]> git.proxmox.com Git - qemu.git/blob - hw/i386/pc.c
pc: add etc/e820 fw_cfg file
[qemu.git] / hw / i386 / pc.c
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 #include "hw/boards.h"
58 #include "hw/pci/pci_host.h"
59 #include "acpi-build.h"
60
61 /* debug PC/ISA interrupts */
62 //#define DEBUG_IRQ
63
64 #ifdef DEBUG_IRQ
65 #define DPRINTF(fmt, ...) \
66 do { printf("CPUIRQ: " fmt , ## __VA_ARGS__); } while (0)
67 #else
68 #define DPRINTF(fmt, ...)
69 #endif
70
71 /* Leave a chunk of memory at the top of RAM for the BIOS ACPI tables. */
72 #define ACPI_DATA_SIZE 0x10000
73 #define BIOS_CFG_IOPORT 0x510
74 #define FW_CFG_ACPI_TABLES (FW_CFG_ARCH_LOCAL + 0)
75 #define FW_CFG_SMBIOS_ENTRIES (FW_CFG_ARCH_LOCAL + 1)
76 #define FW_CFG_IRQ0_OVERRIDE (FW_CFG_ARCH_LOCAL + 2)
77 #define FW_CFG_E820_TABLE (FW_CFG_ARCH_LOCAL + 3)
78 #define FW_CFG_HPET (FW_CFG_ARCH_LOCAL + 4)
79
80 #define E820_NR_ENTRIES 16
81
82 struct e820_entry {
83 uint64_t address;
84 uint64_t length;
85 uint32_t type;
86 } QEMU_PACKED __attribute((__aligned__(4)));
87
88 struct e820_table {
89 uint32_t count;
90 struct e820_entry entry[E820_NR_ENTRIES];
91 } QEMU_PACKED __attribute((__aligned__(4)));
92
93 static struct e820_table e820_reserve;
94 static struct e820_entry *e820_table;
95 static unsigned e820_entries;
96 struct hpet_fw_config hpet_cfg = {.count = UINT8_MAX};
97
98 void gsi_handler(void *opaque, int n, int level)
99 {
100 GSIState *s = opaque;
101
102 DPRINTF("pc: %s GSI %d\n", level ? "raising" : "lowering", n);
103 if (n < ISA_NUM_IRQS) {
104 qemu_set_irq(s->i8259_irq[n], level);
105 }
106 qemu_set_irq(s->ioapic_irq[n], level);
107 }
108
109 static void ioport80_write(void *opaque, hwaddr addr, uint64_t data,
110 unsigned size)
111 {
112 }
113
114 static uint64_t ioport80_read(void *opaque, hwaddr addr, unsigned size)
115 {
116 return 0xffffffffffffffffULL;
117 }
118
119 /* MSDOS compatibility mode FPU exception support */
120 static qemu_irq ferr_irq;
121
122 void pc_register_ferr_irq(qemu_irq irq)
123 {
124 ferr_irq = irq;
125 }
126
127 /* XXX: add IGNNE support */
128 void cpu_set_ferr(CPUX86State *s)
129 {
130 qemu_irq_raise(ferr_irq);
131 }
132
133 static void ioportF0_write(void *opaque, hwaddr addr, uint64_t data,
134 unsigned size)
135 {
136 qemu_irq_lower(ferr_irq);
137 }
138
139 static uint64_t ioportF0_read(void *opaque, hwaddr addr, unsigned size)
140 {
141 return 0xffffffffffffffffULL;
142 }
143
144 /* TSC handling */
145 uint64_t cpu_get_tsc(CPUX86State *env)
146 {
147 return cpu_get_ticks();
148 }
149
150 /* SMM support */
151
152 static cpu_set_smm_t smm_set;
153 static void *smm_arg;
154
155 void cpu_smm_register(cpu_set_smm_t callback, void *arg)
156 {
157 assert(smm_set == NULL);
158 assert(smm_arg == NULL);
159 smm_set = callback;
160 smm_arg = arg;
161 }
162
163 void cpu_smm_update(CPUX86State *env)
164 {
165 if (smm_set && smm_arg && CPU(x86_env_get_cpu(env)) == first_cpu) {
166 smm_set(!!(env->hflags & HF_SMM_MASK), smm_arg);
167 }
168 }
169
170
171 /* IRQ handling */
172 int cpu_get_pic_interrupt(CPUX86State *env)
173 {
174 int intno;
175
176 intno = apic_get_interrupt(env->apic_state);
177 if (intno >= 0) {
178 return intno;
179 }
180 /* read the irq from the PIC */
181 if (!apic_accept_pic_intr(env->apic_state)) {
182 return -1;
183 }
184
185 intno = pic_read_irq(isa_pic);
186 return intno;
187 }
188
189 static void pic_irq_request(void *opaque, int irq, int level)
190 {
191 CPUState *cs = first_cpu;
192 X86CPU *cpu = X86_CPU(cs);
193 CPUX86State *env = &cpu->env;
194
195 DPRINTF("pic_irqs: %s irq %d\n", level? "raise" : "lower", irq);
196 if (env->apic_state) {
197 CPU_FOREACH(cs) {
198 cpu = X86_CPU(cs);
199 env = &cpu->env;
200 if (apic_accept_pic_intr(env->apic_state)) {
201 apic_deliver_pic_intr(env->apic_state, level);
202 }
203 }
204 } else {
205 if (level) {
206 cpu_interrupt(cs, CPU_INTERRUPT_HARD);
207 } else {
208 cpu_reset_interrupt(cs, CPU_INTERRUPT_HARD);
209 }
210 }
211 }
212
213 /* PC cmos mappings */
214
215 #define REG_EQUIPMENT_BYTE 0x14
216
217 static int cmos_get_fd_drive_type(FDriveType fd0)
218 {
219 int val;
220
221 switch (fd0) {
222 case FDRIVE_DRV_144:
223 /* 1.44 Mb 3"5 drive */
224 val = 4;
225 break;
226 case FDRIVE_DRV_288:
227 /* 2.88 Mb 3"5 drive */
228 val = 5;
229 break;
230 case FDRIVE_DRV_120:
231 /* 1.2 Mb 5"5 drive */
232 val = 2;
233 break;
234 case FDRIVE_DRV_NONE:
235 default:
236 val = 0;
237 break;
238 }
239 return val;
240 }
241
242 static void cmos_init_hd(ISADevice *s, int type_ofs, int info_ofs,
243 int16_t cylinders, int8_t heads, int8_t sectors)
244 {
245 rtc_set_memory(s, type_ofs, 47);
246 rtc_set_memory(s, info_ofs, cylinders);
247 rtc_set_memory(s, info_ofs + 1, cylinders >> 8);
248 rtc_set_memory(s, info_ofs + 2, heads);
249 rtc_set_memory(s, info_ofs + 3, 0xff);
250 rtc_set_memory(s, info_ofs + 4, 0xff);
251 rtc_set_memory(s, info_ofs + 5, 0xc0 | ((heads > 8) << 3));
252 rtc_set_memory(s, info_ofs + 6, cylinders);
253 rtc_set_memory(s, info_ofs + 7, cylinders >> 8);
254 rtc_set_memory(s, info_ofs + 8, sectors);
255 }
256
257 /* convert boot_device letter to something recognizable by the bios */
258 static int boot_device2nibble(char boot_device)
259 {
260 switch(boot_device) {
261 case 'a':
262 case 'b':
263 return 0x01; /* floppy boot */
264 case 'c':
265 return 0x02; /* hard drive boot */
266 case 'd':
267 return 0x03; /* CD-ROM boot */
268 case 'n':
269 return 0x04; /* Network boot */
270 }
271 return 0;
272 }
273
274 static int set_boot_dev(ISADevice *s, const char *boot_device)
275 {
276 #define PC_MAX_BOOT_DEVICES 3
277 int nbds, bds[3] = { 0, };
278 int i;
279
280 nbds = strlen(boot_device);
281 if (nbds > PC_MAX_BOOT_DEVICES) {
282 error_report("Too many boot devices for PC");
283 return(1);
284 }
285 for (i = 0; i < nbds; i++) {
286 bds[i] = boot_device2nibble(boot_device[i]);
287 if (bds[i] == 0) {
288 error_report("Invalid boot device for PC: '%c'",
289 boot_device[i]);
290 return(1);
291 }
292 }
293 rtc_set_memory(s, 0x3d, (bds[1] << 4) | bds[0]);
294 rtc_set_memory(s, 0x38, (bds[2] << 4) | (fd_bootchk ? 0x0 : 0x1));
295 return(0);
296 }
297
298 static int pc_boot_set(void *opaque, const char *boot_device)
299 {
300 return set_boot_dev(opaque, boot_device);
301 }
302
303 typedef struct pc_cmos_init_late_arg {
304 ISADevice *rtc_state;
305 BusState *idebus[2];
306 } pc_cmos_init_late_arg;
307
308 static void pc_cmos_init_late(void *opaque)
309 {
310 pc_cmos_init_late_arg *arg = opaque;
311 ISADevice *s = arg->rtc_state;
312 int16_t cylinders;
313 int8_t heads, sectors;
314 int val;
315 int i, trans;
316
317 val = 0;
318 if (ide_get_geometry(arg->idebus[0], 0,
319 &cylinders, &heads, &sectors) >= 0) {
320 cmos_init_hd(s, 0x19, 0x1b, cylinders, heads, sectors);
321 val |= 0xf0;
322 }
323 if (ide_get_geometry(arg->idebus[0], 1,
324 &cylinders, &heads, &sectors) >= 0) {
325 cmos_init_hd(s, 0x1a, 0x24, cylinders, heads, sectors);
326 val |= 0x0f;
327 }
328 rtc_set_memory(s, 0x12, val);
329
330 val = 0;
331 for (i = 0; i < 4; i++) {
332 /* NOTE: ide_get_geometry() returns the physical
333 geometry. It is always such that: 1 <= sects <= 63, 1
334 <= heads <= 16, 1 <= cylinders <= 16383. The BIOS
335 geometry can be different if a translation is done. */
336 if (ide_get_geometry(arg->idebus[i / 2], i % 2,
337 &cylinders, &heads, &sectors) >= 0) {
338 trans = ide_get_bios_chs_trans(arg->idebus[i / 2], i % 2) - 1;
339 assert((trans & ~3) == 0);
340 val |= trans << (i * 2);
341 }
342 }
343 rtc_set_memory(s, 0x39, val);
344
345 qemu_unregister_reset(pc_cmos_init_late, opaque);
346 }
347
348 typedef struct RTCCPUHotplugArg {
349 Notifier cpu_added_notifier;
350 ISADevice *rtc_state;
351 } RTCCPUHotplugArg;
352
353 static void rtc_notify_cpu_added(Notifier *notifier, void *data)
354 {
355 RTCCPUHotplugArg *arg = container_of(notifier, RTCCPUHotplugArg,
356 cpu_added_notifier);
357 ISADevice *s = arg->rtc_state;
358
359 /* increment the number of CPUs */
360 rtc_set_memory(s, 0x5f, rtc_get_memory(s, 0x5f) + 1);
361 }
362
363 void pc_cmos_init(ram_addr_t ram_size, ram_addr_t above_4g_mem_size,
364 const char *boot_device,
365 ISADevice *floppy, BusState *idebus0, BusState *idebus1,
366 ISADevice *s)
367 {
368 int val, nb, i;
369 FDriveType fd_type[2] = { FDRIVE_DRV_NONE, FDRIVE_DRV_NONE };
370 static pc_cmos_init_late_arg arg;
371 static RTCCPUHotplugArg cpu_hotplug_cb;
372
373 /* various important CMOS locations needed by PC/Bochs bios */
374
375 /* memory size */
376 /* base memory (first MiB) */
377 val = MIN(ram_size / 1024, 640);
378 rtc_set_memory(s, 0x15, val);
379 rtc_set_memory(s, 0x16, val >> 8);
380 /* extended memory (next 64MiB) */
381 if (ram_size > 1024 * 1024) {
382 val = (ram_size - 1024 * 1024) / 1024;
383 } else {
384 val = 0;
385 }
386 if (val > 65535)
387 val = 65535;
388 rtc_set_memory(s, 0x17, val);
389 rtc_set_memory(s, 0x18, val >> 8);
390 rtc_set_memory(s, 0x30, val);
391 rtc_set_memory(s, 0x31, val >> 8);
392 /* memory between 16MiB and 4GiB */
393 if (ram_size > 16 * 1024 * 1024) {
394 val = (ram_size - 16 * 1024 * 1024) / 65536;
395 } else {
396 val = 0;
397 }
398 if (val > 65535)
399 val = 65535;
400 rtc_set_memory(s, 0x34, val);
401 rtc_set_memory(s, 0x35, val >> 8);
402 /* memory above 4GiB */
403 val = above_4g_mem_size / 65536;
404 rtc_set_memory(s, 0x5b, val);
405 rtc_set_memory(s, 0x5c, val >> 8);
406 rtc_set_memory(s, 0x5d, val >> 16);
407
408 /* set the number of CPU */
409 rtc_set_memory(s, 0x5f, smp_cpus - 1);
410 /* init CPU hotplug notifier */
411 cpu_hotplug_cb.rtc_state = s;
412 cpu_hotplug_cb.cpu_added_notifier.notify = rtc_notify_cpu_added;
413 qemu_register_cpu_added_notifier(&cpu_hotplug_cb.cpu_added_notifier);
414
415 if (set_boot_dev(s, boot_device)) {
416 exit(1);
417 }
418
419 /* floppy type */
420 if (floppy) {
421 for (i = 0; i < 2; i++) {
422 fd_type[i] = isa_fdc_get_drive_type(floppy, i);
423 }
424 }
425 val = (cmos_get_fd_drive_type(fd_type[0]) << 4) |
426 cmos_get_fd_drive_type(fd_type[1]);
427 rtc_set_memory(s, 0x10, val);
428
429 val = 0;
430 nb = 0;
431 if (fd_type[0] < FDRIVE_DRV_NONE) {
432 nb++;
433 }
434 if (fd_type[1] < FDRIVE_DRV_NONE) {
435 nb++;
436 }
437 switch (nb) {
438 case 0:
439 break;
440 case 1:
441 val |= 0x01; /* 1 drive, ready for boot */
442 break;
443 case 2:
444 val |= 0x41; /* 2 drives, ready for boot */
445 break;
446 }
447 val |= 0x02; /* FPU is there */
448 val |= 0x04; /* PS/2 mouse installed */
449 rtc_set_memory(s, REG_EQUIPMENT_BYTE, val);
450
451 /* hard drives */
452 arg.rtc_state = s;
453 arg.idebus[0] = idebus0;
454 arg.idebus[1] = idebus1;
455 qemu_register_reset(pc_cmos_init_late, &arg);
456 }
457
458 #define TYPE_PORT92 "port92"
459 #define PORT92(obj) OBJECT_CHECK(Port92State, (obj), TYPE_PORT92)
460
461 /* port 92 stuff: could be split off */
462 typedef struct Port92State {
463 ISADevice parent_obj;
464
465 MemoryRegion io;
466 uint8_t outport;
467 qemu_irq *a20_out;
468 } Port92State;
469
470 static void port92_write(void *opaque, hwaddr addr, uint64_t val,
471 unsigned size)
472 {
473 Port92State *s = opaque;
474
475 DPRINTF("port92: write 0x%02x\n", val);
476 s->outport = val;
477 qemu_set_irq(*s->a20_out, (val >> 1) & 1);
478 if (val & 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 .minimum_version_id_old = 1,
506 .fields = (VMStateField []) {
507 VMSTATE_UINT8(outport, Port92State),
508 VMSTATE_END_OF_LIST()
509 }
510 };
511
512 static void port92_reset(DeviceState *d)
513 {
514 Port92State *s = PORT92(d);
515
516 s->outport &= ~1;
517 }
518
519 static const MemoryRegionOps port92_ops = {
520 .read = port92_read,
521 .write = port92_write,
522 .impl = {
523 .min_access_size = 1,
524 .max_access_size = 1,
525 },
526 .endianness = DEVICE_LITTLE_ENDIAN,
527 };
528
529 static void port92_initfn(Object *obj)
530 {
531 Port92State *s = PORT92(obj);
532
533 memory_region_init_io(&s->io, OBJECT(s), &port92_ops, s, "port92", 1);
534
535 s->outport = 0;
536 }
537
538 static void port92_realizefn(DeviceState *dev, Error **errp)
539 {
540 ISADevice *isadev = ISA_DEVICE(dev);
541 Port92State *s = PORT92(dev);
542
543 isa_register_ioport(isadev, &s->io, 0x92);
544 }
545
546 static void port92_class_initfn(ObjectClass *klass, void *data)
547 {
548 DeviceClass *dc = DEVICE_CLASS(klass);
549
550 dc->no_user = 1;
551 dc->realize = port92_realizefn;
552 dc->reset = port92_reset;
553 dc->vmsd = &vmstate_port92_isa;
554 }
555
556 static const TypeInfo port92_info = {
557 .name = TYPE_PORT92,
558 .parent = TYPE_ISA_DEVICE,
559 .instance_size = sizeof(Port92State),
560 .instance_init = port92_initfn,
561 .class_init = port92_class_initfn,
562 };
563
564 static void port92_register_types(void)
565 {
566 type_register_static(&port92_info);
567 }
568
569 type_init(port92_register_types)
570
571 static void handle_a20_line_change(void *opaque, int irq, int level)
572 {
573 X86CPU *cpu = opaque;
574
575 /* XXX: send to all CPUs ? */
576 /* XXX: add logic to handle multiple A20 line sources */
577 x86_cpu_set_a20(cpu, level);
578 }
579
580 int e820_add_entry(uint64_t address, uint64_t length, uint32_t type)
581 {
582 int index = le32_to_cpu(e820_reserve.count);
583 struct e820_entry *entry;
584
585 if (type != E820_RAM) {
586 /* old FW_CFG_E820_TABLE entry -- reservations only */
587 if (index >= E820_NR_ENTRIES) {
588 return -EBUSY;
589 }
590 entry = &e820_reserve.entry[index++];
591
592 entry->address = cpu_to_le64(address);
593 entry->length = cpu_to_le64(length);
594 entry->type = cpu_to_le32(type);
595
596 e820_reserve.count = cpu_to_le32(index);
597 }
598
599 /* new "etc/e820" file -- include ram too */
600 e820_table = g_realloc(e820_table,
601 sizeof(struct e820_entry) * (e820_entries+1));
602 e820_table[e820_entries].address = cpu_to_le64(address);
603 e820_table[e820_entries].length = cpu_to_le64(length);
604 e820_table[e820_entries].type = cpu_to_le32(type);
605 e820_entries++;
606
607 return e820_entries;
608 }
609
610 /* Calculates the limit to CPU APIC ID values
611 *
612 * This function returns the limit for the APIC ID value, so that all
613 * CPU APIC IDs are < pc_apic_id_limit().
614 *
615 * This is used for FW_CFG_MAX_CPUS. See comments on bochs_bios_init().
616 */
617 static unsigned int pc_apic_id_limit(unsigned int max_cpus)
618 {
619 return x86_cpu_apic_id_from_index(max_cpus - 1) + 1;
620 }
621
622 static FWCfgState *bochs_bios_init(void)
623 {
624 FWCfgState *fw_cfg;
625 uint8_t *smbios_table;
626 size_t smbios_len;
627 uint64_t *numa_fw_cfg;
628 int i, j;
629 unsigned int apic_id_limit = pc_apic_id_limit(max_cpus);
630
631 fw_cfg = fw_cfg_init(BIOS_CFG_IOPORT, BIOS_CFG_IOPORT + 1, 0, 0);
632 /* FW_CFG_MAX_CPUS is a bit confusing/problematic on x86:
633 *
634 * SeaBIOS needs FW_CFG_MAX_CPUS for CPU hotplug, but the CPU hotplug
635 * QEMU<->SeaBIOS interface is not based on the "CPU index", but on the APIC
636 * ID of hotplugged CPUs[1]. This means that FW_CFG_MAX_CPUS is not the
637 * "maximum number of CPUs", but the "limit to the APIC ID values SeaBIOS
638 * may see".
639 *
640 * So, this means we must not use max_cpus, here, but the maximum possible
641 * APIC ID value, plus one.
642 *
643 * [1] The only kind of "CPU identifier" used between SeaBIOS and QEMU is
644 * the APIC ID, not the "CPU index"
645 */
646 fw_cfg_add_i16(fw_cfg, FW_CFG_MAX_CPUS, (uint16_t)apic_id_limit);
647 fw_cfg_add_i32(fw_cfg, FW_CFG_ID, 1);
648 fw_cfg_add_i64(fw_cfg, FW_CFG_RAM_SIZE, (uint64_t)ram_size);
649 fw_cfg_add_bytes(fw_cfg, FW_CFG_ACPI_TABLES,
650 acpi_tables, acpi_tables_len);
651 fw_cfg_add_i32(fw_cfg, FW_CFG_IRQ0_OVERRIDE, kvm_allows_irq0_override());
652
653 smbios_table = smbios_get_table(&smbios_len);
654 if (smbios_table)
655 fw_cfg_add_bytes(fw_cfg, FW_CFG_SMBIOS_ENTRIES,
656 smbios_table, smbios_len);
657 fw_cfg_add_bytes(fw_cfg, FW_CFG_E820_TABLE,
658 &e820_reserve, sizeof(e820_reserve));
659 fw_cfg_add_file(fw_cfg, "etc/e820", e820_table,
660 sizeof(struct e820_entry) * e820_entries);
661
662 fw_cfg_add_bytes(fw_cfg, FW_CFG_HPET, &hpet_cfg, sizeof(hpet_cfg));
663 /* allocate memory for the NUMA channel: one (64bit) word for the number
664 * of nodes, one word for each VCPU->node and one word for each node to
665 * hold the amount of memory.
666 */
667 numa_fw_cfg = g_new0(uint64_t, 1 + apic_id_limit + nb_numa_nodes);
668 numa_fw_cfg[0] = cpu_to_le64(nb_numa_nodes);
669 for (i = 0; i < max_cpus; i++) {
670 unsigned int apic_id = x86_cpu_apic_id_from_index(i);
671 assert(apic_id < apic_id_limit);
672 for (j = 0; j < nb_numa_nodes; j++) {
673 if (test_bit(i, node_cpumask[j])) {
674 numa_fw_cfg[apic_id + 1] = cpu_to_le64(j);
675 break;
676 }
677 }
678 }
679 for (i = 0; i < nb_numa_nodes; i++) {
680 numa_fw_cfg[apic_id_limit + 1 + i] = cpu_to_le64(node_mem[i]);
681 }
682 fw_cfg_add_bytes(fw_cfg, FW_CFG_NUMA, numa_fw_cfg,
683 (1 + apic_id_limit + nb_numa_nodes) *
684 sizeof(*numa_fw_cfg));
685
686 return fw_cfg;
687 }
688
689 static long get_file_size(FILE *f)
690 {
691 long where, size;
692
693 /* XXX: on Unix systems, using fstat() probably makes more sense */
694
695 where = ftell(f);
696 fseek(f, 0, SEEK_END);
697 size = ftell(f);
698 fseek(f, where, SEEK_SET);
699
700 return size;
701 }
702
703 static void load_linux(FWCfgState *fw_cfg,
704 const char *kernel_filename,
705 const char *initrd_filename,
706 const char *kernel_cmdline,
707 hwaddr max_ram_size)
708 {
709 uint16_t protocol;
710 int setup_size, kernel_size, initrd_size = 0, cmdline_size;
711 uint32_t initrd_max;
712 uint8_t header[8192], *setup, *kernel, *initrd_data;
713 hwaddr real_addr, prot_addr, cmdline_addr, initrd_addr = 0;
714 FILE *f;
715 char *vmode;
716
717 /* Align to 16 bytes as a paranoia measure */
718 cmdline_size = (strlen(kernel_cmdline)+16) & ~15;
719
720 /* load the kernel header */
721 f = fopen(kernel_filename, "rb");
722 if (!f || !(kernel_size = get_file_size(f)) ||
723 fread(header, 1, MIN(ARRAY_SIZE(header), kernel_size), f) !=
724 MIN(ARRAY_SIZE(header), kernel_size)) {
725 fprintf(stderr, "qemu: could not load kernel '%s': %s\n",
726 kernel_filename, strerror(errno));
727 exit(1);
728 }
729
730 /* kernel protocol version */
731 #if 0
732 fprintf(stderr, "header magic: %#x\n", ldl_p(header+0x202));
733 #endif
734 if (ldl_p(header+0x202) == 0x53726448) {
735 protocol = lduw_p(header+0x206);
736 } else {
737 /* This looks like a multiboot kernel. If it is, let's stop
738 treating it like a Linux kernel. */
739 if (load_multiboot(fw_cfg, f, kernel_filename, initrd_filename,
740 kernel_cmdline, kernel_size, header)) {
741 return;
742 }
743 protocol = 0;
744 }
745
746 if (protocol < 0x200 || !(header[0x211] & 0x01)) {
747 /* Low kernel */
748 real_addr = 0x90000;
749 cmdline_addr = 0x9a000 - cmdline_size;
750 prot_addr = 0x10000;
751 } else if (protocol < 0x202) {
752 /* High but ancient kernel */
753 real_addr = 0x90000;
754 cmdline_addr = 0x9a000 - cmdline_size;
755 prot_addr = 0x100000;
756 } else {
757 /* High and recent kernel */
758 real_addr = 0x10000;
759 cmdline_addr = 0x20000;
760 prot_addr = 0x100000;
761 }
762
763 #if 0
764 fprintf(stderr,
765 "qemu: real_addr = 0x" TARGET_FMT_plx "\n"
766 "qemu: cmdline_addr = 0x" TARGET_FMT_plx "\n"
767 "qemu: prot_addr = 0x" TARGET_FMT_plx "\n",
768 real_addr,
769 cmdline_addr,
770 prot_addr);
771 #endif
772
773 /* highest address for loading the initrd */
774 if (protocol >= 0x203) {
775 initrd_max = ldl_p(header+0x22c);
776 } else {
777 initrd_max = 0x37ffffff;
778 }
779
780 if (initrd_max >= max_ram_size-ACPI_DATA_SIZE)
781 initrd_max = max_ram_size-ACPI_DATA_SIZE-1;
782
783 fw_cfg_add_i32(fw_cfg, FW_CFG_CMDLINE_ADDR, cmdline_addr);
784 fw_cfg_add_i32(fw_cfg, FW_CFG_CMDLINE_SIZE, strlen(kernel_cmdline)+1);
785 fw_cfg_add_string(fw_cfg, FW_CFG_CMDLINE_DATA, kernel_cmdline);
786
787 if (protocol >= 0x202) {
788 stl_p(header+0x228, cmdline_addr);
789 } else {
790 stw_p(header+0x20, 0xA33F);
791 stw_p(header+0x22, cmdline_addr-real_addr);
792 }
793
794 /* handle vga= parameter */
795 vmode = strstr(kernel_cmdline, "vga=");
796 if (vmode) {
797 unsigned int video_mode;
798 /* skip "vga=" */
799 vmode += 4;
800 if (!strncmp(vmode, "normal", 6)) {
801 video_mode = 0xffff;
802 } else if (!strncmp(vmode, "ext", 3)) {
803 video_mode = 0xfffe;
804 } else if (!strncmp(vmode, "ask", 3)) {
805 video_mode = 0xfffd;
806 } else {
807 video_mode = strtol(vmode, NULL, 0);
808 }
809 stw_p(header+0x1fa, video_mode);
810 }
811
812 /* loader type */
813 /* High nybble = B reserved for QEMU; low nybble is revision number.
814 If this code is substantially changed, you may want to consider
815 incrementing the revision. */
816 if (protocol >= 0x200) {
817 header[0x210] = 0xB0;
818 }
819 /* heap */
820 if (protocol >= 0x201) {
821 header[0x211] |= 0x80; /* CAN_USE_HEAP */
822 stw_p(header+0x224, cmdline_addr-real_addr-0x200);
823 }
824
825 /* load initrd */
826 if (initrd_filename) {
827 if (protocol < 0x200) {
828 fprintf(stderr, "qemu: linux kernel too old to load a ram disk\n");
829 exit(1);
830 }
831
832 initrd_size = get_image_size(initrd_filename);
833 if (initrd_size < 0) {
834 fprintf(stderr, "qemu: error reading initrd %s\n",
835 initrd_filename);
836 exit(1);
837 }
838
839 initrd_addr = (initrd_max-initrd_size) & ~4095;
840
841 initrd_data = g_malloc(initrd_size);
842 load_image(initrd_filename, initrd_data);
843
844 fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_ADDR, initrd_addr);
845 fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_SIZE, initrd_size);
846 fw_cfg_add_bytes(fw_cfg, FW_CFG_INITRD_DATA, initrd_data, initrd_size);
847
848 stl_p(header+0x218, initrd_addr);
849 stl_p(header+0x21c, initrd_size);
850 }
851
852 /* load kernel and setup */
853 setup_size = header[0x1f1];
854 if (setup_size == 0) {
855 setup_size = 4;
856 }
857 setup_size = (setup_size+1)*512;
858 kernel_size -= setup_size;
859
860 setup = g_malloc(setup_size);
861 kernel = g_malloc(kernel_size);
862 fseek(f, 0, SEEK_SET);
863 if (fread(setup, 1, setup_size, f) != setup_size) {
864 fprintf(stderr, "fread() failed\n");
865 exit(1);
866 }
867 if (fread(kernel, 1, kernel_size, f) != kernel_size) {
868 fprintf(stderr, "fread() failed\n");
869 exit(1);
870 }
871 fclose(f);
872 memcpy(setup, header, MIN(sizeof(header), setup_size));
873
874 fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_ADDR, prot_addr);
875 fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_SIZE, kernel_size);
876 fw_cfg_add_bytes(fw_cfg, FW_CFG_KERNEL_DATA, kernel, kernel_size);
877
878 fw_cfg_add_i32(fw_cfg, FW_CFG_SETUP_ADDR, real_addr);
879 fw_cfg_add_i32(fw_cfg, FW_CFG_SETUP_SIZE, setup_size);
880 fw_cfg_add_bytes(fw_cfg, FW_CFG_SETUP_DATA, setup, setup_size);
881
882 option_rom[nb_option_roms].name = "linuxboot.bin";
883 option_rom[nb_option_roms].bootindex = 0;
884 nb_option_roms++;
885 }
886
887 #define NE2000_NB_MAX 6
888
889 static const int ne2000_io[NE2000_NB_MAX] = { 0x300, 0x320, 0x340, 0x360,
890 0x280, 0x380 };
891 static const int ne2000_irq[NE2000_NB_MAX] = { 9, 10, 11, 3, 4, 5 };
892
893 static const int parallel_io[MAX_PARALLEL_PORTS] = { 0x378, 0x278, 0x3bc };
894 static const int parallel_irq[MAX_PARALLEL_PORTS] = { 7, 7, 7 };
895
896 void pc_init_ne2k_isa(ISABus *bus, NICInfo *nd)
897 {
898 static int nb_ne2k = 0;
899
900 if (nb_ne2k == NE2000_NB_MAX)
901 return;
902 isa_ne2000_init(bus, ne2000_io[nb_ne2k],
903 ne2000_irq[nb_ne2k], nd);
904 nb_ne2k++;
905 }
906
907 DeviceState *cpu_get_current_apic(void)
908 {
909 if (current_cpu) {
910 X86CPU *cpu = X86_CPU(current_cpu);
911 return cpu->env.apic_state;
912 } else {
913 return NULL;
914 }
915 }
916
917 void pc_acpi_smi_interrupt(void *opaque, int irq, int level)
918 {
919 X86CPU *cpu = opaque;
920
921 if (level) {
922 cpu_interrupt(CPU(cpu), CPU_INTERRUPT_SMI);
923 }
924 }
925
926 static X86CPU *pc_new_cpu(const char *cpu_model, int64_t apic_id,
927 DeviceState *icc_bridge, Error **errp)
928 {
929 X86CPU *cpu;
930 Error *local_err = NULL;
931
932 cpu = cpu_x86_create(cpu_model, icc_bridge, &local_err);
933 if (local_err != NULL) {
934 error_propagate(errp, local_err);
935 return NULL;
936 }
937
938 object_property_set_int(OBJECT(cpu), apic_id, "apic-id", &local_err);
939 object_property_set_bool(OBJECT(cpu), true, "realized", &local_err);
940
941 if (local_err) {
942 error_propagate(errp, local_err);
943 object_unref(OBJECT(cpu));
944 cpu = NULL;
945 }
946 return cpu;
947 }
948
949 static const char *current_cpu_model;
950
951 void pc_hot_add_cpu(const int64_t id, Error **errp)
952 {
953 DeviceState *icc_bridge;
954 int64_t apic_id = x86_cpu_apic_id_from_index(id);
955
956 if (id < 0) {
957 error_setg(errp, "Invalid CPU id: %" PRIi64, id);
958 return;
959 }
960
961 if (cpu_exists(apic_id)) {
962 error_setg(errp, "Unable to add CPU: %" PRIi64
963 ", it already exists", id);
964 return;
965 }
966
967 if (id >= max_cpus) {
968 error_setg(errp, "Unable to add CPU: %" PRIi64
969 ", max allowed: %d", id, max_cpus - 1);
970 return;
971 }
972
973 icc_bridge = DEVICE(object_resolve_path_type("icc-bridge",
974 TYPE_ICC_BRIDGE, NULL));
975 pc_new_cpu(current_cpu_model, apic_id, icc_bridge, errp);
976 }
977
978 void pc_cpus_init(const char *cpu_model, DeviceState *icc_bridge)
979 {
980 int i;
981 X86CPU *cpu = NULL;
982 Error *error = NULL;
983
984 /* init CPUs */
985 if (cpu_model == NULL) {
986 #ifdef TARGET_X86_64
987 cpu_model = "qemu64";
988 #else
989 cpu_model = "qemu32";
990 #endif
991 }
992 current_cpu_model = cpu_model;
993
994 for (i = 0; i < smp_cpus; i++) {
995 cpu = pc_new_cpu(cpu_model, x86_cpu_apic_id_from_index(i),
996 icc_bridge, &error);
997 if (error) {
998 error_report("%s", error_get_pretty(error));
999 error_free(error);
1000 exit(1);
1001 }
1002 }
1003
1004 /* map APIC MMIO area if CPU has APIC */
1005 if (cpu && cpu->env.apic_state) {
1006 /* XXX: what if the base changes? */
1007 sysbus_mmio_map_overlap(SYS_BUS_DEVICE(icc_bridge), 0,
1008 APIC_DEFAULT_ADDRESS, 0x1000);
1009 }
1010 }
1011
1012 /* pci-info ROM file. Little endian format */
1013 typedef struct PcRomPciInfo {
1014 uint64_t w32_min;
1015 uint64_t w32_max;
1016 uint64_t w64_min;
1017 uint64_t w64_max;
1018 } PcRomPciInfo;
1019
1020 static void pc_fw_cfg_guest_info(PcGuestInfo *guest_info)
1021 {
1022 PcRomPciInfo *info;
1023 Object *pci_info;
1024 bool ambiguous = false;
1025
1026 if (!guest_info->has_pci_info || !guest_info->fw_cfg) {
1027 return;
1028 }
1029 pci_info = object_resolve_path_type("", TYPE_PCI_HOST_BRIDGE, &ambiguous);
1030 g_assert(!ambiguous);
1031 if (!pci_info) {
1032 return;
1033 }
1034
1035 info = g_malloc(sizeof *info);
1036 info->w32_min = cpu_to_le64(object_property_get_int(pci_info,
1037 PCI_HOST_PROP_PCI_HOLE_START, NULL));
1038 info->w32_max = cpu_to_le64(object_property_get_int(pci_info,
1039 PCI_HOST_PROP_PCI_HOLE_END, NULL));
1040 info->w64_min = cpu_to_le64(object_property_get_int(pci_info,
1041 PCI_HOST_PROP_PCI_HOLE64_START, NULL));
1042 info->w64_max = cpu_to_le64(object_property_get_int(pci_info,
1043 PCI_HOST_PROP_PCI_HOLE64_END, NULL));
1044 /* Pass PCI hole info to guest via a side channel.
1045 * Required so guest PCI enumeration does the right thing. */
1046 fw_cfg_add_file(guest_info->fw_cfg, "etc/pci-info", info, sizeof *info);
1047 }
1048
1049 typedef struct PcGuestInfoState {
1050 PcGuestInfo info;
1051 Notifier machine_done;
1052 } PcGuestInfoState;
1053
1054 static
1055 void pc_guest_info_machine_done(Notifier *notifier, void *data)
1056 {
1057 PcGuestInfoState *guest_info_state = container_of(notifier,
1058 PcGuestInfoState,
1059 machine_done);
1060 pc_fw_cfg_guest_info(&guest_info_state->info);
1061 acpi_setup(&guest_info_state->info);
1062 }
1063
1064 PcGuestInfo *pc_guest_info_init(ram_addr_t below_4g_mem_size,
1065 ram_addr_t above_4g_mem_size)
1066 {
1067 PcGuestInfoState *guest_info_state = g_malloc0(sizeof *guest_info_state);
1068 PcGuestInfo *guest_info = &guest_info_state->info;
1069 int i, j;
1070
1071 guest_info->ram_size = below_4g_mem_size + above_4g_mem_size;
1072 guest_info->apic_id_limit = pc_apic_id_limit(max_cpus);
1073 guest_info->apic_xrupt_override = kvm_allows_irq0_override();
1074 guest_info->numa_nodes = nb_numa_nodes;
1075 guest_info->node_mem = g_memdup(node_mem, guest_info->numa_nodes *
1076 sizeof *guest_info->node_mem);
1077 guest_info->node_cpu = g_malloc0(guest_info->apic_id_limit *
1078 sizeof *guest_info->node_cpu);
1079
1080 for (i = 0; i < max_cpus; i++) {
1081 unsigned int apic_id = x86_cpu_apic_id_from_index(i);
1082 assert(apic_id < guest_info->apic_id_limit);
1083 for (j = 0; j < nb_numa_nodes; j++) {
1084 if (test_bit(i, node_cpumask[j])) {
1085 guest_info->node_cpu[apic_id] = j;
1086 break;
1087 }
1088 }
1089 }
1090
1091 guest_info_state->machine_done.notify = pc_guest_info_machine_done;
1092 qemu_add_machine_init_done_notifier(&guest_info_state->machine_done);
1093 return guest_info;
1094 }
1095
1096 void pc_init_pci64_hole(PcPciInfo *pci_info, uint64_t pci_hole64_start,
1097 uint64_t pci_hole64_size)
1098 {
1099 if ((sizeof(hwaddr) == 4) || (!pci_hole64_size)) {
1100 return;
1101 }
1102 /*
1103 * BIOS does not set MTRR entries for the 64 bit window, so no need to
1104 * align address to power of two. Align address at 1G, this makes sure
1105 * it can be exactly covered with a PAT entry even when using huge
1106 * pages.
1107 */
1108 pci_info->w64.begin = ROUND_UP(pci_hole64_start, 0x1ULL << 30);
1109 pci_info->w64.end = pci_info->w64.begin + pci_hole64_size;
1110 assert(pci_info->w64.begin <= pci_info->w64.end);
1111 }
1112
1113 void pc_acpi_init(const char *default_dsdt)
1114 {
1115 char *filename;
1116
1117 if (acpi_tables != NULL) {
1118 /* manually set via -acpitable, leave it alone */
1119 return;
1120 }
1121
1122 filename = qemu_find_file(QEMU_FILE_TYPE_BIOS, default_dsdt);
1123 if (filename == NULL) {
1124 fprintf(stderr, "WARNING: failed to find %s\n", default_dsdt);
1125 } else {
1126 char *arg;
1127 QemuOpts *opts;
1128 Error *err = NULL;
1129
1130 arg = g_strdup_printf("file=%s", filename);
1131
1132 /* creates a deep copy of "arg" */
1133 opts = qemu_opts_parse(qemu_find_opts("acpi"), arg, 0);
1134 g_assert(opts != NULL);
1135
1136 acpi_table_add_builtin(opts, &err);
1137 if (err) {
1138 error_report("WARNING: failed to load %s: %s", filename,
1139 error_get_pretty(err));
1140 error_free(err);
1141 }
1142 g_free(arg);
1143 g_free(filename);
1144 }
1145 }
1146
1147 FWCfgState *pc_memory_init(MemoryRegion *system_memory,
1148 const char *kernel_filename,
1149 const char *kernel_cmdline,
1150 const char *initrd_filename,
1151 ram_addr_t below_4g_mem_size,
1152 ram_addr_t above_4g_mem_size,
1153 MemoryRegion *rom_memory,
1154 MemoryRegion **ram_memory,
1155 PcGuestInfo *guest_info)
1156 {
1157 int linux_boot, i;
1158 MemoryRegion *ram, *option_rom_mr;
1159 MemoryRegion *ram_below_4g, *ram_above_4g;
1160 FWCfgState *fw_cfg;
1161
1162 linux_boot = (kernel_filename != NULL);
1163
1164 /* Allocate RAM. We allocate it as a single memory region and use
1165 * aliases to address portions of it, mostly for backwards compatibility
1166 * with older qemus that used qemu_ram_alloc().
1167 */
1168 ram = g_malloc(sizeof(*ram));
1169 memory_region_init_ram(ram, NULL, "pc.ram",
1170 below_4g_mem_size + above_4g_mem_size);
1171 vmstate_register_ram_global(ram);
1172 *ram_memory = ram;
1173 ram_below_4g = g_malloc(sizeof(*ram_below_4g));
1174 memory_region_init_alias(ram_below_4g, NULL, "ram-below-4g", ram,
1175 0, below_4g_mem_size);
1176 memory_region_add_subregion(system_memory, 0, ram_below_4g);
1177 if (0) {
1178 /*
1179 * Ideally we should do that too, but that would ruin the e820
1180 * reservations added by seabios before initializing fw_cfg.
1181 */
1182 e820_add_entry(0, below_4g_mem_size, E820_RAM);
1183 }
1184 if (above_4g_mem_size > 0) {
1185 ram_above_4g = g_malloc(sizeof(*ram_above_4g));
1186 memory_region_init_alias(ram_above_4g, NULL, "ram-above-4g", ram,
1187 below_4g_mem_size, above_4g_mem_size);
1188 memory_region_add_subregion(system_memory, 0x100000000ULL,
1189 ram_above_4g);
1190 e820_add_entry(0x100000000ULL, above_4g_mem_size, E820_RAM);
1191 }
1192
1193
1194 /* Initialize PC system firmware */
1195 pc_system_firmware_init(rom_memory, guest_info->isapc_ram_fw);
1196
1197 option_rom_mr = g_malloc(sizeof(*option_rom_mr));
1198 memory_region_init_ram(option_rom_mr, NULL, "pc.rom", PC_ROM_SIZE);
1199 vmstate_register_ram_global(option_rom_mr);
1200 memory_region_add_subregion_overlap(rom_memory,
1201 PC_ROM_MIN_VGA,
1202 option_rom_mr,
1203 1);
1204
1205 fw_cfg = bochs_bios_init();
1206 rom_set_fw(fw_cfg);
1207
1208 if (linux_boot) {
1209 load_linux(fw_cfg, kernel_filename, initrd_filename, kernel_cmdline, below_4g_mem_size);
1210 }
1211
1212 for (i = 0; i < nb_option_roms; i++) {
1213 rom_add_option(option_rom[i].name, option_rom[i].bootindex);
1214 }
1215 guest_info->fw_cfg = fw_cfg;
1216 return fw_cfg;
1217 }
1218
1219 qemu_irq *pc_allocate_cpu_irq(void)
1220 {
1221 return qemu_allocate_irqs(pic_irq_request, NULL, 1);
1222 }
1223
1224 DeviceState *pc_vga_init(ISABus *isa_bus, PCIBus *pci_bus)
1225 {
1226 DeviceState *dev = NULL;
1227
1228 if (pci_bus) {
1229 PCIDevice *pcidev = pci_vga_init(pci_bus);
1230 dev = pcidev ? &pcidev->qdev : NULL;
1231 } else if (isa_bus) {
1232 ISADevice *isadev = isa_vga_init(isa_bus);
1233 dev = isadev ? DEVICE(isadev) : NULL;
1234 }
1235 return dev;
1236 }
1237
1238 static void cpu_request_exit(void *opaque, int irq, int level)
1239 {
1240 CPUState *cpu = current_cpu;
1241
1242 if (cpu && level) {
1243 cpu_exit(cpu);
1244 }
1245 }
1246
1247 static const MemoryRegionOps ioport80_io_ops = {
1248 .write = ioport80_write,
1249 .read = ioport80_read,
1250 .endianness = DEVICE_NATIVE_ENDIAN,
1251 .impl = {
1252 .min_access_size = 1,
1253 .max_access_size = 1,
1254 },
1255 };
1256
1257 static const MemoryRegionOps ioportF0_io_ops = {
1258 .write = ioportF0_write,
1259 .read = ioportF0_read,
1260 .endianness = DEVICE_NATIVE_ENDIAN,
1261 .impl = {
1262 .min_access_size = 1,
1263 .max_access_size = 1,
1264 },
1265 };
1266
1267 void pc_basic_device_init(ISABus *isa_bus, qemu_irq *gsi,
1268 ISADevice **rtc_state,
1269 ISADevice **floppy,
1270 bool no_vmport)
1271 {
1272 int i;
1273 DriveInfo *fd[MAX_FD];
1274 DeviceState *hpet = NULL;
1275 int pit_isa_irq = 0;
1276 qemu_irq pit_alt_irq = NULL;
1277 qemu_irq rtc_irq = NULL;
1278 qemu_irq *a20_line;
1279 ISADevice *i8042, *port92, *vmmouse, *pit = NULL;
1280 qemu_irq *cpu_exit_irq;
1281 MemoryRegion *ioport80_io = g_new(MemoryRegion, 1);
1282 MemoryRegion *ioportF0_io = g_new(MemoryRegion, 1);
1283
1284 memory_region_init_io(ioport80_io, NULL, &ioport80_io_ops, NULL, "ioport80", 1);
1285 memory_region_add_subregion(isa_bus->address_space_io, 0x80, ioport80_io);
1286
1287 memory_region_init_io(ioportF0_io, NULL, &ioportF0_io_ops, NULL, "ioportF0", 1);
1288 memory_region_add_subregion(isa_bus->address_space_io, 0xf0, ioportF0_io);
1289
1290 /*
1291 * Check if an HPET shall be created.
1292 *
1293 * Without KVM_CAP_PIT_STATE2, we cannot switch off the in-kernel PIT
1294 * when the HPET wants to take over. Thus we have to disable the latter.
1295 */
1296 if (!no_hpet && (!kvm_irqchip_in_kernel() || kvm_has_pit_state2())) {
1297 hpet = sysbus_try_create_simple("hpet", HPET_BASE, NULL);
1298
1299 if (hpet) {
1300 for (i = 0; i < GSI_NUM_PINS; i++) {
1301 sysbus_connect_irq(SYS_BUS_DEVICE(hpet), i, gsi[i]);
1302 }
1303 pit_isa_irq = -1;
1304 pit_alt_irq = qdev_get_gpio_in(hpet, HPET_LEGACY_PIT_INT);
1305 rtc_irq = qdev_get_gpio_in(hpet, HPET_LEGACY_RTC_INT);
1306 }
1307 }
1308 *rtc_state = rtc_init(isa_bus, 2000, rtc_irq);
1309
1310 qemu_register_boot_set(pc_boot_set, *rtc_state);
1311
1312 if (!xen_enabled()) {
1313 if (kvm_irqchip_in_kernel()) {
1314 pit = kvm_pit_init(isa_bus, 0x40);
1315 } else {
1316 pit = pit_init(isa_bus, 0x40, pit_isa_irq, pit_alt_irq);
1317 }
1318 if (hpet) {
1319 /* connect PIT to output control line of the HPET */
1320 qdev_connect_gpio_out(hpet, 0, qdev_get_gpio_in(DEVICE(pit), 0));
1321 }
1322 pcspk_init(isa_bus, pit);
1323 }
1324
1325 for(i = 0; i < MAX_SERIAL_PORTS; i++) {
1326 if (serial_hds[i]) {
1327 serial_isa_init(isa_bus, i, serial_hds[i]);
1328 }
1329 }
1330
1331 for(i = 0; i < MAX_PARALLEL_PORTS; i++) {
1332 if (parallel_hds[i]) {
1333 parallel_init(isa_bus, i, parallel_hds[i]);
1334 }
1335 }
1336
1337 a20_line = qemu_allocate_irqs(handle_a20_line_change, first_cpu, 2);
1338 i8042 = isa_create_simple(isa_bus, "i8042");
1339 i8042_setup_a20_line(i8042, &a20_line[0]);
1340 if (!no_vmport) {
1341 vmport_init(isa_bus);
1342 vmmouse = isa_try_create(isa_bus, "vmmouse");
1343 } else {
1344 vmmouse = NULL;
1345 }
1346 if (vmmouse) {
1347 DeviceState *dev = DEVICE(vmmouse);
1348 qdev_prop_set_ptr(dev, "ps2_mouse", i8042);
1349 qdev_init_nofail(dev);
1350 }
1351 port92 = isa_create_simple(isa_bus, "port92");
1352 port92_init(port92, &a20_line[1]);
1353
1354 cpu_exit_irq = qemu_allocate_irqs(cpu_request_exit, NULL, 1);
1355 DMA_init(0, cpu_exit_irq);
1356
1357 for(i = 0; i < MAX_FD; i++) {
1358 fd[i] = drive_get(IF_FLOPPY, 0, i);
1359 }
1360 *floppy = fdctrl_init_isa(isa_bus, fd);
1361 }
1362
1363 void pc_nic_init(ISABus *isa_bus, PCIBus *pci_bus)
1364 {
1365 int i;
1366
1367 for (i = 0; i < nb_nics; i++) {
1368 NICInfo *nd = &nd_table[i];
1369
1370 if (!pci_bus || (nd->model && strcmp(nd->model, "ne2k_isa") == 0)) {
1371 pc_init_ne2k_isa(isa_bus, nd);
1372 } else {
1373 pci_nic_init_nofail(nd, pci_bus, "e1000", NULL);
1374 }
1375 }
1376 }
1377
1378 void pc_pci_device_init(PCIBus *pci_bus)
1379 {
1380 int max_bus;
1381 int bus;
1382
1383 max_bus = drive_get_max_bus(IF_SCSI);
1384 for (bus = 0; bus <= max_bus; bus++) {
1385 pci_create_simple(pci_bus, -1, "lsi53c895a");
1386 }
1387 }
1388
1389 void ioapic_init_gsi(GSIState *gsi_state, const char *parent_name)
1390 {
1391 DeviceState *dev;
1392 SysBusDevice *d;
1393 unsigned int i;
1394
1395 if (kvm_irqchip_in_kernel()) {
1396 dev = qdev_create(NULL, "kvm-ioapic");
1397 } else {
1398 dev = qdev_create(NULL, "ioapic");
1399 }
1400 if (parent_name) {
1401 object_property_add_child(object_resolve_path(parent_name, NULL),
1402 "ioapic", OBJECT(dev), NULL);
1403 }
1404 qdev_init_nofail(dev);
1405 d = SYS_BUS_DEVICE(dev);
1406 sysbus_mmio_map(d, 0, IO_APIC_DEFAULT_ADDRESS);
1407
1408 for (i = 0; i < IOAPIC_NUM_PINS; i++) {
1409 gsi_state->ioapic_irq[i] = qdev_get_gpio_in(dev, i);
1410 }
1411 }