]> git.proxmox.com Git - qemu.git/blob - hw/i386/pc.c
Merge remote-tracking branch 'sstabellini/xen_fixes_20130603' into staging
[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
59 /* debug PC/ISA interrupts */
60 //#define DEBUG_IRQ
61
62 #ifdef DEBUG_IRQ
63 #define DPRINTF(fmt, ...) \
64 do { printf("CPUIRQ: " fmt , ## __VA_ARGS__); } while (0)
65 #else
66 #define DPRINTF(fmt, ...)
67 #endif
68
69 /* Leave a chunk of memory at the top of RAM for the BIOS ACPI tables. */
70 #define ACPI_DATA_SIZE 0x10000
71 #define BIOS_CFG_IOPORT 0x510
72 #define FW_CFG_ACPI_TABLES (FW_CFG_ARCH_LOCAL + 0)
73 #define FW_CFG_SMBIOS_ENTRIES (FW_CFG_ARCH_LOCAL + 1)
74 #define FW_CFG_IRQ0_OVERRIDE (FW_CFG_ARCH_LOCAL + 2)
75 #define FW_CFG_E820_TABLE (FW_CFG_ARCH_LOCAL + 3)
76 #define FW_CFG_HPET (FW_CFG_ARCH_LOCAL + 4)
77
78 #define IO_APIC_DEFAULT_ADDRESS 0xfec00000
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_table;
94 struct hpet_fw_config hpet_cfg = {.count = UINT8_MAX};
95
96 void gsi_handler(void *opaque, int n, int level)
97 {
98 GSIState *s = opaque;
99
100 DPRINTF("pc: %s GSI %d\n", level ? "raising" : "lowering", n);
101 if (n < ISA_NUM_IRQS) {
102 qemu_set_irq(s->i8259_irq[n], level);
103 }
104 qemu_set_irq(s->ioapic_irq[n], level);
105 }
106
107 static void ioport80_write(void *opaque, hwaddr addr, uint64_t data,
108 unsigned size)
109 {
110 }
111
112 static uint64_t ioport80_read(void *opaque, hwaddr addr, unsigned size)
113 {
114 return 0xffffffffffffffffULL;
115 }
116
117 /* MSDOS compatibility mode FPU exception support */
118 static qemu_irq ferr_irq;
119
120 void pc_register_ferr_irq(qemu_irq irq)
121 {
122 ferr_irq = irq;
123 }
124
125 /* XXX: add IGNNE support */
126 void cpu_set_ferr(CPUX86State *s)
127 {
128 qemu_irq_raise(ferr_irq);
129 }
130
131 static void ioportF0_write(void *opaque, hwaddr addr, uint64_t data,
132 unsigned size)
133 {
134 qemu_irq_lower(ferr_irq);
135 }
136
137 static uint64_t ioportF0_read(void *opaque, hwaddr addr, unsigned size)
138 {
139 return 0xffffffffffffffffULL;
140 }
141
142 /* TSC handling */
143 uint64_t cpu_get_tsc(CPUX86State *env)
144 {
145 return cpu_get_ticks();
146 }
147
148 /* SMM support */
149
150 static cpu_set_smm_t smm_set;
151 static void *smm_arg;
152
153 void cpu_smm_register(cpu_set_smm_t callback, void *arg)
154 {
155 assert(smm_set == NULL);
156 assert(smm_arg == NULL);
157 smm_set = callback;
158 smm_arg = arg;
159 }
160
161 void cpu_smm_update(CPUX86State *env)
162 {
163 if (smm_set && smm_arg && env == first_cpu)
164 smm_set(!!(env->hflags & HF_SMM_MASK), smm_arg);
165 }
166
167
168 /* IRQ handling */
169 int cpu_get_pic_interrupt(CPUX86State *env)
170 {
171 int intno;
172
173 intno = apic_get_interrupt(env->apic_state);
174 if (intno >= 0) {
175 return intno;
176 }
177 /* read the irq from the PIC */
178 if (!apic_accept_pic_intr(env->apic_state)) {
179 return -1;
180 }
181
182 intno = pic_read_irq(isa_pic);
183 return intno;
184 }
185
186 static void pic_irq_request(void *opaque, int irq, int level)
187 {
188 CPUX86State *env = first_cpu;
189
190 DPRINTF("pic_irqs: %s irq %d\n", level? "raise" : "lower", irq);
191 if (env->apic_state) {
192 while (env) {
193 if (apic_accept_pic_intr(env->apic_state)) {
194 apic_deliver_pic_intr(env->apic_state, level);
195 }
196 env = env->next_cpu;
197 }
198 } else {
199 CPUState *cs = CPU(x86_env_get_cpu(env));
200 if (level) {
201 cpu_interrupt(cs, CPU_INTERRUPT_HARD);
202 } else {
203 cpu_reset_interrupt(cs, CPU_INTERRUPT_HARD);
204 }
205 }
206 }
207
208 /* PC cmos mappings */
209
210 #define REG_EQUIPMENT_BYTE 0x14
211
212 static int cmos_get_fd_drive_type(FDriveType fd0)
213 {
214 int val;
215
216 switch (fd0) {
217 case FDRIVE_DRV_144:
218 /* 1.44 Mb 3"5 drive */
219 val = 4;
220 break;
221 case FDRIVE_DRV_288:
222 /* 2.88 Mb 3"5 drive */
223 val = 5;
224 break;
225 case FDRIVE_DRV_120:
226 /* 1.2 Mb 5"5 drive */
227 val = 2;
228 break;
229 case FDRIVE_DRV_NONE:
230 default:
231 val = 0;
232 break;
233 }
234 return val;
235 }
236
237 static void cmos_init_hd(ISADevice *s, int type_ofs, int info_ofs,
238 int16_t cylinders, int8_t heads, int8_t sectors)
239 {
240 rtc_set_memory(s, type_ofs, 47);
241 rtc_set_memory(s, info_ofs, cylinders);
242 rtc_set_memory(s, info_ofs + 1, cylinders >> 8);
243 rtc_set_memory(s, info_ofs + 2, heads);
244 rtc_set_memory(s, info_ofs + 3, 0xff);
245 rtc_set_memory(s, info_ofs + 4, 0xff);
246 rtc_set_memory(s, info_ofs + 5, 0xc0 | ((heads > 8) << 3));
247 rtc_set_memory(s, info_ofs + 6, cylinders);
248 rtc_set_memory(s, info_ofs + 7, cylinders >> 8);
249 rtc_set_memory(s, info_ofs + 8, sectors);
250 }
251
252 /* convert boot_device letter to something recognizable by the bios */
253 static int boot_device2nibble(char boot_device)
254 {
255 switch(boot_device) {
256 case 'a':
257 case 'b':
258 return 0x01; /* floppy boot */
259 case 'c':
260 return 0x02; /* hard drive boot */
261 case 'd':
262 return 0x03; /* CD-ROM boot */
263 case 'n':
264 return 0x04; /* Network boot */
265 }
266 return 0;
267 }
268
269 static int set_boot_dev(ISADevice *s, const char *boot_device, int fd_bootchk)
270 {
271 #define PC_MAX_BOOT_DEVICES 3
272 int nbds, bds[3] = { 0, };
273 int i;
274
275 nbds = strlen(boot_device);
276 if (nbds > PC_MAX_BOOT_DEVICES) {
277 error_report("Too many boot devices for PC");
278 return(1);
279 }
280 for (i = 0; i < nbds; i++) {
281 bds[i] = boot_device2nibble(boot_device[i]);
282 if (bds[i] == 0) {
283 error_report("Invalid boot device for PC: '%c'",
284 boot_device[i]);
285 return(1);
286 }
287 }
288 rtc_set_memory(s, 0x3d, (bds[1] << 4) | bds[0]);
289 rtc_set_memory(s, 0x38, (bds[2] << 4) | (fd_bootchk ? 0x0 : 0x1));
290 return(0);
291 }
292
293 static int pc_boot_set(void *opaque, const char *boot_device)
294 {
295 return set_boot_dev(opaque, boot_device, 0);
296 }
297
298 typedef struct pc_cmos_init_late_arg {
299 ISADevice *rtc_state;
300 BusState *idebus[2];
301 } pc_cmos_init_late_arg;
302
303 static void pc_cmos_init_late(void *opaque)
304 {
305 pc_cmos_init_late_arg *arg = opaque;
306 ISADevice *s = arg->rtc_state;
307 int16_t cylinders;
308 int8_t heads, sectors;
309 int val;
310 int i, trans;
311
312 val = 0;
313 if (ide_get_geometry(arg->idebus[0], 0,
314 &cylinders, &heads, &sectors) >= 0) {
315 cmos_init_hd(s, 0x19, 0x1b, cylinders, heads, sectors);
316 val |= 0xf0;
317 }
318 if (ide_get_geometry(arg->idebus[0], 1,
319 &cylinders, &heads, &sectors) >= 0) {
320 cmos_init_hd(s, 0x1a, 0x24, cylinders, heads, sectors);
321 val |= 0x0f;
322 }
323 rtc_set_memory(s, 0x12, val);
324
325 val = 0;
326 for (i = 0; i < 4; i++) {
327 /* NOTE: ide_get_geometry() returns the physical
328 geometry. It is always such that: 1 <= sects <= 63, 1
329 <= heads <= 16, 1 <= cylinders <= 16383. The BIOS
330 geometry can be different if a translation is done. */
331 if (ide_get_geometry(arg->idebus[i / 2], i % 2,
332 &cylinders, &heads, &sectors) >= 0) {
333 trans = ide_get_bios_chs_trans(arg->idebus[i / 2], i % 2) - 1;
334 assert((trans & ~3) == 0);
335 val |= trans << (i * 2);
336 }
337 }
338 rtc_set_memory(s, 0x39, val);
339
340 qemu_unregister_reset(pc_cmos_init_late, opaque);
341 }
342
343 typedef struct RTCCPUHotplugArg {
344 Notifier cpu_added_notifier;
345 ISADevice *rtc_state;
346 } RTCCPUHotplugArg;
347
348 static void rtc_notify_cpu_added(Notifier *notifier, void *data)
349 {
350 RTCCPUHotplugArg *arg = container_of(notifier, RTCCPUHotplugArg,
351 cpu_added_notifier);
352 ISADevice *s = arg->rtc_state;
353
354 /* increment the number of CPUs */
355 rtc_set_memory(s, 0x5f, rtc_get_memory(s, 0x5f) + 1);
356 }
357
358 void pc_cmos_init(ram_addr_t ram_size, ram_addr_t above_4g_mem_size,
359 const char *boot_device,
360 ISADevice *floppy, BusState *idebus0, BusState *idebus1,
361 ISADevice *s)
362 {
363 int val, nb, i;
364 FDriveType fd_type[2] = { FDRIVE_DRV_NONE, FDRIVE_DRV_NONE };
365 static pc_cmos_init_late_arg arg;
366 static RTCCPUHotplugArg cpu_hotplug_cb;
367
368 /* various important CMOS locations needed by PC/Bochs bios */
369
370 /* memory size */
371 /* base memory (first MiB) */
372 val = MIN(ram_size / 1024, 640);
373 rtc_set_memory(s, 0x15, val);
374 rtc_set_memory(s, 0x16, val >> 8);
375 /* extended memory (next 64MiB) */
376 if (ram_size > 1024 * 1024) {
377 val = (ram_size - 1024 * 1024) / 1024;
378 } else {
379 val = 0;
380 }
381 if (val > 65535)
382 val = 65535;
383 rtc_set_memory(s, 0x17, val);
384 rtc_set_memory(s, 0x18, val >> 8);
385 rtc_set_memory(s, 0x30, val);
386 rtc_set_memory(s, 0x31, val >> 8);
387 /* memory between 16MiB and 4GiB */
388 if (ram_size > 16 * 1024 * 1024) {
389 val = (ram_size - 16 * 1024 * 1024) / 65536;
390 } else {
391 val = 0;
392 }
393 if (val > 65535)
394 val = 65535;
395 rtc_set_memory(s, 0x34, val);
396 rtc_set_memory(s, 0x35, val >> 8);
397 /* memory above 4GiB */
398 val = above_4g_mem_size / 65536;
399 rtc_set_memory(s, 0x5b, val);
400 rtc_set_memory(s, 0x5c, val >> 8);
401 rtc_set_memory(s, 0x5d, val >> 16);
402
403 /* set the number of CPU */
404 rtc_set_memory(s, 0x5f, smp_cpus - 1);
405 /* init CPU hotplug notifier */
406 cpu_hotplug_cb.rtc_state = s;
407 cpu_hotplug_cb.cpu_added_notifier.notify = rtc_notify_cpu_added;
408 qemu_register_cpu_added_notifier(&cpu_hotplug_cb.cpu_added_notifier);
409
410 /* set boot devices, and disable floppy signature check if requested */
411 if (set_boot_dev(s, boot_device, fd_bootchk)) {
412 exit(1);
413 }
414
415 /* floppy type */
416 if (floppy) {
417 for (i = 0; i < 2; i++) {
418 fd_type[i] = isa_fdc_get_drive_type(floppy, i);
419 }
420 }
421 val = (cmos_get_fd_drive_type(fd_type[0]) << 4) |
422 cmos_get_fd_drive_type(fd_type[1]);
423 rtc_set_memory(s, 0x10, val);
424
425 val = 0;
426 nb = 0;
427 if (fd_type[0] < FDRIVE_DRV_NONE) {
428 nb++;
429 }
430 if (fd_type[1] < FDRIVE_DRV_NONE) {
431 nb++;
432 }
433 switch (nb) {
434 case 0:
435 break;
436 case 1:
437 val |= 0x01; /* 1 drive, ready for boot */
438 break;
439 case 2:
440 val |= 0x41; /* 2 drives, ready for boot */
441 break;
442 }
443 val |= 0x02; /* FPU is there */
444 val |= 0x04; /* PS/2 mouse installed */
445 rtc_set_memory(s, REG_EQUIPMENT_BYTE, val);
446
447 /* hard drives */
448 arg.rtc_state = s;
449 arg.idebus[0] = idebus0;
450 arg.idebus[1] = idebus1;
451 qemu_register_reset(pc_cmos_init_late, &arg);
452 }
453
454 #define TYPE_PORT92 "port92"
455 #define PORT92(obj) OBJECT_CHECK(Port92State, (obj), TYPE_PORT92)
456
457 /* port 92 stuff: could be split off */
458 typedef struct Port92State {
459 ISADevice parent_obj;
460
461 MemoryRegion io;
462 uint8_t outport;
463 qemu_irq *a20_out;
464 } Port92State;
465
466 static void port92_write(void *opaque, hwaddr addr, uint64_t val,
467 unsigned size)
468 {
469 Port92State *s = opaque;
470
471 DPRINTF("port92: write 0x%02x\n", val);
472 s->outport = val;
473 qemu_set_irq(*s->a20_out, (val >> 1) & 1);
474 if (val & 1) {
475 qemu_system_reset_request();
476 }
477 }
478
479 static uint64_t port92_read(void *opaque, hwaddr addr,
480 unsigned size)
481 {
482 Port92State *s = opaque;
483 uint32_t ret;
484
485 ret = s->outport;
486 DPRINTF("port92: read 0x%02x\n", ret);
487 return ret;
488 }
489
490 static void port92_init(ISADevice *dev, qemu_irq *a20_out)
491 {
492 Port92State *s = PORT92(dev);
493
494 s->a20_out = a20_out;
495 }
496
497 static const VMStateDescription vmstate_port92_isa = {
498 .name = "port92",
499 .version_id = 1,
500 .minimum_version_id = 1,
501 .minimum_version_id_old = 1,
502 .fields = (VMStateField []) {
503 VMSTATE_UINT8(outport, Port92State),
504 VMSTATE_END_OF_LIST()
505 }
506 };
507
508 static void port92_reset(DeviceState *d)
509 {
510 Port92State *s = PORT92(d);
511
512 s->outport &= ~1;
513 }
514
515 static const MemoryRegionOps port92_ops = {
516 .read = port92_read,
517 .write = port92_write,
518 .impl = {
519 .min_access_size = 1,
520 .max_access_size = 1,
521 },
522 .endianness = DEVICE_LITTLE_ENDIAN,
523 };
524
525 static int port92_initfn(ISADevice *dev)
526 {
527 Port92State *s = PORT92(dev);
528
529 memory_region_init_io(&s->io, &port92_ops, s, "port92", 1);
530 isa_register_ioport(dev, &s->io, 0x92);
531
532 s->outport = 0;
533 return 0;
534 }
535
536 static void port92_class_initfn(ObjectClass *klass, void *data)
537 {
538 DeviceClass *dc = DEVICE_CLASS(klass);
539 ISADeviceClass *ic = ISA_DEVICE_CLASS(klass);
540 ic->init = port92_initfn;
541 dc->no_user = 1;
542 dc->reset = port92_reset;
543 dc->vmsd = &vmstate_port92_isa;
544 }
545
546 static const TypeInfo port92_info = {
547 .name = TYPE_PORT92,
548 .parent = TYPE_ISA_DEVICE,
549 .instance_size = sizeof(Port92State),
550 .class_init = port92_class_initfn,
551 };
552
553 static void port92_register_types(void)
554 {
555 type_register_static(&port92_info);
556 }
557
558 type_init(port92_register_types)
559
560 static void handle_a20_line_change(void *opaque, int irq, int level)
561 {
562 X86CPU *cpu = opaque;
563
564 /* XXX: send to all CPUs ? */
565 /* XXX: add logic to handle multiple A20 line sources */
566 x86_cpu_set_a20(cpu, level);
567 }
568
569 int e820_add_entry(uint64_t address, uint64_t length, uint32_t type)
570 {
571 int index = le32_to_cpu(e820_table.count);
572 struct e820_entry *entry;
573
574 if (index >= E820_NR_ENTRIES)
575 return -EBUSY;
576 entry = &e820_table.entry[index++];
577
578 entry->address = cpu_to_le64(address);
579 entry->length = cpu_to_le64(length);
580 entry->type = cpu_to_le32(type);
581
582 e820_table.count = cpu_to_le32(index);
583 return index;
584 }
585
586 /* Calculates the limit to CPU APIC ID values
587 *
588 * This function returns the limit for the APIC ID value, so that all
589 * CPU APIC IDs are < pc_apic_id_limit().
590 *
591 * This is used for FW_CFG_MAX_CPUS. See comments on bochs_bios_init().
592 */
593 static unsigned int pc_apic_id_limit(unsigned int max_cpus)
594 {
595 return x86_cpu_apic_id_from_index(max_cpus - 1) + 1;
596 }
597
598 static FWCfgState *bochs_bios_init(void)
599 {
600 FWCfgState *fw_cfg;
601 uint8_t *smbios_table;
602 size_t smbios_len;
603 uint64_t *numa_fw_cfg;
604 int i, j;
605 unsigned int apic_id_limit = pc_apic_id_limit(max_cpus);
606
607 fw_cfg = fw_cfg_init(BIOS_CFG_IOPORT, BIOS_CFG_IOPORT + 1, 0, 0);
608 /* FW_CFG_MAX_CPUS is a bit confusing/problematic on x86:
609 *
610 * SeaBIOS needs FW_CFG_MAX_CPUS for CPU hotplug, but the CPU hotplug
611 * QEMU<->SeaBIOS interface is not based on the "CPU index", but on the APIC
612 * ID of hotplugged CPUs[1]. This means that FW_CFG_MAX_CPUS is not the
613 * "maximum number of CPUs", but the "limit to the APIC ID values SeaBIOS
614 * may see".
615 *
616 * So, this means we must not use max_cpus, here, but the maximum possible
617 * APIC ID value, plus one.
618 *
619 * [1] The only kind of "CPU identifier" used between SeaBIOS and QEMU is
620 * the APIC ID, not the "CPU index"
621 */
622 fw_cfg_add_i16(fw_cfg, FW_CFG_MAX_CPUS, (uint16_t)apic_id_limit);
623 fw_cfg_add_i32(fw_cfg, FW_CFG_ID, 1);
624 fw_cfg_add_i64(fw_cfg, FW_CFG_RAM_SIZE, (uint64_t)ram_size);
625 fw_cfg_add_bytes(fw_cfg, FW_CFG_ACPI_TABLES,
626 acpi_tables, acpi_tables_len);
627 fw_cfg_add_i32(fw_cfg, FW_CFG_IRQ0_OVERRIDE, kvm_allows_irq0_override());
628
629 smbios_table = smbios_get_table(&smbios_len);
630 if (smbios_table)
631 fw_cfg_add_bytes(fw_cfg, FW_CFG_SMBIOS_ENTRIES,
632 smbios_table, smbios_len);
633 fw_cfg_add_bytes(fw_cfg, FW_CFG_E820_TABLE,
634 &e820_table, sizeof(e820_table));
635
636 fw_cfg_add_bytes(fw_cfg, FW_CFG_HPET, &hpet_cfg, sizeof(hpet_cfg));
637 /* allocate memory for the NUMA channel: one (64bit) word for the number
638 * of nodes, one word for each VCPU->node and one word for each node to
639 * hold the amount of memory.
640 */
641 numa_fw_cfg = g_new0(uint64_t, 1 + apic_id_limit + nb_numa_nodes);
642 numa_fw_cfg[0] = cpu_to_le64(nb_numa_nodes);
643 for (i = 0; i < max_cpus; i++) {
644 unsigned int apic_id = x86_cpu_apic_id_from_index(i);
645 assert(apic_id < apic_id_limit);
646 for (j = 0; j < nb_numa_nodes; j++) {
647 if (test_bit(i, node_cpumask[j])) {
648 numa_fw_cfg[apic_id + 1] = cpu_to_le64(j);
649 break;
650 }
651 }
652 }
653 for (i = 0; i < nb_numa_nodes; i++) {
654 numa_fw_cfg[apic_id_limit + 1 + i] = cpu_to_le64(node_mem[i]);
655 }
656 fw_cfg_add_bytes(fw_cfg, FW_CFG_NUMA, numa_fw_cfg,
657 (1 + apic_id_limit + nb_numa_nodes) *
658 sizeof(*numa_fw_cfg));
659
660 return fw_cfg;
661 }
662
663 static long get_file_size(FILE *f)
664 {
665 long where, size;
666
667 /* XXX: on Unix systems, using fstat() probably makes more sense */
668
669 where = ftell(f);
670 fseek(f, 0, SEEK_END);
671 size = ftell(f);
672 fseek(f, where, SEEK_SET);
673
674 return size;
675 }
676
677 static void load_linux(FWCfgState *fw_cfg,
678 const char *kernel_filename,
679 const char *initrd_filename,
680 const char *kernel_cmdline,
681 hwaddr max_ram_size)
682 {
683 uint16_t protocol;
684 int setup_size, kernel_size, initrd_size = 0, cmdline_size;
685 uint32_t initrd_max;
686 uint8_t header[8192], *setup, *kernel, *initrd_data;
687 hwaddr real_addr, prot_addr, cmdline_addr, initrd_addr = 0;
688 FILE *f;
689 char *vmode;
690
691 /* Align to 16 bytes as a paranoia measure */
692 cmdline_size = (strlen(kernel_cmdline)+16) & ~15;
693
694 /* load the kernel header */
695 f = fopen(kernel_filename, "rb");
696 if (!f || !(kernel_size = get_file_size(f)) ||
697 fread(header, 1, MIN(ARRAY_SIZE(header), kernel_size), f) !=
698 MIN(ARRAY_SIZE(header), kernel_size)) {
699 fprintf(stderr, "qemu: could not load kernel '%s': %s\n",
700 kernel_filename, strerror(errno));
701 exit(1);
702 }
703
704 /* kernel protocol version */
705 #if 0
706 fprintf(stderr, "header magic: %#x\n", ldl_p(header+0x202));
707 #endif
708 if (ldl_p(header+0x202) == 0x53726448) {
709 protocol = lduw_p(header+0x206);
710 } else {
711 /* This looks like a multiboot kernel. If it is, let's stop
712 treating it like a Linux kernel. */
713 if (load_multiboot(fw_cfg, f, kernel_filename, initrd_filename,
714 kernel_cmdline, kernel_size, header)) {
715 return;
716 }
717 protocol = 0;
718 }
719
720 if (protocol < 0x200 || !(header[0x211] & 0x01)) {
721 /* Low kernel */
722 real_addr = 0x90000;
723 cmdline_addr = 0x9a000 - cmdline_size;
724 prot_addr = 0x10000;
725 } else if (protocol < 0x202) {
726 /* High but ancient kernel */
727 real_addr = 0x90000;
728 cmdline_addr = 0x9a000 - cmdline_size;
729 prot_addr = 0x100000;
730 } else {
731 /* High and recent kernel */
732 real_addr = 0x10000;
733 cmdline_addr = 0x20000;
734 prot_addr = 0x100000;
735 }
736
737 #if 0
738 fprintf(stderr,
739 "qemu: real_addr = 0x" TARGET_FMT_plx "\n"
740 "qemu: cmdline_addr = 0x" TARGET_FMT_plx "\n"
741 "qemu: prot_addr = 0x" TARGET_FMT_plx "\n",
742 real_addr,
743 cmdline_addr,
744 prot_addr);
745 #endif
746
747 /* highest address for loading the initrd */
748 if (protocol >= 0x203) {
749 initrd_max = ldl_p(header+0x22c);
750 } else {
751 initrd_max = 0x37ffffff;
752 }
753
754 if (initrd_max >= max_ram_size-ACPI_DATA_SIZE)
755 initrd_max = max_ram_size-ACPI_DATA_SIZE-1;
756
757 fw_cfg_add_i32(fw_cfg, FW_CFG_CMDLINE_ADDR, cmdline_addr);
758 fw_cfg_add_i32(fw_cfg, FW_CFG_CMDLINE_SIZE, strlen(kernel_cmdline)+1);
759 fw_cfg_add_string(fw_cfg, FW_CFG_CMDLINE_DATA, kernel_cmdline);
760
761 if (protocol >= 0x202) {
762 stl_p(header+0x228, cmdline_addr);
763 } else {
764 stw_p(header+0x20, 0xA33F);
765 stw_p(header+0x22, cmdline_addr-real_addr);
766 }
767
768 /* handle vga= parameter */
769 vmode = strstr(kernel_cmdline, "vga=");
770 if (vmode) {
771 unsigned int video_mode;
772 /* skip "vga=" */
773 vmode += 4;
774 if (!strncmp(vmode, "normal", 6)) {
775 video_mode = 0xffff;
776 } else if (!strncmp(vmode, "ext", 3)) {
777 video_mode = 0xfffe;
778 } else if (!strncmp(vmode, "ask", 3)) {
779 video_mode = 0xfffd;
780 } else {
781 video_mode = strtol(vmode, NULL, 0);
782 }
783 stw_p(header+0x1fa, video_mode);
784 }
785
786 /* loader type */
787 /* High nybble = B reserved for QEMU; low nybble is revision number.
788 If this code is substantially changed, you may want to consider
789 incrementing the revision. */
790 if (protocol >= 0x200) {
791 header[0x210] = 0xB0;
792 }
793 /* heap */
794 if (protocol >= 0x201) {
795 header[0x211] |= 0x80; /* CAN_USE_HEAP */
796 stw_p(header+0x224, cmdline_addr-real_addr-0x200);
797 }
798
799 /* load initrd */
800 if (initrd_filename) {
801 if (protocol < 0x200) {
802 fprintf(stderr, "qemu: linux kernel too old to load a ram disk\n");
803 exit(1);
804 }
805
806 initrd_size = get_image_size(initrd_filename);
807 if (initrd_size < 0) {
808 fprintf(stderr, "qemu: error reading initrd %s\n",
809 initrd_filename);
810 exit(1);
811 }
812
813 initrd_addr = (initrd_max-initrd_size) & ~4095;
814
815 initrd_data = g_malloc(initrd_size);
816 load_image(initrd_filename, initrd_data);
817
818 fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_ADDR, initrd_addr);
819 fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_SIZE, initrd_size);
820 fw_cfg_add_bytes(fw_cfg, FW_CFG_INITRD_DATA, initrd_data, initrd_size);
821
822 stl_p(header+0x218, initrd_addr);
823 stl_p(header+0x21c, initrd_size);
824 }
825
826 /* load kernel and setup */
827 setup_size = header[0x1f1];
828 if (setup_size == 0) {
829 setup_size = 4;
830 }
831 setup_size = (setup_size+1)*512;
832 kernel_size -= setup_size;
833
834 setup = g_malloc(setup_size);
835 kernel = g_malloc(kernel_size);
836 fseek(f, 0, SEEK_SET);
837 if (fread(setup, 1, setup_size, f) != setup_size) {
838 fprintf(stderr, "fread() failed\n");
839 exit(1);
840 }
841 if (fread(kernel, 1, kernel_size, f) != kernel_size) {
842 fprintf(stderr, "fread() failed\n");
843 exit(1);
844 }
845 fclose(f);
846 memcpy(setup, header, MIN(sizeof(header), setup_size));
847
848 fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_ADDR, prot_addr);
849 fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_SIZE, kernel_size);
850 fw_cfg_add_bytes(fw_cfg, FW_CFG_KERNEL_DATA, kernel, kernel_size);
851
852 fw_cfg_add_i32(fw_cfg, FW_CFG_SETUP_ADDR, real_addr);
853 fw_cfg_add_i32(fw_cfg, FW_CFG_SETUP_SIZE, setup_size);
854 fw_cfg_add_bytes(fw_cfg, FW_CFG_SETUP_DATA, setup, setup_size);
855
856 option_rom[nb_option_roms].name = "linuxboot.bin";
857 option_rom[nb_option_roms].bootindex = 0;
858 nb_option_roms++;
859 }
860
861 #define NE2000_NB_MAX 6
862
863 static const int ne2000_io[NE2000_NB_MAX] = { 0x300, 0x320, 0x340, 0x360,
864 0x280, 0x380 };
865 static const int ne2000_irq[NE2000_NB_MAX] = { 9, 10, 11, 3, 4, 5 };
866
867 static const int parallel_io[MAX_PARALLEL_PORTS] = { 0x378, 0x278, 0x3bc };
868 static const int parallel_irq[MAX_PARALLEL_PORTS] = { 7, 7, 7 };
869
870 void pc_init_ne2k_isa(ISABus *bus, NICInfo *nd)
871 {
872 static int nb_ne2k = 0;
873
874 if (nb_ne2k == NE2000_NB_MAX)
875 return;
876 isa_ne2000_init(bus, ne2000_io[nb_ne2k],
877 ne2000_irq[nb_ne2k], nd);
878 nb_ne2k++;
879 }
880
881 DeviceState *cpu_get_current_apic(void)
882 {
883 if (cpu_single_env) {
884 return cpu_single_env->apic_state;
885 } else {
886 return NULL;
887 }
888 }
889
890 void pc_acpi_smi_interrupt(void *opaque, int irq, int level)
891 {
892 X86CPU *cpu = opaque;
893
894 if (level) {
895 cpu_interrupt(CPU(cpu), CPU_INTERRUPT_SMI);
896 }
897 }
898
899 static X86CPU *pc_new_cpu(const char *cpu_model, int64_t apic_id,
900 DeviceState *icc_bridge, Error **errp)
901 {
902 X86CPU *cpu;
903 Error *local_err = NULL;
904
905 cpu = cpu_x86_create(cpu_model, icc_bridge, errp);
906 if (!cpu) {
907 return cpu;
908 }
909
910 object_property_set_int(OBJECT(cpu), apic_id, "apic-id", &local_err);
911 object_property_set_bool(OBJECT(cpu), true, "realized", &local_err);
912
913 if (local_err) {
914 if (cpu != NULL) {
915 object_unref(OBJECT(cpu));
916 cpu = NULL;
917 }
918 error_propagate(errp, local_err);
919 }
920 return cpu;
921 }
922
923 static const char *current_cpu_model;
924
925 void pc_hot_add_cpu(const int64_t id, Error **errp)
926 {
927 DeviceState *icc_bridge;
928 int64_t apic_id = x86_cpu_apic_id_from_index(id);
929
930 if (cpu_exists(apic_id)) {
931 error_setg(errp, "Unable to add CPU: %" PRIi64
932 ", it already exists", id);
933 return;
934 }
935
936 if (id >= max_cpus) {
937 error_setg(errp, "Unable to add CPU: %" PRIi64
938 ", max allowed: %d", id, max_cpus - 1);
939 return;
940 }
941
942 icc_bridge = DEVICE(object_resolve_path_type("icc-bridge",
943 TYPE_ICC_BRIDGE, NULL));
944 pc_new_cpu(current_cpu_model, apic_id, icc_bridge, errp);
945 }
946
947 void pc_cpus_init(const char *cpu_model, DeviceState *icc_bridge)
948 {
949 int i;
950 X86CPU *cpu = NULL;
951 Error *error = NULL;
952
953 /* init CPUs */
954 if (cpu_model == NULL) {
955 #ifdef TARGET_X86_64
956 cpu_model = "qemu64";
957 #else
958 cpu_model = "qemu32";
959 #endif
960 }
961 current_cpu_model = cpu_model;
962
963 for (i = 0; i < smp_cpus; i++) {
964 cpu = pc_new_cpu(cpu_model, x86_cpu_apic_id_from_index(i),
965 icc_bridge, &error);
966 if (error) {
967 fprintf(stderr, "%s\n", error_get_pretty(error));
968 error_free(error);
969 exit(1);
970 }
971 }
972
973 /* map APIC MMIO area if CPU has APIC */
974 if (cpu && cpu->env.apic_state) {
975 /* XXX: what if the base changes? */
976 sysbus_mmio_map_overlap(SYS_BUS_DEVICE(icc_bridge), 0,
977 APIC_DEFAULT_ADDRESS, 0x1000);
978 }
979 }
980
981 void pc_acpi_init(const char *default_dsdt)
982 {
983 char *filename;
984
985 if (acpi_tables != NULL) {
986 /* manually set via -acpitable, leave it alone */
987 return;
988 }
989
990 filename = qemu_find_file(QEMU_FILE_TYPE_BIOS, default_dsdt);
991 if (filename == NULL) {
992 fprintf(stderr, "WARNING: failed to find %s\n", default_dsdt);
993 } else {
994 char *arg;
995 QemuOpts *opts;
996 Error *err = NULL;
997
998 arg = g_strdup_printf("file=%s", filename);
999
1000 /* creates a deep copy of "arg" */
1001 opts = qemu_opts_parse(qemu_find_opts("acpi"), arg, 0);
1002 g_assert(opts != NULL);
1003
1004 acpi_table_add(opts, &err);
1005 if (err) {
1006 fprintf(stderr, "WARNING: failed to load %s: %s\n", filename,
1007 error_get_pretty(err));
1008 error_free(err);
1009 }
1010 g_free(arg);
1011 g_free(filename);
1012 }
1013 }
1014
1015 FWCfgState *pc_memory_init(MemoryRegion *system_memory,
1016 const char *kernel_filename,
1017 const char *kernel_cmdline,
1018 const char *initrd_filename,
1019 ram_addr_t below_4g_mem_size,
1020 ram_addr_t above_4g_mem_size,
1021 MemoryRegion *rom_memory,
1022 MemoryRegion **ram_memory)
1023 {
1024 int linux_boot, i;
1025 MemoryRegion *ram, *option_rom_mr;
1026 MemoryRegion *ram_below_4g, *ram_above_4g;
1027 FWCfgState *fw_cfg;
1028
1029 linux_boot = (kernel_filename != NULL);
1030
1031 /* Allocate RAM. We allocate it as a single memory region and use
1032 * aliases to address portions of it, mostly for backwards compatibility
1033 * with older qemus that used qemu_ram_alloc().
1034 */
1035 ram = g_malloc(sizeof(*ram));
1036 memory_region_init_ram(ram, "pc.ram",
1037 below_4g_mem_size + above_4g_mem_size);
1038 vmstate_register_ram_global(ram);
1039 *ram_memory = ram;
1040 ram_below_4g = g_malloc(sizeof(*ram_below_4g));
1041 memory_region_init_alias(ram_below_4g, "ram-below-4g", ram,
1042 0, below_4g_mem_size);
1043 memory_region_add_subregion(system_memory, 0, ram_below_4g);
1044 if (above_4g_mem_size > 0) {
1045 ram_above_4g = g_malloc(sizeof(*ram_above_4g));
1046 memory_region_init_alias(ram_above_4g, "ram-above-4g", ram,
1047 below_4g_mem_size, above_4g_mem_size);
1048 memory_region_add_subregion(system_memory, 0x100000000ULL,
1049 ram_above_4g);
1050 }
1051
1052
1053 /* Initialize PC system firmware */
1054 pc_system_firmware_init(rom_memory);
1055
1056 option_rom_mr = g_malloc(sizeof(*option_rom_mr));
1057 memory_region_init_ram(option_rom_mr, "pc.rom", PC_ROM_SIZE);
1058 vmstate_register_ram_global(option_rom_mr);
1059 memory_region_add_subregion_overlap(rom_memory,
1060 PC_ROM_MIN_VGA,
1061 option_rom_mr,
1062 1);
1063
1064 fw_cfg = bochs_bios_init();
1065 rom_set_fw(fw_cfg);
1066
1067 if (linux_boot) {
1068 load_linux(fw_cfg, kernel_filename, initrd_filename, kernel_cmdline, below_4g_mem_size);
1069 }
1070
1071 for (i = 0; i < nb_option_roms; i++) {
1072 rom_add_option(option_rom[i].name, option_rom[i].bootindex);
1073 }
1074 return fw_cfg;
1075 }
1076
1077 qemu_irq *pc_allocate_cpu_irq(void)
1078 {
1079 return qemu_allocate_irqs(pic_irq_request, NULL, 1);
1080 }
1081
1082 DeviceState *pc_vga_init(ISABus *isa_bus, PCIBus *pci_bus)
1083 {
1084 DeviceState *dev = NULL;
1085
1086 if (pci_bus) {
1087 PCIDevice *pcidev = pci_vga_init(pci_bus);
1088 dev = pcidev ? &pcidev->qdev : NULL;
1089 } else if (isa_bus) {
1090 ISADevice *isadev = isa_vga_init(isa_bus);
1091 dev = isadev ? &isadev->qdev : NULL;
1092 }
1093 return dev;
1094 }
1095
1096 static void cpu_request_exit(void *opaque, int irq, int level)
1097 {
1098 CPUX86State *env = cpu_single_env;
1099
1100 if (env && level) {
1101 cpu_exit(env);
1102 }
1103 }
1104
1105 static const MemoryRegionOps ioport80_io_ops = {
1106 .write = ioport80_write,
1107 .read = ioport80_read,
1108 .endianness = DEVICE_NATIVE_ENDIAN,
1109 .impl = {
1110 .min_access_size = 1,
1111 .max_access_size = 1,
1112 },
1113 };
1114
1115 static const MemoryRegionOps ioportF0_io_ops = {
1116 .write = ioportF0_write,
1117 .read = ioportF0_read,
1118 .endianness = DEVICE_NATIVE_ENDIAN,
1119 .impl = {
1120 .min_access_size = 1,
1121 .max_access_size = 1,
1122 },
1123 };
1124
1125 void pc_basic_device_init(ISABus *isa_bus, qemu_irq *gsi,
1126 ISADevice **rtc_state,
1127 ISADevice **floppy,
1128 bool no_vmport)
1129 {
1130 int i;
1131 DriveInfo *fd[MAX_FD];
1132 DeviceState *hpet = NULL;
1133 int pit_isa_irq = 0;
1134 qemu_irq pit_alt_irq = NULL;
1135 qemu_irq rtc_irq = NULL;
1136 qemu_irq *a20_line;
1137 ISADevice *i8042, *port92, *vmmouse, *pit = NULL;
1138 qemu_irq *cpu_exit_irq;
1139 MemoryRegion *ioport80_io = g_new(MemoryRegion, 1);
1140 MemoryRegion *ioportF0_io = g_new(MemoryRegion, 1);
1141
1142 memory_region_init_io(ioport80_io, &ioport80_io_ops, NULL, "ioport80", 1);
1143 memory_region_add_subregion(isa_bus->address_space_io, 0x80, ioport80_io);
1144
1145 memory_region_init_io(ioportF0_io, &ioportF0_io_ops, NULL, "ioportF0", 1);
1146 memory_region_add_subregion(isa_bus->address_space_io, 0xf0, ioportF0_io);
1147
1148 /*
1149 * Check if an HPET shall be created.
1150 *
1151 * Without KVM_CAP_PIT_STATE2, we cannot switch off the in-kernel PIT
1152 * when the HPET wants to take over. Thus we have to disable the latter.
1153 */
1154 if (!no_hpet && (!kvm_irqchip_in_kernel() || kvm_has_pit_state2())) {
1155 hpet = sysbus_try_create_simple("hpet", HPET_BASE, NULL);
1156
1157 if (hpet) {
1158 for (i = 0; i < GSI_NUM_PINS; i++) {
1159 sysbus_connect_irq(SYS_BUS_DEVICE(hpet), i, gsi[i]);
1160 }
1161 pit_isa_irq = -1;
1162 pit_alt_irq = qdev_get_gpio_in(hpet, HPET_LEGACY_PIT_INT);
1163 rtc_irq = qdev_get_gpio_in(hpet, HPET_LEGACY_RTC_INT);
1164 }
1165 }
1166 *rtc_state = rtc_init(isa_bus, 2000, rtc_irq);
1167
1168 qemu_register_boot_set(pc_boot_set, *rtc_state);
1169
1170 if (!xen_enabled()) {
1171 if (kvm_irqchip_in_kernel()) {
1172 pit = kvm_pit_init(isa_bus, 0x40);
1173 } else {
1174 pit = pit_init(isa_bus, 0x40, pit_isa_irq, pit_alt_irq);
1175 }
1176 if (hpet) {
1177 /* connect PIT to output control line of the HPET */
1178 qdev_connect_gpio_out(hpet, 0, qdev_get_gpio_in(&pit->qdev, 0));
1179 }
1180 pcspk_init(isa_bus, pit);
1181 }
1182
1183 for(i = 0; i < MAX_SERIAL_PORTS; i++) {
1184 if (serial_hds[i]) {
1185 serial_isa_init(isa_bus, i, serial_hds[i]);
1186 }
1187 }
1188
1189 for(i = 0; i < MAX_PARALLEL_PORTS; i++) {
1190 if (parallel_hds[i]) {
1191 parallel_init(isa_bus, i, parallel_hds[i]);
1192 }
1193 }
1194
1195 a20_line = qemu_allocate_irqs(handle_a20_line_change,
1196 x86_env_get_cpu(first_cpu), 2);
1197 i8042 = isa_create_simple(isa_bus, "i8042");
1198 i8042_setup_a20_line(i8042, &a20_line[0]);
1199 if (!no_vmport) {
1200 vmport_init(isa_bus);
1201 vmmouse = isa_try_create(isa_bus, "vmmouse");
1202 } else {
1203 vmmouse = NULL;
1204 }
1205 if (vmmouse) {
1206 qdev_prop_set_ptr(&vmmouse->qdev, "ps2_mouse", i8042);
1207 qdev_init_nofail(&vmmouse->qdev);
1208 }
1209 port92 = isa_create_simple(isa_bus, "port92");
1210 port92_init(port92, &a20_line[1]);
1211
1212 cpu_exit_irq = qemu_allocate_irqs(cpu_request_exit, NULL, 1);
1213 DMA_init(0, cpu_exit_irq);
1214
1215 for(i = 0; i < MAX_FD; i++) {
1216 fd[i] = drive_get(IF_FLOPPY, 0, i);
1217 }
1218 *floppy = fdctrl_init_isa(isa_bus, fd);
1219 }
1220
1221 void pc_nic_init(ISABus *isa_bus, PCIBus *pci_bus)
1222 {
1223 int i;
1224
1225 for (i = 0; i < nb_nics; i++) {
1226 NICInfo *nd = &nd_table[i];
1227
1228 if (!pci_bus || (nd->model && strcmp(nd->model, "ne2k_isa") == 0)) {
1229 pc_init_ne2k_isa(isa_bus, nd);
1230 } else {
1231 pci_nic_init_nofail(nd, "e1000", NULL);
1232 }
1233 }
1234 }
1235
1236 void pc_pci_device_init(PCIBus *pci_bus)
1237 {
1238 int max_bus;
1239 int bus;
1240
1241 max_bus = drive_get_max_bus(IF_SCSI);
1242 for (bus = 0; bus <= max_bus; bus++) {
1243 pci_create_simple(pci_bus, -1, "lsi53c895a");
1244 }
1245 }
1246
1247 void ioapic_init_gsi(GSIState *gsi_state, const char *parent_name)
1248 {
1249 DeviceState *dev;
1250 SysBusDevice *d;
1251 unsigned int i;
1252
1253 if (kvm_irqchip_in_kernel()) {
1254 dev = qdev_create(NULL, "kvm-ioapic");
1255 } else {
1256 dev = qdev_create(NULL, "ioapic");
1257 }
1258 if (parent_name) {
1259 object_property_add_child(object_resolve_path(parent_name, NULL),
1260 "ioapic", OBJECT(dev), NULL);
1261 }
1262 qdev_init_nofail(dev);
1263 d = SYS_BUS_DEVICE(dev);
1264 sysbus_mmio_map(d, 0, IO_APIC_DEFAULT_ADDRESS);
1265
1266 for (i = 0; i < IOAPIC_NUM_PINS; i++) {
1267 gsi_state->ioapic_irq[i] = qdev_get_gpio_in(dev, i);
1268 }
1269 }