xref: /qemu/hw/i386/pc.c (revision 53a89e262bd3e97b2da3afec0a60e5466770ae8c)
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/blockdev.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/cpu/icc_bus.h"
57 
58 /* debug PC/ISA interrupts */
59 //#define DEBUG_IRQ
60 
61 #ifdef DEBUG_IRQ
62 #define DPRINTF(fmt, ...)                                       \
63     do { printf("CPUIRQ: " fmt , ## __VA_ARGS__); } while (0)
64 #else
65 #define DPRINTF(fmt, ...)
66 #endif
67 
68 /* Leave a chunk of memory at the top of RAM for the BIOS ACPI tables.  */
69 #define ACPI_DATA_SIZE       0x10000
70 #define BIOS_CFG_IOPORT 0x510
71 #define FW_CFG_ACPI_TABLES (FW_CFG_ARCH_LOCAL + 0)
72 #define FW_CFG_SMBIOS_ENTRIES (FW_CFG_ARCH_LOCAL + 1)
73 #define FW_CFG_IRQ0_OVERRIDE (FW_CFG_ARCH_LOCAL + 2)
74 #define FW_CFG_E820_TABLE (FW_CFG_ARCH_LOCAL + 3)
75 #define FW_CFG_HPET (FW_CFG_ARCH_LOCAL + 4)
76 
77 #define IO_APIC_DEFAULT_ADDRESS 0xfec00000
78 
79 #define E820_NR_ENTRIES		16
80 
81 struct e820_entry {
82     uint64_t address;
83     uint64_t length;
84     uint32_t type;
85 } QEMU_PACKED __attribute((__aligned__(4)));
86 
87 struct e820_table {
88     uint32_t count;
89     struct e820_entry entry[E820_NR_ENTRIES];
90 } QEMU_PACKED __attribute((__aligned__(4)));
91 
92 static struct e820_table e820_table;
93 struct hpet_fw_config hpet_cfg = {.count = UINT8_MAX};
94 
95 void gsi_handler(void *opaque, int n, int level)
96 {
97     GSIState *s = opaque;
98 
99     DPRINTF("pc: %s GSI %d\n", level ? "raising" : "lowering", n);
100     if (n < ISA_NUM_IRQS) {
101         qemu_set_irq(s->i8259_irq[n], level);
102     }
103     qemu_set_irq(s->ioapic_irq[n], level);
104 }
105 
106 static void ioport80_write(void *opaque, hwaddr addr, uint64_t data,
107                            unsigned size)
108 {
109 }
110 
111 static uint64_t ioport80_read(void *opaque, hwaddr addr, unsigned size)
112 {
113     return 0xffffffffffffffffULL;
114 }
115 
116 /* MSDOS compatibility mode FPU exception support */
117 static qemu_irq ferr_irq;
118 
119 void pc_register_ferr_irq(qemu_irq irq)
120 {
121     ferr_irq = irq;
122 }
123 
124 /* XXX: add IGNNE support */
125 void cpu_set_ferr(CPUX86State *s)
126 {
127     qemu_irq_raise(ferr_irq);
128 }
129 
130 static void ioportF0_write(void *opaque, hwaddr addr, uint64_t data,
131                            unsigned size)
132 {
133     qemu_irq_lower(ferr_irq);
134 }
135 
136 static uint64_t ioportF0_read(void *opaque, hwaddr addr, unsigned size)
137 {
138     return 0xffffffffffffffffULL;
139 }
140 
141 /* TSC handling */
142 uint64_t cpu_get_tsc(CPUX86State *env)
143 {
144     return cpu_get_ticks();
145 }
146 
147 /* SMM support */
148 
149 static cpu_set_smm_t smm_set;
150 static void *smm_arg;
151 
152 void cpu_smm_register(cpu_set_smm_t callback, void *arg)
153 {
154     assert(smm_set == NULL);
155     assert(smm_arg == NULL);
156     smm_set = callback;
157     smm_arg = arg;
158 }
159 
160 void cpu_smm_update(CPUX86State *env)
161 {
162     if (smm_set && smm_arg && env == first_cpu)
163         smm_set(!!(env->hflags & HF_SMM_MASK), smm_arg);
164 }
165 
166 
167 /* IRQ handling */
168 int cpu_get_pic_interrupt(CPUX86State *env)
169 {
170     int intno;
171 
172     intno = apic_get_interrupt(env->apic_state);
173     if (intno >= 0) {
174         return intno;
175     }
176     /* read the irq from the PIC */
177     if (!apic_accept_pic_intr(env->apic_state)) {
178         return -1;
179     }
180 
181     intno = pic_read_irq(isa_pic);
182     return intno;
183 }
184 
185 static void pic_irq_request(void *opaque, int irq, int level)
186 {
187     CPUX86State *env = first_cpu;
188 
189     DPRINTF("pic_irqs: %s irq %d\n", level? "raise" : "lower", irq);
190     if (env->apic_state) {
191         while (env) {
192             if (apic_accept_pic_intr(env->apic_state)) {
193                 apic_deliver_pic_intr(env->apic_state, level);
194             }
195             env = env->next_cpu;
196         }
197     } else {
198         CPUState *cs = CPU(x86_env_get_cpu(env));
199         if (level) {
200             cpu_interrupt(cs, CPU_INTERRUPT_HARD);
201         } else {
202             cpu_reset_interrupt(cs, CPU_INTERRUPT_HARD);
203         }
204     }
205 }
206 
207 /* PC cmos mappings */
208 
209 #define REG_EQUIPMENT_BYTE          0x14
210 
211 static int cmos_get_fd_drive_type(FDriveType fd0)
212 {
213     int val;
214 
215     switch (fd0) {
216     case FDRIVE_DRV_144:
217         /* 1.44 Mb 3"5 drive */
218         val = 4;
219         break;
220     case FDRIVE_DRV_288:
221         /* 2.88 Mb 3"5 drive */
222         val = 5;
223         break;
224     case FDRIVE_DRV_120:
225         /* 1.2 Mb 5"5 drive */
226         val = 2;
227         break;
228     case FDRIVE_DRV_NONE:
229     default:
230         val = 0;
231         break;
232     }
233     return val;
234 }
235 
236 static void cmos_init_hd(ISADevice *s, int type_ofs, int info_ofs,
237                          int16_t cylinders, int8_t heads, int8_t sectors)
238 {
239     rtc_set_memory(s, type_ofs, 47);
240     rtc_set_memory(s, info_ofs, cylinders);
241     rtc_set_memory(s, info_ofs + 1, cylinders >> 8);
242     rtc_set_memory(s, info_ofs + 2, heads);
243     rtc_set_memory(s, info_ofs + 3, 0xff);
244     rtc_set_memory(s, info_ofs + 4, 0xff);
245     rtc_set_memory(s, info_ofs + 5, 0xc0 | ((heads > 8) << 3));
246     rtc_set_memory(s, info_ofs + 6, cylinders);
247     rtc_set_memory(s, info_ofs + 7, cylinders >> 8);
248     rtc_set_memory(s, info_ofs + 8, sectors);
249 }
250 
251 /* convert boot_device letter to something recognizable by the bios */
252 static int boot_device2nibble(char boot_device)
253 {
254     switch(boot_device) {
255     case 'a':
256     case 'b':
257         return 0x01; /* floppy boot */
258     case 'c':
259         return 0x02; /* hard drive boot */
260     case 'd':
261         return 0x03; /* CD-ROM boot */
262     case 'n':
263         return 0x04; /* Network boot */
264     }
265     return 0;
266 }
267 
268 static int set_boot_dev(ISADevice *s, const char *boot_device, int fd_bootchk)
269 {
270 #define PC_MAX_BOOT_DEVICES 3
271     int nbds, bds[3] = { 0, };
272     int i;
273 
274     nbds = strlen(boot_device);
275     if (nbds > PC_MAX_BOOT_DEVICES) {
276         error_report("Too many boot devices for PC");
277         return(1);
278     }
279     for (i = 0; i < nbds; i++) {
280         bds[i] = boot_device2nibble(boot_device[i]);
281         if (bds[i] == 0) {
282             error_report("Invalid boot device for PC: '%c'",
283                          boot_device[i]);
284             return(1);
285         }
286     }
287     rtc_set_memory(s, 0x3d, (bds[1] << 4) | bds[0]);
288     rtc_set_memory(s, 0x38, (bds[2] << 4) | (fd_bootchk ? 0x0 : 0x1));
289     return(0);
290 }
291 
292 static int pc_boot_set(void *opaque, const char *boot_device)
293 {
294     return set_boot_dev(opaque, boot_device, 0);
295 }
296 
297 typedef struct pc_cmos_init_late_arg {
298     ISADevice *rtc_state;
299     BusState *idebus[2];
300 } pc_cmos_init_late_arg;
301 
302 static void pc_cmos_init_late(void *opaque)
303 {
304     pc_cmos_init_late_arg *arg = opaque;
305     ISADevice *s = arg->rtc_state;
306     int16_t cylinders;
307     int8_t heads, sectors;
308     int val;
309     int i, trans;
310 
311     val = 0;
312     if (ide_get_geometry(arg->idebus[0], 0,
313                          &cylinders, &heads, &sectors) >= 0) {
314         cmos_init_hd(s, 0x19, 0x1b, cylinders, heads, sectors);
315         val |= 0xf0;
316     }
317     if (ide_get_geometry(arg->idebus[0], 1,
318                          &cylinders, &heads, &sectors) >= 0) {
319         cmos_init_hd(s, 0x1a, 0x24, cylinders, heads, sectors);
320         val |= 0x0f;
321     }
322     rtc_set_memory(s, 0x12, val);
323 
324     val = 0;
325     for (i = 0; i < 4; i++) {
326         /* NOTE: ide_get_geometry() returns the physical
327            geometry.  It is always such that: 1 <= sects <= 63, 1
328            <= heads <= 16, 1 <= cylinders <= 16383. The BIOS
329            geometry can be different if a translation is done. */
330         if (ide_get_geometry(arg->idebus[i / 2], i % 2,
331                              &cylinders, &heads, &sectors) >= 0) {
332             trans = ide_get_bios_chs_trans(arg->idebus[i / 2], i % 2) - 1;
333             assert((trans & ~3) == 0);
334             val |= trans << (i * 2);
335         }
336     }
337     rtc_set_memory(s, 0x39, val);
338 
339     qemu_unregister_reset(pc_cmos_init_late, opaque);
340 }
341 
342 typedef struct RTCCPUHotplugArg {
343     Notifier cpu_added_notifier;
344     ISADevice *rtc_state;
345 } RTCCPUHotplugArg;
346 
347 static void rtc_notify_cpu_added(Notifier *notifier, void *data)
348 {
349     RTCCPUHotplugArg *arg = container_of(notifier, RTCCPUHotplugArg,
350                                          cpu_added_notifier);
351     ISADevice *s = arg->rtc_state;
352 
353     /* increment the number of CPUs */
354     rtc_set_memory(s, 0x5f, rtc_get_memory(s, 0x5f) + 1);
355 }
356 
357 void pc_cmos_init(ram_addr_t ram_size, ram_addr_t above_4g_mem_size,
358                   const char *boot_device,
359                   ISADevice *floppy, BusState *idebus0, BusState *idebus1,
360                   ISADevice *s)
361 {
362     int val, nb, i;
363     FDriveType fd_type[2] = { FDRIVE_DRV_NONE, FDRIVE_DRV_NONE };
364     static pc_cmos_init_late_arg arg;
365     static RTCCPUHotplugArg cpu_hotplug_cb;
366 
367     /* various important CMOS locations needed by PC/Bochs bios */
368 
369     /* memory size */
370     /* base memory (first MiB) */
371     val = MIN(ram_size / 1024, 640);
372     rtc_set_memory(s, 0x15, val);
373     rtc_set_memory(s, 0x16, val >> 8);
374     /* extended memory (next 64MiB) */
375     if (ram_size > 1024 * 1024) {
376         val = (ram_size - 1024 * 1024) / 1024;
377     } else {
378         val = 0;
379     }
380     if (val > 65535)
381         val = 65535;
382     rtc_set_memory(s, 0x17, val);
383     rtc_set_memory(s, 0x18, val >> 8);
384     rtc_set_memory(s, 0x30, val);
385     rtc_set_memory(s, 0x31, val >> 8);
386     /* memory between 16MiB and 4GiB */
387     if (ram_size > 16 * 1024 * 1024) {
388         val = (ram_size - 16 * 1024 * 1024) / 65536;
389     } else {
390         val = 0;
391     }
392     if (val > 65535)
393         val = 65535;
394     rtc_set_memory(s, 0x34, val);
395     rtc_set_memory(s, 0x35, val >> 8);
396     /* memory above 4GiB */
397     val = above_4g_mem_size / 65536;
398     rtc_set_memory(s, 0x5b, val);
399     rtc_set_memory(s, 0x5c, val >> 8);
400     rtc_set_memory(s, 0x5d, val >> 16);
401 
402     /* set the number of CPU */
403     rtc_set_memory(s, 0x5f, smp_cpus - 1);
404     /* init CPU hotplug notifier */
405     cpu_hotplug_cb.rtc_state = s;
406     cpu_hotplug_cb.cpu_added_notifier.notify = rtc_notify_cpu_added;
407     qemu_register_cpu_added_notifier(&cpu_hotplug_cb.cpu_added_notifier);
408 
409     /* set boot devices, and disable floppy signature check if requested */
410     if (set_boot_dev(s, boot_device, fd_bootchk)) {
411         exit(1);
412     }
413 
414     /* floppy type */
415     if (floppy) {
416         for (i = 0; i < 2; i++) {
417             fd_type[i] = isa_fdc_get_drive_type(floppy, i);
418         }
419     }
420     val = (cmos_get_fd_drive_type(fd_type[0]) << 4) |
421         cmos_get_fd_drive_type(fd_type[1]);
422     rtc_set_memory(s, 0x10, val);
423 
424     val = 0;
425     nb = 0;
426     if (fd_type[0] < FDRIVE_DRV_NONE) {
427         nb++;
428     }
429     if (fd_type[1] < FDRIVE_DRV_NONE) {
430         nb++;
431     }
432     switch (nb) {
433     case 0:
434         break;
435     case 1:
436         val |= 0x01; /* 1 drive, ready for boot */
437         break;
438     case 2:
439         val |= 0x41; /* 2 drives, ready for boot */
440         break;
441     }
442     val |= 0x02; /* FPU is there */
443     val |= 0x04; /* PS/2 mouse installed */
444     rtc_set_memory(s, REG_EQUIPMENT_BYTE, val);
445 
446     /* hard drives */
447     arg.rtc_state = s;
448     arg.idebus[0] = idebus0;
449     arg.idebus[1] = idebus1;
450     qemu_register_reset(pc_cmos_init_late, &arg);
451 }
452 
453 #define TYPE_PORT92 "port92"
454 #define PORT92(obj) OBJECT_CHECK(Port92State, (obj), TYPE_PORT92)
455 
456 /* port 92 stuff: could be split off */
457 typedef struct Port92State {
458     ISADevice parent_obj;
459 
460     MemoryRegion io;
461     uint8_t outport;
462     qemu_irq *a20_out;
463 } Port92State;
464 
465 static void port92_write(void *opaque, hwaddr addr, uint64_t val,
466                          unsigned size)
467 {
468     Port92State *s = opaque;
469 
470     DPRINTF("port92: write 0x%02x\n", val);
471     s->outport = val;
472     qemu_set_irq(*s->a20_out, (val >> 1) & 1);
473     if (val & 1) {
474         qemu_system_reset_request();
475     }
476 }
477 
478 static uint64_t port92_read(void *opaque, hwaddr addr,
479                             unsigned size)
480 {
481     Port92State *s = opaque;
482     uint32_t ret;
483 
484     ret = s->outport;
485     DPRINTF("port92: read 0x%02x\n", ret);
486     return ret;
487 }
488 
489 static void port92_init(ISADevice *dev, qemu_irq *a20_out)
490 {
491     Port92State *s = PORT92(dev);
492 
493     s->a20_out = a20_out;
494 }
495 
496 static const VMStateDescription vmstate_port92_isa = {
497     .name = "port92",
498     .version_id = 1,
499     .minimum_version_id = 1,
500     .minimum_version_id_old = 1,
501     .fields      = (VMStateField []) {
502         VMSTATE_UINT8(outport, Port92State),
503         VMSTATE_END_OF_LIST()
504     }
505 };
506 
507 static void port92_reset(DeviceState *d)
508 {
509     Port92State *s = PORT92(d);
510 
511     s->outport &= ~1;
512 }
513 
514 static const MemoryRegionOps port92_ops = {
515     .read = port92_read,
516     .write = port92_write,
517     .impl = {
518         .min_access_size = 1,
519         .max_access_size = 1,
520     },
521     .endianness = DEVICE_LITTLE_ENDIAN,
522 };
523 
524 static int port92_initfn(ISADevice *dev)
525 {
526     Port92State *s = PORT92(dev);
527 
528     memory_region_init_io(&s->io, &port92_ops, s, "port92", 1);
529     isa_register_ioport(dev, &s->io, 0x92);
530 
531     s->outport = 0;
532     return 0;
533 }
534 
535 static void port92_class_initfn(ObjectClass *klass, void *data)
536 {
537     DeviceClass *dc = DEVICE_CLASS(klass);
538     ISADeviceClass *ic = ISA_DEVICE_CLASS(klass);
539     ic->init = port92_initfn;
540     dc->no_user = 1;
541     dc->reset = port92_reset;
542     dc->vmsd = &vmstate_port92_isa;
543 }
544 
545 static const TypeInfo port92_info = {
546     .name          = TYPE_PORT92,
547     .parent        = TYPE_ISA_DEVICE,
548     .instance_size = sizeof(Port92State),
549     .class_init    = port92_class_initfn,
550 };
551 
552 static void port92_register_types(void)
553 {
554     type_register_static(&port92_info);
555 }
556 
557 type_init(port92_register_types)
558 
559 static void handle_a20_line_change(void *opaque, int irq, int level)
560 {
561     X86CPU *cpu = opaque;
562 
563     /* XXX: send to all CPUs ? */
564     /* XXX: add logic to handle multiple A20 line sources */
565     x86_cpu_set_a20(cpu, level);
566 }
567 
568 int e820_add_entry(uint64_t address, uint64_t length, uint32_t type)
569 {
570     int index = le32_to_cpu(e820_table.count);
571     struct e820_entry *entry;
572 
573     if (index >= E820_NR_ENTRIES)
574         return -EBUSY;
575     entry = &e820_table.entry[index++];
576 
577     entry->address = cpu_to_le64(address);
578     entry->length = cpu_to_le64(length);
579     entry->type = cpu_to_le32(type);
580 
581     e820_table.count = cpu_to_le32(index);
582     return index;
583 }
584 
585 /* Calculates the limit to CPU APIC ID values
586  *
587  * This function returns the limit for the APIC ID value, so that all
588  * CPU APIC IDs are < pc_apic_id_limit().
589  *
590  * This is used for FW_CFG_MAX_CPUS. See comments on bochs_bios_init().
591  */
592 static unsigned int pc_apic_id_limit(unsigned int max_cpus)
593 {
594     return x86_cpu_apic_id_from_index(max_cpus - 1) + 1;
595 }
596 
597 static void *bochs_bios_init(void)
598 {
599     void *fw_cfg;
600     uint8_t *smbios_table;
601     size_t smbios_len;
602     uint64_t *numa_fw_cfg;
603     int i, j;
604     unsigned int apic_id_limit = pc_apic_id_limit(max_cpus);
605 
606     fw_cfg = fw_cfg_init(BIOS_CFG_IOPORT, BIOS_CFG_IOPORT + 1, 0, 0);
607     /* FW_CFG_MAX_CPUS is a bit confusing/problematic on x86:
608      *
609      * SeaBIOS needs FW_CFG_MAX_CPUS for CPU hotplug, but the CPU hotplug
610      * QEMU<->SeaBIOS interface is not based on the "CPU index", but on the APIC
611      * ID of hotplugged CPUs[1]. This means that FW_CFG_MAX_CPUS is not the
612      * "maximum number of CPUs", but the "limit to the APIC ID values SeaBIOS
613      * may see".
614      *
615      * So, this means we must not use max_cpus, here, but the maximum possible
616      * APIC ID value, plus one.
617      *
618      * [1] The only kind of "CPU identifier" used between SeaBIOS and QEMU is
619      *     the APIC ID, not the "CPU index"
620      */
621     fw_cfg_add_i16(fw_cfg, FW_CFG_MAX_CPUS, (uint16_t)apic_id_limit);
622     fw_cfg_add_i32(fw_cfg, FW_CFG_ID, 1);
623     fw_cfg_add_i64(fw_cfg, FW_CFG_RAM_SIZE, (uint64_t)ram_size);
624     fw_cfg_add_bytes(fw_cfg, FW_CFG_ACPI_TABLES,
625                      acpi_tables, acpi_tables_len);
626     fw_cfg_add_i32(fw_cfg, FW_CFG_IRQ0_OVERRIDE, kvm_allows_irq0_override());
627 
628     smbios_table = smbios_get_table(&smbios_len);
629     if (smbios_table)
630         fw_cfg_add_bytes(fw_cfg, FW_CFG_SMBIOS_ENTRIES,
631                          smbios_table, smbios_len);
632     fw_cfg_add_bytes(fw_cfg, FW_CFG_E820_TABLE,
633                      &e820_table, sizeof(e820_table));
634 
635     fw_cfg_add_bytes(fw_cfg, FW_CFG_HPET, &hpet_cfg, sizeof(hpet_cfg));
636     /* allocate memory for the NUMA channel: one (64bit) word for the number
637      * of nodes, one word for each VCPU->node and one word for each node to
638      * hold the amount of memory.
639      */
640     numa_fw_cfg = g_new0(uint64_t, 1 + apic_id_limit + nb_numa_nodes);
641     numa_fw_cfg[0] = cpu_to_le64(nb_numa_nodes);
642     for (i = 0; i < max_cpus; i++) {
643         unsigned int apic_id = x86_cpu_apic_id_from_index(i);
644         assert(apic_id < apic_id_limit);
645         for (j = 0; j < nb_numa_nodes; j++) {
646             if (test_bit(i, node_cpumask[j])) {
647                 numa_fw_cfg[apic_id + 1] = cpu_to_le64(j);
648                 break;
649             }
650         }
651     }
652     for (i = 0; i < nb_numa_nodes; i++) {
653         numa_fw_cfg[apic_id_limit + 1 + i] = cpu_to_le64(node_mem[i]);
654     }
655     fw_cfg_add_bytes(fw_cfg, FW_CFG_NUMA, numa_fw_cfg,
656                      (1 + apic_id_limit + nb_numa_nodes) *
657                      sizeof(*numa_fw_cfg));
658 
659     return fw_cfg;
660 }
661 
662 static long get_file_size(FILE *f)
663 {
664     long where, size;
665 
666     /* XXX: on Unix systems, using fstat() probably makes more sense */
667 
668     where = ftell(f);
669     fseek(f, 0, SEEK_END);
670     size = ftell(f);
671     fseek(f, where, SEEK_SET);
672 
673     return size;
674 }
675 
676 static void load_linux(void *fw_cfg,
677                        const char *kernel_filename,
678                        const char *initrd_filename,
679                        const char *kernel_cmdline,
680                        hwaddr max_ram_size)
681 {
682     uint16_t protocol;
683     int setup_size, kernel_size, initrd_size = 0, cmdline_size;
684     uint32_t initrd_max;
685     uint8_t header[8192], *setup, *kernel, *initrd_data;
686     hwaddr real_addr, prot_addr, cmdline_addr, initrd_addr = 0;
687     FILE *f;
688     char *vmode;
689 
690     /* Align to 16 bytes as a paranoia measure */
691     cmdline_size = (strlen(kernel_cmdline)+16) & ~15;
692 
693     /* load the kernel header */
694     f = fopen(kernel_filename, "rb");
695     if (!f || !(kernel_size = get_file_size(f)) ||
696         fread(header, 1, MIN(ARRAY_SIZE(header), kernel_size), f) !=
697         MIN(ARRAY_SIZE(header), kernel_size)) {
698         fprintf(stderr, "qemu: could not load kernel '%s': %s\n",
699                 kernel_filename, strerror(errno));
700         exit(1);
701     }
702 
703     /* kernel protocol version */
704 #if 0
705     fprintf(stderr, "header magic: %#x\n", ldl_p(header+0x202));
706 #endif
707     if (ldl_p(header+0x202) == 0x53726448) {
708         protocol = lduw_p(header+0x206);
709     } else {
710         /* This looks like a multiboot kernel. If it is, let's stop
711            treating it like a Linux kernel. */
712         if (load_multiboot(fw_cfg, f, kernel_filename, initrd_filename,
713                            kernel_cmdline, kernel_size, header)) {
714             return;
715         }
716         protocol = 0;
717     }
718 
719     if (protocol < 0x200 || !(header[0x211] & 0x01)) {
720         /* Low kernel */
721         real_addr    = 0x90000;
722         cmdline_addr = 0x9a000 - cmdline_size;
723         prot_addr    = 0x10000;
724     } else if (protocol < 0x202) {
725         /* High but ancient kernel */
726         real_addr    = 0x90000;
727         cmdline_addr = 0x9a000 - cmdline_size;
728         prot_addr    = 0x100000;
729     } else {
730         /* High and recent kernel */
731         real_addr    = 0x10000;
732         cmdline_addr = 0x20000;
733         prot_addr    = 0x100000;
734     }
735 
736 #if 0
737     fprintf(stderr,
738             "qemu: real_addr     = 0x" TARGET_FMT_plx "\n"
739             "qemu: cmdline_addr  = 0x" TARGET_FMT_plx "\n"
740             "qemu: prot_addr     = 0x" TARGET_FMT_plx "\n",
741             real_addr,
742             cmdline_addr,
743             prot_addr);
744 #endif
745 
746     /* highest address for loading the initrd */
747     if (protocol >= 0x203) {
748         initrd_max = ldl_p(header+0x22c);
749     } else {
750         initrd_max = 0x37ffffff;
751     }
752 
753     if (initrd_max >= max_ram_size-ACPI_DATA_SIZE)
754     	initrd_max = max_ram_size-ACPI_DATA_SIZE-1;
755 
756     fw_cfg_add_i32(fw_cfg, FW_CFG_CMDLINE_ADDR, cmdline_addr);
757     fw_cfg_add_i32(fw_cfg, FW_CFG_CMDLINE_SIZE, strlen(kernel_cmdline)+1);
758     fw_cfg_add_string(fw_cfg, FW_CFG_CMDLINE_DATA, kernel_cmdline);
759 
760     if (protocol >= 0x202) {
761         stl_p(header+0x228, cmdline_addr);
762     } else {
763         stw_p(header+0x20, 0xA33F);
764         stw_p(header+0x22, cmdline_addr-real_addr);
765     }
766 
767     /* handle vga= parameter */
768     vmode = strstr(kernel_cmdline, "vga=");
769     if (vmode) {
770         unsigned int video_mode;
771         /* skip "vga=" */
772         vmode += 4;
773         if (!strncmp(vmode, "normal", 6)) {
774             video_mode = 0xffff;
775         } else if (!strncmp(vmode, "ext", 3)) {
776             video_mode = 0xfffe;
777         } else if (!strncmp(vmode, "ask", 3)) {
778             video_mode = 0xfffd;
779         } else {
780             video_mode = strtol(vmode, NULL, 0);
781         }
782         stw_p(header+0x1fa, video_mode);
783     }
784 
785     /* loader type */
786     /* High nybble = B reserved for QEMU; low nybble is revision number.
787        If this code is substantially changed, you may want to consider
788        incrementing the revision. */
789     if (protocol >= 0x200) {
790         header[0x210] = 0xB0;
791     }
792     /* heap */
793     if (protocol >= 0x201) {
794         header[0x211] |= 0x80;	/* CAN_USE_HEAP */
795         stw_p(header+0x224, cmdline_addr-real_addr-0x200);
796     }
797 
798     /* load initrd */
799     if (initrd_filename) {
800         if (protocol < 0x200) {
801             fprintf(stderr, "qemu: linux kernel too old to load a ram disk\n");
802             exit(1);
803         }
804 
805         initrd_size = get_image_size(initrd_filename);
806         if (initrd_size < 0) {
807             fprintf(stderr, "qemu: error reading initrd %s\n",
808                     initrd_filename);
809             exit(1);
810         }
811 
812         initrd_addr = (initrd_max-initrd_size) & ~4095;
813 
814         initrd_data = g_malloc(initrd_size);
815         load_image(initrd_filename, initrd_data);
816 
817         fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_ADDR, initrd_addr);
818         fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_SIZE, initrd_size);
819         fw_cfg_add_bytes(fw_cfg, FW_CFG_INITRD_DATA, initrd_data, initrd_size);
820 
821         stl_p(header+0x218, initrd_addr);
822         stl_p(header+0x21c, initrd_size);
823     }
824 
825     /* load kernel and setup */
826     setup_size = header[0x1f1];
827     if (setup_size == 0) {
828         setup_size = 4;
829     }
830     setup_size = (setup_size+1)*512;
831     kernel_size -= setup_size;
832 
833     setup  = g_malloc(setup_size);
834     kernel = g_malloc(kernel_size);
835     fseek(f, 0, SEEK_SET);
836     if (fread(setup, 1, setup_size, f) != setup_size) {
837         fprintf(stderr, "fread() failed\n");
838         exit(1);
839     }
840     if (fread(kernel, 1, kernel_size, f) != kernel_size) {
841         fprintf(stderr, "fread() failed\n");
842         exit(1);
843     }
844     fclose(f);
845     memcpy(setup, header, MIN(sizeof(header), setup_size));
846 
847     fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_ADDR, prot_addr);
848     fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_SIZE, kernel_size);
849     fw_cfg_add_bytes(fw_cfg, FW_CFG_KERNEL_DATA, kernel, kernel_size);
850 
851     fw_cfg_add_i32(fw_cfg, FW_CFG_SETUP_ADDR, real_addr);
852     fw_cfg_add_i32(fw_cfg, FW_CFG_SETUP_SIZE, setup_size);
853     fw_cfg_add_bytes(fw_cfg, FW_CFG_SETUP_DATA, setup, setup_size);
854 
855     option_rom[nb_option_roms].name = "linuxboot.bin";
856     option_rom[nb_option_roms].bootindex = 0;
857     nb_option_roms++;
858 }
859 
860 #define NE2000_NB_MAX 6
861 
862 static const int ne2000_io[NE2000_NB_MAX] = { 0x300, 0x320, 0x340, 0x360,
863                                               0x280, 0x380 };
864 static const int ne2000_irq[NE2000_NB_MAX] = { 9, 10, 11, 3, 4, 5 };
865 
866 static const int parallel_io[MAX_PARALLEL_PORTS] = { 0x378, 0x278, 0x3bc };
867 static const int parallel_irq[MAX_PARALLEL_PORTS] = { 7, 7, 7 };
868 
869 void pc_init_ne2k_isa(ISABus *bus, NICInfo *nd)
870 {
871     static int nb_ne2k = 0;
872 
873     if (nb_ne2k == NE2000_NB_MAX)
874         return;
875     isa_ne2000_init(bus, ne2000_io[nb_ne2k],
876                     ne2000_irq[nb_ne2k], nd);
877     nb_ne2k++;
878 }
879 
880 DeviceState *cpu_get_current_apic(void)
881 {
882     if (cpu_single_env) {
883         return cpu_single_env->apic_state;
884     } else {
885         return NULL;
886     }
887 }
888 
889 void pc_acpi_smi_interrupt(void *opaque, int irq, int level)
890 {
891     X86CPU *cpu = opaque;
892 
893     if (level) {
894         cpu_interrupt(CPU(cpu), CPU_INTERRUPT_SMI);
895     }
896 }
897 
898 static X86CPU *pc_new_cpu(const char *cpu_model, int64_t apic_id,
899                           DeviceState *icc_bridge, Error **errp)
900 {
901     X86CPU *cpu;
902     Error *local_err = NULL;
903 
904     cpu = cpu_x86_create(cpu_model, icc_bridge, errp);
905     if (!cpu) {
906         return cpu;
907     }
908 
909     object_property_set_int(OBJECT(cpu), apic_id, "apic-id", &local_err);
910     object_property_set_bool(OBJECT(cpu), true, "realized", &local_err);
911 
912     if (local_err) {
913         if (cpu != NULL) {
914             object_unref(OBJECT(cpu));
915             cpu = NULL;
916         }
917         error_propagate(errp, local_err);
918     }
919     return cpu;
920 }
921 
922 void pc_cpus_init(const char *cpu_model, DeviceState *icc_bridge)
923 {
924     int i;
925     X86CPU *cpu = NULL;
926     Error *error = NULL;
927 
928     /* init CPUs */
929     if (cpu_model == NULL) {
930 #ifdef TARGET_X86_64
931         cpu_model = "qemu64";
932 #else
933         cpu_model = "qemu32";
934 #endif
935     }
936 
937     for (i = 0; i < smp_cpus; i++) {
938         cpu = pc_new_cpu(cpu_model, x86_cpu_apic_id_from_index(i),
939                          icc_bridge, &error);
940         if (error) {
941             fprintf(stderr, "%s\n", error_get_pretty(error));
942             error_free(error);
943             exit(1);
944         }
945     }
946 
947     /* map APIC MMIO area if CPU has APIC */
948     if (cpu && cpu->env.apic_state) {
949         /* XXX: what if the base changes? */
950         sysbus_mmio_map_overlap(SYS_BUS_DEVICE(icc_bridge), 0,
951                                 APIC_DEFAULT_ADDRESS, 0x1000);
952     }
953 }
954 
955 void pc_acpi_init(const char *default_dsdt)
956 {
957     char *filename;
958 
959     if (acpi_tables != NULL) {
960         /* manually set via -acpitable, leave it alone */
961         return;
962     }
963 
964     filename = qemu_find_file(QEMU_FILE_TYPE_BIOS, default_dsdt);
965     if (filename == NULL) {
966         fprintf(stderr, "WARNING: failed to find %s\n", default_dsdt);
967     } else {
968         char *arg;
969         QemuOpts *opts;
970         Error *err = NULL;
971 
972         arg = g_strdup_printf("file=%s", filename);
973 
974         /* creates a deep copy of "arg" */
975         opts = qemu_opts_parse(qemu_find_opts("acpi"), arg, 0);
976         g_assert(opts != NULL);
977 
978         acpi_table_add(opts, &err);
979         if (err) {
980             fprintf(stderr, "WARNING: failed to load %s: %s\n", filename,
981                     error_get_pretty(err));
982             error_free(err);
983         }
984         g_free(arg);
985         g_free(filename);
986     }
987 }
988 
989 void *pc_memory_init(MemoryRegion *system_memory,
990                     const char *kernel_filename,
991                     const char *kernel_cmdline,
992                     const char *initrd_filename,
993                     ram_addr_t below_4g_mem_size,
994                     ram_addr_t above_4g_mem_size,
995                     MemoryRegion *rom_memory,
996                     MemoryRegion **ram_memory)
997 {
998     int linux_boot, i;
999     MemoryRegion *ram, *option_rom_mr;
1000     MemoryRegion *ram_below_4g, *ram_above_4g;
1001     void *fw_cfg;
1002 
1003     linux_boot = (kernel_filename != NULL);
1004 
1005     /* Allocate RAM.  We allocate it as a single memory region and use
1006      * aliases to address portions of it, mostly for backwards compatibility
1007      * with older qemus that used qemu_ram_alloc().
1008      */
1009     ram = g_malloc(sizeof(*ram));
1010     memory_region_init_ram(ram, "pc.ram",
1011                            below_4g_mem_size + above_4g_mem_size);
1012     vmstate_register_ram_global(ram);
1013     *ram_memory = ram;
1014     ram_below_4g = g_malloc(sizeof(*ram_below_4g));
1015     memory_region_init_alias(ram_below_4g, "ram-below-4g", ram,
1016                              0, below_4g_mem_size);
1017     memory_region_add_subregion(system_memory, 0, ram_below_4g);
1018     if (above_4g_mem_size > 0) {
1019         ram_above_4g = g_malloc(sizeof(*ram_above_4g));
1020         memory_region_init_alias(ram_above_4g, "ram-above-4g", ram,
1021                                  below_4g_mem_size, above_4g_mem_size);
1022         memory_region_add_subregion(system_memory, 0x100000000ULL,
1023                                     ram_above_4g);
1024     }
1025 
1026 
1027     /* Initialize PC system firmware */
1028     pc_system_firmware_init(rom_memory);
1029 
1030     option_rom_mr = g_malloc(sizeof(*option_rom_mr));
1031     memory_region_init_ram(option_rom_mr, "pc.rom", PC_ROM_SIZE);
1032     vmstate_register_ram_global(option_rom_mr);
1033     memory_region_add_subregion_overlap(rom_memory,
1034                                         PC_ROM_MIN_VGA,
1035                                         option_rom_mr,
1036                                         1);
1037 
1038     fw_cfg = bochs_bios_init();
1039     rom_set_fw(fw_cfg);
1040 
1041     if (linux_boot) {
1042         load_linux(fw_cfg, kernel_filename, initrd_filename, kernel_cmdline, below_4g_mem_size);
1043     }
1044 
1045     for (i = 0; i < nb_option_roms; i++) {
1046         rom_add_option(option_rom[i].name, option_rom[i].bootindex);
1047     }
1048     return fw_cfg;
1049 }
1050 
1051 qemu_irq *pc_allocate_cpu_irq(void)
1052 {
1053     return qemu_allocate_irqs(pic_irq_request, NULL, 1);
1054 }
1055 
1056 DeviceState *pc_vga_init(ISABus *isa_bus, PCIBus *pci_bus)
1057 {
1058     DeviceState *dev = NULL;
1059 
1060     if (pci_bus) {
1061         PCIDevice *pcidev = pci_vga_init(pci_bus);
1062         dev = pcidev ? &pcidev->qdev : NULL;
1063     } else if (isa_bus) {
1064         ISADevice *isadev = isa_vga_init(isa_bus);
1065         dev = isadev ? &isadev->qdev : NULL;
1066     }
1067     return dev;
1068 }
1069 
1070 static void cpu_request_exit(void *opaque, int irq, int level)
1071 {
1072     CPUX86State *env = cpu_single_env;
1073 
1074     if (env && level) {
1075         cpu_exit(env);
1076     }
1077 }
1078 
1079 static const MemoryRegionOps ioport80_io_ops = {
1080     .write = ioport80_write,
1081     .read = ioport80_read,
1082     .endianness = DEVICE_NATIVE_ENDIAN,
1083     .impl = {
1084         .min_access_size = 1,
1085         .max_access_size = 1,
1086     },
1087 };
1088 
1089 static const MemoryRegionOps ioportF0_io_ops = {
1090     .write = ioportF0_write,
1091     .read = ioportF0_read,
1092     .endianness = DEVICE_NATIVE_ENDIAN,
1093     .impl = {
1094         .min_access_size = 1,
1095         .max_access_size = 1,
1096     },
1097 };
1098 
1099 void pc_basic_device_init(ISABus *isa_bus, qemu_irq *gsi,
1100                           ISADevice **rtc_state,
1101                           ISADevice **floppy,
1102                           bool no_vmport)
1103 {
1104     int i;
1105     DriveInfo *fd[MAX_FD];
1106     DeviceState *hpet = NULL;
1107     int pit_isa_irq = 0;
1108     qemu_irq pit_alt_irq = NULL;
1109     qemu_irq rtc_irq = NULL;
1110     qemu_irq *a20_line;
1111     ISADevice *i8042, *port92, *vmmouse, *pit = NULL;
1112     qemu_irq *cpu_exit_irq;
1113     MemoryRegion *ioport80_io = g_new(MemoryRegion, 1);
1114     MemoryRegion *ioportF0_io = g_new(MemoryRegion, 1);
1115 
1116     memory_region_init_io(ioport80_io, &ioport80_io_ops, NULL, "ioport80", 1);
1117     memory_region_add_subregion(isa_bus->address_space_io, 0x80, ioport80_io);
1118 
1119     memory_region_init_io(ioportF0_io, &ioportF0_io_ops, NULL, "ioportF0", 1);
1120     memory_region_add_subregion(isa_bus->address_space_io, 0xf0, ioportF0_io);
1121 
1122     /*
1123      * Check if an HPET shall be created.
1124      *
1125      * Without KVM_CAP_PIT_STATE2, we cannot switch off the in-kernel PIT
1126      * when the HPET wants to take over. Thus we have to disable the latter.
1127      */
1128     if (!no_hpet && (!kvm_irqchip_in_kernel() || kvm_has_pit_state2())) {
1129         hpet = sysbus_try_create_simple("hpet", HPET_BASE, NULL);
1130 
1131         if (hpet) {
1132             for (i = 0; i < GSI_NUM_PINS; i++) {
1133                 sysbus_connect_irq(SYS_BUS_DEVICE(hpet), i, gsi[i]);
1134             }
1135             pit_isa_irq = -1;
1136             pit_alt_irq = qdev_get_gpio_in(hpet, HPET_LEGACY_PIT_INT);
1137             rtc_irq = qdev_get_gpio_in(hpet, HPET_LEGACY_RTC_INT);
1138         }
1139     }
1140     *rtc_state = rtc_init(isa_bus, 2000, rtc_irq);
1141 
1142     qemu_register_boot_set(pc_boot_set, *rtc_state);
1143 
1144     if (!xen_enabled()) {
1145         if (kvm_irqchip_in_kernel()) {
1146             pit = kvm_pit_init(isa_bus, 0x40);
1147         } else {
1148             pit = pit_init(isa_bus, 0x40, pit_isa_irq, pit_alt_irq);
1149         }
1150         if (hpet) {
1151             /* connect PIT to output control line of the HPET */
1152             qdev_connect_gpio_out(hpet, 0, qdev_get_gpio_in(&pit->qdev, 0));
1153         }
1154         pcspk_init(isa_bus, pit);
1155     }
1156 
1157     for(i = 0; i < MAX_SERIAL_PORTS; i++) {
1158         if (serial_hds[i]) {
1159             serial_isa_init(isa_bus, i, serial_hds[i]);
1160         }
1161     }
1162 
1163     for(i = 0; i < MAX_PARALLEL_PORTS; i++) {
1164         if (parallel_hds[i]) {
1165             parallel_init(isa_bus, i, parallel_hds[i]);
1166         }
1167     }
1168 
1169     a20_line = qemu_allocate_irqs(handle_a20_line_change,
1170                                   x86_env_get_cpu(first_cpu), 2);
1171     i8042 = isa_create_simple(isa_bus, "i8042");
1172     i8042_setup_a20_line(i8042, &a20_line[0]);
1173     if (!no_vmport) {
1174         vmport_init(isa_bus);
1175         vmmouse = isa_try_create(isa_bus, "vmmouse");
1176     } else {
1177         vmmouse = NULL;
1178     }
1179     if (vmmouse) {
1180         qdev_prop_set_ptr(&vmmouse->qdev, "ps2_mouse", i8042);
1181         qdev_init_nofail(&vmmouse->qdev);
1182     }
1183     port92 = isa_create_simple(isa_bus, "port92");
1184     port92_init(port92, &a20_line[1]);
1185 
1186     cpu_exit_irq = qemu_allocate_irqs(cpu_request_exit, NULL, 1);
1187     DMA_init(0, cpu_exit_irq);
1188 
1189     for(i = 0; i < MAX_FD; i++) {
1190         fd[i] = drive_get(IF_FLOPPY, 0, i);
1191     }
1192     *floppy = fdctrl_init_isa(isa_bus, fd);
1193 }
1194 
1195 void pc_nic_init(ISABus *isa_bus, PCIBus *pci_bus)
1196 {
1197     int i;
1198 
1199     for (i = 0; i < nb_nics; i++) {
1200         NICInfo *nd = &nd_table[i];
1201 
1202         if (!pci_bus || (nd->model && strcmp(nd->model, "ne2k_isa") == 0)) {
1203             pc_init_ne2k_isa(isa_bus, nd);
1204         } else {
1205             pci_nic_init_nofail(nd, "e1000", NULL);
1206         }
1207     }
1208 }
1209 
1210 void pc_pci_device_init(PCIBus *pci_bus)
1211 {
1212     int max_bus;
1213     int bus;
1214 
1215     max_bus = drive_get_max_bus(IF_SCSI);
1216     for (bus = 0; bus <= max_bus; bus++) {
1217         pci_create_simple(pci_bus, -1, "lsi53c895a");
1218     }
1219 }
1220 
1221 void ioapic_init_gsi(GSIState *gsi_state, const char *parent_name)
1222 {
1223     DeviceState *dev;
1224     SysBusDevice *d;
1225     unsigned int i;
1226 
1227     if (kvm_irqchip_in_kernel()) {
1228         dev = qdev_create(NULL, "kvm-ioapic");
1229     } else {
1230         dev = qdev_create(NULL, "ioapic");
1231     }
1232     if (parent_name) {
1233         object_property_add_child(object_resolve_path(parent_name, NULL),
1234                                   "ioapic", OBJECT(dev), NULL);
1235     }
1236     qdev_init_nofail(dev);
1237     d = SYS_BUS_DEVICE(dev);
1238     sysbus_mmio_map(d, 0, IO_APIC_DEFAULT_ADDRESS);
1239 
1240     for (i = 0; i < IOAPIC_NUM_PINS; i++) {
1241         gsi_state->ioapic_irq[i] = qdev_get_gpio_in(dev, i);
1242     }
1243 }
1244