xref: /openbmc/qemu/hw/i386/pc.c (revision 228aa992)
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/block/fdc.h"
29 #include "hw/ide.h"
30 #include "hw/pci/pci.h"
31 #include "monitor/monitor.h"
32 #include "hw/nvram/fw_cfg.h"
33 #include "hw/timer/hpet.h"
34 #include "hw/i386/smbios.h"
35 #include "hw/loader.h"
36 #include "elf.h"
37 #include "multiboot.h"
38 #include "hw/timer/mc146818rtc.h"
39 #include "hw/timer/i8254.h"
40 #include "hw/audio/pcspk.h"
41 #include "hw/pci/msi.h"
42 #include "hw/sysbus.h"
43 #include "sysemu/sysemu.h"
44 #include "sysemu/kvm.h"
45 #include "kvm_i386.h"
46 #include "hw/xen/xen.h"
47 #include "sysemu/block-backend.h"
48 #include "hw/block/block.h"
49 #include "ui/qemu-spice.h"
50 #include "exec/memory.h"
51 #include "exec/address-spaces.h"
52 #include "sysemu/arch_init.h"
53 #include "qemu/bitmap.h"
54 #include "qemu/config-file.h"
55 #include "hw/acpi/acpi.h"
56 #include "hw/acpi/cpu_hotplug.h"
57 #include "hw/cpu/icc_bus.h"
58 #include "hw/boards.h"
59 #include "hw/pci/pci_host.h"
60 #include "acpi-build.h"
61 #include "hw/mem/pc-dimm.h"
62 #include "trace.h"
63 #include "qapi/visitor.h"
64 
65 /* debug PC/ISA interrupts */
66 //#define DEBUG_IRQ
67 
68 #ifdef DEBUG_IRQ
69 #define DPRINTF(fmt, ...)                                       \
70     do { printf("CPUIRQ: " fmt , ## __VA_ARGS__); } while (0)
71 #else
72 #define DPRINTF(fmt, ...)
73 #endif
74 
75 /* Leave a chunk of memory at the top of RAM for the BIOS ACPI tables
76  * (128K) and other BIOS datastructures (less than 4K reported to be used at
77  * the moment, 32K should be enough for a while).  */
78 static unsigned acpi_data_size = 0x20000 + 0x8000;
79 void pc_set_legacy_acpi_data_size(void)
80 {
81     acpi_data_size = 0x10000;
82 }
83 
84 #define BIOS_CFG_IOPORT 0x510
85 #define FW_CFG_ACPI_TABLES (FW_CFG_ARCH_LOCAL + 0)
86 #define FW_CFG_SMBIOS_ENTRIES (FW_CFG_ARCH_LOCAL + 1)
87 #define FW_CFG_IRQ0_OVERRIDE (FW_CFG_ARCH_LOCAL + 2)
88 #define FW_CFG_E820_TABLE (FW_CFG_ARCH_LOCAL + 3)
89 #define FW_CFG_HPET (FW_CFG_ARCH_LOCAL + 4)
90 
91 #define E820_NR_ENTRIES		16
92 
93 struct e820_entry {
94     uint64_t address;
95     uint64_t length;
96     uint32_t type;
97 } QEMU_PACKED __attribute((__aligned__(4)));
98 
99 struct e820_table {
100     uint32_t count;
101     struct e820_entry entry[E820_NR_ENTRIES];
102 } QEMU_PACKED __attribute((__aligned__(4)));
103 
104 static struct e820_table e820_reserve;
105 static struct e820_entry *e820_table;
106 static unsigned e820_entries;
107 struct hpet_fw_config hpet_cfg = {.count = UINT8_MAX};
108 
109 void gsi_handler(void *opaque, int n, int level)
110 {
111     GSIState *s = opaque;
112 
113     DPRINTF("pc: %s GSI %d\n", level ? "raising" : "lowering", n);
114     if (n < ISA_NUM_IRQS) {
115         qemu_set_irq(s->i8259_irq[n], level);
116     }
117     qemu_set_irq(s->ioapic_irq[n], level);
118 }
119 
120 static void ioport80_write(void *opaque, hwaddr addr, uint64_t data,
121                            unsigned size)
122 {
123 }
124 
125 static uint64_t ioport80_read(void *opaque, hwaddr addr, unsigned size)
126 {
127     return 0xffffffffffffffffULL;
128 }
129 
130 /* MSDOS compatibility mode FPU exception support */
131 static qemu_irq ferr_irq;
132 
133 void pc_register_ferr_irq(qemu_irq irq)
134 {
135     ferr_irq = irq;
136 }
137 
138 /* XXX: add IGNNE support */
139 void cpu_set_ferr(CPUX86State *s)
140 {
141     qemu_irq_raise(ferr_irq);
142 }
143 
144 static void ioportF0_write(void *opaque, hwaddr addr, uint64_t data,
145                            unsigned size)
146 {
147     qemu_irq_lower(ferr_irq);
148 }
149 
150 static uint64_t ioportF0_read(void *opaque, hwaddr addr, unsigned size)
151 {
152     return 0xffffffffffffffffULL;
153 }
154 
155 /* TSC handling */
156 uint64_t cpu_get_tsc(CPUX86State *env)
157 {
158     return cpu_get_ticks();
159 }
160 
161 /* SMM support */
162 
163 static cpu_set_smm_t smm_set;
164 static void *smm_arg;
165 
166 void cpu_smm_register(cpu_set_smm_t callback, void *arg)
167 {
168     assert(smm_set == NULL);
169     assert(smm_arg == NULL);
170     smm_set = callback;
171     smm_arg = arg;
172 }
173 
174 void cpu_smm_update(CPUX86State *env)
175 {
176     if (smm_set && smm_arg && CPU(x86_env_get_cpu(env)) == first_cpu) {
177         smm_set(!!(env->hflags & HF_SMM_MASK), smm_arg);
178     }
179 }
180 
181 
182 /* IRQ handling */
183 int cpu_get_pic_interrupt(CPUX86State *env)
184 {
185     X86CPU *cpu = x86_env_get_cpu(env);
186     int intno;
187 
188     intno = apic_get_interrupt(cpu->apic_state);
189     if (intno >= 0) {
190         return intno;
191     }
192     /* read the irq from the PIC */
193     if (!apic_accept_pic_intr(cpu->apic_state)) {
194         return -1;
195     }
196 
197     intno = pic_read_irq(isa_pic);
198     return intno;
199 }
200 
201 static void pic_irq_request(void *opaque, int irq, int level)
202 {
203     CPUState *cs = first_cpu;
204     X86CPU *cpu = X86_CPU(cs);
205 
206     DPRINTF("pic_irqs: %s irq %d\n", level? "raise" : "lower", irq);
207     if (cpu->apic_state) {
208         CPU_FOREACH(cs) {
209             cpu = X86_CPU(cs);
210             if (apic_accept_pic_intr(cpu->apic_state)) {
211                 apic_deliver_pic_intr(cpu->apic_state, level);
212             }
213         }
214     } else {
215         if (level) {
216             cpu_interrupt(cs, CPU_INTERRUPT_HARD);
217         } else {
218             cpu_reset_interrupt(cs, CPU_INTERRUPT_HARD);
219         }
220     }
221 }
222 
223 /* PC cmos mappings */
224 
225 #define REG_EQUIPMENT_BYTE          0x14
226 
227 static int cmos_get_fd_drive_type(FDriveType fd0)
228 {
229     int val;
230 
231     switch (fd0) {
232     case FDRIVE_DRV_144:
233         /* 1.44 Mb 3"5 drive */
234         val = 4;
235         break;
236     case FDRIVE_DRV_288:
237         /* 2.88 Mb 3"5 drive */
238         val = 5;
239         break;
240     case FDRIVE_DRV_120:
241         /* 1.2 Mb 5"5 drive */
242         val = 2;
243         break;
244     case FDRIVE_DRV_NONE:
245     default:
246         val = 0;
247         break;
248     }
249     return val;
250 }
251 
252 static void cmos_init_hd(ISADevice *s, int type_ofs, int info_ofs,
253                          int16_t cylinders, int8_t heads, int8_t sectors)
254 {
255     rtc_set_memory(s, type_ofs, 47);
256     rtc_set_memory(s, info_ofs, cylinders);
257     rtc_set_memory(s, info_ofs + 1, cylinders >> 8);
258     rtc_set_memory(s, info_ofs + 2, heads);
259     rtc_set_memory(s, info_ofs + 3, 0xff);
260     rtc_set_memory(s, info_ofs + 4, 0xff);
261     rtc_set_memory(s, info_ofs + 5, 0xc0 | ((heads > 8) << 3));
262     rtc_set_memory(s, info_ofs + 6, cylinders);
263     rtc_set_memory(s, info_ofs + 7, cylinders >> 8);
264     rtc_set_memory(s, info_ofs + 8, sectors);
265 }
266 
267 /* convert boot_device letter to something recognizable by the bios */
268 static int boot_device2nibble(char boot_device)
269 {
270     switch(boot_device) {
271     case 'a':
272     case 'b':
273         return 0x01; /* floppy boot */
274     case 'c':
275         return 0x02; /* hard drive boot */
276     case 'd':
277         return 0x03; /* CD-ROM boot */
278     case 'n':
279         return 0x04; /* Network boot */
280     }
281     return 0;
282 }
283 
284 static int set_boot_dev(ISADevice *s, const char *boot_device)
285 {
286 #define PC_MAX_BOOT_DEVICES 3
287     int nbds, bds[3] = { 0, };
288     int i;
289 
290     nbds = strlen(boot_device);
291     if (nbds > PC_MAX_BOOT_DEVICES) {
292         error_report("Too many boot devices for PC");
293         return(1);
294     }
295     for (i = 0; i < nbds; i++) {
296         bds[i] = boot_device2nibble(boot_device[i]);
297         if (bds[i] == 0) {
298             error_report("Invalid boot device for PC: '%c'",
299                          boot_device[i]);
300             return(1);
301         }
302     }
303     rtc_set_memory(s, 0x3d, (bds[1] << 4) | bds[0]);
304     rtc_set_memory(s, 0x38, (bds[2] << 4) | (fd_bootchk ? 0x0 : 0x1));
305     return(0);
306 }
307 
308 static int pc_boot_set(void *opaque, const char *boot_device)
309 {
310     return set_boot_dev(opaque, boot_device);
311 }
312 
313 typedef struct pc_cmos_init_late_arg {
314     ISADevice *rtc_state;
315     BusState *idebus[2];
316 } pc_cmos_init_late_arg;
317 
318 static void pc_cmos_init_late(void *opaque)
319 {
320     pc_cmos_init_late_arg *arg = opaque;
321     ISADevice *s = arg->rtc_state;
322     int16_t cylinders;
323     int8_t heads, sectors;
324     int val;
325     int i, trans;
326 
327     val = 0;
328     if (ide_get_geometry(arg->idebus[0], 0,
329                          &cylinders, &heads, &sectors) >= 0) {
330         cmos_init_hd(s, 0x19, 0x1b, cylinders, heads, sectors);
331         val |= 0xf0;
332     }
333     if (ide_get_geometry(arg->idebus[0], 1,
334                          &cylinders, &heads, &sectors) >= 0) {
335         cmos_init_hd(s, 0x1a, 0x24, cylinders, heads, sectors);
336         val |= 0x0f;
337     }
338     rtc_set_memory(s, 0x12, val);
339 
340     val = 0;
341     for (i = 0; i < 4; i++) {
342         /* NOTE: ide_get_geometry() returns the physical
343            geometry.  It is always such that: 1 <= sects <= 63, 1
344            <= heads <= 16, 1 <= cylinders <= 16383. The BIOS
345            geometry can be different if a translation is done. */
346         if (ide_get_geometry(arg->idebus[i / 2], i % 2,
347                              &cylinders, &heads, &sectors) >= 0) {
348             trans = ide_get_bios_chs_trans(arg->idebus[i / 2], i % 2) - 1;
349             assert((trans & ~3) == 0);
350             val |= trans << (i * 2);
351         }
352     }
353     rtc_set_memory(s, 0x39, val);
354 
355     qemu_unregister_reset(pc_cmos_init_late, opaque);
356 }
357 
358 void pc_cmos_init(ram_addr_t ram_size, ram_addr_t above_4g_mem_size,
359                   const char *boot_device, MachineState *machine,
360                   ISADevice *floppy, BusState *idebus0, BusState *idebus1,
361                   ISADevice *s)
362 {
363     int val, nb, i;
364     FDriveType fd_type[2] = { FDRIVE_DRV_NONE, FDRIVE_DRV_NONE };
365     static pc_cmos_init_late_arg arg;
366     PCMachineState *pc_machine = PC_MACHINE(machine);
367 
368     /* various important CMOS locations needed by PC/Bochs bios */
369 
370     /* memory size */
371     /* base memory (first MiB) */
372     val = MIN(ram_size / 1024, 640);
373     rtc_set_memory(s, 0x15, val);
374     rtc_set_memory(s, 0x16, val >> 8);
375     /* extended memory (next 64MiB) */
376     if (ram_size > 1024 * 1024) {
377         val = (ram_size - 1024 * 1024) / 1024;
378     } else {
379         val = 0;
380     }
381     if (val > 65535)
382         val = 65535;
383     rtc_set_memory(s, 0x17, val);
384     rtc_set_memory(s, 0x18, val >> 8);
385     rtc_set_memory(s, 0x30, val);
386     rtc_set_memory(s, 0x31, val >> 8);
387     /* memory between 16MiB and 4GiB */
388     if (ram_size > 16 * 1024 * 1024) {
389         val = (ram_size - 16 * 1024 * 1024) / 65536;
390     } else {
391         val = 0;
392     }
393     if (val > 65535)
394         val = 65535;
395     rtc_set_memory(s, 0x34, val);
396     rtc_set_memory(s, 0x35, val >> 8);
397     /* memory above 4GiB */
398     val = above_4g_mem_size / 65536;
399     rtc_set_memory(s, 0x5b, val);
400     rtc_set_memory(s, 0x5c, val >> 8);
401     rtc_set_memory(s, 0x5d, val >> 16);
402 
403     /* set the number of CPU */
404     rtc_set_memory(s, 0x5f, smp_cpus - 1);
405 
406     object_property_add_link(OBJECT(machine), "rtc_state",
407                              TYPE_ISA_DEVICE,
408                              (Object **)&pc_machine->rtc,
409                              object_property_allow_set_link,
410                              OBJ_PROP_LINK_UNREF_ON_RELEASE, &error_abort);
411     object_property_set_link(OBJECT(machine), OBJECT(s),
412                              "rtc_state", &error_abort);
413 
414     if (set_boot_dev(s, boot_device)) {
415         exit(1);
416     }
417 
418     /* floppy type */
419     if (floppy) {
420         for (i = 0; i < 2; i++) {
421             fd_type[i] = isa_fdc_get_drive_type(floppy, i);
422         }
423     }
424     val = (cmos_get_fd_drive_type(fd_type[0]) << 4) |
425         cmos_get_fd_drive_type(fd_type[1]);
426     rtc_set_memory(s, 0x10, val);
427 
428     val = 0;
429     nb = 0;
430     if (fd_type[0] < FDRIVE_DRV_NONE) {
431         nb++;
432     }
433     if (fd_type[1] < FDRIVE_DRV_NONE) {
434         nb++;
435     }
436     switch (nb) {
437     case 0:
438         break;
439     case 1:
440         val |= 0x01; /* 1 drive, ready for boot */
441         break;
442     case 2:
443         val |= 0x41; /* 2 drives, ready for boot */
444         break;
445     }
446     val |= 0x02; /* FPU is there */
447     val |= 0x04; /* PS/2 mouse installed */
448     rtc_set_memory(s, REG_EQUIPMENT_BYTE, val);
449 
450     /* hard drives */
451     arg.rtc_state = s;
452     arg.idebus[0] = idebus0;
453     arg.idebus[1] = idebus1;
454     qemu_register_reset(pc_cmos_init_late, &arg);
455 }
456 
457 #define TYPE_PORT92 "port92"
458 #define PORT92(obj) OBJECT_CHECK(Port92State, (obj), TYPE_PORT92)
459 
460 /* port 92 stuff: could be split off */
461 typedef struct Port92State {
462     ISADevice parent_obj;
463 
464     MemoryRegion io;
465     uint8_t outport;
466     qemu_irq *a20_out;
467 } Port92State;
468 
469 static void port92_write(void *opaque, hwaddr addr, uint64_t val,
470                          unsigned size)
471 {
472     Port92State *s = opaque;
473     int oldval = s->outport;
474 
475     DPRINTF("port92: write 0x%02" PRIx64 "\n", val);
476     s->outport = val;
477     qemu_set_irq(*s->a20_out, (val >> 1) & 1);
478     if ((val & 1) && !(oldval & 1)) {
479         qemu_system_reset_request();
480     }
481 }
482 
483 static uint64_t port92_read(void *opaque, hwaddr addr,
484                             unsigned size)
485 {
486     Port92State *s = opaque;
487     uint32_t ret;
488 
489     ret = s->outport;
490     DPRINTF("port92: read 0x%02x\n", ret);
491     return ret;
492 }
493 
494 static void port92_init(ISADevice *dev, qemu_irq *a20_out)
495 {
496     Port92State *s = PORT92(dev);
497 
498     s->a20_out = a20_out;
499 }
500 
501 static const VMStateDescription vmstate_port92_isa = {
502     .name = "port92",
503     .version_id = 1,
504     .minimum_version_id = 1,
505     .fields = (VMStateField[]) {
506         VMSTATE_UINT8(outport, Port92State),
507         VMSTATE_END_OF_LIST()
508     }
509 };
510 
511 static void port92_reset(DeviceState *d)
512 {
513     Port92State *s = PORT92(d);
514 
515     s->outport &= ~1;
516 }
517 
518 static const MemoryRegionOps port92_ops = {
519     .read = port92_read,
520     .write = port92_write,
521     .impl = {
522         .min_access_size = 1,
523         .max_access_size = 1,
524     },
525     .endianness = DEVICE_LITTLE_ENDIAN,
526 };
527 
528 static void port92_initfn(Object *obj)
529 {
530     Port92State *s = PORT92(obj);
531 
532     memory_region_init_io(&s->io, OBJECT(s), &port92_ops, s, "port92", 1);
533 
534     s->outport = 0;
535 }
536 
537 static void port92_realizefn(DeviceState *dev, Error **errp)
538 {
539     ISADevice *isadev = ISA_DEVICE(dev);
540     Port92State *s = PORT92(dev);
541 
542     isa_register_ioport(isadev, &s->io, 0x92);
543 }
544 
545 static void port92_class_initfn(ObjectClass *klass, void *data)
546 {
547     DeviceClass *dc = DEVICE_CLASS(klass);
548 
549     dc->realize = port92_realizefn;
550     dc->reset = port92_reset;
551     dc->vmsd = &vmstate_port92_isa;
552     /*
553      * Reason: unlike ordinary ISA devices, this one needs additional
554      * wiring: its A20 output line needs to be wired up by
555      * port92_init().
556      */
557     dc->cannot_instantiate_with_device_add_yet = true;
558 }
559 
560 static const TypeInfo port92_info = {
561     .name          = TYPE_PORT92,
562     .parent        = TYPE_ISA_DEVICE,
563     .instance_size = sizeof(Port92State),
564     .instance_init = port92_initfn,
565     .class_init    = port92_class_initfn,
566 };
567 
568 static void port92_register_types(void)
569 {
570     type_register_static(&port92_info);
571 }
572 
573 type_init(port92_register_types)
574 
575 static void handle_a20_line_change(void *opaque, int irq, int level)
576 {
577     X86CPU *cpu = opaque;
578 
579     /* XXX: send to all CPUs ? */
580     /* XXX: add logic to handle multiple A20 line sources */
581     x86_cpu_set_a20(cpu, level);
582 }
583 
584 int e820_add_entry(uint64_t address, uint64_t length, uint32_t type)
585 {
586     int index = le32_to_cpu(e820_reserve.count);
587     struct e820_entry *entry;
588 
589     if (type != E820_RAM) {
590         /* old FW_CFG_E820_TABLE entry -- reservations only */
591         if (index >= E820_NR_ENTRIES) {
592             return -EBUSY;
593         }
594         entry = &e820_reserve.entry[index++];
595 
596         entry->address = cpu_to_le64(address);
597         entry->length = cpu_to_le64(length);
598         entry->type = cpu_to_le32(type);
599 
600         e820_reserve.count = cpu_to_le32(index);
601     }
602 
603     /* new "etc/e820" file -- include ram too */
604     e820_table = g_realloc(e820_table,
605                            sizeof(struct e820_entry) * (e820_entries+1));
606     e820_table[e820_entries].address = cpu_to_le64(address);
607     e820_table[e820_entries].length = cpu_to_le64(length);
608     e820_table[e820_entries].type = cpu_to_le32(type);
609     e820_entries++;
610 
611     return e820_entries;
612 }
613 
614 int e820_get_num_entries(void)
615 {
616     return e820_entries;
617 }
618 
619 bool e820_get_entry(int idx, uint32_t type, uint64_t *address, uint64_t *length)
620 {
621     if (idx < e820_entries && e820_table[idx].type == cpu_to_le32(type)) {
622         *address = le64_to_cpu(e820_table[idx].address);
623         *length = le64_to_cpu(e820_table[idx].length);
624         return true;
625     }
626     return false;
627 }
628 
629 /* Calculates the limit to CPU APIC ID values
630  *
631  * This function returns the limit for the APIC ID value, so that all
632  * CPU APIC IDs are < pc_apic_id_limit().
633  *
634  * This is used for FW_CFG_MAX_CPUS. See comments on bochs_bios_init().
635  */
636 static unsigned int pc_apic_id_limit(unsigned int max_cpus)
637 {
638     return x86_cpu_apic_id_from_index(max_cpus - 1) + 1;
639 }
640 
641 static FWCfgState *bochs_bios_init(void)
642 {
643     FWCfgState *fw_cfg;
644     uint8_t *smbios_tables, *smbios_anchor;
645     size_t smbios_tables_len, smbios_anchor_len;
646     uint64_t *numa_fw_cfg;
647     int i, j;
648     unsigned int apic_id_limit = pc_apic_id_limit(max_cpus);
649 
650     fw_cfg = fw_cfg_init(BIOS_CFG_IOPORT, BIOS_CFG_IOPORT + 1, 0, 0);
651     /* FW_CFG_MAX_CPUS is a bit confusing/problematic on x86:
652      *
653      * SeaBIOS needs FW_CFG_MAX_CPUS for CPU hotplug, but the CPU hotplug
654      * QEMU<->SeaBIOS interface is not based on the "CPU index", but on the APIC
655      * ID of hotplugged CPUs[1]. This means that FW_CFG_MAX_CPUS is not the
656      * "maximum number of CPUs", but the "limit to the APIC ID values SeaBIOS
657      * may see".
658      *
659      * So, this means we must not use max_cpus, here, but the maximum possible
660      * APIC ID value, plus one.
661      *
662      * [1] The only kind of "CPU identifier" used between SeaBIOS and QEMU is
663      *     the APIC ID, not the "CPU index"
664      */
665     fw_cfg_add_i16(fw_cfg, FW_CFG_MAX_CPUS, (uint16_t)apic_id_limit);
666     fw_cfg_add_i32(fw_cfg, FW_CFG_ID, 1);
667     fw_cfg_add_i64(fw_cfg, FW_CFG_RAM_SIZE, (uint64_t)ram_size);
668     fw_cfg_add_bytes(fw_cfg, FW_CFG_ACPI_TABLES,
669                      acpi_tables, acpi_tables_len);
670     fw_cfg_add_i32(fw_cfg, FW_CFG_IRQ0_OVERRIDE, kvm_allows_irq0_override());
671 
672     smbios_tables = smbios_get_table_legacy(&smbios_tables_len);
673     if (smbios_tables) {
674         fw_cfg_add_bytes(fw_cfg, FW_CFG_SMBIOS_ENTRIES,
675                          smbios_tables, smbios_tables_len);
676     }
677 
678     smbios_get_tables(&smbios_tables, &smbios_tables_len,
679                       &smbios_anchor, &smbios_anchor_len);
680     if (smbios_anchor) {
681         fw_cfg_add_file(fw_cfg, "etc/smbios/smbios-tables",
682                         smbios_tables, smbios_tables_len);
683         fw_cfg_add_file(fw_cfg, "etc/smbios/smbios-anchor",
684                         smbios_anchor, smbios_anchor_len);
685     }
686 
687     fw_cfg_add_bytes(fw_cfg, FW_CFG_E820_TABLE,
688                      &e820_reserve, sizeof(e820_reserve));
689     fw_cfg_add_file(fw_cfg, "etc/e820", e820_table,
690                     sizeof(struct e820_entry) * e820_entries);
691 
692     fw_cfg_add_bytes(fw_cfg, FW_CFG_HPET, &hpet_cfg, sizeof(hpet_cfg));
693     /* allocate memory for the NUMA channel: one (64bit) word for the number
694      * of nodes, one word for each VCPU->node and one word for each node to
695      * hold the amount of memory.
696      */
697     numa_fw_cfg = g_new0(uint64_t, 1 + apic_id_limit + nb_numa_nodes);
698     numa_fw_cfg[0] = cpu_to_le64(nb_numa_nodes);
699     for (i = 0; i < max_cpus; i++) {
700         unsigned int apic_id = x86_cpu_apic_id_from_index(i);
701         assert(apic_id < apic_id_limit);
702         for (j = 0; j < nb_numa_nodes; j++) {
703             if (test_bit(i, numa_info[j].node_cpu)) {
704                 numa_fw_cfg[apic_id + 1] = cpu_to_le64(j);
705                 break;
706             }
707         }
708     }
709     for (i = 0; i < nb_numa_nodes; i++) {
710         numa_fw_cfg[apic_id_limit + 1 + i] = cpu_to_le64(numa_info[i].node_mem);
711     }
712     fw_cfg_add_bytes(fw_cfg, FW_CFG_NUMA, numa_fw_cfg,
713                      (1 + apic_id_limit + nb_numa_nodes) *
714                      sizeof(*numa_fw_cfg));
715 
716     return fw_cfg;
717 }
718 
719 static long get_file_size(FILE *f)
720 {
721     long where, size;
722 
723     /* XXX: on Unix systems, using fstat() probably makes more sense */
724 
725     where = ftell(f);
726     fseek(f, 0, SEEK_END);
727     size = ftell(f);
728     fseek(f, where, SEEK_SET);
729 
730     return size;
731 }
732 
733 static void load_linux(FWCfgState *fw_cfg,
734                        const char *kernel_filename,
735                        const char *initrd_filename,
736                        const char *kernel_cmdline,
737                        hwaddr max_ram_size)
738 {
739     uint16_t protocol;
740     int setup_size, kernel_size, initrd_size = 0, cmdline_size;
741     uint32_t initrd_max;
742     uint8_t header[8192], *setup, *kernel, *initrd_data;
743     hwaddr real_addr, prot_addr, cmdline_addr, initrd_addr = 0;
744     FILE *f;
745     char *vmode;
746 
747     /* Align to 16 bytes as a paranoia measure */
748     cmdline_size = (strlen(kernel_cmdline)+16) & ~15;
749 
750     /* load the kernel header */
751     f = fopen(kernel_filename, "rb");
752     if (!f || !(kernel_size = get_file_size(f)) ||
753         fread(header, 1, MIN(ARRAY_SIZE(header), kernel_size), f) !=
754         MIN(ARRAY_SIZE(header), kernel_size)) {
755         fprintf(stderr, "qemu: could not load kernel '%s': %s\n",
756                 kernel_filename, strerror(errno));
757         exit(1);
758     }
759 
760     /* kernel protocol version */
761 #if 0
762     fprintf(stderr, "header magic: %#x\n", ldl_p(header+0x202));
763 #endif
764     if (ldl_p(header+0x202) == 0x53726448) {
765         protocol = lduw_p(header+0x206);
766     } else {
767         /* This looks like a multiboot kernel. If it is, let's stop
768            treating it like a Linux kernel. */
769         if (load_multiboot(fw_cfg, f, kernel_filename, initrd_filename,
770                            kernel_cmdline, kernel_size, header)) {
771             return;
772         }
773         protocol = 0;
774     }
775 
776     if (protocol < 0x200 || !(header[0x211] & 0x01)) {
777         /* Low kernel */
778         real_addr    = 0x90000;
779         cmdline_addr = 0x9a000 - cmdline_size;
780         prot_addr    = 0x10000;
781     } else if (protocol < 0x202) {
782         /* High but ancient kernel */
783         real_addr    = 0x90000;
784         cmdline_addr = 0x9a000 - cmdline_size;
785         prot_addr    = 0x100000;
786     } else {
787         /* High and recent kernel */
788         real_addr    = 0x10000;
789         cmdline_addr = 0x20000;
790         prot_addr    = 0x100000;
791     }
792 
793 #if 0
794     fprintf(stderr,
795             "qemu: real_addr     = 0x" TARGET_FMT_plx "\n"
796             "qemu: cmdline_addr  = 0x" TARGET_FMT_plx "\n"
797             "qemu: prot_addr     = 0x" TARGET_FMT_plx "\n",
798             real_addr,
799             cmdline_addr,
800             prot_addr);
801 #endif
802 
803     /* highest address for loading the initrd */
804     if (protocol >= 0x203) {
805         initrd_max = ldl_p(header+0x22c);
806     } else {
807         initrd_max = 0x37ffffff;
808     }
809 
810     if (initrd_max >= max_ram_size - acpi_data_size) {
811         initrd_max = max_ram_size - acpi_data_size - 1;
812     }
813 
814     fw_cfg_add_i32(fw_cfg, FW_CFG_CMDLINE_ADDR, cmdline_addr);
815     fw_cfg_add_i32(fw_cfg, FW_CFG_CMDLINE_SIZE, strlen(kernel_cmdline)+1);
816     fw_cfg_add_string(fw_cfg, FW_CFG_CMDLINE_DATA, kernel_cmdline);
817 
818     if (protocol >= 0x202) {
819         stl_p(header+0x228, cmdline_addr);
820     } else {
821         stw_p(header+0x20, 0xA33F);
822         stw_p(header+0x22, cmdline_addr-real_addr);
823     }
824 
825     /* handle vga= parameter */
826     vmode = strstr(kernel_cmdline, "vga=");
827     if (vmode) {
828         unsigned int video_mode;
829         /* skip "vga=" */
830         vmode += 4;
831         if (!strncmp(vmode, "normal", 6)) {
832             video_mode = 0xffff;
833         } else if (!strncmp(vmode, "ext", 3)) {
834             video_mode = 0xfffe;
835         } else if (!strncmp(vmode, "ask", 3)) {
836             video_mode = 0xfffd;
837         } else {
838             video_mode = strtol(vmode, NULL, 0);
839         }
840         stw_p(header+0x1fa, video_mode);
841     }
842 
843     /* loader type */
844     /* High nybble = B reserved for QEMU; low nybble is revision number.
845        If this code is substantially changed, you may want to consider
846        incrementing the revision. */
847     if (protocol >= 0x200) {
848         header[0x210] = 0xB0;
849     }
850     /* heap */
851     if (protocol >= 0x201) {
852         header[0x211] |= 0x80;	/* CAN_USE_HEAP */
853         stw_p(header+0x224, cmdline_addr-real_addr-0x200);
854     }
855 
856     /* load initrd */
857     if (initrd_filename) {
858         if (protocol < 0x200) {
859             fprintf(stderr, "qemu: linux kernel too old to load a ram disk\n");
860             exit(1);
861         }
862 
863         initrd_size = get_image_size(initrd_filename);
864         if (initrd_size < 0) {
865             fprintf(stderr, "qemu: error reading initrd %s: %s\n",
866                     initrd_filename, strerror(errno));
867             exit(1);
868         }
869 
870         initrd_addr = (initrd_max-initrd_size) & ~4095;
871 
872         initrd_data = g_malloc(initrd_size);
873         load_image(initrd_filename, initrd_data);
874 
875         fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_ADDR, initrd_addr);
876         fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_SIZE, initrd_size);
877         fw_cfg_add_bytes(fw_cfg, FW_CFG_INITRD_DATA, initrd_data, initrd_size);
878 
879         stl_p(header+0x218, initrd_addr);
880         stl_p(header+0x21c, initrd_size);
881     }
882 
883     /* load kernel and setup */
884     setup_size = header[0x1f1];
885     if (setup_size == 0) {
886         setup_size = 4;
887     }
888     setup_size = (setup_size+1)*512;
889     kernel_size -= setup_size;
890 
891     setup  = g_malloc(setup_size);
892     kernel = g_malloc(kernel_size);
893     fseek(f, 0, SEEK_SET);
894     if (fread(setup, 1, setup_size, f) != setup_size) {
895         fprintf(stderr, "fread() failed\n");
896         exit(1);
897     }
898     if (fread(kernel, 1, kernel_size, f) != kernel_size) {
899         fprintf(stderr, "fread() failed\n");
900         exit(1);
901     }
902     fclose(f);
903     memcpy(setup, header, MIN(sizeof(header), setup_size));
904 
905     fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_ADDR, prot_addr);
906     fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_SIZE, kernel_size);
907     fw_cfg_add_bytes(fw_cfg, FW_CFG_KERNEL_DATA, kernel, kernel_size);
908 
909     fw_cfg_add_i32(fw_cfg, FW_CFG_SETUP_ADDR, real_addr);
910     fw_cfg_add_i32(fw_cfg, FW_CFG_SETUP_SIZE, setup_size);
911     fw_cfg_add_bytes(fw_cfg, FW_CFG_SETUP_DATA, setup, setup_size);
912 
913     option_rom[nb_option_roms].name = "linuxboot.bin";
914     option_rom[nb_option_roms].bootindex = 0;
915     nb_option_roms++;
916 }
917 
918 #define NE2000_NB_MAX 6
919 
920 static const int ne2000_io[NE2000_NB_MAX] = { 0x300, 0x320, 0x340, 0x360,
921                                               0x280, 0x380 };
922 static const int ne2000_irq[NE2000_NB_MAX] = { 9, 10, 11, 3, 4, 5 };
923 
924 void pc_init_ne2k_isa(ISABus *bus, NICInfo *nd)
925 {
926     static int nb_ne2k = 0;
927 
928     if (nb_ne2k == NE2000_NB_MAX)
929         return;
930     isa_ne2000_init(bus, ne2000_io[nb_ne2k],
931                     ne2000_irq[nb_ne2k], nd);
932     nb_ne2k++;
933 }
934 
935 DeviceState *cpu_get_current_apic(void)
936 {
937     if (current_cpu) {
938         X86CPU *cpu = X86_CPU(current_cpu);
939         return cpu->apic_state;
940     } else {
941         return NULL;
942     }
943 }
944 
945 void pc_acpi_smi_interrupt(void *opaque, int irq, int level)
946 {
947     X86CPU *cpu = opaque;
948 
949     if (level) {
950         cpu_interrupt(CPU(cpu), CPU_INTERRUPT_SMI);
951     }
952 }
953 
954 static X86CPU *pc_new_cpu(const char *cpu_model, int64_t apic_id,
955                           DeviceState *icc_bridge, Error **errp)
956 {
957     X86CPU *cpu;
958     Error *local_err = NULL;
959 
960     cpu = cpu_x86_create(cpu_model, icc_bridge, &local_err);
961     if (local_err != NULL) {
962         error_propagate(errp, local_err);
963         return NULL;
964     }
965 
966     object_property_set_int(OBJECT(cpu), apic_id, "apic-id", &local_err);
967     object_property_set_bool(OBJECT(cpu), true, "realized", &local_err);
968 
969     if (local_err) {
970         error_propagate(errp, local_err);
971         object_unref(OBJECT(cpu));
972         cpu = NULL;
973     }
974     return cpu;
975 }
976 
977 static const char *current_cpu_model;
978 
979 void pc_hot_add_cpu(const int64_t id, Error **errp)
980 {
981     DeviceState *icc_bridge;
982     int64_t apic_id = x86_cpu_apic_id_from_index(id);
983 
984     if (id < 0) {
985         error_setg(errp, "Invalid CPU id: %" PRIi64, id);
986         return;
987     }
988 
989     if (cpu_exists(apic_id)) {
990         error_setg(errp, "Unable to add CPU: %" PRIi64
991                    ", it already exists", id);
992         return;
993     }
994 
995     if (id >= max_cpus) {
996         error_setg(errp, "Unable to add CPU: %" PRIi64
997                    ", max allowed: %d", id, max_cpus - 1);
998         return;
999     }
1000 
1001     if (apic_id >= ACPI_CPU_HOTPLUG_ID_LIMIT) {
1002         error_setg(errp, "Unable to add CPU: %" PRIi64
1003                    ", resulting APIC ID (%" PRIi64 ") is too large",
1004                    id, apic_id);
1005         return;
1006     }
1007 
1008     icc_bridge = DEVICE(object_resolve_path_type("icc-bridge",
1009                                                  TYPE_ICC_BRIDGE, NULL));
1010     pc_new_cpu(current_cpu_model, apic_id, icc_bridge, errp);
1011 }
1012 
1013 void pc_cpus_init(const char *cpu_model, DeviceState *icc_bridge)
1014 {
1015     int i;
1016     X86CPU *cpu = NULL;
1017     Error *error = NULL;
1018     unsigned long apic_id_limit;
1019 
1020     /* init CPUs */
1021     if (cpu_model == NULL) {
1022 #ifdef TARGET_X86_64
1023         cpu_model = "qemu64";
1024 #else
1025         cpu_model = "qemu32";
1026 #endif
1027     }
1028     current_cpu_model = cpu_model;
1029 
1030     apic_id_limit = pc_apic_id_limit(max_cpus);
1031     if (apic_id_limit > ACPI_CPU_HOTPLUG_ID_LIMIT) {
1032         error_report("max_cpus is too large. APIC ID of last CPU is %lu",
1033                      apic_id_limit - 1);
1034         exit(1);
1035     }
1036 
1037     for (i = 0; i < smp_cpus; i++) {
1038         cpu = pc_new_cpu(cpu_model, x86_cpu_apic_id_from_index(i),
1039                          icc_bridge, &error);
1040         if (error) {
1041             error_report("%s", error_get_pretty(error));
1042             error_free(error);
1043             exit(1);
1044         }
1045     }
1046 
1047     /* map APIC MMIO area if CPU has APIC */
1048     if (cpu && cpu->apic_state) {
1049         /* XXX: what if the base changes? */
1050         sysbus_mmio_map_overlap(SYS_BUS_DEVICE(icc_bridge), 0,
1051                                 APIC_DEFAULT_ADDRESS, 0x1000);
1052     }
1053 
1054     /* tell smbios about cpuid version and features */
1055     smbios_set_cpuid(cpu->env.cpuid_version, cpu->env.features[FEAT_1_EDX]);
1056 }
1057 
1058 /* pci-info ROM file. Little endian format */
1059 typedef struct PcRomPciInfo {
1060     uint64_t w32_min;
1061     uint64_t w32_max;
1062     uint64_t w64_min;
1063     uint64_t w64_max;
1064 } PcRomPciInfo;
1065 
1066 typedef struct PcGuestInfoState {
1067     PcGuestInfo info;
1068     Notifier machine_done;
1069 } PcGuestInfoState;
1070 
1071 static
1072 void pc_guest_info_machine_done(Notifier *notifier, void *data)
1073 {
1074     PcGuestInfoState *guest_info_state = container_of(notifier,
1075                                                       PcGuestInfoState,
1076                                                       machine_done);
1077     acpi_setup(&guest_info_state->info);
1078 }
1079 
1080 PcGuestInfo *pc_guest_info_init(ram_addr_t below_4g_mem_size,
1081                                 ram_addr_t above_4g_mem_size)
1082 {
1083     PcGuestInfoState *guest_info_state = g_malloc0(sizeof *guest_info_state);
1084     PcGuestInfo *guest_info = &guest_info_state->info;
1085     int i, j;
1086 
1087     guest_info->ram_size_below_4g = below_4g_mem_size;
1088     guest_info->ram_size = below_4g_mem_size + above_4g_mem_size;
1089     guest_info->apic_id_limit = pc_apic_id_limit(max_cpus);
1090     guest_info->apic_xrupt_override = kvm_allows_irq0_override();
1091     guest_info->numa_nodes = nb_numa_nodes;
1092     guest_info->node_mem = g_malloc0(guest_info->numa_nodes *
1093                                     sizeof *guest_info->node_mem);
1094     for (i = 0; i < nb_numa_nodes; i++) {
1095         guest_info->node_mem[i] = numa_info[i].node_mem;
1096     }
1097 
1098     guest_info->node_cpu = g_malloc0(guest_info->apic_id_limit *
1099                                      sizeof *guest_info->node_cpu);
1100 
1101     for (i = 0; i < max_cpus; i++) {
1102         unsigned int apic_id = x86_cpu_apic_id_from_index(i);
1103         assert(apic_id < guest_info->apic_id_limit);
1104         for (j = 0; j < nb_numa_nodes; j++) {
1105             if (test_bit(i, numa_info[j].node_cpu)) {
1106                 guest_info->node_cpu[apic_id] = j;
1107                 break;
1108             }
1109         }
1110     }
1111 
1112     guest_info_state->machine_done.notify = pc_guest_info_machine_done;
1113     qemu_add_machine_init_done_notifier(&guest_info_state->machine_done);
1114     return guest_info;
1115 }
1116 
1117 /* setup pci memory address space mapping into system address space */
1118 void pc_pci_as_mapping_init(Object *owner, MemoryRegion *system_memory,
1119                             MemoryRegion *pci_address_space)
1120 {
1121     /* Set to lower priority than RAM */
1122     memory_region_add_subregion_overlap(system_memory, 0x0,
1123                                         pci_address_space, -1);
1124 }
1125 
1126 void pc_acpi_init(const char *default_dsdt)
1127 {
1128     char *filename;
1129 
1130     if (acpi_tables != NULL) {
1131         /* manually set via -acpitable, leave it alone */
1132         return;
1133     }
1134 
1135     filename = qemu_find_file(QEMU_FILE_TYPE_BIOS, default_dsdt);
1136     if (filename == NULL) {
1137         fprintf(stderr, "WARNING: failed to find %s\n", default_dsdt);
1138     } else {
1139         char *arg;
1140         QemuOpts *opts;
1141         Error *err = NULL;
1142 
1143         arg = g_strdup_printf("file=%s", filename);
1144 
1145         /* creates a deep copy of "arg" */
1146         opts = qemu_opts_parse(qemu_find_opts("acpi"), arg, 0);
1147         g_assert(opts != NULL);
1148 
1149         acpi_table_add_builtin(opts, &err);
1150         if (err) {
1151             error_report("WARNING: failed to load %s: %s", filename,
1152                          error_get_pretty(err));
1153             error_free(err);
1154         }
1155         g_free(arg);
1156         g_free(filename);
1157     }
1158 }
1159 
1160 FWCfgState *xen_load_linux(const char *kernel_filename,
1161                            const char *kernel_cmdline,
1162                            const char *initrd_filename,
1163                            ram_addr_t below_4g_mem_size,
1164                            PcGuestInfo *guest_info)
1165 {
1166     int i;
1167     FWCfgState *fw_cfg;
1168 
1169     assert(kernel_filename != NULL);
1170 
1171     fw_cfg = fw_cfg_init(BIOS_CFG_IOPORT, BIOS_CFG_IOPORT + 1, 0, 0);
1172     rom_set_fw(fw_cfg);
1173 
1174     load_linux(fw_cfg, kernel_filename, initrd_filename,
1175                kernel_cmdline, below_4g_mem_size);
1176     for (i = 0; i < nb_option_roms; i++) {
1177         assert(!strcmp(option_rom[i].name, "linuxboot.bin") ||
1178                !strcmp(option_rom[i].name, "multiboot.bin"));
1179         rom_add_option(option_rom[i].name, option_rom[i].bootindex);
1180     }
1181     guest_info->fw_cfg = fw_cfg;
1182     return fw_cfg;
1183 }
1184 
1185 FWCfgState *pc_memory_init(MachineState *machine,
1186                            MemoryRegion *system_memory,
1187                            ram_addr_t below_4g_mem_size,
1188                            ram_addr_t above_4g_mem_size,
1189                            MemoryRegion *rom_memory,
1190                            MemoryRegion **ram_memory,
1191                            PcGuestInfo *guest_info)
1192 {
1193     int linux_boot, i;
1194     MemoryRegion *ram, *option_rom_mr;
1195     MemoryRegion *ram_below_4g, *ram_above_4g;
1196     FWCfgState *fw_cfg;
1197     PCMachineState *pcms = PC_MACHINE(machine);
1198 
1199     assert(machine->ram_size == below_4g_mem_size + above_4g_mem_size);
1200 
1201     linux_boot = (machine->kernel_filename != NULL);
1202 
1203     /* Allocate RAM.  We allocate it as a single memory region and use
1204      * aliases to address portions of it, mostly for backwards compatibility
1205      * with older qemus that used qemu_ram_alloc().
1206      */
1207     ram = g_malloc(sizeof(*ram));
1208     memory_region_allocate_system_memory(ram, NULL, "pc.ram",
1209                                          machine->ram_size);
1210     *ram_memory = ram;
1211     ram_below_4g = g_malloc(sizeof(*ram_below_4g));
1212     memory_region_init_alias(ram_below_4g, NULL, "ram-below-4g", ram,
1213                              0, below_4g_mem_size);
1214     memory_region_add_subregion(system_memory, 0, ram_below_4g);
1215     e820_add_entry(0, below_4g_mem_size, E820_RAM);
1216     if (above_4g_mem_size > 0) {
1217         ram_above_4g = g_malloc(sizeof(*ram_above_4g));
1218         memory_region_init_alias(ram_above_4g, NULL, "ram-above-4g", ram,
1219                                  below_4g_mem_size, above_4g_mem_size);
1220         memory_region_add_subregion(system_memory, 0x100000000ULL,
1221                                     ram_above_4g);
1222         e820_add_entry(0x100000000ULL, above_4g_mem_size, E820_RAM);
1223     }
1224 
1225     if (!guest_info->has_reserved_memory &&
1226         (machine->ram_slots ||
1227          (machine->maxram_size > machine->ram_size))) {
1228         MachineClass *mc = MACHINE_GET_CLASS(machine);
1229 
1230         error_report("\"-memory 'slots|maxmem'\" is not supported by: %s",
1231                      mc->name);
1232         exit(EXIT_FAILURE);
1233     }
1234 
1235     /* initialize hotplug memory address space */
1236     if (guest_info->has_reserved_memory &&
1237         (machine->ram_size < machine->maxram_size)) {
1238         ram_addr_t hotplug_mem_size =
1239             machine->maxram_size - machine->ram_size;
1240 
1241         if (machine->ram_slots > ACPI_MAX_RAM_SLOTS) {
1242             error_report("unsupported amount of memory slots: %"PRIu64,
1243                          machine->ram_slots);
1244             exit(EXIT_FAILURE);
1245         }
1246 
1247         pcms->hotplug_memory_base =
1248             ROUND_UP(0x100000000ULL + above_4g_mem_size, 1ULL << 30);
1249 
1250         if ((pcms->hotplug_memory_base + hotplug_mem_size) <
1251             hotplug_mem_size) {
1252             error_report("unsupported amount of maximum memory: " RAM_ADDR_FMT,
1253                          machine->maxram_size);
1254             exit(EXIT_FAILURE);
1255         }
1256 
1257         memory_region_init(&pcms->hotplug_memory, OBJECT(pcms),
1258                            "hotplug-memory", hotplug_mem_size);
1259         memory_region_add_subregion(system_memory, pcms->hotplug_memory_base,
1260                                     &pcms->hotplug_memory);
1261     }
1262 
1263     /* Initialize PC system firmware */
1264     pc_system_firmware_init(rom_memory, guest_info->isapc_ram_fw);
1265 
1266     option_rom_mr = g_malloc(sizeof(*option_rom_mr));
1267     memory_region_init_ram(option_rom_mr, NULL, "pc.rom", PC_ROM_SIZE,
1268                            &error_abort);
1269     vmstate_register_ram_global(option_rom_mr);
1270     memory_region_add_subregion_overlap(rom_memory,
1271                                         PC_ROM_MIN_VGA,
1272                                         option_rom_mr,
1273                                         1);
1274 
1275     fw_cfg = bochs_bios_init();
1276     rom_set_fw(fw_cfg);
1277 
1278     if (guest_info->has_reserved_memory && pcms->hotplug_memory_base) {
1279         uint64_t *val = g_malloc(sizeof(*val));
1280         *val = cpu_to_le64(ROUND_UP(pcms->hotplug_memory_base, 0x1ULL << 30));
1281         fw_cfg_add_file(fw_cfg, "etc/reserved-memory-end", val, sizeof(*val));
1282     }
1283 
1284     if (linux_boot) {
1285         load_linux(fw_cfg, machine->kernel_filename, machine->initrd_filename,
1286                    machine->kernel_cmdline, below_4g_mem_size);
1287     }
1288 
1289     for (i = 0; i < nb_option_roms; i++) {
1290         rom_add_option(option_rom[i].name, option_rom[i].bootindex);
1291     }
1292     guest_info->fw_cfg = fw_cfg;
1293     return fw_cfg;
1294 }
1295 
1296 qemu_irq *pc_allocate_cpu_irq(void)
1297 {
1298     return qemu_allocate_irqs(pic_irq_request, NULL, 1);
1299 }
1300 
1301 DeviceState *pc_vga_init(ISABus *isa_bus, PCIBus *pci_bus)
1302 {
1303     DeviceState *dev = NULL;
1304 
1305     if (pci_bus) {
1306         PCIDevice *pcidev = pci_vga_init(pci_bus);
1307         dev = pcidev ? &pcidev->qdev : NULL;
1308     } else if (isa_bus) {
1309         ISADevice *isadev = isa_vga_init(isa_bus);
1310         dev = isadev ? DEVICE(isadev) : NULL;
1311     }
1312     return dev;
1313 }
1314 
1315 static void cpu_request_exit(void *opaque, int irq, int level)
1316 {
1317     CPUState *cpu = current_cpu;
1318 
1319     if (cpu && level) {
1320         cpu_exit(cpu);
1321     }
1322 }
1323 
1324 static const MemoryRegionOps ioport80_io_ops = {
1325     .write = ioport80_write,
1326     .read = ioport80_read,
1327     .endianness = DEVICE_NATIVE_ENDIAN,
1328     .impl = {
1329         .min_access_size = 1,
1330         .max_access_size = 1,
1331     },
1332 };
1333 
1334 static const MemoryRegionOps ioportF0_io_ops = {
1335     .write = ioportF0_write,
1336     .read = ioportF0_read,
1337     .endianness = DEVICE_NATIVE_ENDIAN,
1338     .impl = {
1339         .min_access_size = 1,
1340         .max_access_size = 1,
1341     },
1342 };
1343 
1344 void pc_basic_device_init(ISABus *isa_bus, qemu_irq *gsi,
1345                           ISADevice **rtc_state,
1346                           ISADevice **floppy,
1347                           bool no_vmport,
1348                           uint32 hpet_irqs)
1349 {
1350     int i;
1351     DriveInfo *fd[MAX_FD];
1352     DeviceState *hpet = NULL;
1353     int pit_isa_irq = 0;
1354     qemu_irq pit_alt_irq = NULL;
1355     qemu_irq rtc_irq = NULL;
1356     qemu_irq *a20_line;
1357     ISADevice *i8042, *port92, *vmmouse, *pit = NULL;
1358     qemu_irq *cpu_exit_irq;
1359     MemoryRegion *ioport80_io = g_new(MemoryRegion, 1);
1360     MemoryRegion *ioportF0_io = g_new(MemoryRegion, 1);
1361 
1362     memory_region_init_io(ioport80_io, NULL, &ioport80_io_ops, NULL, "ioport80", 1);
1363     memory_region_add_subregion(isa_bus->address_space_io, 0x80, ioport80_io);
1364 
1365     memory_region_init_io(ioportF0_io, NULL, &ioportF0_io_ops, NULL, "ioportF0", 1);
1366     memory_region_add_subregion(isa_bus->address_space_io, 0xf0, ioportF0_io);
1367 
1368     /*
1369      * Check if an HPET shall be created.
1370      *
1371      * Without KVM_CAP_PIT_STATE2, we cannot switch off the in-kernel PIT
1372      * when the HPET wants to take over. Thus we have to disable the latter.
1373      */
1374     if (!no_hpet && (!kvm_irqchip_in_kernel() || kvm_has_pit_state2())) {
1375         /* In order to set property, here not using sysbus_try_create_simple */
1376         hpet = qdev_try_create(NULL, TYPE_HPET);
1377         if (hpet) {
1378             /* For pc-piix-*, hpet's intcap is always IRQ2. For pc-q35-1.7
1379              * and earlier, use IRQ2 for compat. Otherwise, use IRQ16~23,
1380              * IRQ8 and IRQ2.
1381              */
1382             uint8_t compat = object_property_get_int(OBJECT(hpet),
1383                     HPET_INTCAP, NULL);
1384             if (!compat) {
1385                 qdev_prop_set_uint32(hpet, HPET_INTCAP, hpet_irqs);
1386             }
1387             qdev_init_nofail(hpet);
1388             sysbus_mmio_map(SYS_BUS_DEVICE(hpet), 0, HPET_BASE);
1389 
1390             for (i = 0; i < GSI_NUM_PINS; i++) {
1391                 sysbus_connect_irq(SYS_BUS_DEVICE(hpet), i, gsi[i]);
1392             }
1393             pit_isa_irq = -1;
1394             pit_alt_irq = qdev_get_gpio_in(hpet, HPET_LEGACY_PIT_INT);
1395             rtc_irq = qdev_get_gpio_in(hpet, HPET_LEGACY_RTC_INT);
1396         }
1397     }
1398     *rtc_state = rtc_init(isa_bus, 2000, rtc_irq);
1399 
1400     qemu_register_boot_set(pc_boot_set, *rtc_state);
1401 
1402     if (!xen_enabled()) {
1403         if (kvm_irqchip_in_kernel()) {
1404             pit = kvm_pit_init(isa_bus, 0x40);
1405         } else {
1406             pit = pit_init(isa_bus, 0x40, pit_isa_irq, pit_alt_irq);
1407         }
1408         if (hpet) {
1409             /* connect PIT to output control line of the HPET */
1410             qdev_connect_gpio_out(hpet, 0, qdev_get_gpio_in(DEVICE(pit), 0));
1411         }
1412         pcspk_init(isa_bus, pit);
1413     }
1414 
1415     for(i = 0; i < MAX_SERIAL_PORTS; i++) {
1416         if (serial_hds[i]) {
1417             serial_isa_init(isa_bus, i, serial_hds[i]);
1418         }
1419     }
1420 
1421     for(i = 0; i < MAX_PARALLEL_PORTS; i++) {
1422         if (parallel_hds[i]) {
1423             parallel_init(isa_bus, i, parallel_hds[i]);
1424         }
1425     }
1426 
1427     a20_line = qemu_allocate_irqs(handle_a20_line_change, first_cpu, 2);
1428     i8042 = isa_create_simple(isa_bus, "i8042");
1429     i8042_setup_a20_line(i8042, &a20_line[0]);
1430     if (!no_vmport) {
1431         vmport_init(isa_bus);
1432         vmmouse = isa_try_create(isa_bus, "vmmouse");
1433     } else {
1434         vmmouse = NULL;
1435     }
1436     if (vmmouse) {
1437         DeviceState *dev = DEVICE(vmmouse);
1438         qdev_prop_set_ptr(dev, "ps2_mouse", i8042);
1439         qdev_init_nofail(dev);
1440     }
1441     port92 = isa_create_simple(isa_bus, "port92");
1442     port92_init(port92, &a20_line[1]);
1443 
1444     cpu_exit_irq = qemu_allocate_irqs(cpu_request_exit, NULL, 1);
1445     DMA_init(0, cpu_exit_irq);
1446 
1447     for(i = 0; i < MAX_FD; i++) {
1448         fd[i] = drive_get(IF_FLOPPY, 0, i);
1449     }
1450     *floppy = fdctrl_init_isa(isa_bus, fd);
1451 }
1452 
1453 void pc_nic_init(ISABus *isa_bus, PCIBus *pci_bus)
1454 {
1455     int i;
1456 
1457     for (i = 0; i < nb_nics; i++) {
1458         NICInfo *nd = &nd_table[i];
1459 
1460         if (!pci_bus || (nd->model && strcmp(nd->model, "ne2k_isa") == 0)) {
1461             pc_init_ne2k_isa(isa_bus, nd);
1462         } else {
1463             pci_nic_init_nofail(nd, pci_bus, "e1000", NULL);
1464         }
1465     }
1466 }
1467 
1468 void pc_pci_device_init(PCIBus *pci_bus)
1469 {
1470     int max_bus;
1471     int bus;
1472 
1473     max_bus = drive_get_max_bus(IF_SCSI);
1474     for (bus = 0; bus <= max_bus; bus++) {
1475         pci_create_simple(pci_bus, -1, "lsi53c895a");
1476     }
1477 }
1478 
1479 void ioapic_init_gsi(GSIState *gsi_state, const char *parent_name)
1480 {
1481     DeviceState *dev;
1482     SysBusDevice *d;
1483     unsigned int i;
1484 
1485     if (kvm_irqchip_in_kernel()) {
1486         dev = qdev_create(NULL, "kvm-ioapic");
1487     } else {
1488         dev = qdev_create(NULL, "ioapic");
1489     }
1490     if (parent_name) {
1491         object_property_add_child(object_resolve_path(parent_name, NULL),
1492                                   "ioapic", OBJECT(dev), NULL);
1493     }
1494     qdev_init_nofail(dev);
1495     d = SYS_BUS_DEVICE(dev);
1496     sysbus_mmio_map(d, 0, IO_APIC_DEFAULT_ADDRESS);
1497 
1498     for (i = 0; i < IOAPIC_NUM_PINS; i++) {
1499         gsi_state->ioapic_irq[i] = qdev_get_gpio_in(dev, i);
1500     }
1501 }
1502 
1503 static void pc_generic_machine_class_init(ObjectClass *oc, void *data)
1504 {
1505     MachineClass *mc = MACHINE_CLASS(oc);
1506     QEMUMachine *qm = data;
1507 
1508     mc->family = qm->family;
1509     mc->name = qm->name;
1510     mc->alias = qm->alias;
1511     mc->desc = qm->desc;
1512     mc->init = qm->init;
1513     mc->reset = qm->reset;
1514     mc->hot_add_cpu = qm->hot_add_cpu;
1515     mc->kvm_type = qm->kvm_type;
1516     mc->block_default_type = qm->block_default_type;
1517     mc->units_per_default_bus = qm->units_per_default_bus;
1518     mc->max_cpus = qm->max_cpus;
1519     mc->no_serial = qm->no_serial;
1520     mc->no_parallel = qm->no_parallel;
1521     mc->use_virtcon = qm->use_virtcon;
1522     mc->use_sclp = qm->use_sclp;
1523     mc->no_floppy = qm->no_floppy;
1524     mc->no_cdrom = qm->no_cdrom;
1525     mc->no_sdcard = qm->no_sdcard;
1526     mc->is_default = qm->is_default;
1527     mc->default_machine_opts = qm->default_machine_opts;
1528     mc->default_boot_order = qm->default_boot_order;
1529     mc->default_display = qm->default_display;
1530     mc->compat_props = qm->compat_props;
1531     mc->hw_version = qm->hw_version;
1532 }
1533 
1534 void qemu_register_pc_machine(QEMUMachine *m)
1535 {
1536     char *name = g_strconcat(m->name, TYPE_MACHINE_SUFFIX, NULL);
1537     TypeInfo ti = {
1538         .name       = name,
1539         .parent     = TYPE_PC_MACHINE,
1540         .class_init = pc_generic_machine_class_init,
1541         .class_data = (void *)m,
1542     };
1543 
1544     type_register(&ti);
1545     g_free(name);
1546 }
1547 
1548 static void pc_dimm_plug(HotplugHandler *hotplug_dev,
1549                          DeviceState *dev, Error **errp)
1550 {
1551     int slot;
1552     HotplugHandlerClass *hhc;
1553     Error *local_err = NULL;
1554     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1555     MachineState *machine = MACHINE(hotplug_dev);
1556     PCDIMMDevice *dimm = PC_DIMM(dev);
1557     PCDIMMDeviceClass *ddc = PC_DIMM_GET_CLASS(dimm);
1558     MemoryRegion *mr = ddc->get_memory_region(dimm);
1559     uint64_t addr = object_property_get_int(OBJECT(dimm), PC_DIMM_ADDR_PROP,
1560                                             &local_err);
1561     if (local_err) {
1562         goto out;
1563     }
1564 
1565     addr = pc_dimm_get_free_addr(pcms->hotplug_memory_base,
1566                                  memory_region_size(&pcms->hotplug_memory),
1567                                  !addr ? NULL : &addr,
1568                                  memory_region_size(mr), &local_err);
1569     if (local_err) {
1570         goto out;
1571     }
1572 
1573     object_property_set_int(OBJECT(dev), addr, PC_DIMM_ADDR_PROP, &local_err);
1574     if (local_err) {
1575         goto out;
1576     }
1577     trace_mhp_pc_dimm_assigned_address(addr);
1578 
1579     slot = object_property_get_int(OBJECT(dev), PC_DIMM_SLOT_PROP, &local_err);
1580     if (local_err) {
1581         goto out;
1582     }
1583 
1584     slot = pc_dimm_get_free_slot(slot == PC_DIMM_UNASSIGNED_SLOT ? NULL : &slot,
1585                                  machine->ram_slots, &local_err);
1586     if (local_err) {
1587         goto out;
1588     }
1589     object_property_set_int(OBJECT(dev), slot, PC_DIMM_SLOT_PROP, &local_err);
1590     if (local_err) {
1591         goto out;
1592     }
1593     trace_mhp_pc_dimm_assigned_slot(slot);
1594 
1595     if (!pcms->acpi_dev) {
1596         error_setg(&local_err,
1597                    "memory hotplug is not enabled: missing acpi device");
1598         goto out;
1599     }
1600 
1601     memory_region_add_subregion(&pcms->hotplug_memory,
1602                                 addr - pcms->hotplug_memory_base, mr);
1603     vmstate_register_ram(mr, dev);
1604 
1605     hhc = HOTPLUG_HANDLER_GET_CLASS(pcms->acpi_dev);
1606     hhc->plug(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &local_err);
1607 out:
1608     error_propagate(errp, local_err);
1609 }
1610 
1611 static void pc_cpu_plug(HotplugHandler *hotplug_dev,
1612                         DeviceState *dev, Error **errp)
1613 {
1614     HotplugHandlerClass *hhc;
1615     Error *local_err = NULL;
1616     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1617 
1618     if (!dev->hotplugged) {
1619         goto out;
1620     }
1621 
1622     if (!pcms->acpi_dev) {
1623         error_setg(&local_err,
1624                    "cpu hotplug is not enabled: missing acpi device");
1625         goto out;
1626     }
1627 
1628     hhc = HOTPLUG_HANDLER_GET_CLASS(pcms->acpi_dev);
1629     hhc->plug(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &local_err);
1630     if (local_err) {
1631         goto out;
1632     }
1633 
1634     /* increment the number of CPUs */
1635     rtc_set_memory(pcms->rtc, 0x5f, rtc_get_memory(pcms->rtc, 0x5f) + 1);
1636 out:
1637     error_propagate(errp, local_err);
1638 }
1639 
1640 static void pc_machine_device_plug_cb(HotplugHandler *hotplug_dev,
1641                                       DeviceState *dev, Error **errp)
1642 {
1643     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
1644         pc_dimm_plug(hotplug_dev, dev, errp);
1645     } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
1646         pc_cpu_plug(hotplug_dev, dev, errp);
1647     }
1648 }
1649 
1650 static HotplugHandler *pc_get_hotpug_handler(MachineState *machine,
1651                                              DeviceState *dev)
1652 {
1653     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(machine);
1654 
1655     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM) ||
1656         object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
1657         return HOTPLUG_HANDLER(machine);
1658     }
1659 
1660     return pcmc->get_hotplug_handler ?
1661         pcmc->get_hotplug_handler(machine, dev) : NULL;
1662 }
1663 
1664 static void
1665 pc_machine_get_hotplug_memory_region_size(Object *obj, Visitor *v, void *opaque,
1666                                           const char *name, Error **errp)
1667 {
1668     PCMachineState *pcms = PC_MACHINE(obj);
1669     int64_t value = memory_region_size(&pcms->hotplug_memory);
1670 
1671     visit_type_int(v, &value, name, errp);
1672 }
1673 
1674 static void pc_machine_get_max_ram_below_4g(Object *obj, Visitor *v,
1675                                          void *opaque, const char *name,
1676                                          Error **errp)
1677 {
1678     PCMachineState *pcms = PC_MACHINE(obj);
1679     uint64_t value = pcms->max_ram_below_4g;
1680 
1681     visit_type_size(v, &value, name, errp);
1682 }
1683 
1684 static void pc_machine_set_max_ram_below_4g(Object *obj, Visitor *v,
1685                                          void *opaque, const char *name,
1686                                          Error **errp)
1687 {
1688     PCMachineState *pcms = PC_MACHINE(obj);
1689     Error *error = NULL;
1690     uint64_t value;
1691 
1692     visit_type_size(v, &value, name, &error);
1693     if (error) {
1694         error_propagate(errp, error);
1695         return;
1696     }
1697     if (value > (1ULL << 32)) {
1698         error_set(&error, ERROR_CLASS_GENERIC_ERROR,
1699                   "Machine option 'max-ram-below-4g=%"PRIu64
1700                   "' expects size less than or equal to 4G", value);
1701         error_propagate(errp, error);
1702         return;
1703     }
1704 
1705     if (value < (1ULL << 20)) {
1706         error_report("Warning: small max_ram_below_4g(%"PRIu64
1707                      ") less than 1M.  BIOS may not work..",
1708                      value);
1709     }
1710 
1711     pcms->max_ram_below_4g = value;
1712 }
1713 
1714 static bool pc_machine_get_vmport(Object *obj, Error **errp)
1715 {
1716     PCMachineState *pcms = PC_MACHINE(obj);
1717 
1718     return pcms->vmport;
1719 }
1720 
1721 static void pc_machine_set_vmport(Object *obj, bool value, Error **errp)
1722 {
1723     PCMachineState *pcms = PC_MACHINE(obj);
1724 
1725     pcms->vmport = value;
1726 }
1727 
1728 static void pc_machine_initfn(Object *obj)
1729 {
1730     PCMachineState *pcms = PC_MACHINE(obj);
1731 
1732     object_property_add(obj, PC_MACHINE_MEMHP_REGION_SIZE, "int",
1733                         pc_machine_get_hotplug_memory_region_size,
1734                         NULL, NULL, NULL, NULL);
1735     pcms->max_ram_below_4g = 1ULL << 32; /* 4G */
1736     object_property_add(obj, PC_MACHINE_MAX_RAM_BELOW_4G, "size",
1737                         pc_machine_get_max_ram_below_4g,
1738                         pc_machine_set_max_ram_below_4g,
1739                         NULL, NULL, NULL);
1740     pcms->vmport = !xen_enabled();
1741     object_property_add_bool(obj, PC_MACHINE_VMPORT,
1742                              pc_machine_get_vmport,
1743                              pc_machine_set_vmport,
1744                              NULL);
1745 }
1746 
1747 static void pc_machine_class_init(ObjectClass *oc, void *data)
1748 {
1749     MachineClass *mc = MACHINE_CLASS(oc);
1750     PCMachineClass *pcmc = PC_MACHINE_CLASS(oc);
1751     HotplugHandlerClass *hc = HOTPLUG_HANDLER_CLASS(oc);
1752 
1753     pcmc->get_hotplug_handler = mc->get_hotplug_handler;
1754     mc->get_hotplug_handler = pc_get_hotpug_handler;
1755     hc->plug = pc_machine_device_plug_cb;
1756 }
1757 
1758 static const TypeInfo pc_machine_info = {
1759     .name = TYPE_PC_MACHINE,
1760     .parent = TYPE_MACHINE,
1761     .abstract = true,
1762     .instance_size = sizeof(PCMachineState),
1763     .instance_init = pc_machine_initfn,
1764     .class_size = sizeof(PCMachineClass),
1765     .class_init = pc_machine_class_init,
1766     .interfaces = (InterfaceInfo[]) {
1767          { TYPE_HOTPLUG_HANDLER },
1768          { }
1769     },
1770 };
1771 
1772 static void pc_machine_register_types(void)
1773 {
1774     type_register_static(&pc_machine_info);
1775 }
1776 
1777 type_init(pc_machine_register_types)
1778