANDLER },
1980          { }
1981     },
1982 };
1983 
1984 static void pc_machine_register_types(void)
1985 {
1986     type_register_static(&pc_machine_info);
1987 }
1988 
1989 type_init(pc_machine_register_types)
1990