xref: /openbmc/qemu/hw/i386/pc.c (revision ece5e5bfa1377546d5f94e1bb04298e48ce60c1c)
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_renew(struct e820_entry, e820_table, e820_entries + 1);
605     e820_table[e820_entries].address = cpu_to_le64(address);
606     e820_table[e820_entries].length = cpu_to_le64(length);
607     e820_table[e820_entries].type = cpu_to_le32(type);
608     e820_entries++;
609 
610     return e820_entries;
611 }
612 
613 int e820_get_num_entries(void)
614 {
615     return e820_entries;
616 }
617 
618 bool e820_get_entry(int idx, uint32_t type, uint64_t *address, uint64_t *length)
619 {
620     if (idx < e820_entries && e820_table[idx].type == cpu_to_le32(type)) {
621         *address = le64_to_cpu(e820_table[idx].address);
622         *length = le64_to_cpu(e820_table[idx].length);
623         return true;
624     }
625     return false;
626 }
627 
628 /* Calculates the limit to CPU APIC ID values
629  *
630  * This function returns the limit for the APIC ID value, so that all
631  * CPU APIC IDs are < pc_apic_id_limit().
632  *
633  * This is used for FW_CFG_MAX_CPUS. See comments on bochs_bios_init().
634  */
635 static unsigned int pc_apic_id_limit(unsigned int max_cpus)
636 {
637     return x86_cpu_apic_id_from_index(max_cpus - 1) + 1;
638 }
639 
640 static FWCfgState *bochs_bios_init(void)
641 {
642     FWCfgState *fw_cfg;
643     uint8_t *smbios_tables, *smbios_anchor;
644     size_t smbios_tables_len, smbios_anchor_len;
645     uint64_t *numa_fw_cfg;
646     int i, j;
647     unsigned int apic_id_limit = pc_apic_id_limit(max_cpus);
648 
649     fw_cfg = fw_cfg_init(BIOS_CFG_IOPORT, BIOS_CFG_IOPORT + 1, 0, 0);
650     /* FW_CFG_MAX_CPUS is a bit confusing/problematic on x86:
651      *
652      * SeaBIOS needs FW_CFG_MAX_CPUS for CPU hotplug, but the CPU hotplug
653      * QEMU<->SeaBIOS interface is not based on the "CPU index", but on the APIC
654      * ID of hotplugged CPUs[1]. This means that FW_CFG_MAX_CPUS is not the
655      * "maximum number of CPUs", but the "limit to the APIC ID values SeaBIOS
656      * may see".
657      *
658      * So, this means we must not use max_cpus, here, but the maximum possible
659      * APIC ID value, plus one.
660      *
661      * [1] The only kind of "CPU identifier" used between SeaBIOS and QEMU is
662      *     the APIC ID, not the "CPU index"
663      */
664     fw_cfg_add_i16(fw_cfg, FW_CFG_MAX_CPUS, (uint16_t)apic_id_limit);
665     fw_cfg_add_i32(fw_cfg, FW_CFG_ID, 1);
666     fw_cfg_add_i64(fw_cfg, FW_CFG_RAM_SIZE, (uint64_t)ram_size);
667     fw_cfg_add_bytes(fw_cfg, FW_CFG_ACPI_TABLES,
668                      acpi_tables, acpi_tables_len);
669     fw_cfg_add_i32(fw_cfg, FW_CFG_IRQ0_OVERRIDE, kvm_allows_irq0_override());
670 
671     smbios_tables = smbios_get_table_legacy(&smbios_tables_len);
672     if (smbios_tables) {
673         fw_cfg_add_bytes(fw_cfg, FW_CFG_SMBIOS_ENTRIES,
674                          smbios_tables, smbios_tables_len);
675     }
676 
677     smbios_get_tables(&smbios_tables, &smbios_tables_len,
678                       &smbios_anchor, &smbios_anchor_len);
679     if (smbios_anchor) {
680         fw_cfg_add_file(fw_cfg, "etc/smbios/smbios-tables",
681                         smbios_tables, smbios_tables_len);
682         fw_cfg_add_file(fw_cfg, "etc/smbios/smbios-anchor",
683                         smbios_anchor, smbios_anchor_len);
684     }
685 
686     fw_cfg_add_bytes(fw_cfg, FW_CFG_E820_TABLE,
687                      &e820_reserve, sizeof(e820_reserve));
688     fw_cfg_add_file(fw_cfg, "etc/e820", e820_table,
689                     sizeof(struct e820_entry) * e820_entries);
690 
691     fw_cfg_add_bytes(fw_cfg, FW_CFG_HPET, &hpet_cfg, sizeof(hpet_cfg));
692     /* allocate memory for the NUMA channel: one (64bit) word for the number
693      * of nodes, one word for each VCPU->node and one word for each node to
694      * hold the amount of memory.
695      */
696     numa_fw_cfg = g_new0(uint64_t, 1 + apic_id_limit + nb_numa_nodes);
697     numa_fw_cfg[0] = cpu_to_le64(nb_numa_nodes);
698     for (i = 0; i < max_cpus; i++) {
699         unsigned int apic_id = x86_cpu_apic_id_from_index(i);
700         assert(apic_id < apic_id_limit);
701         for (j = 0; j < nb_numa_nodes; j++) {
702             if (test_bit(i, numa_info[j].node_cpu)) {
703                 numa_fw_cfg[apic_id + 1] = cpu_to_le64(j);
704                 break;
705             }
706         }
707     }
708     for (i = 0; i < nb_numa_nodes; i++) {
709         numa_fw_cfg[apic_id_limit + 1 + i] = cpu_to_le64(numa_info[i].node_mem);
710     }
711     fw_cfg_add_bytes(fw_cfg, FW_CFG_NUMA, numa_fw_cfg,
712                      (1 + apic_id_limit + nb_numa_nodes) *
713                      sizeof(*numa_fw_cfg));
714 
715     return fw_cfg;
716 }
717 
718 static long get_file_size(FILE *f)
719 {
720     long where, size;
721 
722     /* XXX: on Unix systems, using fstat() probably makes more sense */
723 
724     where = ftell(f);
725     fseek(f, 0, SEEK_END);
726     size = ftell(f);
727     fseek(f, where, SEEK_SET);
728 
729     return size;
730 }
731 
732 static void load_linux(FWCfgState *fw_cfg,
733                        const char *kernel_filename,
734                        const char *initrd_filename,
735                        const char *kernel_cmdline,
736                        hwaddr max_ram_size)
737 {
738     uint16_t protocol;
739     int setup_size, kernel_size, initrd_size = 0, cmdline_size;
740     uint32_t initrd_max;
741     uint8_t header[8192], *setup, *kernel, *initrd_data;
742     hwaddr real_addr, prot_addr, cmdline_addr, initrd_addr = 0;
743     FILE *f;
744     char *vmode;
745 
746     /* Align to 16 bytes as a paranoia measure */
747     cmdline_size = (strlen(kernel_cmdline)+16) & ~15;
748 
749     /* load the kernel header */
750     f = fopen(kernel_filename, "rb");
751     if (!f || !(kernel_size = get_file_size(f)) ||
752         fread(header, 1, MIN(ARRAY_SIZE(header), kernel_size), f) !=
753         MIN(ARRAY_SIZE(header), kernel_size)) {
754         fprintf(stderr, "qemu: could not load kernel '%s': %s\n",
755                 kernel_filename, strerror(errno));
756         exit(1);
757     }
758 
759     /* kernel protocol version */
760 #if 0
761     fprintf(stderr, "header magic: %#x\n", ldl_p(header+0x202));
762 #endif
763     if (ldl_p(header+0x202) == 0x53726448) {
764         protocol = lduw_p(header+0x206);
765     } else {
766         /* This looks like a multiboot kernel. If it is, let's stop
767            treating it like a Linux kernel. */
768         if (load_multiboot(fw_cfg, f, kernel_filename, initrd_filename,
769                            kernel_cmdline, kernel_size, header)) {
770             return;
771         }
772         protocol = 0;
773     }
774 
775     if (protocol < 0x200 || !(header[0x211] & 0x01)) {
776         /* Low kernel */
777         real_addr    = 0x90000;
778         cmdline_addr = 0x9a000 - cmdline_size;
779         prot_addr    = 0x10000;
780     } else if (protocol < 0x202) {
781         /* High but ancient kernel */
782         real_addr    = 0x90000;
783         cmdline_addr = 0x9a000 - cmdline_size;
784         prot_addr    = 0x100000;
785     } else {
786         /* High and recent kernel */
787         real_addr    = 0x10000;
788         cmdline_addr = 0x20000;
789         prot_addr    = 0x100000;
790     }
791 
792 #if 0
793     fprintf(stderr,
794             "qemu: real_addr     = 0x" TARGET_FMT_plx "\n"
795             "qemu: cmdline_addr  = 0x" TARGET_FMT_plx "\n"
796             "qemu: prot_addr     = 0x" TARGET_FMT_plx "\n",
797             real_addr,
798             cmdline_addr,
799             prot_addr);
800 #endif
801 
802     /* highest address for loading the initrd */
803     if (protocol >= 0x203) {
804         initrd_max = ldl_p(header+0x22c);
805     } else {
806         initrd_max = 0x37ffffff;
807     }
808 
809     if (initrd_max >= max_ram_size - acpi_data_size) {
810         initrd_max = max_ram_size - acpi_data_size - 1;
811     }
812 
813     fw_cfg_add_i32(fw_cfg, FW_CFG_CMDLINE_ADDR, cmdline_addr);
814     fw_cfg_add_i32(fw_cfg, FW_CFG_CMDLINE_SIZE, strlen(kernel_cmdline)+1);
815     fw_cfg_add_string(fw_cfg, FW_CFG_CMDLINE_DATA, kernel_cmdline);
816 
817     if (protocol >= 0x202) {
818         stl_p(header+0x228, cmdline_addr);
819     } else {
820         stw_p(header+0x20, 0xA33F);
821         stw_p(header+0x22, cmdline_addr-real_addr);
822     }
823 
824     /* handle vga= parameter */
825     vmode = strstr(kernel_cmdline, "vga=");
826     if (vmode) {
827         unsigned int video_mode;
828         /* skip "vga=" */
829         vmode += 4;
830         if (!strncmp(vmode, "normal", 6)) {
831             video_mode = 0xffff;
832         } else if (!strncmp(vmode, "ext", 3)) {
833             video_mode = 0xfffe;
834         } else if (!strncmp(vmode, "ask", 3)) {
835             video_mode = 0xfffd;
836         } else {
837             video_mode = strtol(vmode, NULL, 0);
838         }
839         stw_p(header+0x1fa, video_mode);
840     }
841 
842     /* loader type */
843     /* High nybble = B reserved for QEMU; low nybble is revision number.
844        If this code is substantially changed, you may want to consider
845        incrementing the revision. */
846     if (protocol >= 0x200) {
847         header[0x210] = 0xB0;
848     }
849     /* heap */
850     if (protocol >= 0x201) {
851         header[0x211] |= 0x80;	/* CAN_USE_HEAP */
852         stw_p(header+0x224, cmdline_addr-real_addr-0x200);
853     }
854 
855     /* load initrd */
856     if (initrd_filename) {
857         if (protocol < 0x200) {
858             fprintf(stderr, "qemu: linux kernel too old to load a ram disk\n");
859             exit(1);
860         }
861 
862         initrd_size = get_image_size(initrd_filename);
863         if (initrd_size < 0) {
864             fprintf(stderr, "qemu: error reading initrd %s: %s\n",
865                     initrd_filename, strerror(errno));
866             exit(1);
867         }
868 
869         initrd_addr = (initrd_max-initrd_size) & ~4095;
870 
871         initrd_data = g_malloc(initrd_size);
872         load_image(initrd_filename, initrd_data);
873 
874         fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_ADDR, initrd_addr);
875         fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_SIZE, initrd_size);
876         fw_cfg_add_bytes(fw_cfg, FW_CFG_INITRD_DATA, initrd_data, initrd_size);
877 
878         stl_p(header+0x218, initrd_addr);
879         stl_p(header+0x21c, initrd_size);
880     }
881 
882     /* load kernel and setup */
883     setup_size = header[0x1f1];
884     if (setup_size == 0) {
885         setup_size = 4;
886     }
887     setup_size = (setup_size+1)*512;
888     kernel_size -= setup_size;
889 
890     setup  = g_malloc(setup_size);
891     kernel = g_malloc(kernel_size);
892     fseek(f, 0, SEEK_SET);
893     if (fread(setup, 1, setup_size, f) != setup_size) {
894         fprintf(stderr, "fread() failed\n");
895         exit(1);
896     }
897     if (fread(kernel, 1, kernel_size, f) != kernel_size) {
898         fprintf(stderr, "fread() failed\n");
899         exit(1);
900     }
901     fclose(f);
902     memcpy(setup, header, MIN(sizeof(header), setup_size));
903 
904     fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_ADDR, prot_addr);
905     fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_SIZE, kernel_size);
906     fw_cfg_add_bytes(fw_cfg, FW_CFG_KERNEL_DATA, kernel, kernel_size);
907 
908     fw_cfg_add_i32(fw_cfg, FW_CFG_SETUP_ADDR, real_addr);
909     fw_cfg_add_i32(fw_cfg, FW_CFG_SETUP_SIZE, setup_size);
910     fw_cfg_add_bytes(fw_cfg, FW_CFG_SETUP_DATA, setup, setup_size);
911 
912     option_rom[nb_option_roms].name = "linuxboot.bin";
913     option_rom[nb_option_roms].bootindex = 0;
914     nb_option_roms++;
915 }
916 
917 #define NE2000_NB_MAX 6
918 
919 static const int ne2000_io[NE2000_NB_MAX] = { 0x300, 0x320, 0x340, 0x360,
920                                               0x280, 0x380 };
921 static const int ne2000_irq[NE2000_NB_MAX] = { 9, 10, 11, 3, 4, 5 };
922 
923 void pc_init_ne2k_isa(ISABus *bus, NICInfo *nd)
924 {
925     static int nb_ne2k = 0;
926 
927     if (nb_ne2k == NE2000_NB_MAX)
928         return;
929     isa_ne2000_init(bus, ne2000_io[nb_ne2k],
930                     ne2000_irq[nb_ne2k], nd);
931     nb_ne2k++;
932 }
933 
934 DeviceState *cpu_get_current_apic(void)
935 {
936     if (current_cpu) {
937         X86CPU *cpu = X86_CPU(current_cpu);
938         return cpu->apic_state;
939     } else {
940         return NULL;
941     }
942 }
943 
944 void pc_acpi_smi_interrupt(void *opaque, int irq, int level)
945 {
946     X86CPU *cpu = opaque;
947 
948     if (level) {
949         cpu_interrupt(CPU(cpu), CPU_INTERRUPT_SMI);
950     }
951 }
952 
953 static X86CPU *pc_new_cpu(const char *cpu_model, int64_t apic_id,
954                           DeviceState *icc_bridge, Error **errp)
955 {
956     X86CPU *cpu;
957     Error *local_err = NULL;
958 
959     cpu = cpu_x86_create(cpu_model, icc_bridge, &local_err);
960     if (local_err != NULL) {
961         error_propagate(errp, local_err);
962         return NULL;
963     }
964 
965     object_property_set_int(OBJECT(cpu), apic_id, "apic-id", &local_err);
966     object_property_set_bool(OBJECT(cpu), true, "realized", &local_err);
967 
968     if (local_err) {
969         error_propagate(errp, local_err);
970         object_unref(OBJECT(cpu));
971         cpu = NULL;
972     }
973     return cpu;
974 }
975 
976 static const char *current_cpu_model;
977 
978 void pc_hot_add_cpu(const int64_t id, Error **errp)
979 {
980     DeviceState *icc_bridge;
981     int64_t apic_id = x86_cpu_apic_id_from_index(id);
982 
983     if (id < 0) {
984         error_setg(errp, "Invalid CPU id: %" PRIi64, id);
985         return;
986     }
987 
988     if (cpu_exists(apic_id)) {
989         error_setg(errp, "Unable to add CPU: %" PRIi64
990                    ", it already exists", id);
991         return;
992     }
993 
994     if (id >= max_cpus) {
995         error_setg(errp, "Unable to add CPU: %" PRIi64
996                    ", max allowed: %d", id, max_cpus - 1);
997         return;
998     }
999 
1000     if (apic_id >= ACPI_CPU_HOTPLUG_ID_LIMIT) {
1001         error_setg(errp, "Unable to add CPU: %" PRIi64
1002                    ", resulting APIC ID (%" PRIi64 ") is too large",
1003                    id, apic_id);
1004         return;
1005     }
1006 
1007     icc_bridge = DEVICE(object_resolve_path_type("icc-bridge",
1008                                                  TYPE_ICC_BRIDGE, NULL));
1009     pc_new_cpu(current_cpu_model, apic_id, icc_bridge, errp);
1010 }
1011 
1012 void pc_cpus_init(const char *cpu_model, DeviceState *icc_bridge)
1013 {
1014     int i;
1015     X86CPU *cpu = NULL;
1016     Error *error = NULL;
1017     unsigned long apic_id_limit;
1018 
1019     /* init CPUs */
1020     if (cpu_model == NULL) {
1021 #ifdef TARGET_X86_64
1022         cpu_model = "qemu64";
1023 #else
1024         cpu_model = "qemu32";
1025 #endif
1026     }
1027     current_cpu_model = cpu_model;
1028 
1029     apic_id_limit = pc_apic_id_limit(max_cpus);
1030     if (apic_id_limit > ACPI_CPU_HOTPLUG_ID_LIMIT) {
1031         error_report("max_cpus is too large. APIC ID of last CPU is %lu",
1032                      apic_id_limit - 1);
1033         exit(1);
1034     }
1035 
1036     for (i = 0; i < smp_cpus; i++) {
1037         cpu = pc_new_cpu(cpu_model, x86_cpu_apic_id_from_index(i),
1038                          icc_bridge, &error);
1039         if (error) {
1040             error_report("%s", error_get_pretty(error));
1041             error_free(error);
1042             exit(1);
1043         }
1044     }
1045 
1046     /* map APIC MMIO area if CPU has APIC */
1047     if (cpu && cpu->apic_state) {
1048         /* XXX: what if the base changes? */
1049         sysbus_mmio_map_overlap(SYS_BUS_DEVICE(icc_bridge), 0,
1050                                 APIC_DEFAULT_ADDRESS, 0x1000);
1051     }
1052 
1053     /* tell smbios about cpuid version and features */
1054     smbios_set_cpuid(cpu->env.cpuid_version, cpu->env.features[FEAT_1_EDX]);
1055 }
1056 
1057 /* pci-info ROM file. Little endian format */
1058 typedef struct PcRomPciInfo {
1059     uint64_t w32_min;
1060     uint64_t w32_max;
1061     uint64_t w64_min;
1062     uint64_t w64_max;
1063 } PcRomPciInfo;
1064 
1065 typedef struct PcGuestInfoState {
1066     PcGuestInfo info;
1067     Notifier machine_done;
1068 } PcGuestInfoState;
1069 
1070 static
1071 void pc_guest_info_machine_done(Notifier *notifier, void *data)
1072 {
1073     PcGuestInfoState *guest_info_state = container_of(notifier,
1074                                                       PcGuestInfoState,
1075                                                       machine_done);
1076     acpi_setup(&guest_info_state->info);
1077 }
1078 
1079 PcGuestInfo *pc_guest_info_init(ram_addr_t below_4g_mem_size,
1080                                 ram_addr_t above_4g_mem_size)
1081 {
1082     PcGuestInfoState *guest_info_state = g_malloc0(sizeof *guest_info_state);
1083     PcGuestInfo *guest_info = &guest_info_state->info;
1084     int i, j;
1085 
1086     guest_info->ram_size_below_4g = below_4g_mem_size;
1087     guest_info->ram_size = below_4g_mem_size + above_4g_mem_size;
1088     guest_info->apic_id_limit = pc_apic_id_limit(max_cpus);
1089     guest_info->apic_xrupt_override = kvm_allows_irq0_override();
1090     guest_info->numa_nodes = nb_numa_nodes;
1091     guest_info->node_mem = g_malloc0(guest_info->numa_nodes *
1092                                     sizeof *guest_info->node_mem);
1093     for (i = 0; i < nb_numa_nodes; i++) {
1094         guest_info->node_mem[i] = numa_info[i].node_mem;
1095     }
1096 
1097     guest_info->node_cpu = g_malloc0(guest_info->apic_id_limit *
1098                                      sizeof *guest_info->node_cpu);
1099 
1100     for (i = 0; i < max_cpus; i++) {
1101         unsigned int apic_id = x86_cpu_apic_id_from_index(i);
1102         assert(apic_id < guest_info->apic_id_limit);
1103         for (j = 0; j < nb_numa_nodes; j++) {
1104             if (test_bit(i, numa_info[j].node_cpu)) {
1105                 guest_info->node_cpu[apic_id] = j;
1106                 break;
1107             }
1108         }
1109     }
1110 
1111     guest_info_state->machine_done.notify = pc_guest_info_machine_done;
1112     qemu_add_machine_init_done_notifier(&guest_info_state->machine_done);
1113     return guest_info;
1114 }
1115 
1116 /* setup pci memory address space mapping into system address space */
1117 void pc_pci_as_mapping_init(Object *owner, MemoryRegion *system_memory,
1118                             MemoryRegion *pci_address_space)
1119 {
1120     /* Set to lower priority than RAM */
1121     memory_region_add_subregion_overlap(system_memory, 0x0,
1122                                         pci_address_space, -1);
1123 }
1124 
1125 void pc_acpi_init(const char *default_dsdt)
1126 {
1127     char *filename;
1128 
1129     if (acpi_tables != NULL) {
1130         /* manually set via -acpitable, leave it alone */
1131         return;
1132     }
1133 
1134     filename = qemu_find_file(QEMU_FILE_TYPE_BIOS, default_dsdt);
1135     if (filename == NULL) {
1136         fprintf(stderr, "WARNING: failed to find %s\n", default_dsdt);
1137     } else {
1138         char *arg;
1139         QemuOpts *opts;
1140         Error *err = NULL;
1141 
1142         arg = g_strdup_printf("file=%s", filename);
1143 
1144         /* creates a deep copy of "arg" */
1145         opts = qemu_opts_parse(qemu_find_opts("acpi"), arg, 0);
1146         g_assert(opts != NULL);
1147 
1148         acpi_table_add_builtin(opts, &err);
1149         if (err) {
1150             error_report("WARNING: failed to load %s: %s", filename,
1151                          error_get_pretty(err));
1152             error_free(err);
1153         }
1154         g_free(arg);
1155         g_free(filename);
1156     }
1157 }
1158 
1159 FWCfgState *xen_load_linux(const char *kernel_filename,
1160                            const char *kernel_cmdline,
1161                            const char *initrd_filename,
1162                            ram_addr_t below_4g_mem_size,
1163                            PcGuestInfo *guest_info)
1164 {
1165     int i;
1166     FWCfgState *fw_cfg;
1167 
1168     assert(kernel_filename != NULL);
1169 
1170     fw_cfg = fw_cfg_init(BIOS_CFG_IOPORT, BIOS_CFG_IOPORT + 1, 0, 0);
1171     rom_set_fw(fw_cfg);
1172 
1173     load_linux(fw_cfg, kernel_filename, initrd_filename,
1174                kernel_cmdline, below_4g_mem_size);
1175     for (i = 0; i < nb_option_roms; i++) {
1176         assert(!strcmp(option_rom[i].name, "linuxboot.bin") ||
1177                !strcmp(option_rom[i].name, "multiboot.bin"));
1178         rom_add_option(option_rom[i].name, option_rom[i].bootindex);
1179     }
1180     guest_info->fw_cfg = fw_cfg;
1181     return fw_cfg;
1182 }
1183 
1184 FWCfgState *pc_memory_init(MachineState *machine,
1185                            MemoryRegion *system_memory,
1186                            ram_addr_t below_4g_mem_size,
1187                            ram_addr_t above_4g_mem_size,
1188                            MemoryRegion *rom_memory,
1189                            MemoryRegion **ram_memory,
1190                            PcGuestInfo *guest_info)
1191 {
1192     int linux_boot, i;
1193     MemoryRegion *ram, *option_rom_mr;
1194     MemoryRegion *ram_below_4g, *ram_above_4g;
1195     FWCfgState *fw_cfg;
1196     PCMachineState *pcms = PC_MACHINE(machine);
1197 
1198     assert(machine->ram_size == below_4g_mem_size + above_4g_mem_size);
1199 
1200     linux_boot = (machine->kernel_filename != NULL);
1201 
1202     /* Allocate RAM.  We allocate it as a single memory region and use
1203      * aliases to address portions of it, mostly for backwards compatibility
1204      * with older qemus that used qemu_ram_alloc().
1205      */
1206     ram = g_malloc(sizeof(*ram));
1207     memory_region_allocate_system_memory(ram, NULL, "pc.ram",
1208                                          machine->ram_size);
1209     *ram_memory = ram;
1210     ram_below_4g = g_malloc(sizeof(*ram_below_4g));
1211     memory_region_init_alias(ram_below_4g, NULL, "ram-below-4g", ram,
1212                              0, below_4g_mem_size);
1213     memory_region_add_subregion(system_memory, 0, ram_below_4g);
1214     e820_add_entry(0, below_4g_mem_size, E820_RAM);
1215     if (above_4g_mem_size > 0) {
1216         ram_above_4g = g_malloc(sizeof(*ram_above_4g));
1217         memory_region_init_alias(ram_above_4g, NULL, "ram-above-4g", ram,
1218                                  below_4g_mem_size, above_4g_mem_size);
1219         memory_region_add_subregion(system_memory, 0x100000000ULL,
1220                                     ram_above_4g);
1221         e820_add_entry(0x100000000ULL, above_4g_mem_size, E820_RAM);
1222     }
1223 
1224     if (!guest_info->has_reserved_memory &&
1225         (machine->ram_slots ||
1226          (machine->maxram_size > machine->ram_size))) {
1227         MachineClass *mc = MACHINE_GET_CLASS(machine);
1228 
1229         error_report("\"-memory 'slots|maxmem'\" is not supported by: %s",
1230                      mc->name);
1231         exit(EXIT_FAILURE);
1232     }
1233 
1234     /* initialize hotplug memory address space */
1235     if (guest_info->has_reserved_memory &&
1236         (machine->ram_size < machine->maxram_size)) {
1237         ram_addr_t hotplug_mem_size =
1238             machine->maxram_size - machine->ram_size;
1239 
1240         if (machine->ram_slots > ACPI_MAX_RAM_SLOTS) {
1241             error_report("unsupported amount of memory slots: %"PRIu64,
1242                          machine->ram_slots);
1243             exit(EXIT_FAILURE);
1244         }
1245 
1246         pcms->hotplug_memory_base =
1247             ROUND_UP(0x100000000ULL + above_4g_mem_size, 1ULL << 30);
1248 
1249         if (pcms->enforce_aligned_dimm) {
1250             /* size hotplug region assuming 1G page max alignment per slot */
1251             hotplug_mem_size += (1ULL << 30) * machine->ram_slots;
1252         }
1253 
1254         if ((pcms->hotplug_memory_base + hotplug_mem_size) <
1255             hotplug_mem_size) {
1256             error_report("unsupported amount of maximum memory: " RAM_ADDR_FMT,
1257                          machine->maxram_size);
1258             exit(EXIT_FAILURE);
1259         }
1260 
1261         memory_region_init(&pcms->hotplug_memory, OBJECT(pcms),
1262                            "hotplug-memory", hotplug_mem_size);
1263         memory_region_add_subregion(system_memory, pcms->hotplug_memory_base,
1264                                     &pcms->hotplug_memory);
1265     }
1266 
1267     /* Initialize PC system firmware */
1268     pc_system_firmware_init(rom_memory, guest_info->isapc_ram_fw);
1269 
1270     option_rom_mr = g_malloc(sizeof(*option_rom_mr));
1271     memory_region_init_ram(option_rom_mr, NULL, "pc.rom", PC_ROM_SIZE,
1272                            &error_abort);
1273     vmstate_register_ram_global(option_rom_mr);
1274     memory_region_add_subregion_overlap(rom_memory,
1275                                         PC_ROM_MIN_VGA,
1276                                         option_rom_mr,
1277                                         1);
1278 
1279     fw_cfg = bochs_bios_init();
1280     rom_set_fw(fw_cfg);
1281 
1282     if (guest_info->has_reserved_memory && pcms->hotplug_memory_base) {
1283         uint64_t *val = g_malloc(sizeof(*val));
1284         *val = cpu_to_le64(ROUND_UP(pcms->hotplug_memory_base, 0x1ULL << 30));
1285         fw_cfg_add_file(fw_cfg, "etc/reserved-memory-end", val, sizeof(*val));
1286     }
1287 
1288     if (linux_boot) {
1289         load_linux(fw_cfg, machine->kernel_filename, machine->initrd_filename,
1290                    machine->kernel_cmdline, below_4g_mem_size);
1291     }
1292 
1293     for (i = 0; i < nb_option_roms; i++) {
1294         rom_add_option(option_rom[i].name, option_rom[i].bootindex);
1295     }
1296     guest_info->fw_cfg = fw_cfg;
1297     return fw_cfg;
1298 }
1299 
1300 qemu_irq *pc_allocate_cpu_irq(void)
1301 {
1302     return qemu_allocate_irqs(pic_irq_request, NULL, 1);
1303 }
1304 
1305 DeviceState *pc_vga_init(ISABus *isa_bus, PCIBus *pci_bus)
1306 {
1307     DeviceState *dev = NULL;
1308 
1309     if (pci_bus) {
1310         PCIDevice *pcidev = pci_vga_init(pci_bus);
1311         dev = pcidev ? &pcidev->qdev : NULL;
1312     } else if (isa_bus) {
1313         ISADevice *isadev = isa_vga_init(isa_bus);
1314         dev = isadev ? DEVICE(isadev) : NULL;
1315     }
1316     return dev;
1317 }
1318 
1319 static void cpu_request_exit(void *opaque, int irq, int level)
1320 {
1321     CPUState *cpu = current_cpu;
1322 
1323     if (cpu && level) {
1324         cpu_exit(cpu);
1325     }
1326 }
1327 
1328 static const MemoryRegionOps ioport80_io_ops = {
1329     .write = ioport80_write,
1330     .read = ioport80_read,
1331     .endianness = DEVICE_NATIVE_ENDIAN,
1332     .impl = {
1333         .min_access_size = 1,
1334         .max_access_size = 1,
1335     },
1336 };
1337 
1338 static const MemoryRegionOps ioportF0_io_ops = {
1339     .write = ioportF0_write,
1340     .read = ioportF0_read,
1341     .endianness = DEVICE_NATIVE_ENDIAN,
1342     .impl = {
1343         .min_access_size = 1,
1344         .max_access_size = 1,
1345     },
1346 };
1347 
1348 void pc_basic_device_init(ISABus *isa_bus, qemu_irq *gsi,
1349                           ISADevice **rtc_state,
1350                           ISADevice **floppy,
1351                           bool no_vmport,
1352                           uint32 hpet_irqs)
1353 {
1354     int i;
1355     DriveInfo *fd[MAX_FD];
1356     DeviceState *hpet = NULL;
1357     int pit_isa_irq = 0;
1358     qemu_irq pit_alt_irq = NULL;
1359     qemu_irq rtc_irq = NULL;
1360     qemu_irq *a20_line;
1361     ISADevice *i8042, *port92, *vmmouse, *pit = NULL;
1362     qemu_irq *cpu_exit_irq;
1363     MemoryRegion *ioport80_io = g_new(MemoryRegion, 1);
1364     MemoryRegion *ioportF0_io = g_new(MemoryRegion, 1);
1365 
1366     memory_region_init_io(ioport80_io, NULL, &ioport80_io_ops, NULL, "ioport80", 1);
1367     memory_region_add_subregion(isa_bus->address_space_io, 0x80, ioport80_io);
1368 
1369     memory_region_init_io(ioportF0_io, NULL, &ioportF0_io_ops, NULL, "ioportF0", 1);
1370     memory_region_add_subregion(isa_bus->address_space_io, 0xf0, ioportF0_io);
1371 
1372     /*
1373      * Check if an HPET shall be created.
1374      *
1375      * Without KVM_CAP_PIT_STATE2, we cannot switch off the in-kernel PIT
1376      * when the HPET wants to take over. Thus we have to disable the latter.
1377      */
1378     if (!no_hpet && (!kvm_irqchip_in_kernel() || kvm_has_pit_state2())) {
1379         /* In order to set property, here not using sysbus_try_create_simple */
1380         hpet = qdev_try_create(NULL, TYPE_HPET);
1381         if (hpet) {
1382             /* For pc-piix-*, hpet's intcap is always IRQ2. For pc-q35-1.7
1383              * and earlier, use IRQ2 for compat. Otherwise, use IRQ16~23,
1384              * IRQ8 and IRQ2.
1385              */
1386             uint8_t compat = object_property_get_int(OBJECT(hpet),
1387                     HPET_INTCAP, NULL);
1388             if (!compat) {
1389                 qdev_prop_set_uint32(hpet, HPET_INTCAP, hpet_irqs);
1390             }
1391             qdev_init_nofail(hpet);
1392             sysbus_mmio_map(SYS_BUS_DEVICE(hpet), 0, HPET_BASE);
1393 
1394             for (i = 0; i < GSI_NUM_PINS; i++) {
1395                 sysbus_connect_irq(SYS_BUS_DEVICE(hpet), i, gsi[i]);
1396             }
1397             pit_isa_irq = -1;
1398             pit_alt_irq = qdev_get_gpio_in(hpet, HPET_LEGACY_PIT_INT);
1399             rtc_irq = qdev_get_gpio_in(hpet, HPET_LEGACY_RTC_INT);
1400         }
1401     }
1402     *rtc_state = rtc_init(isa_bus, 2000, rtc_irq);
1403 
1404     qemu_register_boot_set(pc_boot_set, *rtc_state);
1405 
1406     if (!xen_enabled()) {
1407         if (kvm_irqchip_in_kernel()) {
1408             pit = kvm_pit_init(isa_bus, 0x40);
1409         } else {
1410             pit = pit_init(isa_bus, 0x40, pit_isa_irq, pit_alt_irq);
1411         }
1412         if (hpet) {
1413             /* connect PIT to output control line of the HPET */
1414             qdev_connect_gpio_out(hpet, 0, qdev_get_gpio_in(DEVICE(pit), 0));
1415         }
1416         pcspk_init(isa_bus, pit);
1417     }
1418 
1419     for(i = 0; i < MAX_SERIAL_PORTS; i++) {
1420         if (serial_hds[i]) {
1421             serial_isa_init(isa_bus, i, serial_hds[i]);
1422         }
1423     }
1424 
1425     for(i = 0; i < MAX_PARALLEL_PORTS; i++) {
1426         if (parallel_hds[i]) {
1427             parallel_init(isa_bus, i, parallel_hds[i]);
1428         }
1429     }
1430 
1431     a20_line = qemu_allocate_irqs(handle_a20_line_change, first_cpu, 2);
1432     i8042 = isa_create_simple(isa_bus, "i8042");
1433     i8042_setup_a20_line(i8042, &a20_line[0]);
1434     if (!no_vmport) {
1435         vmport_init(isa_bus);
1436         vmmouse = isa_try_create(isa_bus, "vmmouse");
1437     } else {
1438         vmmouse = NULL;
1439     }
1440     if (vmmouse) {
1441         DeviceState *dev = DEVICE(vmmouse);
1442         qdev_prop_set_ptr(dev, "ps2_mouse", i8042);
1443         qdev_init_nofail(dev);
1444     }
1445     port92 = isa_create_simple(isa_bus, "port92");
1446     port92_init(port92, &a20_line[1]);
1447 
1448     cpu_exit_irq = qemu_allocate_irqs(cpu_request_exit, NULL, 1);
1449     DMA_init(0, cpu_exit_irq);
1450 
1451     for(i = 0; i < MAX_FD; i++) {
1452         fd[i] = drive_get(IF_FLOPPY, 0, i);
1453     }
1454     *floppy = fdctrl_init_isa(isa_bus, fd);
1455 }
1456 
1457 void pc_nic_init(ISABus *isa_bus, PCIBus *pci_bus)
1458 {
1459     int i;
1460 
1461     for (i = 0; i < nb_nics; i++) {
1462         NICInfo *nd = &nd_table[i];
1463 
1464         if (!pci_bus || (nd->model && strcmp(nd->model, "ne2k_isa") == 0)) {
1465             pc_init_ne2k_isa(isa_bus, nd);
1466         } else {
1467             pci_nic_init_nofail(nd, pci_bus, "e1000", NULL);
1468         }
1469     }
1470 }
1471 
1472 void pc_pci_device_init(PCIBus *pci_bus)
1473 {
1474     int max_bus;
1475     int bus;
1476 
1477     max_bus = drive_get_max_bus(IF_SCSI);
1478     for (bus = 0; bus <= max_bus; bus++) {
1479         pci_create_simple(pci_bus, -1, "lsi53c895a");
1480     }
1481 }
1482 
1483 void ioapic_init_gsi(GSIState *gsi_state, const char *parent_name)
1484 {
1485     DeviceState *dev;
1486     SysBusDevice *d;
1487     unsigned int i;
1488 
1489     if (kvm_irqchip_in_kernel()) {
1490         dev = qdev_create(NULL, "kvm-ioapic");
1491     } else {
1492         dev = qdev_create(NULL, "ioapic");
1493     }
1494     if (parent_name) {
1495         object_property_add_child(object_resolve_path(parent_name, NULL),
1496                                   "ioapic", OBJECT(dev), NULL);
1497     }
1498     qdev_init_nofail(dev);
1499     d = SYS_BUS_DEVICE(dev);
1500     sysbus_mmio_map(d, 0, IO_APIC_DEFAULT_ADDRESS);
1501 
1502     for (i = 0; i < IOAPIC_NUM_PINS; i++) {
1503         gsi_state->ioapic_irq[i] = qdev_get_gpio_in(dev, i);
1504     }
1505 }
1506 
1507 static void pc_generic_machine_class_init(ObjectClass *oc, void *data)
1508 {
1509     MachineClass *mc = MACHINE_CLASS(oc);
1510     QEMUMachine *qm = data;
1511 
1512     mc->family = qm->family;
1513     mc->name = qm->name;
1514     mc->alias = qm->alias;
1515     mc->desc = qm->desc;
1516     mc->init = qm->init;
1517     mc->reset = qm->reset;
1518     mc->hot_add_cpu = qm->hot_add_cpu;
1519     mc->kvm_type = qm->kvm_type;
1520     mc->block_default_type = qm->block_default_type;
1521     mc->units_per_default_bus = qm->units_per_default_bus;
1522     mc->max_cpus = qm->max_cpus;
1523     mc->no_serial = qm->no_serial;
1524     mc->no_parallel = qm->no_parallel;
1525     mc->use_virtcon = qm->use_virtcon;
1526     mc->use_sclp = qm->use_sclp;
1527     mc->no_floppy = qm->no_floppy;
1528     mc->no_cdrom = qm->no_cdrom;
1529     mc->no_sdcard = qm->no_sdcard;
1530     mc->is_default = qm->is_default;
1531     mc->default_machine_opts = qm->default_machine_opts;
1532     mc->default_boot_order = qm->default_boot_order;
1533     mc->default_display = qm->default_display;
1534     mc->compat_props = qm->compat_props;
1535     mc->hw_version = qm->hw_version;
1536 }
1537 
1538 void qemu_register_pc_machine(QEMUMachine *m)
1539 {
1540     char *name = g_strconcat(m->name, TYPE_MACHINE_SUFFIX, NULL);
1541     TypeInfo ti = {
1542         .name       = name,
1543         .parent     = TYPE_PC_MACHINE,
1544         .class_init = pc_generic_machine_class_init,
1545         .class_data = (void *)m,
1546     };
1547 
1548     type_register(&ti);
1549     g_free(name);
1550 }
1551 
1552 static int pc_dimm_count(Object *obj, void *opaque)
1553 {
1554     int *count = opaque;
1555 
1556     if (object_dynamic_cast(obj, TYPE_PC_DIMM)) {
1557         (*count)++;
1558     }
1559 
1560     object_child_foreach(obj, pc_dimm_count, opaque);
1561     return 0;
1562 }
1563 
1564 static int pc_existing_dimms_capacity(Object *obj, void *opaque)
1565 {
1566     Error *local_err = NULL;
1567     uint64_t *size = opaque;
1568 
1569     if (object_dynamic_cast(obj, TYPE_PC_DIMM)) {
1570         (*size) += object_property_get_int(obj, PC_DIMM_SIZE_PROP, &local_err);
1571 
1572         if (local_err) {
1573             qerror_report_err(local_err);
1574             error_free(local_err);
1575             return 1;
1576         }
1577     }
1578 
1579     object_child_foreach(obj, pc_dimm_count, opaque);
1580     return 0;
1581 }
1582 
1583 static void pc_dimm_plug(HotplugHandler *hotplug_dev,
1584                          DeviceState *dev, Error **errp)
1585 {
1586     int slot;
1587     HotplugHandlerClass *hhc;
1588     Error *local_err = NULL;
1589     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1590     MachineState *machine = MACHINE(hotplug_dev);
1591     PCDIMMDevice *dimm = PC_DIMM(dev);
1592     PCDIMMDeviceClass *ddc = PC_DIMM_GET_CLASS(dimm);
1593     MemoryRegion *mr = ddc->get_memory_region(dimm);
1594     uint64_t existing_dimms_capacity = 0;
1595     uint64_t align = TARGET_PAGE_SIZE;
1596     uint64_t addr;
1597 
1598     addr = object_property_get_int(OBJECT(dimm), PC_DIMM_ADDR_PROP, &local_err);
1599     if (local_err) {
1600         goto out;
1601     }
1602 
1603     if (memory_region_get_alignment(mr) && pcms->enforce_aligned_dimm) {
1604         align = memory_region_get_alignment(mr);
1605     }
1606 
1607     addr = pc_dimm_get_free_addr(pcms->hotplug_memory_base,
1608                                  memory_region_size(&pcms->hotplug_memory),
1609                                  !addr ? NULL : &addr, align,
1610                                  memory_region_size(mr), &local_err);
1611     if (local_err) {
1612         goto out;
1613     }
1614 
1615     if (pc_existing_dimms_capacity(OBJECT(machine), &existing_dimms_capacity)) {
1616         error_setg(&local_err, "failed to get total size of existing DIMMs");
1617         goto out;
1618     }
1619 
1620     if (existing_dimms_capacity + memory_region_size(mr) >
1621         machine->maxram_size - machine->ram_size) {
1622         error_setg(&local_err, "not enough space, currently 0x%" PRIx64
1623                    " in use of total 0x" RAM_ADDR_FMT,
1624                    existing_dimms_capacity, machine->maxram_size);
1625         goto out;
1626     }
1627 
1628     object_property_set_int(OBJECT(dev), addr, PC_DIMM_ADDR_PROP, &local_err);
1629     if (local_err) {
1630         goto out;
1631     }
1632     trace_mhp_pc_dimm_assigned_address(addr);
1633 
1634     slot = object_property_get_int(OBJECT(dev), PC_DIMM_SLOT_PROP, &local_err);
1635     if (local_err) {
1636         goto out;
1637     }
1638 
1639     slot = pc_dimm_get_free_slot(slot == PC_DIMM_UNASSIGNED_SLOT ? NULL : &slot,
1640                                  machine->ram_slots, &local_err);
1641     if (local_err) {
1642         goto out;
1643     }
1644     object_property_set_int(OBJECT(dev), slot, PC_DIMM_SLOT_PROP, &local_err);
1645     if (local_err) {
1646         goto out;
1647     }
1648     trace_mhp_pc_dimm_assigned_slot(slot);
1649 
1650     if (!pcms->acpi_dev) {
1651         error_setg(&local_err,
1652                    "memory hotplug is not enabled: missing acpi device");
1653         goto out;
1654     }
1655 
1656     if (kvm_enabled() && !kvm_has_free_slot(machine)) {
1657         error_setg(&local_err, "hypervisor has no free memory slots left");
1658         goto out;
1659     }
1660 
1661     memory_region_add_subregion(&pcms->hotplug_memory,
1662                                 addr - pcms->hotplug_memory_base, mr);
1663     vmstate_register_ram(mr, dev);
1664 
1665     hhc = HOTPLUG_HANDLER_GET_CLASS(pcms->acpi_dev);
1666     hhc->plug(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &local_err);
1667 out:
1668     error_propagate(errp, local_err);
1669 }
1670 
1671 static void pc_cpu_plug(HotplugHandler *hotplug_dev,
1672                         DeviceState *dev, Error **errp)
1673 {
1674     HotplugHandlerClass *hhc;
1675     Error *local_err = NULL;
1676     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1677 
1678     if (!dev->hotplugged) {
1679         goto out;
1680     }
1681 
1682     if (!pcms->acpi_dev) {
1683         error_setg(&local_err,
1684                    "cpu hotplug is not enabled: missing acpi device");
1685         goto out;
1686     }
1687 
1688     hhc = HOTPLUG_HANDLER_GET_CLASS(pcms->acpi_dev);
1689     hhc->plug(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &local_err);
1690     if (local_err) {
1691         goto out;
1692     }
1693 
1694     /* increment the number of CPUs */
1695     rtc_set_memory(pcms->rtc, 0x5f, rtc_get_memory(pcms->rtc, 0x5f) + 1);
1696 out:
1697     error_propagate(errp, local_err);
1698 }
1699 
1700 static void pc_machine_device_plug_cb(HotplugHandler *hotplug_dev,
1701                                       DeviceState *dev, Error **errp)
1702 {
1703     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
1704         pc_dimm_plug(hotplug_dev, dev, errp);
1705     } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
1706         pc_cpu_plug(hotplug_dev, dev, errp);
1707     }
1708 }
1709 
1710 static HotplugHandler *pc_get_hotpug_handler(MachineState *machine,
1711                                              DeviceState *dev)
1712 {
1713     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(machine);
1714 
1715     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM) ||
1716         object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
1717         return HOTPLUG_HANDLER(machine);
1718     }
1719 
1720     return pcmc->get_hotplug_handler ?
1721         pcmc->get_hotplug_handler(machine, dev) : NULL;
1722 }
1723 
1724 static void
1725 pc_machine_get_hotplug_memory_region_size(Object *obj, Visitor *v, void *opaque,
1726                                           const char *name, Error **errp)
1727 {
1728     PCMachineState *pcms = PC_MACHINE(obj);
1729     int64_t value = memory_region_size(&pcms->hotplug_memory);
1730 
1731     visit_type_int(v, &value, name, errp);
1732 }
1733 
1734 static void pc_machine_get_max_ram_below_4g(Object *obj, Visitor *v,
1735                                          void *opaque, const char *name,
1736                                          Error **errp)
1737 {
1738     PCMachineState *pcms = PC_MACHINE(obj);
1739     uint64_t value = pcms->max_ram_below_4g;
1740 
1741     visit_type_size(v, &value, name, errp);
1742 }
1743 
1744 static void pc_machine_set_max_ram_below_4g(Object *obj, Visitor *v,
1745                                          void *opaque, const char *name,
1746                                          Error **errp)
1747 {
1748     PCMachineState *pcms = PC_MACHINE(obj);
1749     Error *error = NULL;
1750     uint64_t value;
1751 
1752     visit_type_size(v, &value, name, &error);
1753     if (error) {
1754         error_propagate(errp, error);
1755         return;
1756     }
1757     if (value > (1ULL << 32)) {
1758         error_set(&error, ERROR_CLASS_GENERIC_ERROR,
1759                   "Machine option 'max-ram-below-4g=%"PRIu64
1760                   "' expects size less than or equal to 4G", value);
1761         error_propagate(errp, error);
1762         return;
1763     }
1764 
1765     if (value < (1ULL << 20)) {
1766         error_report("Warning: small max_ram_below_4g(%"PRIu64
1767                      ") less than 1M.  BIOS may not work..",
1768                      value);
1769     }
1770 
1771     pcms->max_ram_below_4g = value;
1772 }
1773 
1774 static bool pc_machine_get_vmport(Object *obj, Error **errp)
1775 {
1776     PCMachineState *pcms = PC_MACHINE(obj);
1777 
1778     return pcms->vmport;
1779 }
1780 
1781 static void pc_machine_set_vmport(Object *obj, bool value, Error **errp)
1782 {
1783     PCMachineState *pcms = PC_MACHINE(obj);
1784 
1785     pcms->vmport = value;
1786 }
1787 
1788 static bool pc_machine_get_aligned_dimm(Object *obj, Error **errp)
1789 {
1790     PCMachineState *pcms = PC_MACHINE(obj);
1791 
1792     return pcms->enforce_aligned_dimm;
1793 }
1794 
1795 static void pc_machine_initfn(Object *obj)
1796 {
1797     PCMachineState *pcms = PC_MACHINE(obj);
1798 
1799     object_property_add(obj, PC_MACHINE_MEMHP_REGION_SIZE, "int",
1800                         pc_machine_get_hotplug_memory_region_size,
1801                         NULL, NULL, NULL, NULL);
1802     pcms->max_ram_below_4g = 1ULL << 32; /* 4G */
1803     object_property_add(obj, PC_MACHINE_MAX_RAM_BELOW_4G, "size",
1804                         pc_machine_get_max_ram_below_4g,
1805                         pc_machine_set_max_ram_below_4g,
1806                         NULL, NULL, NULL);
1807 
1808     pcms->vmport = !xen_enabled();
1809     object_property_add_bool(obj, PC_MACHINE_VMPORT,
1810                              pc_machine_get_vmport,
1811                              pc_machine_set_vmport,
1812                              NULL);
1813 
1814     pcms->enforce_aligned_dimm = true;
1815     object_property_add_bool(obj, PC_MACHINE_ENFORCE_ALIGNED_DIMM,
1816                              pc_machine_get_aligned_dimm,
1817                              NULL, NULL);
1818 }
1819 
1820 static void pc_machine_class_init(ObjectClass *oc, void *data)
1821 {
1822     MachineClass *mc = MACHINE_CLASS(oc);
1823     PCMachineClass *pcmc = PC_MACHINE_CLASS(oc);
1824     HotplugHandlerClass *hc = HOTPLUG_HANDLER_CLASS(oc);
1825 
1826     pcmc->get_hotplug_handler = mc->get_hotplug_handler;
1827     mc->get_hotplug_handler = pc_get_hotpug_handler;
1828     hc->plug = pc_machine_device_plug_cb;
1829 }
1830 
1831 static const TypeInfo pc_machine_info = {
1832     .name = TYPE_PC_MACHINE,
1833     .parent = TYPE_MACHINE,
1834     .abstract = true,
1835     .instance_size = sizeof(PCMachineState),
1836     .instance_init = pc_machine_initfn,
1837     .class_size = sizeof(PCMachineClass),
1838     .class_init = pc_machine_class_init,
1839     .interfaces = (InterfaceInfo[]) {
1840          { TYPE_HOTPLUG_HANDLER },
1841          { }
1842     },
1843 };
1844 
1845 static void pc_machine_register_types(void)
1846 {
1847     type_register_static(&pc_machine_info);
1848 }
1849 
1850 type_init(pc_machine_register_types)
1851