xref: /qemu/hw/i386/pc.c (revision f0513d2c0156799e0c75a108ab9a049eea4f9607)
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 
57 /* debug PC/ISA interrupts */
58 //#define DEBUG_IRQ
59 
60 #ifdef DEBUG_IRQ
61 #define DPRINTF(fmt, ...)                                       \
62     do { printf("CPUIRQ: " fmt , ## __VA_ARGS__); } while (0)
63 #else
64 #define DPRINTF(fmt, ...)
65 #endif
66 
67 /* Leave a chunk of memory at the top of RAM for the BIOS ACPI tables.  */
68 #define ACPI_DATA_SIZE       0x10000
69 #define BIOS_CFG_IOPORT 0x510
70 #define FW_CFG_ACPI_TABLES (FW_CFG_ARCH_LOCAL + 0)
71 #define FW_CFG_SMBIOS_ENTRIES (FW_CFG_ARCH_LOCAL + 1)
72 #define FW_CFG_IRQ0_OVERRIDE (FW_CFG_ARCH_LOCAL + 2)
73 #define FW_CFG_E820_TABLE (FW_CFG_ARCH_LOCAL + 3)
74 #define FW_CFG_HPET (FW_CFG_ARCH_LOCAL + 4)
75 
76 #define IO_APIC_DEFAULT_ADDRESS 0xfec00000
77 
78 #define E820_NR_ENTRIES		16
79 
80 struct e820_entry {
81     uint64_t address;
82     uint64_t length;
83     uint32_t type;
84 } QEMU_PACKED __attribute((__aligned__(4)));
85 
86 struct e820_table {
87     uint32_t count;
88     struct e820_entry entry[E820_NR_ENTRIES];
89 } QEMU_PACKED __attribute((__aligned__(4)));
90 
91 static struct e820_table e820_table;
92 struct hpet_fw_config hpet_cfg = {.count = UINT8_MAX};
93 
94 void gsi_handler(void *opaque, int n, int level)
95 {
96     GSIState *s = opaque;
97 
98     DPRINTF("pc: %s GSI %d\n", level ? "raising" : "lowering", n);
99     if (n < ISA_NUM_IRQS) {
100         qemu_set_irq(s->i8259_irq[n], level);
101     }
102     qemu_set_irq(s->ioapic_irq[n], level);
103 }
104 
105 static void ioport80_write(void *opaque, hwaddr addr, uint64_t data,
106                            unsigned size)
107 {
108 }
109 
110 static uint64_t ioport80_read(void *opaque, hwaddr addr, unsigned size)
111 {
112     return 0xffffffffffffffffULL;
113 }
114 
115 /* MSDOS compatibility mode FPU exception support */
116 static qemu_irq ferr_irq;
117 
118 void pc_register_ferr_irq(qemu_irq irq)
119 {
120     ferr_irq = irq;
121 }
122 
123 /* XXX: add IGNNE support */
124 void cpu_set_ferr(CPUX86State *s)
125 {
126     qemu_irq_raise(ferr_irq);
127 }
128 
129 static void ioportF0_write(void *opaque, hwaddr addr, uint64_t data,
130                            unsigned size)
131 {
132     qemu_irq_lower(ferr_irq);
133 }
134 
135 static uint64_t ioportF0_read(void *opaque, hwaddr addr, unsigned size)
136 {
137     return 0xffffffffffffffffULL;
138 }
139 
140 /* TSC handling */
141 uint64_t cpu_get_tsc(CPUX86State *env)
142 {
143     return cpu_get_ticks();
144 }
145 
146 /* SMM support */
147 
148 static cpu_set_smm_t smm_set;
149 static void *smm_arg;
150 
151 void cpu_smm_register(cpu_set_smm_t callback, void *arg)
152 {
153     assert(smm_set == NULL);
154     assert(smm_arg == NULL);
155     smm_set = callback;
156     smm_arg = arg;
157 }
158 
159 void cpu_smm_update(CPUX86State *env)
160 {
161     if (smm_set && smm_arg && env == first_cpu)
162         smm_set(!!(env->hflags & HF_SMM_MASK), smm_arg);
163 }
164 
165 
166 /* IRQ handling */
167 int cpu_get_pic_interrupt(CPUX86State *env)
168 {
169     int intno;
170 
171     intno = apic_get_interrupt(env->apic_state);
172     if (intno >= 0) {
173         return intno;
174     }
175     /* read the irq from the PIC */
176     if (!apic_accept_pic_intr(env->apic_state)) {
177         return -1;
178     }
179 
180     intno = pic_read_irq(isa_pic);
181     return intno;
182 }
183 
184 static void pic_irq_request(void *opaque, int irq, int level)
185 {
186     CPUX86State *env = first_cpu;
187 
188     DPRINTF("pic_irqs: %s irq %d\n", level? "raise" : "lower", irq);
189     if (env->apic_state) {
190         while (env) {
191             if (apic_accept_pic_intr(env->apic_state)) {
192                 apic_deliver_pic_intr(env->apic_state, level);
193             }
194             env = env->next_cpu;
195         }
196     } else {
197         CPUState *cs = CPU(x86_env_get_cpu(env));
198         if (level) {
199             cpu_interrupt(cs, CPU_INTERRUPT_HARD);
200         } else {
201             cpu_reset_interrupt(cs, CPU_INTERRUPT_HARD);
202         }
203     }
204 }
205 
206 /* PC cmos mappings */
207 
208 #define REG_EQUIPMENT_BYTE          0x14
209 
210 static int cmos_get_fd_drive_type(FDriveType fd0)
211 {
212     int val;
213 
214     switch (fd0) {
215     case FDRIVE_DRV_144:
216         /* 1.44 Mb 3"5 drive */
217         val = 4;
218         break;
219     case FDRIVE_DRV_288:
220         /* 2.88 Mb 3"5 drive */
221         val = 5;
222         break;
223     case FDRIVE_DRV_120:
224         /* 1.2 Mb 5"5 drive */
225         val = 2;
226         break;
227     case FDRIVE_DRV_NONE:
228     default:
229         val = 0;
230         break;
231     }
232     return val;
233 }
234 
235 static void cmos_init_hd(ISADevice *s, int type_ofs, int info_ofs,
236                          int16_t cylinders, int8_t heads, int8_t sectors)
237 {
238     rtc_set_memory(s, type_ofs, 47);
239     rtc_set_memory(s, info_ofs, cylinders);
240     rtc_set_memory(s, info_ofs + 1, cylinders >> 8);
241     rtc_set_memory(s, info_ofs + 2, heads);
242     rtc_set_memory(s, info_ofs + 3, 0xff);
243     rtc_set_memory(s, info_ofs + 4, 0xff);
244     rtc_set_memory(s, info_ofs + 5, 0xc0 | ((heads > 8) << 3));
245     rtc_set_memory(s, info_ofs + 6, cylinders);
246     rtc_set_memory(s, info_ofs + 7, cylinders >> 8);
247     rtc_set_memory(s, info_ofs + 8, sectors);
248 }
249 
250 /* convert boot_device letter to something recognizable by the bios */
251 static int boot_device2nibble(char boot_device)
252 {
253     switch(boot_device) {
254     case 'a':
255     case 'b':
256         return 0x01; /* floppy boot */
257     case 'c':
258         return 0x02; /* hard drive boot */
259     case 'd':
260         return 0x03; /* CD-ROM boot */
261     case 'n':
262         return 0x04; /* Network boot */
263     }
264     return 0;
265 }
266 
267 static int set_boot_dev(ISADevice *s, const char *boot_device, int fd_bootchk)
268 {
269 #define PC_MAX_BOOT_DEVICES 3
270     int nbds, bds[3] = { 0, };
271     int i;
272 
273     nbds = strlen(boot_device);
274     if (nbds > PC_MAX_BOOT_DEVICES) {
275         error_report("Too many boot devices for PC");
276         return(1);
277     }
278     for (i = 0; i < nbds; i++) {
279         bds[i] = boot_device2nibble(boot_device[i]);
280         if (bds[i] == 0) {
281             error_report("Invalid boot device for PC: '%c'",
282                          boot_device[i]);
283             return(1);
284         }
285     }
286     rtc_set_memory(s, 0x3d, (bds[1] << 4) | bds[0]);
287     rtc_set_memory(s, 0x38, (bds[2] << 4) | (fd_bootchk ? 0x0 : 0x1));
288     return(0);
289 }
290 
291 static int pc_boot_set(void *opaque, const char *boot_device)
292 {
293     return set_boot_dev(opaque, boot_device, 0);
294 }
295 
296 typedef struct pc_cmos_init_late_arg {
297     ISADevice *rtc_state;
298     BusState *idebus[2];
299 } pc_cmos_init_late_arg;
300 
301 static void pc_cmos_init_late(void *opaque)
302 {
303     pc_cmos_init_late_arg *arg = opaque;
304     ISADevice *s = arg->rtc_state;
305     int16_t cylinders;
306     int8_t heads, sectors;
307     int val;
308     int i, trans;
309 
310     val = 0;
311     if (ide_get_geometry(arg->idebus[0], 0,
312                          &cylinders, &heads, &sectors) >= 0) {
313         cmos_init_hd(s, 0x19, 0x1b, cylinders, heads, sectors);
314         val |= 0xf0;
315     }
316     if (ide_get_geometry(arg->idebus[0], 1,
317                          &cylinders, &heads, &sectors) >= 0) {
318         cmos_init_hd(s, 0x1a, 0x24, cylinders, heads, sectors);
319         val |= 0x0f;
320     }
321     rtc_set_memory(s, 0x12, val);
322 
323     val = 0;
324     for (i = 0; i < 4; i++) {
325         /* NOTE: ide_get_geometry() returns the physical
326            geometry.  It is always such that: 1 <= sects <= 63, 1
327            <= heads <= 16, 1 <= cylinders <= 16383. The BIOS
328            geometry can be different if a translation is done. */
329         if (ide_get_geometry(arg->idebus[i / 2], i % 2,
330                              &cylinders, &heads, &sectors) >= 0) {
331             trans = ide_get_bios_chs_trans(arg->idebus[i / 2], i % 2) - 1;
332             assert((trans & ~3) == 0);
333             val |= trans << (i * 2);
334         }
335     }
336     rtc_set_memory(s, 0x39, val);
337 
338     qemu_unregister_reset(pc_cmos_init_late, opaque);
339 }
340 
341 typedef struct RTCCPUHotplugArg {
342     Notifier cpu_added_notifier;
343     ISADevice *rtc_state;
344 } RTCCPUHotplugArg;
345 
346 static void rtc_notify_cpu_added(Notifier *notifier, void *data)
347 {
348     RTCCPUHotplugArg *arg = container_of(notifier, RTCCPUHotplugArg,
349                                          cpu_added_notifier);
350     ISADevice *s = arg->rtc_state;
351 
352     /* increment the number of CPUs */
353     rtc_set_memory(s, 0x5f, rtc_get_memory(s, 0x5f) + 1);
354 }
355 
356 void pc_cmos_init(ram_addr_t ram_size, ram_addr_t above_4g_mem_size,
357                   const char *boot_device,
358                   ISADevice *floppy, BusState *idebus0, BusState *idebus1,
359                   ISADevice *s)
360 {
361     int val, nb, i;
362     FDriveType fd_type[2] = { FDRIVE_DRV_NONE, FDRIVE_DRV_NONE };
363     static pc_cmos_init_late_arg arg;
364     static RTCCPUHotplugArg cpu_hotplug_cb;
365 
366     /* various important CMOS locations needed by PC/Bochs bios */
367 
368     /* memory size */
369     /* base memory (first MiB) */
370     val = MIN(ram_size / 1024, 640);
371     rtc_set_memory(s, 0x15, val);
372     rtc_set_memory(s, 0x16, val >> 8);
373     /* extended memory (next 64MiB) */
374     if (ram_size > 1024 * 1024) {
375         val = (ram_size - 1024 * 1024) / 1024;
376     } else {
377         val = 0;
378     }
379     if (val > 65535)
380         val = 65535;
381     rtc_set_memory(s, 0x17, val);
382     rtc_set_memory(s, 0x18, val >> 8);
383     rtc_set_memory(s, 0x30, val);
384     rtc_set_memory(s, 0x31, val >> 8);
385     /* memory between 16MiB and 4GiB */
386     if (ram_size > 16 * 1024 * 1024) {
387         val = (ram_size - 16 * 1024 * 1024) / 65536;
388     } else {
389         val = 0;
390     }
391     if (val > 65535)
392         val = 65535;
393     rtc_set_memory(s, 0x34, val);
394     rtc_set_memory(s, 0x35, val >> 8);
395     /* memory above 4GiB */
396     val = above_4g_mem_size / 65536;
397     rtc_set_memory(s, 0x5b, val);
398     rtc_set_memory(s, 0x5c, val >> 8);
399     rtc_set_memory(s, 0x5d, val >> 16);
400 
401     /* set the number of CPU */
402     rtc_set_memory(s, 0x5f, smp_cpus - 1);
403     /* init CPU hotplug notifier */
404     cpu_hotplug_cb.rtc_state = s;
405     cpu_hotplug_cb.cpu_added_notifier.notify = rtc_notify_cpu_added;
406     qemu_register_cpu_added_notifier(&cpu_hotplug_cb.cpu_added_notifier);
407 
408     /* set boot devices, and disable floppy signature check if requested */
409     if (set_boot_dev(s, boot_device, fd_bootchk)) {
410         exit(1);
411     }
412 
413     /* floppy type */
414     if (floppy) {
415         for (i = 0; i < 2; i++) {
416             fd_type[i] = isa_fdc_get_drive_type(floppy, i);
417         }
418     }
419     val = (cmos_get_fd_drive_type(fd_type[0]) << 4) |
420         cmos_get_fd_drive_type(fd_type[1]);
421     rtc_set_memory(s, 0x10, val);
422 
423     val = 0;
424     nb = 0;
425     if (fd_type[0] < FDRIVE_DRV_NONE) {
426         nb++;
427     }
428     if (fd_type[1] < FDRIVE_DRV_NONE) {
429         nb++;
430     }
431     switch (nb) {
432     case 0:
433         break;
434     case 1:
435         val |= 0x01; /* 1 drive, ready for boot */
436         break;
437     case 2:
438         val |= 0x41; /* 2 drives, ready for boot */
439         break;
440     }
441     val |= 0x02; /* FPU is there */
442     val |= 0x04; /* PS/2 mouse installed */
443     rtc_set_memory(s, REG_EQUIPMENT_BYTE, val);
444 
445     /* hard drives */
446     arg.rtc_state = s;
447     arg.idebus[0] = idebus0;
448     arg.idebus[1] = idebus1;
449     qemu_register_reset(pc_cmos_init_late, &arg);
450 }
451 
452 #define TYPE_PORT92 "port92"
453 #define PORT92(obj) OBJECT_CHECK(Port92State, (obj), TYPE_PORT92)
454 
455 /* port 92 stuff: could be split off */
456 typedef struct Port92State {
457     ISADevice parent_obj;
458 
459     MemoryRegion io;
460     uint8_t outport;
461     qemu_irq *a20_out;
462 } Port92State;
463 
464 static void port92_write(void *opaque, hwaddr addr, uint64_t val,
465                          unsigned size)
466 {
467     Port92State *s = opaque;
468 
469     DPRINTF("port92: write 0x%02x\n", val);
470     s->outport = val;
471     qemu_set_irq(*s->a20_out, (val >> 1) & 1);
472     if (val & 1) {
473         qemu_system_reset_request();
474     }
475 }
476 
477 static uint64_t port92_read(void *opaque, hwaddr addr,
478                             unsigned size)
479 {
480     Port92State *s = opaque;
481     uint32_t ret;
482 
483     ret = s->outport;
484     DPRINTF("port92: read 0x%02x\n", ret);
485     return ret;
486 }
487 
488 static void port92_init(ISADevice *dev, qemu_irq *a20_out)
489 {
490     Port92State *s = PORT92(dev);
491 
492     s->a20_out = a20_out;
493 }
494 
495 static const VMStateDescription vmstate_port92_isa = {
496     .name = "port92",
497     .version_id = 1,
498     .minimum_version_id = 1,
499     .minimum_version_id_old = 1,
500     .fields      = (VMStateField []) {
501         VMSTATE_UINT8(outport, Port92State),
502         VMSTATE_END_OF_LIST()
503     }
504 };
505 
506 static void port92_reset(DeviceState *d)
507 {
508     Port92State *s = PORT92(d);
509 
510     s->outport &= ~1;
511 }
512 
513 static const MemoryRegionOps port92_ops = {
514     .read = port92_read,
515     .write = port92_write,
516     .impl = {
517         .min_access_size = 1,
518         .max_access_size = 1,
519     },
520     .endianness = DEVICE_LITTLE_ENDIAN,
521 };
522 
523 static int port92_initfn(ISADevice *dev)
524 {
525     Port92State *s = PORT92(dev);
526 
527     memory_region_init_io(&s->io, &port92_ops, s, "port92", 1);
528     isa_register_ioport(dev, &s->io, 0x92);
529 
530     s->outport = 0;
531     return 0;
532 }
533 
534 static void port92_class_initfn(ObjectClass *klass, void *data)
535 {
536     DeviceClass *dc = DEVICE_CLASS(klass);
537     ISADeviceClass *ic = ISA_DEVICE_CLASS(klass);
538     ic->init = port92_initfn;
539     dc->no_user = 1;
540     dc->reset = port92_reset;
541     dc->vmsd = &vmstate_port92_isa;
542 }
543 
544 static const TypeInfo port92_info = {
545     .name          = TYPE_PORT92,
546     .parent        = TYPE_ISA_DEVICE,
547     .instance_size = sizeof(Port92State),
548     .class_init    = port92_class_initfn,
549 };
550 
551 static void port92_register_types(void)
552 {
553     type_register_static(&port92_info);
554 }
555 
556 type_init(port92_register_types)
557 
558 static void handle_a20_line_change(void *opaque, int irq, int level)
559 {
560     X86CPU *cpu = opaque;
561 
562     /* XXX: send to all CPUs ? */
563     /* XXX: add logic to handle multiple A20 line sources */
564     x86_cpu_set_a20(cpu, level);
565 }
566 
567 int e820_add_entry(uint64_t address, uint64_t length, uint32_t type)
568 {
569     int index = le32_to_cpu(e820_table.count);
570     struct e820_entry *entry;
571 
572     if (index >= E820_NR_ENTRIES)
573         return -EBUSY;
574     entry = &e820_table.entry[index++];
575 
576     entry->address = cpu_to_le64(address);
577     entry->length = cpu_to_le64(length);
578     entry->type = cpu_to_le32(type);
579 
580     e820_table.count = cpu_to_le32(index);
581     return index;
582 }
583 
584 /* Calculates the limit to CPU APIC ID values
585  *
586  * This function returns the limit for the APIC ID value, so that all
587  * CPU APIC IDs are < pc_apic_id_limit().
588  *
589  * This is used for FW_CFG_MAX_CPUS. See comments on bochs_bios_init().
590  */
591 static unsigned int pc_apic_id_limit(unsigned int max_cpus)
592 {
593     return x86_cpu_apic_id_from_index(max_cpus - 1) + 1;
594 }
595 
596 static void *bochs_bios_init(void)
597 {
598     void *fw_cfg;
599     uint8_t *smbios_table;
600     size_t smbios_len;
601     uint64_t *numa_fw_cfg;
602     int i, j;
603     unsigned int apic_id_limit = pc_apic_id_limit(max_cpus);
604 
605     fw_cfg = fw_cfg_init(BIOS_CFG_IOPORT, BIOS_CFG_IOPORT + 1, 0, 0);
606     /* FW_CFG_MAX_CPUS is a bit confusing/problematic on x86:
607      *
608      * SeaBIOS needs FW_CFG_MAX_CPUS for CPU hotplug, but the CPU hotplug
609      * QEMU<->SeaBIOS interface is not based on the "CPU index", but on the APIC
610      * ID of hotplugged CPUs[1]. This means that FW_CFG_MAX_CPUS is not the
611      * "maximum number of CPUs", but the "limit to the APIC ID values SeaBIOS
612      * may see".
613      *
614      * So, this means we must not use max_cpus, here, but the maximum possible
615      * APIC ID value, plus one.
616      *
617      * [1] The only kind of "CPU identifier" used between SeaBIOS and QEMU is
618      *     the APIC ID, not the "CPU index"
619      */
620     fw_cfg_add_i16(fw_cfg, FW_CFG_MAX_CPUS, (uint16_t)apic_id_limit);
621     fw_cfg_add_i32(fw_cfg, FW_CFG_ID, 1);
622     fw_cfg_add_i64(fw_cfg, FW_CFG_RAM_SIZE, (uint64_t)ram_size);
623     fw_cfg_add_bytes(fw_cfg, FW_CFG_ACPI_TABLES,
624                      acpi_tables, acpi_tables_len);
625     fw_cfg_add_i32(fw_cfg, FW_CFG_IRQ0_OVERRIDE, kvm_allows_irq0_override());
626 
627     smbios_table = smbios_get_table(&smbios_len);
628     if (smbios_table)
629         fw_cfg_add_bytes(fw_cfg, FW_CFG_SMBIOS_ENTRIES,
630                          smbios_table, smbios_len);
631     fw_cfg_add_bytes(fw_cfg, FW_CFG_E820_TABLE,
632                      &e820_table, sizeof(e820_table));
633 
634     fw_cfg_add_bytes(fw_cfg, FW_CFG_HPET, &hpet_cfg, sizeof(hpet_cfg));
635     /* allocate memory for the NUMA channel: one (64bit) word for the number
636      * of nodes, one word for each VCPU->node and one word for each node to
637      * hold the amount of memory.
638      */
639     numa_fw_cfg = g_new0(uint64_t, 1 + apic_id_limit + nb_numa_nodes);
640     numa_fw_cfg[0] = cpu_to_le64(nb_numa_nodes);
641     for (i = 0; i < max_cpus; i++) {
642         unsigned int apic_id = x86_cpu_apic_id_from_index(i);
643         assert(apic_id < apic_id_limit);
644         for (j = 0; j < nb_numa_nodes; j++) {
645             if (test_bit(i, node_cpumask[j])) {
646                 numa_fw_cfg[apic_id + 1] = cpu_to_le64(j);
647                 break;
648             }
649         }
650     }
651     for (i = 0; i < nb_numa_nodes; i++) {
652         numa_fw_cfg[apic_id_limit + 1 + i] = cpu_to_le64(node_mem[i]);
653     }
654     fw_cfg_add_bytes(fw_cfg, FW_CFG_NUMA, numa_fw_cfg,
655                      (1 + apic_id_limit + nb_numa_nodes) *
656                      sizeof(*numa_fw_cfg));
657 
658     return fw_cfg;
659 }
660 
661 static long get_file_size(FILE *f)
662 {
663     long where, size;
664 
665     /* XXX: on Unix systems, using fstat() probably makes more sense */
666 
667     where = ftell(f);
668     fseek(f, 0, SEEK_END);
669     size = ftell(f);
670     fseek(f, where, SEEK_SET);
671 
672     return size;
673 }
674 
675 static void load_linux(void *fw_cfg,
676                        const char *kernel_filename,
677                        const char *initrd_filename,
678                        const char *kernel_cmdline,
679                        hwaddr max_ram_size)
680 {
681     uint16_t protocol;
682     int setup_size, kernel_size, initrd_size = 0, cmdline_size;
683     uint32_t initrd_max;
684     uint8_t header[8192], *setup, *kernel, *initrd_data;
685     hwaddr real_addr, prot_addr, cmdline_addr, initrd_addr = 0;
686     FILE *f;
687     char *vmode;
688 
689     /* Align to 16 bytes as a paranoia measure */
690     cmdline_size = (strlen(kernel_cmdline)+16) & ~15;
691 
692     /* load the kernel header */
693     f = fopen(kernel_filename, "rb");
694     if (!f || !(kernel_size = get_file_size(f)) ||
695         fread(header, 1, MIN(ARRAY_SIZE(header), kernel_size), f) !=
696         MIN(ARRAY_SIZE(header), kernel_size)) {
697         fprintf(stderr, "qemu: could not load kernel '%s': %s\n",
698                 kernel_filename, strerror(errno));
699         exit(1);
700     }
701 
702     /* kernel protocol version */
703 #if 0
704     fprintf(stderr, "header magic: %#x\n", ldl_p(header+0x202));
705 #endif
706     if (ldl_p(header+0x202) == 0x53726448) {
707         protocol = lduw_p(header+0x206);
708     } else {
709         /* This looks like a multiboot kernel. If it is, let's stop
710            treating it like a Linux kernel. */
711         if (load_multiboot(fw_cfg, f, kernel_filename, initrd_filename,
712                            kernel_cmdline, kernel_size, header)) {
713             return;
714         }
715         protocol = 0;
716     }
717 
718     if (protocol < 0x200 || !(header[0x211] & 0x01)) {
719         /* Low kernel */
720         real_addr    = 0x90000;
721         cmdline_addr = 0x9a000 - cmdline_size;
722         prot_addr    = 0x10000;
723     } else if (protocol < 0x202) {
724         /* High but ancient kernel */
725         real_addr    = 0x90000;
726         cmdline_addr = 0x9a000 - cmdline_size;
727         prot_addr    = 0x100000;
728     } else {
729         /* High and recent kernel */
730         real_addr    = 0x10000;
731         cmdline_addr = 0x20000;
732         prot_addr    = 0x100000;
733     }
734 
735 #if 0
736     fprintf(stderr,
737             "qemu: real_addr     = 0x" TARGET_FMT_plx "\n"
738             "qemu: cmdline_addr  = 0x" TARGET_FMT_plx "\n"
739             "qemu: prot_addr     = 0x" TARGET_FMT_plx "\n",
740             real_addr,
741             cmdline_addr,
742             prot_addr);
743 #endif
744 
745     /* highest address for loading the initrd */
746     if (protocol >= 0x203) {
747         initrd_max = ldl_p(header+0x22c);
748     } else {
749         initrd_max = 0x37ffffff;
750     }
751 
752     if (initrd_max >= max_ram_size-ACPI_DATA_SIZE)
753     	initrd_max = max_ram_size-ACPI_DATA_SIZE-1;
754 
755     fw_cfg_add_i32(fw_cfg, FW_CFG_CMDLINE_ADDR, cmdline_addr);
756     fw_cfg_add_i32(fw_cfg, FW_CFG_CMDLINE_SIZE, strlen(kernel_cmdline)+1);
757     fw_cfg_add_string(fw_cfg, FW_CFG_CMDLINE_DATA, kernel_cmdline);
758 
759     if (protocol >= 0x202) {
760         stl_p(header+0x228, cmdline_addr);
761     } else {
762         stw_p(header+0x20, 0xA33F);
763         stw_p(header+0x22, cmdline_addr-real_addr);
764     }
765 
766     /* handle vga= parameter */
767     vmode = strstr(kernel_cmdline, "vga=");
768     if (vmode) {
769         unsigned int video_mode;
770         /* skip "vga=" */
771         vmode += 4;
772         if (!strncmp(vmode, "normal", 6)) {
773             video_mode = 0xffff;
774         } else if (!strncmp(vmode, "ext", 3)) {
775             video_mode = 0xfffe;
776         } else if (!strncmp(vmode, "ask", 3)) {
777             video_mode = 0xfffd;
778         } else {
779             video_mode = strtol(vmode, NULL, 0);
780         }
781         stw_p(header+0x1fa, video_mode);
782     }
783 
784     /* loader type */
785     /* High nybble = B reserved for QEMU; low nybble is revision number.
786        If this code is substantially changed, you may want to consider
787        incrementing the revision. */
788     if (protocol >= 0x200) {
789         header[0x210] = 0xB0;
790     }
791     /* heap */
792     if (protocol >= 0x201) {
793         header[0x211] |= 0x80;	/* CAN_USE_HEAP */
794         stw_p(header+0x224, cmdline_addr-real_addr-0x200);
795     }
796 
797     /* load initrd */
798     if (initrd_filename) {
799         if (protocol < 0x200) {
800             fprintf(stderr, "qemu: linux kernel too old to load a ram disk\n");
801             exit(1);
802         }
803 
804         initrd_size = get_image_size(initrd_filename);
805         if (initrd_size < 0) {
806             fprintf(stderr, "qemu: error reading initrd %s\n",
807                     initrd_filename);
808             exit(1);
809         }
810 
811         initrd_addr = (initrd_max-initrd_size) & ~4095;
812 
813         initrd_data = g_malloc(initrd_size);
814         load_image(initrd_filename, initrd_data);
815 
816         fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_ADDR, initrd_addr);
817         fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_SIZE, initrd_size);
818         fw_cfg_add_bytes(fw_cfg, FW_CFG_INITRD_DATA, initrd_data, initrd_size);
819 
820         stl_p(header+0x218, initrd_addr);
821         stl_p(header+0x21c, initrd_size);
822     }
823 
824     /* load kernel and setup */
825     setup_size = header[0x1f1];
826     if (setup_size == 0) {
827         setup_size = 4;
828     }
829     setup_size = (setup_size+1)*512;
830     kernel_size -= setup_size;
831 
832     setup  = g_malloc(setup_size);
833     kernel = g_malloc(kernel_size);
834     fseek(f, 0, SEEK_SET);
835     if (fread(setup, 1, setup_size, f) != setup_size) {
836         fprintf(stderr, "fread() failed\n");
837         exit(1);
838     }
839     if (fread(kernel, 1, kernel_size, f) != kernel_size) {
840         fprintf(stderr, "fread() failed\n");
841         exit(1);
842     }
843     fclose(f);
844     memcpy(setup, header, MIN(sizeof(header), setup_size));
845 
846     fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_ADDR, prot_addr);
847     fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_SIZE, kernel_size);
848     fw_cfg_add_bytes(fw_cfg, FW_CFG_KERNEL_DATA, kernel, kernel_size);
849 
850     fw_cfg_add_i32(fw_cfg, FW_CFG_SETUP_ADDR, real_addr);
851     fw_cfg_add_i32(fw_cfg, FW_CFG_SETUP_SIZE, setup_size);
852     fw_cfg_add_bytes(fw_cfg, FW_CFG_SETUP_DATA, setup, setup_size);
853 
854     option_rom[nb_option_roms].name = "linuxboot.bin";
855     option_rom[nb_option_roms].bootindex = 0;
856     nb_option_roms++;
857 }
858 
859 #define NE2000_NB_MAX 6
860 
861 static const int ne2000_io[NE2000_NB_MAX] = { 0x300, 0x320, 0x340, 0x360,
862                                               0x280, 0x380 };
863 static const int ne2000_irq[NE2000_NB_MAX] = { 9, 10, 11, 3, 4, 5 };
864 
865 static const int parallel_io[MAX_PARALLEL_PORTS] = { 0x378, 0x278, 0x3bc };
866 static const int parallel_irq[MAX_PARALLEL_PORTS] = { 7, 7, 7 };
867 
868 void pc_init_ne2k_isa(ISABus *bus, NICInfo *nd)
869 {
870     static int nb_ne2k = 0;
871 
872     if (nb_ne2k == NE2000_NB_MAX)
873         return;
874     isa_ne2000_init(bus, ne2000_io[nb_ne2k],
875                     ne2000_irq[nb_ne2k], nd);
876     nb_ne2k++;
877 }
878 
879 DeviceState *cpu_get_current_apic(void)
880 {
881     if (cpu_single_env) {
882         return cpu_single_env->apic_state;
883     } else {
884         return NULL;
885     }
886 }
887 
888 void pc_acpi_smi_interrupt(void *opaque, int irq, int level)
889 {
890     X86CPU *cpu = opaque;
891 
892     if (level) {
893         cpu_interrupt(CPU(cpu), CPU_INTERRUPT_SMI);
894     }
895 }
896 
897 static X86CPU *pc_new_cpu(const char *cpu_model, int64_t apic_id, Error **errp)
898 {
899     X86CPU *cpu;
900     Error *local_err = NULL;
901 
902     cpu = cpu_x86_create(cpu_model, errp);
903     if (!cpu) {
904         return cpu;
905     }
906 
907     object_property_set_int(OBJECT(cpu), apic_id, "apic-id", &local_err);
908     object_property_set_bool(OBJECT(cpu), true, "realized", &local_err);
909 
910     if (local_err) {
911         if (cpu != NULL) {
912             object_unref(OBJECT(cpu));
913             cpu = NULL;
914         }
915         error_propagate(errp, local_err);
916     }
917     return cpu;
918 }
919 
920 void pc_cpus_init(const char *cpu_model)
921 {
922     int i;
923     Error *error = NULL;
924 
925     /* init CPUs */
926     if (cpu_model == NULL) {
927 #ifdef TARGET_X86_64
928         cpu_model = "qemu64";
929 #else
930         cpu_model = "qemu32";
931 #endif
932     }
933 
934     for (i = 0; i < smp_cpus; i++) {
935         pc_new_cpu(cpu_model, x86_cpu_apic_id_from_index(i), &error);
936         if (error) {
937             fprintf(stderr, "%s\n", error_get_pretty(error));
938             error_free(error);
939             exit(1);
940         }
941     }
942 }
943 
944 void pc_acpi_init(const char *default_dsdt)
945 {
946     char *filename;
947 
948     if (acpi_tables != NULL) {
949         /* manually set via -acpitable, leave it alone */
950         return;
951     }
952 
953     filename = qemu_find_file(QEMU_FILE_TYPE_BIOS, default_dsdt);
954     if (filename == NULL) {
955         fprintf(stderr, "WARNING: failed to find %s\n", default_dsdt);
956     } else {
957         char *arg;
958         QemuOpts *opts;
959         Error *err = NULL;
960 
961         arg = g_strdup_printf("file=%s", filename);
962 
963         /* creates a deep copy of "arg" */
964         opts = qemu_opts_parse(qemu_find_opts("acpi"), arg, 0);
965         g_assert(opts != NULL);
966 
967         acpi_table_add(opts, &err);
968         if (err) {
969             fprintf(stderr, "WARNING: failed to load %s: %s\n", filename,
970                     error_get_pretty(err));
971             error_free(err);
972         }
973         g_free(arg);
974         g_free(filename);
975     }
976 }
977 
978 void *pc_memory_init(MemoryRegion *system_memory,
979                     const char *kernel_filename,
980                     const char *kernel_cmdline,
981                     const char *initrd_filename,
982                     ram_addr_t below_4g_mem_size,
983                     ram_addr_t above_4g_mem_size,
984                     MemoryRegion *rom_memory,
985                     MemoryRegion **ram_memory)
986 {
987     int linux_boot, i;
988     MemoryRegion *ram, *option_rom_mr;
989     MemoryRegion *ram_below_4g, *ram_above_4g;
990     void *fw_cfg;
991 
992     linux_boot = (kernel_filename != NULL);
993 
994     /* Allocate RAM.  We allocate it as a single memory region and use
995      * aliases to address portions of it, mostly for backwards compatibility
996      * with older qemus that used qemu_ram_alloc().
997      */
998     ram = g_malloc(sizeof(*ram));
999     memory_region_init_ram(ram, "pc.ram",
1000                            below_4g_mem_size + above_4g_mem_size);
1001     vmstate_register_ram_global(ram);
1002     *ram_memory = ram;
1003     ram_below_4g = g_malloc(sizeof(*ram_below_4g));
1004     memory_region_init_alias(ram_below_4g, "ram-below-4g", ram,
1005                              0, below_4g_mem_size);
1006     memory_region_add_subregion(system_memory, 0, ram_below_4g);
1007     if (above_4g_mem_size > 0) {
1008         ram_above_4g = g_malloc(sizeof(*ram_above_4g));
1009         memory_region_init_alias(ram_above_4g, "ram-above-4g", ram,
1010                                  below_4g_mem_size, above_4g_mem_size);
1011         memory_region_add_subregion(system_memory, 0x100000000ULL,
1012                                     ram_above_4g);
1013     }
1014 
1015 
1016     /* Initialize PC system firmware */
1017     pc_system_firmware_init(rom_memory);
1018 
1019     option_rom_mr = g_malloc(sizeof(*option_rom_mr));
1020     memory_region_init_ram(option_rom_mr, "pc.rom", PC_ROM_SIZE);
1021     vmstate_register_ram_global(option_rom_mr);
1022     memory_region_add_subregion_overlap(rom_memory,
1023                                         PC_ROM_MIN_VGA,
1024                                         option_rom_mr,
1025                                         1);
1026 
1027     fw_cfg = bochs_bios_init();
1028     rom_set_fw(fw_cfg);
1029 
1030     if (linux_boot) {
1031         load_linux(fw_cfg, kernel_filename, initrd_filename, kernel_cmdline, below_4g_mem_size);
1032     }
1033 
1034     for (i = 0; i < nb_option_roms; i++) {
1035         rom_add_option(option_rom[i].name, option_rom[i].bootindex);
1036     }
1037     return fw_cfg;
1038 }
1039 
1040 qemu_irq *pc_allocate_cpu_irq(void)
1041 {
1042     return qemu_allocate_irqs(pic_irq_request, NULL, 1);
1043 }
1044 
1045 DeviceState *pc_vga_init(ISABus *isa_bus, PCIBus *pci_bus)
1046 {
1047     DeviceState *dev = NULL;
1048 
1049     if (pci_bus) {
1050         PCIDevice *pcidev = pci_vga_init(pci_bus);
1051         dev = pcidev ? &pcidev->qdev : NULL;
1052     } else if (isa_bus) {
1053         ISADevice *isadev = isa_vga_init(isa_bus);
1054         dev = isadev ? &isadev->qdev : NULL;
1055     }
1056     return dev;
1057 }
1058 
1059 static void cpu_request_exit(void *opaque, int irq, int level)
1060 {
1061     CPUX86State *env = cpu_single_env;
1062 
1063     if (env && level) {
1064         cpu_exit(env);
1065     }
1066 }
1067 
1068 static const MemoryRegionOps ioport80_io_ops = {
1069     .write = ioport80_write,
1070     .read = ioport80_read,
1071     .endianness = DEVICE_NATIVE_ENDIAN,
1072     .impl = {
1073         .min_access_size = 1,
1074         .max_access_size = 1,
1075     },
1076 };
1077 
1078 static const MemoryRegionOps ioportF0_io_ops = {
1079     .write = ioportF0_write,
1080     .read = ioportF0_read,
1081     .endianness = DEVICE_NATIVE_ENDIAN,
1082     .impl = {
1083         .min_access_size = 1,
1084         .max_access_size = 1,
1085     },
1086 };
1087 
1088 void pc_basic_device_init(ISABus *isa_bus, qemu_irq *gsi,
1089                           ISADevice **rtc_state,
1090                           ISADevice **floppy,
1091                           bool no_vmport)
1092 {
1093     int i;
1094     DriveInfo *fd[MAX_FD];
1095     DeviceState *hpet = NULL;
1096     int pit_isa_irq = 0;
1097     qemu_irq pit_alt_irq = NULL;
1098     qemu_irq rtc_irq = NULL;
1099     qemu_irq *a20_line;
1100     ISADevice *i8042, *port92, *vmmouse, *pit = NULL;
1101     qemu_irq *cpu_exit_irq;
1102     MemoryRegion *ioport80_io = g_new(MemoryRegion, 1);
1103     MemoryRegion *ioportF0_io = g_new(MemoryRegion, 1);
1104 
1105     memory_region_init_io(ioport80_io, &ioport80_io_ops, NULL, "ioport80", 1);
1106     memory_region_add_subregion(isa_bus->address_space_io, 0x80, ioport80_io);
1107 
1108     memory_region_init_io(ioportF0_io, &ioportF0_io_ops, NULL, "ioportF0", 1);
1109     memory_region_add_subregion(isa_bus->address_space_io, 0xf0, ioportF0_io);
1110 
1111     /*
1112      * Check if an HPET shall be created.
1113      *
1114      * Without KVM_CAP_PIT_STATE2, we cannot switch off the in-kernel PIT
1115      * when the HPET wants to take over. Thus we have to disable the latter.
1116      */
1117     if (!no_hpet && (!kvm_irqchip_in_kernel() || kvm_has_pit_state2())) {
1118         hpet = sysbus_try_create_simple("hpet", HPET_BASE, NULL);
1119 
1120         if (hpet) {
1121             for (i = 0; i < GSI_NUM_PINS; i++) {
1122                 sysbus_connect_irq(SYS_BUS_DEVICE(hpet), i, gsi[i]);
1123             }
1124             pit_isa_irq = -1;
1125             pit_alt_irq = qdev_get_gpio_in(hpet, HPET_LEGACY_PIT_INT);
1126             rtc_irq = qdev_get_gpio_in(hpet, HPET_LEGACY_RTC_INT);
1127         }
1128     }
1129     *rtc_state = rtc_init(isa_bus, 2000, rtc_irq);
1130 
1131     qemu_register_boot_set(pc_boot_set, *rtc_state);
1132 
1133     if (!xen_enabled()) {
1134         if (kvm_irqchip_in_kernel()) {
1135             pit = kvm_pit_init(isa_bus, 0x40);
1136         } else {
1137             pit = pit_init(isa_bus, 0x40, pit_isa_irq, pit_alt_irq);
1138         }
1139         if (hpet) {
1140             /* connect PIT to output control line of the HPET */
1141             qdev_connect_gpio_out(hpet, 0, qdev_get_gpio_in(&pit->qdev, 0));
1142         }
1143         pcspk_init(isa_bus, pit);
1144     }
1145 
1146     for(i = 0; i < MAX_SERIAL_PORTS; i++) {
1147         if (serial_hds[i]) {
1148             serial_isa_init(isa_bus, i, serial_hds[i]);
1149         }
1150     }
1151 
1152     for(i = 0; i < MAX_PARALLEL_PORTS; i++) {
1153         if (parallel_hds[i]) {
1154             parallel_init(isa_bus, i, parallel_hds[i]);
1155         }
1156     }
1157 
1158     a20_line = qemu_allocate_irqs(handle_a20_line_change,
1159                                   x86_env_get_cpu(first_cpu), 2);
1160     i8042 = isa_create_simple(isa_bus, "i8042");
1161     i8042_setup_a20_line(i8042, &a20_line[0]);
1162     if (!no_vmport) {
1163         vmport_init(isa_bus);
1164         vmmouse = isa_try_create(isa_bus, "vmmouse");
1165     } else {
1166         vmmouse = NULL;
1167     }
1168     if (vmmouse) {
1169         qdev_prop_set_ptr(&vmmouse->qdev, "ps2_mouse", i8042);
1170         qdev_init_nofail(&vmmouse->qdev);
1171     }
1172     port92 = isa_create_simple(isa_bus, "port92");
1173     port92_init(port92, &a20_line[1]);
1174 
1175     cpu_exit_irq = qemu_allocate_irqs(cpu_request_exit, NULL, 1);
1176     DMA_init(0, cpu_exit_irq);
1177 
1178     for(i = 0; i < MAX_FD; i++) {
1179         fd[i] = drive_get(IF_FLOPPY, 0, i);
1180     }
1181     *floppy = fdctrl_init_isa(isa_bus, fd);
1182 }
1183 
1184 void pc_nic_init(ISABus *isa_bus, PCIBus *pci_bus)
1185 {
1186     int i;
1187 
1188     for (i = 0; i < nb_nics; i++) {
1189         NICInfo *nd = &nd_table[i];
1190 
1191         if (!pci_bus || (nd->model && strcmp(nd->model, "ne2k_isa") == 0)) {
1192             pc_init_ne2k_isa(isa_bus, nd);
1193         } else {
1194             pci_nic_init_nofail(nd, "e1000", NULL);
1195         }
1196     }
1197 }
1198 
1199 void pc_pci_device_init(PCIBus *pci_bus)
1200 {
1201     int max_bus;
1202     int bus;
1203 
1204     max_bus = drive_get_max_bus(IF_SCSI);
1205     for (bus = 0; bus <= max_bus; bus++) {
1206         pci_create_simple(pci_bus, -1, "lsi53c895a");
1207     }
1208 }
1209 
1210 void ioapic_init_gsi(GSIState *gsi_state, const char *parent_name)
1211 {
1212     DeviceState *dev;
1213     SysBusDevice *d;
1214     unsigned int i;
1215 
1216     if (kvm_irqchip_in_kernel()) {
1217         dev = qdev_create(NULL, "kvm-ioapic");
1218     } else {
1219         dev = qdev_create(NULL, "ioapic");
1220     }
1221     if (parent_name) {
1222         object_property_add_child(object_resolve_path(parent_name, NULL),
1223                                   "ioapic", OBJECT(dev), NULL);
1224     }
1225     qdev_init_nofail(dev);
1226     d = SYS_BUS_DEVICE(dev);
1227     sysbus_mmio_map(d, 0, IO_APIC_DEFAULT_ADDRESS);
1228 
1229     for (i = 0; i < IOAPIC_NUM_PINS; i++) {
1230         gsi_state->ioapic_irq[i] = qdev_get_gpio_in(dev, i);
1231     }
1232 }
1233