Merge remote-tracking branch 'mst/tags/for_upstream' into staging
[qemu/kevin.git] / hw / i386 / pc.c
blob816bfa872ce3de10ccb38fa0410ffdf3c60acfec
1 /*
2 * QEMU PC System Emulator
4 * Copyright (c) 2003-2004 Fabrice Bellard
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:
13 * The above copyright notice and this permission notice shall be included in
14 * all copies or substantial portions of the Software.
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.
24 #include "qemu/osdep.h"
25 #include "hw/hw.h"
26 #include "hw/i386/pc.h"
27 #include "hw/char/serial.h"
28 #include "hw/i386/apic.h"
29 #include "hw/i386/topology.h"
30 #include "sysemu/cpus.h"
31 #include "hw/block/fdc.h"
32 #include "hw/ide.h"
33 #include "hw/pci/pci.h"
34 #include "hw/pci/pci_bus.h"
35 #include "hw/nvram/fw_cfg.h"
36 #include "hw/timer/hpet.h"
37 #include "hw/smbios/smbios.h"
38 #include "hw/loader.h"
39 #include "elf.h"
40 #include "multiboot.h"
41 #include "hw/timer/mc146818rtc.h"
42 #include "hw/timer/i8254.h"
43 #include "hw/audio/pcspk.h"
44 #include "hw/pci/msi.h"
45 #include "hw/sysbus.h"
46 #include "sysemu/sysemu.h"
47 #include "sysemu/numa.h"
48 #include "sysemu/kvm.h"
49 #include "sysemu/qtest.h"
50 #include "kvm_i386.h"
51 #include "hw/xen/xen.h"
52 #include "sysemu/block-backend.h"
53 #include "hw/block/block.h"
54 #include "ui/qemu-spice.h"
55 #include "exec/memory.h"
56 #include "exec/address-spaces.h"
57 #include "sysemu/arch_init.h"
58 #include "qemu/bitmap.h"
59 #include "qemu/config-file.h"
60 #include "qemu/error-report.h"
61 #include "hw/acpi/acpi.h"
62 #include "hw/acpi/cpu_hotplug.h"
63 #include "hw/boards.h"
64 #include "hw/pci/pci_host.h"
65 #include "acpi-build.h"
66 #include "hw/mem/pc-dimm.h"
67 #include "qapi/visitor.h"
68 #include "qapi-visit.h"
69 #include "qom/cpu.h"
70 #include "hw/nmi.h"
71 #include "hw/i386/intel_iommu.h"
73 /* debug PC/ISA interrupts */
74 //#define DEBUG_IRQ
76 #ifdef DEBUG_IRQ
77 #define DPRINTF(fmt, ...) \
78 do { printf("CPUIRQ: " fmt , ## __VA_ARGS__); } while (0)
79 #else
80 #define DPRINTF(fmt, ...)
81 #endif
83 #define FW_CFG_ACPI_TABLES (FW_CFG_ARCH_LOCAL + 0)
84 #define FW_CFG_SMBIOS_ENTRIES (FW_CFG_ARCH_LOCAL + 1)
85 #define FW_CFG_IRQ0_OVERRIDE (FW_CFG_ARCH_LOCAL + 2)
86 #define FW_CFG_E820_TABLE (FW_CFG_ARCH_LOCAL + 3)
87 #define FW_CFG_HPET (FW_CFG_ARCH_LOCAL + 4)
89 #define E820_NR_ENTRIES 16
91 struct e820_entry {
92 uint64_t address;
93 uint64_t length;
94 uint32_t type;
95 } QEMU_PACKED __attribute((__aligned__(4)));
97 struct e820_table {
98 uint32_t count;
99 struct e820_entry entry[E820_NR_ENTRIES];
100 } QEMU_PACKED __attribute((__aligned__(4)));
102 static struct e820_table e820_reserve;
103 static struct e820_entry *e820_table;
104 static unsigned e820_entries;
105 struct hpet_fw_config hpet_cfg = {.count = UINT8_MAX};
107 void gsi_handler(void *opaque, int n, int level)
109 GSIState *s = opaque;
111 DPRINTF("pc: %s GSI %d\n", level ? "raising" : "lowering", n);
112 if (n < ISA_NUM_IRQS) {
113 qemu_set_irq(s->i8259_irq[n], level);
115 qemu_set_irq(s->ioapic_irq[n], level);
118 static void ioport80_write(void *opaque, hwaddr addr, uint64_t data,
119 unsigned size)
123 static uint64_t ioport80_read(void *opaque, hwaddr addr, unsigned size)
125 return 0xffffffffffffffffULL;
128 /* MSDOS compatibility mode FPU exception support */
129 static qemu_irq ferr_irq;
131 void pc_register_ferr_irq(qemu_irq irq)
133 ferr_irq = irq;
136 /* XXX: add IGNNE support */
137 void cpu_set_ferr(CPUX86State *s)
139 qemu_irq_raise(ferr_irq);
142 static void ioportF0_write(void *opaque, hwaddr addr, uint64_t data,
143 unsigned size)
145 qemu_irq_lower(ferr_irq);
148 static uint64_t ioportF0_read(void *opaque, hwaddr addr, unsigned size)
150 return 0xffffffffffffffffULL;
153 /* TSC handling */
154 uint64_t cpu_get_tsc(CPUX86State *env)
156 return cpu_get_ticks();
159 /* IRQ handling */
160 int cpu_get_pic_interrupt(CPUX86State *env)
162 X86CPU *cpu = x86_env_get_cpu(env);
163 int intno;
165 if (!kvm_irqchip_in_kernel()) {
166 intno = apic_get_interrupt(cpu->apic_state);
167 if (intno >= 0) {
168 return intno;
170 /* read the irq from the PIC */
171 if (!apic_accept_pic_intr(cpu->apic_state)) {
172 return -1;
176 intno = pic_read_irq(isa_pic);
177 return intno;
180 static void pic_irq_request(void *opaque, int irq, int level)
182 CPUState *cs = first_cpu;
183 X86CPU *cpu = X86_CPU(cs);
185 DPRINTF("pic_irqs: %s irq %d\n", level? "raise" : "lower", irq);
186 if (cpu->apic_state && !kvm_irqchip_in_kernel()) {
187 CPU_FOREACH(cs) {
188 cpu = X86_CPU(cs);
189 if (apic_accept_pic_intr(cpu->apic_state)) {
190 apic_deliver_pic_intr(cpu->apic_state, level);
193 } else {
194 if (level) {
195 cpu_interrupt(cs, CPU_INTERRUPT_HARD);
196 } else {
197 cpu_reset_interrupt(cs, CPU_INTERRUPT_HARD);
202 /* PC cmos mappings */
204 #define REG_EQUIPMENT_BYTE 0x14
206 int cmos_get_fd_drive_type(FloppyDriveType fd0)
208 int val;
210 switch (fd0) {
211 case FLOPPY_DRIVE_TYPE_144:
212 /* 1.44 Mb 3"5 drive */
213 val = 4;
214 break;
215 case FLOPPY_DRIVE_TYPE_288:
216 /* 2.88 Mb 3"5 drive */
217 val = 5;
218 break;
219 case FLOPPY_DRIVE_TYPE_120:
220 /* 1.2 Mb 5"5 drive */
221 val = 2;
222 break;
223 case FLOPPY_DRIVE_TYPE_NONE:
224 default:
225 val = 0;
226 break;
228 return val;
231 static void cmos_init_hd(ISADevice *s, int type_ofs, int info_ofs,
232 int16_t cylinders, int8_t heads, int8_t sectors)
234 rtc_set_memory(s, type_ofs, 47);
235 rtc_set_memory(s, info_ofs, cylinders);
236 rtc_set_memory(s, info_ofs + 1, cylinders >> 8);
237 rtc_set_memory(s, info_ofs + 2, heads);
238 rtc_set_memory(s, info_ofs + 3, 0xff);
239 rtc_set_memory(s, info_ofs + 4, 0xff);
240 rtc_set_memory(s, info_ofs + 5, 0xc0 | ((heads > 8) << 3));
241 rtc_set_memory(s, info_ofs + 6, cylinders);
242 rtc_set_memory(s, info_ofs + 7, cylinders >> 8);
243 rtc_set_memory(s, info_ofs + 8, sectors);
246 /* convert boot_device letter to something recognizable by the bios */
247 static int boot_device2nibble(char boot_device)
249 switch(boot_device) {
250 case 'a':
251 case 'b':
252 return 0x01; /* floppy boot */
253 case 'c':
254 return 0x02; /* hard drive boot */
255 case 'd':
256 return 0x03; /* CD-ROM boot */
257 case 'n':
258 return 0x04; /* Network boot */
260 return 0;
263 static void set_boot_dev(ISADevice *s, const char *boot_device, Error **errp)
265 #define PC_MAX_BOOT_DEVICES 3
266 int nbds, bds[3] = { 0, };
267 int i;
269 nbds = strlen(boot_device);
270 if (nbds > PC_MAX_BOOT_DEVICES) {
271 error_setg(errp, "Too many boot devices for PC");
272 return;
274 for (i = 0; i < nbds; i++) {
275 bds[i] = boot_device2nibble(boot_device[i]);
276 if (bds[i] == 0) {
277 error_setg(errp, "Invalid boot device for PC: '%c'",
278 boot_device[i]);
279 return;
282 rtc_set_memory(s, 0x3d, (bds[1] << 4) | bds[0]);
283 rtc_set_memory(s, 0x38, (bds[2] << 4) | (fd_bootchk ? 0x0 : 0x1));
286 static void pc_boot_set(void *opaque, const char *boot_device, Error **errp)
288 set_boot_dev(opaque, boot_device, errp);
291 static void pc_cmos_init_floppy(ISADevice *rtc_state, ISADevice *floppy)
293 int val, nb, i;
294 FloppyDriveType fd_type[2] = { FLOPPY_DRIVE_TYPE_NONE,
295 FLOPPY_DRIVE_TYPE_NONE };
297 /* floppy type */
298 if (floppy) {
299 for (i = 0; i < 2; i++) {
300 fd_type[i] = isa_fdc_get_drive_type(floppy, i);
303 val = (cmos_get_fd_drive_type(fd_type[0]) << 4) |
304 cmos_get_fd_drive_type(fd_type[1]);
305 rtc_set_memory(rtc_state, 0x10, val);
307 val = rtc_get_memory(rtc_state, REG_EQUIPMENT_BYTE);
308 nb = 0;
309 if (fd_type[0] != FLOPPY_DRIVE_TYPE_NONE) {
310 nb++;
312 if (fd_type[1] != FLOPPY_DRIVE_TYPE_NONE) {
313 nb++;
315 switch (nb) {
316 case 0:
317 break;
318 case 1:
319 val |= 0x01; /* 1 drive, ready for boot */
320 break;
321 case 2:
322 val |= 0x41; /* 2 drives, ready for boot */
323 break;
325 rtc_set_memory(rtc_state, REG_EQUIPMENT_BYTE, val);
328 typedef struct pc_cmos_init_late_arg {
329 ISADevice *rtc_state;
330 BusState *idebus[2];
331 } pc_cmos_init_late_arg;
333 typedef struct check_fdc_state {
334 ISADevice *floppy;
335 bool multiple;
336 } CheckFdcState;
338 static int check_fdc(Object *obj, void *opaque)
340 CheckFdcState *state = opaque;
341 Object *fdc;
342 uint32_t iobase;
343 Error *local_err = NULL;
345 fdc = object_dynamic_cast(obj, TYPE_ISA_FDC);
346 if (!fdc) {
347 return 0;
350 iobase = object_property_get_int(obj, "iobase", &local_err);
351 if (local_err || iobase != 0x3f0) {
352 error_free(local_err);
353 return 0;
356 if (state->floppy) {
357 state->multiple = true;
358 } else {
359 state->floppy = ISA_DEVICE(obj);
361 return 0;
364 static const char * const fdc_container_path[] = {
365 "/unattached", "/peripheral", "/peripheral-anon"
369 * Locate the FDC at IO address 0x3f0, in order to configure the CMOS registers
370 * and ACPI objects.
372 ISADevice *pc_find_fdc0(void)
374 int i;
375 Object *container;
376 CheckFdcState state = { 0 };
378 for (i = 0; i < ARRAY_SIZE(fdc_container_path); i++) {
379 container = container_get(qdev_get_machine(), fdc_container_path[i]);
380 object_child_foreach(container, check_fdc, &state);
383 if (state.multiple) {
384 error_report("warning: multiple floppy disk controllers with "
385 "iobase=0x3f0 have been found");
386 error_printf("the one being picked for CMOS setup might not reflect "
387 "your intent\n");
390 return state.floppy;
393 static void pc_cmos_init_late(void *opaque)
395 pc_cmos_init_late_arg *arg = opaque;
396 ISADevice *s = arg->rtc_state;
397 int16_t cylinders;
398 int8_t heads, sectors;
399 int val;
400 int i, trans;
402 val = 0;
403 if (arg->idebus[0] && ide_get_geometry(arg->idebus[0], 0,
404 &cylinders, &heads, &sectors) >= 0) {
405 cmos_init_hd(s, 0x19, 0x1b, cylinders, heads, sectors);
406 val |= 0xf0;
408 if (arg->idebus[0] && ide_get_geometry(arg->idebus[0], 1,
409 &cylinders, &heads, &sectors) >= 0) {
410 cmos_init_hd(s, 0x1a, 0x24, cylinders, heads, sectors);
411 val |= 0x0f;
413 rtc_set_memory(s, 0x12, val);
415 val = 0;
416 for (i = 0; i < 4; i++) {
417 /* NOTE: ide_get_geometry() returns the physical
418 geometry. It is always such that: 1 <= sects <= 63, 1
419 <= heads <= 16, 1 <= cylinders <= 16383. The BIOS
420 geometry can be different if a translation is done. */
421 if (arg->idebus[i / 2] &&
422 ide_get_geometry(arg->idebus[i / 2], i % 2,
423 &cylinders, &heads, &sectors) >= 0) {
424 trans = ide_get_bios_chs_trans(arg->idebus[i / 2], i % 2) - 1;
425 assert((trans & ~3) == 0);
426 val |= trans << (i * 2);
429 rtc_set_memory(s, 0x39, val);
431 pc_cmos_init_floppy(s, pc_find_fdc0());
433 qemu_unregister_reset(pc_cmos_init_late, opaque);
436 void pc_cmos_init(PCMachineState *pcms,
437 BusState *idebus0, BusState *idebus1,
438 ISADevice *s)
440 int val;
441 static pc_cmos_init_late_arg arg;
443 /* various important CMOS locations needed by PC/Bochs bios */
445 /* memory size */
446 /* base memory (first MiB) */
447 val = MIN(pcms->below_4g_mem_size / 1024, 640);
448 rtc_set_memory(s, 0x15, val);
449 rtc_set_memory(s, 0x16, val >> 8);
450 /* extended memory (next 64MiB) */
451 if (pcms->below_4g_mem_size > 1024 * 1024) {
452 val = (pcms->below_4g_mem_size - 1024 * 1024) / 1024;
453 } else {
454 val = 0;
456 if (val > 65535)
457 val = 65535;
458 rtc_set_memory(s, 0x17, val);
459 rtc_set_memory(s, 0x18, val >> 8);
460 rtc_set_memory(s, 0x30, val);
461 rtc_set_memory(s, 0x31, val >> 8);
462 /* memory between 16MiB and 4GiB */
463 if (pcms->below_4g_mem_size > 16 * 1024 * 1024) {
464 val = (pcms->below_4g_mem_size - 16 * 1024 * 1024) / 65536;
465 } else {
466 val = 0;
468 if (val > 65535)
469 val = 65535;
470 rtc_set_memory(s, 0x34, val);
471 rtc_set_memory(s, 0x35, val >> 8);
472 /* memory above 4GiB */
473 val = pcms->above_4g_mem_size / 65536;
474 rtc_set_memory(s, 0x5b, val);
475 rtc_set_memory(s, 0x5c, val >> 8);
476 rtc_set_memory(s, 0x5d, val >> 16);
478 object_property_add_link(OBJECT(pcms), "rtc_state",
479 TYPE_ISA_DEVICE,
480 (Object **)&pcms->rtc,
481 object_property_allow_set_link,
482 OBJ_PROP_LINK_UNREF_ON_RELEASE, &error_abort);
483 object_property_set_link(OBJECT(pcms), OBJECT(s),
484 "rtc_state", &error_abort);
486 set_boot_dev(s, MACHINE(pcms)->boot_order, &error_fatal);
488 val = 0;
489 val |= 0x02; /* FPU is there */
490 val |= 0x04; /* PS/2 mouse installed */
491 rtc_set_memory(s, REG_EQUIPMENT_BYTE, val);
493 /* hard drives and FDC */
494 arg.rtc_state = s;
495 arg.idebus[0] = idebus0;
496 arg.idebus[1] = idebus1;
497 qemu_register_reset(pc_cmos_init_late, &arg);
500 #define TYPE_PORT92 "port92"
501 #define PORT92(obj) OBJECT_CHECK(Port92State, (obj), TYPE_PORT92)
503 /* port 92 stuff: could be split off */
504 typedef struct Port92State {
505 ISADevice parent_obj;
507 MemoryRegion io;
508 uint8_t outport;
509 qemu_irq a20_out;
510 } Port92State;
512 static void port92_write(void *opaque, hwaddr addr, uint64_t val,
513 unsigned size)
515 Port92State *s = opaque;
516 int oldval = s->outport;
518 DPRINTF("port92: write 0x%02" PRIx64 "\n", val);
519 s->outport = val;
520 qemu_set_irq(s->a20_out, (val >> 1) & 1);
521 if ((val & 1) && !(oldval & 1)) {
522 qemu_system_reset_request();
526 static uint64_t port92_read(void *opaque, hwaddr addr,
527 unsigned size)
529 Port92State *s = opaque;
530 uint32_t ret;
532 ret = s->outport;
533 DPRINTF("port92: read 0x%02x\n", ret);
534 return ret;
537 static void port92_init(ISADevice *dev, qemu_irq a20_out)
539 qdev_connect_gpio_out_named(DEVICE(dev), PORT92_A20_LINE, 0, a20_out);
542 static const VMStateDescription vmstate_port92_isa = {
543 .name = "port92",
544 .version_id = 1,
545 .minimum_version_id = 1,
546 .fields = (VMStateField[]) {
547 VMSTATE_UINT8(outport, Port92State),
548 VMSTATE_END_OF_LIST()
552 static void port92_reset(DeviceState *d)
554 Port92State *s = PORT92(d);
556 s->outport &= ~1;
559 static const MemoryRegionOps port92_ops = {
560 .read = port92_read,
561 .write = port92_write,
562 .impl = {
563 .min_access_size = 1,
564 .max_access_size = 1,
566 .endianness = DEVICE_LITTLE_ENDIAN,
569 static void port92_initfn(Object *obj)
571 Port92State *s = PORT92(obj);
573 memory_region_init_io(&s->io, OBJECT(s), &port92_ops, s, "port92", 1);
575 s->outport = 0;
577 qdev_init_gpio_out_named(DEVICE(obj), &s->a20_out, PORT92_A20_LINE, 1);
580 static void port92_realizefn(DeviceState *dev, Error **errp)
582 ISADevice *isadev = ISA_DEVICE(dev);
583 Port92State *s = PORT92(dev);
585 isa_register_ioport(isadev, &s->io, 0x92);
588 static void port92_class_initfn(ObjectClass *klass, void *data)
590 DeviceClass *dc = DEVICE_CLASS(klass);
592 dc->realize = port92_realizefn;
593 dc->reset = port92_reset;
594 dc->vmsd = &vmstate_port92_isa;
596 * Reason: unlike ordinary ISA devices, this one needs additional
597 * wiring: its A20 output line needs to be wired up by
598 * port92_init().
600 dc->user_creatable = false;
603 static const TypeInfo port92_info = {
604 .name = TYPE_PORT92,
605 .parent = TYPE_ISA_DEVICE,
606 .instance_size = sizeof(Port92State),
607 .instance_init = port92_initfn,
608 .class_init = port92_class_initfn,
611 static void port92_register_types(void)
613 type_register_static(&port92_info);
616 type_init(port92_register_types)
618 static void handle_a20_line_change(void *opaque, int irq, int level)
620 X86CPU *cpu = opaque;
622 /* XXX: send to all CPUs ? */
623 /* XXX: add logic to handle multiple A20 line sources */
624 x86_cpu_set_a20(cpu, level);
627 int e820_add_entry(uint64_t address, uint64_t length, uint32_t type)
629 int index = le32_to_cpu(e820_reserve.count);
630 struct e820_entry *entry;
632 if (type != E820_RAM) {
633 /* old FW_CFG_E820_TABLE entry -- reservations only */
634 if (index >= E820_NR_ENTRIES) {
635 return -EBUSY;
637 entry = &e820_reserve.entry[index++];
639 entry->address = cpu_to_le64(address);
640 entry->length = cpu_to_le64(length);
641 entry->type = cpu_to_le32(type);
643 e820_reserve.count = cpu_to_le32(index);
646 /* new "etc/e820" file -- include ram too */
647 e820_table = g_renew(struct e820_entry, e820_table, e820_entries + 1);
648 e820_table[e820_entries].address = cpu_to_le64(address);
649 e820_table[e820_entries].length = cpu_to_le64(length);
650 e820_table[e820_entries].type = cpu_to_le32(type);
651 e820_entries++;
653 return e820_entries;
656 int e820_get_num_entries(void)
658 return e820_entries;
661 bool e820_get_entry(int idx, uint32_t type, uint64_t *address, uint64_t *length)
663 if (idx < e820_entries && e820_table[idx].type == cpu_to_le32(type)) {
664 *address = le64_to_cpu(e820_table[idx].address);
665 *length = le64_to_cpu(e820_table[idx].length);
666 return true;
668 return false;
671 /* Enables contiguous-apic-ID mode, for compatibility */
672 static bool compat_apic_id_mode;
674 void enable_compat_apic_id_mode(void)
676 compat_apic_id_mode = true;
679 /* Calculates initial APIC ID for a specific CPU index
681 * Currently we need to be able to calculate the APIC ID from the CPU index
682 * alone (without requiring a CPU object), as the QEMU<->Seabios interfaces have
683 * no concept of "CPU index", and the NUMA tables on fw_cfg need the APIC ID of
684 * all CPUs up to max_cpus.
686 static uint32_t x86_cpu_apic_id_from_index(unsigned int cpu_index)
688 uint32_t correct_id;
689 static bool warned;
691 correct_id = x86_apicid_from_cpu_idx(smp_cores, smp_threads, cpu_index);
692 if (compat_apic_id_mode) {
693 if (cpu_index != correct_id && !warned && !qtest_enabled()) {
694 error_report("APIC IDs set in compatibility mode, "
695 "CPU topology won't match the configuration");
696 warned = true;
698 return cpu_index;
699 } else {
700 return correct_id;
704 static void pc_build_smbios(PCMachineState *pcms)
706 uint8_t *smbios_tables, *smbios_anchor;
707 size_t smbios_tables_len, smbios_anchor_len;
708 struct smbios_phys_mem_area *mem_array;
709 unsigned i, array_count;
710 MachineState *ms = MACHINE(pcms);
711 X86CPU *cpu = X86_CPU(ms->possible_cpus->cpus[0].cpu);
713 /* tell smbios about cpuid version and features */
714 smbios_set_cpuid(cpu->env.cpuid_version, cpu->env.features[FEAT_1_EDX]);
716 smbios_tables = smbios_get_table_legacy(&smbios_tables_len);
717 if (smbios_tables) {
718 fw_cfg_add_bytes(pcms->fw_cfg, FW_CFG_SMBIOS_ENTRIES,
719 smbios_tables, smbios_tables_len);
722 /* build the array of physical mem area from e820 table */
723 mem_array = g_malloc0(sizeof(*mem_array) * e820_get_num_entries());
724 for (i = 0, array_count = 0; i < e820_get_num_entries(); i++) {
725 uint64_t addr, len;
727 if (e820_get_entry(i, E820_RAM, &addr, &len)) {
728 mem_array[array_count].address = addr;
729 mem_array[array_count].length = len;
730 array_count++;
733 smbios_get_tables(mem_array, array_count,
734 &smbios_tables, &smbios_tables_len,
735 &smbios_anchor, &smbios_anchor_len);
736 g_free(mem_array);
738 if (smbios_anchor) {
739 fw_cfg_add_file(pcms->fw_cfg, "etc/smbios/smbios-tables",
740 smbios_tables, smbios_tables_len);
741 fw_cfg_add_file(pcms->fw_cfg, "etc/smbios/smbios-anchor",
742 smbios_anchor, smbios_anchor_len);
746 static FWCfgState *bochs_bios_init(AddressSpace *as, PCMachineState *pcms)
748 FWCfgState *fw_cfg;
749 uint64_t *numa_fw_cfg;
750 int i;
751 const CPUArchIdList *cpus;
752 MachineClass *mc = MACHINE_GET_CLASS(pcms);
754 fw_cfg = fw_cfg_init_io_dma(FW_CFG_IO_BASE, FW_CFG_IO_BASE + 4, as);
755 fw_cfg_add_i16(fw_cfg, FW_CFG_NB_CPUS, pcms->boot_cpus);
757 /* FW_CFG_MAX_CPUS is a bit confusing/problematic on x86:
759 * For machine types prior to 1.8, SeaBIOS needs FW_CFG_MAX_CPUS for
760 * building MPTable, ACPI MADT, ACPI CPU hotplug and ACPI SRAT table,
761 * that tables are based on xAPIC ID and QEMU<->SeaBIOS interface
762 * for CPU hotplug also uses APIC ID and not "CPU index".
763 * This means that FW_CFG_MAX_CPUS is not the "maximum number of CPUs",
764 * but the "limit to the APIC ID values SeaBIOS may see".
766 * So for compatibility reasons with old BIOSes we are stuck with
767 * "etc/max-cpus" actually being apic_id_limit
769 fw_cfg_add_i16(fw_cfg, FW_CFG_MAX_CPUS, (uint16_t)pcms->apic_id_limit);
770 fw_cfg_add_i64(fw_cfg, FW_CFG_RAM_SIZE, (uint64_t)ram_size);
771 fw_cfg_add_bytes(fw_cfg, FW_CFG_ACPI_TABLES,
772 acpi_tables, acpi_tables_len);
773 fw_cfg_add_i32(fw_cfg, FW_CFG_IRQ0_OVERRIDE, kvm_allows_irq0_override());
775 fw_cfg_add_bytes(fw_cfg, FW_CFG_E820_TABLE,
776 &e820_reserve, sizeof(e820_reserve));
777 fw_cfg_add_file(fw_cfg, "etc/e820", e820_table,
778 sizeof(struct e820_entry) * e820_entries);
780 fw_cfg_add_bytes(fw_cfg, FW_CFG_HPET, &hpet_cfg, sizeof(hpet_cfg));
781 /* allocate memory for the NUMA channel: one (64bit) word for the number
782 * of nodes, one word for each VCPU->node and one word for each node to
783 * hold the amount of memory.
785 numa_fw_cfg = g_new0(uint64_t, 1 + pcms->apic_id_limit + nb_numa_nodes);
786 numa_fw_cfg[0] = cpu_to_le64(nb_numa_nodes);
787 cpus = mc->possible_cpu_arch_ids(MACHINE(pcms));
788 for (i = 0; i < cpus->len; i++) {
789 unsigned int apic_id = cpus->cpus[i].arch_id;
790 assert(apic_id < pcms->apic_id_limit);
791 if (cpus->cpus[i].props.has_node_id) {
792 numa_fw_cfg[apic_id + 1] = cpu_to_le64(cpus->cpus[i].props.node_id);
795 for (i = 0; i < nb_numa_nodes; i++) {
796 numa_fw_cfg[pcms->apic_id_limit + 1 + i] =
797 cpu_to_le64(numa_info[i].node_mem);
799 fw_cfg_add_bytes(fw_cfg, FW_CFG_NUMA, numa_fw_cfg,
800 (1 + pcms->apic_id_limit + nb_numa_nodes) *
801 sizeof(*numa_fw_cfg));
803 return fw_cfg;
806 static long get_file_size(FILE *f)
808 long where, size;
810 /* XXX: on Unix systems, using fstat() probably makes more sense */
812 where = ftell(f);
813 fseek(f, 0, SEEK_END);
814 size = ftell(f);
815 fseek(f, where, SEEK_SET);
817 return size;
820 /* setup_data types */
821 #define SETUP_NONE 0
822 #define SETUP_E820_EXT 1
823 #define SETUP_DTB 2
824 #define SETUP_PCI 3
825 #define SETUP_EFI 4
827 struct setup_data {
828 uint64_t next;
829 uint32_t type;
830 uint32_t len;
831 uint8_t data[0];
832 } __attribute__((packed));
834 static void load_linux(PCMachineState *pcms,
835 FWCfgState *fw_cfg)
837 uint16_t protocol;
838 int setup_size, kernel_size, initrd_size = 0, cmdline_size;
839 int dtb_size, setup_data_offset;
840 uint32_t initrd_max;
841 uint8_t header[8192], *setup, *kernel, *initrd_data;
842 hwaddr real_addr, prot_addr, cmdline_addr, initrd_addr = 0;
843 FILE *f;
844 char *vmode;
845 MachineState *machine = MACHINE(pcms);
846 PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
847 struct setup_data *setup_data;
848 const char *kernel_filename = machine->kernel_filename;
849 const char *initrd_filename = machine->initrd_filename;
850 const char *dtb_filename = machine->dtb;
851 const char *kernel_cmdline = machine->kernel_cmdline;
853 /* Align to 16 bytes as a paranoia measure */
854 cmdline_size = (strlen(kernel_cmdline)+16) & ~15;
856 /* load the kernel header */
857 f = fopen(kernel_filename, "rb");
858 if (!f || !(kernel_size = get_file_size(f)) ||
859 fread(header, 1, MIN(ARRAY_SIZE(header), kernel_size), f) !=
860 MIN(ARRAY_SIZE(header), kernel_size)) {
861 fprintf(stderr, "qemu: could not load kernel '%s': %s\n",
862 kernel_filename, strerror(errno));
863 exit(1);
866 /* kernel protocol version */
867 #if 0
868 fprintf(stderr, "header magic: %#x\n", ldl_p(header+0x202));
869 #endif
870 if (ldl_p(header+0x202) == 0x53726448) {
871 protocol = lduw_p(header+0x206);
872 } else {
873 /* This looks like a multiboot kernel. If it is, let's stop
874 treating it like a Linux kernel. */
875 if (load_multiboot(fw_cfg, f, kernel_filename, initrd_filename,
876 kernel_cmdline, kernel_size, header)) {
877 return;
879 protocol = 0;
882 if (protocol < 0x200 || !(header[0x211] & 0x01)) {
883 /* Low kernel */
884 real_addr = 0x90000;
885 cmdline_addr = 0x9a000 - cmdline_size;
886 prot_addr = 0x10000;
887 } else if (protocol < 0x202) {
888 /* High but ancient kernel */
889 real_addr = 0x90000;
890 cmdline_addr = 0x9a000 - cmdline_size;
891 prot_addr = 0x100000;
892 } else {
893 /* High and recent kernel */
894 real_addr = 0x10000;
895 cmdline_addr = 0x20000;
896 prot_addr = 0x100000;
899 #if 0
900 fprintf(stderr,
901 "qemu: real_addr = 0x" TARGET_FMT_plx "\n"
902 "qemu: cmdline_addr = 0x" TARGET_FMT_plx "\n"
903 "qemu: prot_addr = 0x" TARGET_FMT_plx "\n",
904 real_addr,
905 cmdline_addr,
906 prot_addr);
907 #endif
909 /* highest address for loading the initrd */
910 if (protocol >= 0x203) {
911 initrd_max = ldl_p(header+0x22c);
912 } else {
913 initrd_max = 0x37ffffff;
916 if (initrd_max >= pcms->below_4g_mem_size - pcmc->acpi_data_size) {
917 initrd_max = pcms->below_4g_mem_size - pcmc->acpi_data_size - 1;
920 fw_cfg_add_i32(fw_cfg, FW_CFG_CMDLINE_ADDR, cmdline_addr);
921 fw_cfg_add_i32(fw_cfg, FW_CFG_CMDLINE_SIZE, strlen(kernel_cmdline)+1);
922 fw_cfg_add_string(fw_cfg, FW_CFG_CMDLINE_DATA, kernel_cmdline);
924 if (protocol >= 0x202) {
925 stl_p(header+0x228, cmdline_addr);
926 } else {
927 stw_p(header+0x20, 0xA33F);
928 stw_p(header+0x22, cmdline_addr-real_addr);
931 /* handle vga= parameter */
932 vmode = strstr(kernel_cmdline, "vga=");
933 if (vmode) {
934 unsigned int video_mode;
935 /* skip "vga=" */
936 vmode += 4;
937 if (!strncmp(vmode, "normal", 6)) {
938 video_mode = 0xffff;
939 } else if (!strncmp(vmode, "ext", 3)) {
940 video_mode = 0xfffe;
941 } else if (!strncmp(vmode, "ask", 3)) {
942 video_mode = 0xfffd;
943 } else {
944 video_mode = strtol(vmode, NULL, 0);
946 stw_p(header+0x1fa, video_mode);
949 /* loader type */
950 /* High nybble = B reserved for QEMU; low nybble is revision number.
951 If this code is substantially changed, you may want to consider
952 incrementing the revision. */
953 if (protocol >= 0x200) {
954 header[0x210] = 0xB0;
956 /* heap */
957 if (protocol >= 0x201) {
958 header[0x211] |= 0x80; /* CAN_USE_HEAP */
959 stw_p(header+0x224, cmdline_addr-real_addr-0x200);
962 /* load initrd */
963 if (initrd_filename) {
964 if (protocol < 0x200) {
965 fprintf(stderr, "qemu: linux kernel too old to load a ram disk\n");
966 exit(1);
969 initrd_size = get_image_size(initrd_filename);
970 if (initrd_size < 0) {
971 fprintf(stderr, "qemu: error reading initrd %s: %s\n",
972 initrd_filename, strerror(errno));
973 exit(1);
976 initrd_addr = (initrd_max-initrd_size) & ~4095;
978 initrd_data = g_malloc(initrd_size);
979 load_image(initrd_filename, initrd_data);
981 fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_ADDR, initrd_addr);
982 fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_SIZE, initrd_size);
983 fw_cfg_add_bytes(fw_cfg, FW_CFG_INITRD_DATA, initrd_data, initrd_size);
985 stl_p(header+0x218, initrd_addr);
986 stl_p(header+0x21c, initrd_size);
989 /* load kernel and setup */
990 setup_size = header[0x1f1];
991 if (setup_size == 0) {
992 setup_size = 4;
994 setup_size = (setup_size+1)*512;
995 if (setup_size > kernel_size) {
996 fprintf(stderr, "qemu: invalid kernel header\n");
997 exit(1);
999 kernel_size -= setup_size;
1001 setup = g_malloc(setup_size);
1002 kernel = g_malloc(kernel_size);
1003 fseek(f, 0, SEEK_SET);
1004 if (fread(setup, 1, setup_size, f) != setup_size) {
1005 fprintf(stderr, "fread() failed\n");
1006 exit(1);
1008 if (fread(kernel, 1, kernel_size, f) != kernel_size) {
1009 fprintf(stderr, "fread() failed\n");
1010 exit(1);
1012 fclose(f);
1014 /* append dtb to kernel */
1015 if (dtb_filename) {
1016 if (protocol < 0x209) {
1017 fprintf(stderr, "qemu: Linux kernel too old to load a dtb\n");
1018 exit(1);
1021 dtb_size = get_image_size(dtb_filename);
1022 if (dtb_size <= 0) {
1023 fprintf(stderr, "qemu: error reading dtb %s: %s\n",
1024 dtb_filename, strerror(errno));
1025 exit(1);
1028 setup_data_offset = QEMU_ALIGN_UP(kernel_size, 16);
1029 kernel_size = setup_data_offset + sizeof(struct setup_data) + dtb_size;
1030 kernel = g_realloc(kernel, kernel_size);
1032 stq_p(header+0x250, prot_addr + setup_data_offset);
1034 setup_data = (struct setup_data *)(kernel + setup_data_offset);
1035 setup_data->next = 0;
1036 setup_data->type = cpu_to_le32(SETUP_DTB);
1037 setup_data->len = cpu_to_le32(dtb_size);
1039 load_image_size(dtb_filename, setup_data->data, dtb_size);
1042 memcpy(setup, header, MIN(sizeof(header), setup_size));
1044 fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_ADDR, prot_addr);
1045 fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_SIZE, kernel_size);
1046 fw_cfg_add_bytes(fw_cfg, FW_CFG_KERNEL_DATA, kernel, kernel_size);
1048 fw_cfg_add_i32(fw_cfg, FW_CFG_SETUP_ADDR, real_addr);
1049 fw_cfg_add_i32(fw_cfg, FW_CFG_SETUP_SIZE, setup_size);
1050 fw_cfg_add_bytes(fw_cfg, FW_CFG_SETUP_DATA, setup, setup_size);
1052 option_rom[nb_option_roms].bootindex = 0;
1053 option_rom[nb_option_roms].name = "linuxboot.bin";
1054 if (pcmc->linuxboot_dma_enabled && fw_cfg_dma_enabled(fw_cfg)) {
1055 option_rom[nb_option_roms].name = "linuxboot_dma.bin";
1057 nb_option_roms++;
1060 #define NE2000_NB_MAX 6
1062 static const int ne2000_io[NE2000_NB_MAX] = { 0x300, 0x320, 0x340, 0x360,
1063 0x280, 0x380 };
1064 static const int ne2000_irq[NE2000_NB_MAX] = { 9, 10, 11, 3, 4, 5 };
1066 void pc_init_ne2k_isa(ISABus *bus, NICInfo *nd)
1068 static int nb_ne2k = 0;
1070 if (nb_ne2k == NE2000_NB_MAX)
1071 return;
1072 isa_ne2000_init(bus, ne2000_io[nb_ne2k],
1073 ne2000_irq[nb_ne2k], nd);
1074 nb_ne2k++;
1077 DeviceState *cpu_get_current_apic(void)
1079 if (current_cpu) {
1080 X86CPU *cpu = X86_CPU(current_cpu);
1081 return cpu->apic_state;
1082 } else {
1083 return NULL;
1087 void pc_acpi_smi_interrupt(void *opaque, int irq, int level)
1089 X86CPU *cpu = opaque;
1091 if (level) {
1092 cpu_interrupt(CPU(cpu), CPU_INTERRUPT_SMI);
1096 static void pc_new_cpu(const char *typename, int64_t apic_id, Error **errp)
1098 Object *cpu = NULL;
1099 Error *local_err = NULL;
1101 cpu = object_new(typename);
1103 object_property_set_int(cpu, apic_id, "apic-id", &local_err);
1104 object_property_set_bool(cpu, true, "realized", &local_err);
1106 object_unref(cpu);
1107 error_propagate(errp, local_err);
1110 void pc_hot_add_cpu(const int64_t id, Error **errp)
1112 ObjectClass *oc;
1113 MachineState *ms = MACHINE(qdev_get_machine());
1114 int64_t apic_id = x86_cpu_apic_id_from_index(id);
1115 Error *local_err = NULL;
1117 if (id < 0) {
1118 error_setg(errp, "Invalid CPU id: %" PRIi64, id);
1119 return;
1122 if (apic_id >= ACPI_CPU_HOTPLUG_ID_LIMIT) {
1123 error_setg(errp, "Unable to add CPU: %" PRIi64
1124 ", resulting APIC ID (%" PRIi64 ") is too large",
1125 id, apic_id);
1126 return;
1129 assert(ms->possible_cpus->cpus[0].cpu); /* BSP is always present */
1130 oc = OBJECT_CLASS(CPU_GET_CLASS(ms->possible_cpus->cpus[0].cpu));
1131 pc_new_cpu(object_class_get_name(oc), apic_id, &local_err);
1132 if (local_err) {
1133 error_propagate(errp, local_err);
1134 return;
1138 void pc_cpus_init(PCMachineState *pcms)
1140 int i;
1141 CPUClass *cc;
1142 ObjectClass *oc;
1143 const char *typename;
1144 gchar **model_pieces;
1145 const CPUArchIdList *possible_cpus;
1146 MachineState *machine = MACHINE(pcms);
1147 MachineClass *mc = MACHINE_GET_CLASS(pcms);
1149 /* init CPUs */
1150 if (machine->cpu_model == NULL) {
1151 #ifdef TARGET_X86_64
1152 machine->cpu_model = "qemu64";
1153 #else
1154 machine->cpu_model = "qemu32";
1155 #endif
1158 model_pieces = g_strsplit(machine->cpu_model, ",", 2);
1159 if (!model_pieces[0]) {
1160 error_report("Invalid/empty CPU model name");
1161 exit(1);
1164 oc = cpu_class_by_name(TYPE_X86_CPU, model_pieces[0]);
1165 if (oc == NULL) {
1166 error_report("Unable to find CPU definition: %s", model_pieces[0]);
1167 exit(1);
1169 typename = object_class_get_name(oc);
1170 cc = CPU_CLASS(oc);
1171 cc->parse_features(typename, model_pieces[1], &error_fatal);
1172 g_strfreev(model_pieces);
1174 /* Calculates the limit to CPU APIC ID values
1176 * Limit for the APIC ID value, so that all
1177 * CPU APIC IDs are < pcms->apic_id_limit.
1179 * This is used for FW_CFG_MAX_CPUS. See comments on bochs_bios_init().
1181 pcms->apic_id_limit = x86_cpu_apic_id_from_index(max_cpus - 1) + 1;
1182 possible_cpus = mc->possible_cpu_arch_ids(machine);
1183 for (i = 0; i < smp_cpus; i++) {
1184 pc_new_cpu(typename, possible_cpus->cpus[i].arch_id, &error_fatal);
1188 static void pc_build_feature_control_file(PCMachineState *pcms)
1190 MachineState *ms = MACHINE(pcms);
1191 X86CPU *cpu = X86_CPU(ms->possible_cpus->cpus[0].cpu);
1192 CPUX86State *env = &cpu->env;
1193 uint32_t unused, ecx, edx;
1194 uint64_t feature_control_bits = 0;
1195 uint64_t *val;
1197 cpu_x86_cpuid(env, 1, 0, &unused, &unused, &ecx, &edx);
1198 if (ecx & CPUID_EXT_VMX) {
1199 feature_control_bits |= FEATURE_CONTROL_VMXON_ENABLED_OUTSIDE_SMX;
1202 if ((edx & (CPUID_EXT2_MCE | CPUID_EXT2_MCA)) ==
1203 (CPUID_EXT2_MCE | CPUID_EXT2_MCA) &&
1204 (env->mcg_cap & MCG_LMCE_P)) {
1205 feature_control_bits |= FEATURE_CONTROL_LMCE;
1208 if (!feature_control_bits) {
1209 return;
1212 val = g_malloc(sizeof(*val));
1213 *val = cpu_to_le64(feature_control_bits | FEATURE_CONTROL_LOCKED);
1214 fw_cfg_add_file(pcms->fw_cfg, "etc/msr_feature_control", val, sizeof(*val));
1217 static void rtc_set_cpus_count(ISADevice *rtc, uint16_t cpus_count)
1219 if (cpus_count > 0xff) {
1220 /* If the number of CPUs can't be represented in 8 bits, the
1221 * BIOS must use "FW_CFG_NB_CPUS". Set RTC field to 0 just
1222 * to make old BIOSes fail more predictably.
1224 rtc_set_memory(rtc, 0x5f, 0);
1225 } else {
1226 rtc_set_memory(rtc, 0x5f, cpus_count - 1);
1230 static
1231 void pc_machine_done(Notifier *notifier, void *data)
1233 PCMachineState *pcms = container_of(notifier,
1234 PCMachineState, machine_done);
1235 PCIBus *bus = pcms->bus;
1237 /* set the number of CPUs */
1238 rtc_set_cpus_count(pcms->rtc, pcms->boot_cpus);
1240 if (bus) {
1241 int extra_hosts = 0;
1243 QLIST_FOREACH(bus, &bus->child, sibling) {
1244 /* look for expander root buses */
1245 if (pci_bus_is_root(bus)) {
1246 extra_hosts++;
1249 if (extra_hosts && pcms->fw_cfg) {
1250 uint64_t *val = g_malloc(sizeof(*val));
1251 *val = cpu_to_le64(extra_hosts);
1252 fw_cfg_add_file(pcms->fw_cfg,
1253 "etc/extra-pci-roots", val, sizeof(*val));
1257 acpi_setup();
1258 if (pcms->fw_cfg) {
1259 pc_build_smbios(pcms);
1260 pc_build_feature_control_file(pcms);
1261 /* update FW_CFG_NB_CPUS to account for -device added CPUs */
1262 fw_cfg_modify_i16(pcms->fw_cfg, FW_CFG_NB_CPUS, pcms->boot_cpus);
1265 if (pcms->apic_id_limit > 255) {
1266 IntelIOMMUState *iommu = INTEL_IOMMU_DEVICE(x86_iommu_get_default());
1268 if (!iommu || !iommu->x86_iommu.intr_supported ||
1269 iommu->intr_eim != ON_OFF_AUTO_ON) {
1270 error_report("current -smp configuration requires "
1271 "Extended Interrupt Mode enabled. "
1272 "You can add an IOMMU using: "
1273 "-device intel-iommu,intremap=on,eim=on");
1274 exit(EXIT_FAILURE);
1279 void pc_guest_info_init(PCMachineState *pcms)
1281 int i;
1283 pcms->apic_xrupt_override = kvm_allows_irq0_override();
1284 pcms->numa_nodes = nb_numa_nodes;
1285 pcms->node_mem = g_malloc0(pcms->numa_nodes *
1286 sizeof *pcms->node_mem);
1287 for (i = 0; i < nb_numa_nodes; i++) {
1288 pcms->node_mem[i] = numa_info[i].node_mem;
1291 pcms->machine_done.notify = pc_machine_done;
1292 qemu_add_machine_init_done_notifier(&pcms->machine_done);
1295 /* setup pci memory address space mapping into system address space */
1296 void pc_pci_as_mapping_init(Object *owner, MemoryRegion *system_memory,
1297 MemoryRegion *pci_address_space)
1299 /* Set to lower priority than RAM */
1300 memory_region_add_subregion_overlap(system_memory, 0x0,
1301 pci_address_space, -1);
1304 void pc_acpi_init(const char *default_dsdt)
1306 char *filename;
1308 if (acpi_tables != NULL) {
1309 /* manually set via -acpitable, leave it alone */
1310 return;
1313 filename = qemu_find_file(QEMU_FILE_TYPE_BIOS, default_dsdt);
1314 if (filename == NULL) {
1315 fprintf(stderr, "WARNING: failed to find %s\n", default_dsdt);
1316 } else {
1317 QemuOpts *opts = qemu_opts_create(qemu_find_opts("acpi"), NULL, 0,
1318 &error_abort);
1319 Error *err = NULL;
1321 qemu_opt_set(opts, "file", filename, &error_abort);
1323 acpi_table_add_builtin(opts, &err);
1324 if (err) {
1325 error_reportf_err(err, "WARNING: failed to load %s: ",
1326 filename);
1328 g_free(filename);
1332 void xen_load_linux(PCMachineState *pcms)
1334 int i;
1335 FWCfgState *fw_cfg;
1337 assert(MACHINE(pcms)->kernel_filename != NULL);
1339 fw_cfg = fw_cfg_init_io(FW_CFG_IO_BASE);
1340 fw_cfg_add_i16(fw_cfg, FW_CFG_NB_CPUS, pcms->boot_cpus);
1341 rom_set_fw(fw_cfg);
1343 load_linux(pcms, fw_cfg);
1344 for (i = 0; i < nb_option_roms; i++) {
1345 assert(!strcmp(option_rom[i].name, "linuxboot.bin") ||
1346 !strcmp(option_rom[i].name, "linuxboot_dma.bin") ||
1347 !strcmp(option_rom[i].name, "multiboot.bin"));
1348 rom_add_option(option_rom[i].name, option_rom[i].bootindex);
1350 pcms->fw_cfg = fw_cfg;
1353 void pc_memory_init(PCMachineState *pcms,
1354 MemoryRegion *system_memory,
1355 MemoryRegion *rom_memory,
1356 MemoryRegion **ram_memory)
1358 int linux_boot, i;
1359 MemoryRegion *ram, *option_rom_mr;
1360 MemoryRegion *ram_below_4g, *ram_above_4g;
1361 FWCfgState *fw_cfg;
1362 MachineState *machine = MACHINE(pcms);
1363 PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
1365 assert(machine->ram_size == pcms->below_4g_mem_size +
1366 pcms->above_4g_mem_size);
1368 linux_boot = (machine->kernel_filename != NULL);
1370 /* Allocate RAM. We allocate it as a single memory region and use
1371 * aliases to address portions of it, mostly for backwards compatibility
1372 * with older qemus that used qemu_ram_alloc().
1374 ram = g_malloc(sizeof(*ram));
1375 memory_region_allocate_system_memory(ram, NULL, "pc.ram",
1376 machine->ram_size);
1377 *ram_memory = ram;
1378 ram_below_4g = g_malloc(sizeof(*ram_below_4g));
1379 memory_region_init_alias(ram_below_4g, NULL, "ram-below-4g", ram,
1380 0, pcms->below_4g_mem_size);
1381 memory_region_add_subregion(system_memory, 0, ram_below_4g);
1382 e820_add_entry(0, pcms->below_4g_mem_size, E820_RAM);
1383 if (pcms->above_4g_mem_size > 0) {
1384 ram_above_4g = g_malloc(sizeof(*ram_above_4g));
1385 memory_region_init_alias(ram_above_4g, NULL, "ram-above-4g", ram,
1386 pcms->below_4g_mem_size,
1387 pcms->above_4g_mem_size);
1388 memory_region_add_subregion(system_memory, 0x100000000ULL,
1389 ram_above_4g);
1390 e820_add_entry(0x100000000ULL, pcms->above_4g_mem_size, E820_RAM);
1393 if (!pcmc->has_reserved_memory &&
1394 (machine->ram_slots ||
1395 (machine->maxram_size > machine->ram_size))) {
1396 MachineClass *mc = MACHINE_GET_CLASS(machine);
1398 error_report("\"-memory 'slots|maxmem'\" is not supported by: %s",
1399 mc->name);
1400 exit(EXIT_FAILURE);
1403 /* initialize hotplug memory address space */
1404 if (pcmc->has_reserved_memory &&
1405 (machine->ram_size < machine->maxram_size)) {
1406 ram_addr_t hotplug_mem_size =
1407 machine->maxram_size - machine->ram_size;
1409 if (machine->ram_slots > ACPI_MAX_RAM_SLOTS) {
1410 error_report("unsupported amount of memory slots: %"PRIu64,
1411 machine->ram_slots);
1412 exit(EXIT_FAILURE);
1415 if (QEMU_ALIGN_UP(machine->maxram_size,
1416 TARGET_PAGE_SIZE) != machine->maxram_size) {
1417 error_report("maximum memory size must by aligned to multiple of "
1418 "%d bytes", TARGET_PAGE_SIZE);
1419 exit(EXIT_FAILURE);
1422 pcms->hotplug_memory.base =
1423 ROUND_UP(0x100000000ULL + pcms->above_4g_mem_size, 1ULL << 30);
1425 if (pcmc->enforce_aligned_dimm) {
1426 /* size hotplug region assuming 1G page max alignment per slot */
1427 hotplug_mem_size += (1ULL << 30) * machine->ram_slots;
1430 if ((pcms->hotplug_memory.base + hotplug_mem_size) <
1431 hotplug_mem_size) {
1432 error_report("unsupported amount of maximum memory: " RAM_ADDR_FMT,
1433 machine->maxram_size);
1434 exit(EXIT_FAILURE);
1437 memory_region_init(&pcms->hotplug_memory.mr, OBJECT(pcms),
1438 "hotplug-memory", hotplug_mem_size);
1439 memory_region_add_subregion(system_memory, pcms->hotplug_memory.base,
1440 &pcms->hotplug_memory.mr);
1443 /* Initialize PC system firmware */
1444 pc_system_firmware_init(rom_memory, !pcmc->pci_enabled);
1446 option_rom_mr = g_malloc(sizeof(*option_rom_mr));
1447 memory_region_init_ram(option_rom_mr, NULL, "pc.rom", PC_ROM_SIZE,
1448 &error_fatal);
1449 vmstate_register_ram_global(option_rom_mr);
1450 memory_region_add_subregion_overlap(rom_memory,
1451 PC_ROM_MIN_VGA,
1452 option_rom_mr,
1455 fw_cfg = bochs_bios_init(&address_space_memory, pcms);
1457 rom_set_fw(fw_cfg);
1459 if (pcmc->has_reserved_memory && pcms->hotplug_memory.base) {
1460 uint64_t *val = g_malloc(sizeof(*val));
1461 PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
1462 uint64_t res_mem_end = pcms->hotplug_memory.base;
1464 if (!pcmc->broken_reserved_end) {
1465 res_mem_end += memory_region_size(&pcms->hotplug_memory.mr);
1467 *val = cpu_to_le64(ROUND_UP(res_mem_end, 0x1ULL << 30));
1468 fw_cfg_add_file(fw_cfg, "etc/reserved-memory-end", val, sizeof(*val));
1471 if (linux_boot) {
1472 load_linux(pcms, fw_cfg);
1475 for (i = 0; i < nb_option_roms; i++) {
1476 rom_add_option(option_rom[i].name, option_rom[i].bootindex);
1478 pcms->fw_cfg = fw_cfg;
1480 /* Init default IOAPIC address space */
1481 pcms->ioapic_as = &address_space_memory;
1484 qemu_irq pc_allocate_cpu_irq(void)
1486 return qemu_allocate_irq(pic_irq_request, NULL, 0);
1489 DeviceState *pc_vga_init(ISABus *isa_bus, PCIBus *pci_bus)
1491 DeviceState *dev = NULL;
1493 rom_set_order_override(FW_CFG_ORDER_OVERRIDE_VGA);
1494 if (pci_bus) {
1495 PCIDevice *pcidev = pci_vga_init(pci_bus);
1496 dev = pcidev ? &pcidev->qdev : NULL;
1497 } else if (isa_bus) {
1498 ISADevice *isadev = isa_vga_init(isa_bus);
1499 dev = isadev ? DEVICE(isadev) : NULL;
1501 rom_reset_order_override();
1502 return dev;
1505 static const MemoryRegionOps ioport80_io_ops = {
1506 .write = ioport80_write,
1507 .read = ioport80_read,
1508 .endianness = DEVICE_NATIVE_ENDIAN,
1509 .impl = {
1510 .min_access_size = 1,
1511 .max_access_size = 1,
1515 static const MemoryRegionOps ioportF0_io_ops = {
1516 .write = ioportF0_write,
1517 .read = ioportF0_read,
1518 .endianness = DEVICE_NATIVE_ENDIAN,
1519 .impl = {
1520 .min_access_size = 1,
1521 .max_access_size = 1,
1525 void pc_basic_device_init(ISABus *isa_bus, qemu_irq *gsi,
1526 ISADevice **rtc_state,
1527 bool create_fdctrl,
1528 bool no_vmport,
1529 bool has_pit,
1530 uint32_t hpet_irqs)
1532 int i;
1533 DriveInfo *fd[MAX_FD];
1534 DeviceState *hpet = NULL;
1535 int pit_isa_irq = 0;
1536 qemu_irq pit_alt_irq = NULL;
1537 qemu_irq rtc_irq = NULL;
1538 qemu_irq *a20_line;
1539 ISADevice *i8042, *port92, *vmmouse, *pit = NULL;
1540 MemoryRegion *ioport80_io = g_new(MemoryRegion, 1);
1541 MemoryRegion *ioportF0_io = g_new(MemoryRegion, 1);
1543 memory_region_init_io(ioport80_io, NULL, &ioport80_io_ops, NULL, "ioport80", 1);
1544 memory_region_add_subregion(isa_bus->address_space_io, 0x80, ioport80_io);
1546 memory_region_init_io(ioportF0_io, NULL, &ioportF0_io_ops, NULL, "ioportF0", 1);
1547 memory_region_add_subregion(isa_bus->address_space_io, 0xf0, ioportF0_io);
1550 * Check if an HPET shall be created.
1552 * Without KVM_CAP_PIT_STATE2, we cannot switch off the in-kernel PIT
1553 * when the HPET wants to take over. Thus we have to disable the latter.
1555 if (!no_hpet && (!kvm_irqchip_in_kernel() || kvm_has_pit_state2())) {
1556 /* In order to set property, here not using sysbus_try_create_simple */
1557 hpet = qdev_try_create(NULL, TYPE_HPET);
1558 if (hpet) {
1559 /* For pc-piix-*, hpet's intcap is always IRQ2. For pc-q35-1.7
1560 * and earlier, use IRQ2 for compat. Otherwise, use IRQ16~23,
1561 * IRQ8 and IRQ2.
1563 uint8_t compat = object_property_get_int(OBJECT(hpet),
1564 HPET_INTCAP, NULL);
1565 if (!compat) {
1566 qdev_prop_set_uint32(hpet, HPET_INTCAP, hpet_irqs);
1568 qdev_init_nofail(hpet);
1569 sysbus_mmio_map(SYS_BUS_DEVICE(hpet), 0, HPET_BASE);
1571 for (i = 0; i < GSI_NUM_PINS; i++) {
1572 sysbus_connect_irq(SYS_BUS_DEVICE(hpet), i, gsi[i]);
1574 pit_isa_irq = -1;
1575 pit_alt_irq = qdev_get_gpio_in(hpet, HPET_LEGACY_PIT_INT);
1576 rtc_irq = qdev_get_gpio_in(hpet, HPET_LEGACY_RTC_INT);
1579 *rtc_state = rtc_init(isa_bus, 2000, rtc_irq);
1581 qemu_register_boot_set(pc_boot_set, *rtc_state);
1583 if (!xen_enabled() && has_pit) {
1584 if (kvm_pit_in_kernel()) {
1585 pit = kvm_pit_init(isa_bus, 0x40);
1586 } else {
1587 pit = pit_init(isa_bus, 0x40, pit_isa_irq, pit_alt_irq);
1589 if (hpet) {
1590 /* connect PIT to output control line of the HPET */
1591 qdev_connect_gpio_out(hpet, 0, qdev_get_gpio_in(DEVICE(pit), 0));
1593 pcspk_init(isa_bus, pit);
1596 serial_hds_isa_init(isa_bus, 0, MAX_SERIAL_PORTS);
1597 parallel_hds_isa_init(isa_bus, MAX_PARALLEL_PORTS);
1599 a20_line = qemu_allocate_irqs(handle_a20_line_change, first_cpu, 2);
1600 i8042 = isa_create_simple(isa_bus, "i8042");
1601 i8042_setup_a20_line(i8042, a20_line[0]);
1602 if (!no_vmport) {
1603 vmport_init(isa_bus);
1604 vmmouse = isa_try_create(isa_bus, "vmmouse");
1605 } else {
1606 vmmouse = NULL;
1608 if (vmmouse) {
1609 DeviceState *dev = DEVICE(vmmouse);
1610 qdev_prop_set_ptr(dev, "ps2_mouse", i8042);
1611 qdev_init_nofail(dev);
1613 port92 = isa_create_simple(isa_bus, "port92");
1614 port92_init(port92, a20_line[1]);
1615 g_free(a20_line);
1617 DMA_init(isa_bus, 0);
1619 for(i = 0; i < MAX_FD; i++) {
1620 fd[i] = drive_get(IF_FLOPPY, 0, i);
1621 create_fdctrl |= !!fd[i];
1623 if (create_fdctrl) {
1624 fdctrl_init_isa(isa_bus, fd);
1628 void pc_nic_init(ISABus *isa_bus, PCIBus *pci_bus)
1630 int i;
1632 rom_set_order_override(FW_CFG_ORDER_OVERRIDE_NIC);
1633 for (i = 0; i < nb_nics; i++) {
1634 NICInfo *nd = &nd_table[i];
1636 if (!pci_bus || (nd->model && strcmp(nd->model, "ne2k_isa") == 0)) {
1637 pc_init_ne2k_isa(isa_bus, nd);
1638 } else {
1639 pci_nic_init_nofail(nd, pci_bus, "e1000", NULL);
1642 rom_reset_order_override();
1645 void pc_pci_device_init(PCIBus *pci_bus)
1647 int max_bus;
1648 int bus;
1650 /* Note: if=scsi is deprecated with PC machine types */
1651 max_bus = drive_get_max_bus(IF_SCSI);
1652 for (bus = 0; bus <= max_bus; bus++) {
1653 pci_create_simple(pci_bus, -1, "lsi53c895a");
1655 * By not creating frontends here, we make
1656 * scsi_legacy_handle_cmdline() create them, and warn that
1657 * this usage is deprecated.
1662 void ioapic_init_gsi(GSIState *gsi_state, const char *parent_name)
1664 DeviceState *dev;
1665 SysBusDevice *d;
1666 unsigned int i;
1668 if (kvm_ioapic_in_kernel()) {
1669 dev = qdev_create(NULL, "kvm-ioapic");
1670 } else {
1671 dev = qdev_create(NULL, "ioapic");
1673 if (parent_name) {
1674 object_property_add_child(object_resolve_path(parent_name, NULL),
1675 "ioapic", OBJECT(dev), NULL);
1677 qdev_init_nofail(dev);
1678 d = SYS_BUS_DEVICE(dev);
1679 sysbus_mmio_map(d, 0, IO_APIC_DEFAULT_ADDRESS);
1681 for (i = 0; i < IOAPIC_NUM_PINS; i++) {
1682 gsi_state->ioapic_irq[i] = qdev_get_gpio_in(dev, i);
1686 static void pc_dimm_plug(HotplugHandler *hotplug_dev,
1687 DeviceState *dev, Error **errp)
1689 HotplugHandlerClass *hhc;
1690 Error *local_err = NULL;
1691 PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1692 PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
1693 PCDIMMDevice *dimm = PC_DIMM(dev);
1694 PCDIMMDeviceClass *ddc = PC_DIMM_GET_CLASS(dimm);
1695 MemoryRegion *mr = ddc->get_memory_region(dimm);
1696 uint64_t align = TARGET_PAGE_SIZE;
1698 if (memory_region_get_alignment(mr) && pcmc->enforce_aligned_dimm) {
1699 align = memory_region_get_alignment(mr);
1702 if (!pcms->acpi_dev) {
1703 error_setg(&local_err,
1704 "memory hotplug is not enabled: missing acpi device");
1705 goto out;
1708 pc_dimm_memory_plug(dev, &pcms->hotplug_memory, mr, align, &local_err);
1709 if (local_err) {
1710 goto out;
1713 if (object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM)) {
1714 if (!pcms->acpi_nvdimm_state.is_enabled) {
1715 error_setg(&local_err,
1716 "nvdimm is not enabled: missing 'nvdimm' in '-M'");
1717 goto out;
1719 nvdimm_plug(&pcms->acpi_nvdimm_state);
1722 hhc = HOTPLUG_HANDLER_GET_CLASS(pcms->acpi_dev);
1723 hhc->plug(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &error_abort);
1724 out:
1725 error_propagate(errp, local_err);
1728 static void pc_dimm_unplug_request(HotplugHandler *hotplug_dev,
1729 DeviceState *dev, Error **errp)
1731 HotplugHandlerClass *hhc;
1732 Error *local_err = NULL;
1733 PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1735 if (!pcms->acpi_dev) {
1736 error_setg(&local_err,
1737 "memory hotplug is not enabled: missing acpi device");
1738 goto out;
1741 if (object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM)) {
1742 error_setg(&local_err,
1743 "nvdimm device hot unplug is not supported yet.");
1744 goto out;
1747 hhc = HOTPLUG_HANDLER_GET_CLASS(pcms->acpi_dev);
1748 hhc->unplug_request(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &local_err);
1750 out:
1751 error_propagate(errp, local_err);
1754 static void pc_dimm_unplug(HotplugHandler *hotplug_dev,
1755 DeviceState *dev, Error **errp)
1757 PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1758 PCDIMMDevice *dimm = PC_DIMM(dev);
1759 PCDIMMDeviceClass *ddc = PC_DIMM_GET_CLASS(dimm);
1760 MemoryRegion *mr = ddc->get_memory_region(dimm);
1761 HotplugHandlerClass *hhc;
1762 Error *local_err = NULL;
1764 hhc = HOTPLUG_HANDLER_GET_CLASS(pcms->acpi_dev);
1765 hhc->unplug(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &local_err);
1767 if (local_err) {
1768 goto out;
1771 pc_dimm_memory_unplug(dev, &pcms->hotplug_memory, mr);
1772 object_unparent(OBJECT(dev));
1774 out:
1775 error_propagate(errp, local_err);
1778 static int pc_apic_cmp(const void *a, const void *b)
1780 CPUArchId *apic_a = (CPUArchId *)a;
1781 CPUArchId *apic_b = (CPUArchId *)b;
1783 return apic_a->arch_id - apic_b->arch_id;
1786 /* returns pointer to CPUArchId descriptor that matches CPU's apic_id
1787 * in ms->possible_cpus->cpus, if ms->possible_cpus->cpus has no
1788 * entry corresponding to CPU's apic_id returns NULL.
1790 static CPUArchId *pc_find_cpu_slot(MachineState *ms, uint32_t id, int *idx)
1792 CPUArchId apic_id, *found_cpu;
1794 apic_id.arch_id = id;
1795 found_cpu = bsearch(&apic_id, ms->possible_cpus->cpus,
1796 ms->possible_cpus->len, sizeof(*ms->possible_cpus->cpus),
1797 pc_apic_cmp);
1798 if (found_cpu && idx) {
1799 *idx = found_cpu - ms->possible_cpus->cpus;
1801 return found_cpu;
1804 static void pc_cpu_plug(HotplugHandler *hotplug_dev,
1805 DeviceState *dev, Error **errp)
1807 CPUArchId *found_cpu;
1808 HotplugHandlerClass *hhc;
1809 Error *local_err = NULL;
1810 X86CPU *cpu = X86_CPU(dev);
1811 PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1813 if (pcms->acpi_dev) {
1814 hhc = HOTPLUG_HANDLER_GET_CLASS(pcms->acpi_dev);
1815 hhc->plug(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &local_err);
1816 if (local_err) {
1817 goto out;
1821 /* increment the number of CPUs */
1822 pcms->boot_cpus++;
1823 if (pcms->rtc) {
1824 rtc_set_cpus_count(pcms->rtc, pcms->boot_cpus);
1826 if (pcms->fw_cfg) {
1827 fw_cfg_modify_i16(pcms->fw_cfg, FW_CFG_NB_CPUS, pcms->boot_cpus);
1830 found_cpu = pc_find_cpu_slot(MACHINE(pcms), cpu->apic_id, NULL);
1831 found_cpu->cpu = OBJECT(dev);
1832 out:
1833 error_propagate(errp, local_err);
1835 static void pc_cpu_unplug_request_cb(HotplugHandler *hotplug_dev,
1836 DeviceState *dev, Error **errp)
1838 int idx = -1;
1839 HotplugHandlerClass *hhc;
1840 Error *local_err = NULL;
1841 X86CPU *cpu = X86_CPU(dev);
1842 PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1844 pc_find_cpu_slot(MACHINE(pcms), cpu->apic_id, &idx);
1845 assert(idx != -1);
1846 if (idx == 0) {
1847 error_setg(&local_err, "Boot CPU is unpluggable");
1848 goto out;
1851 hhc = HOTPLUG_HANDLER_GET_CLASS(pcms->acpi_dev);
1852 hhc->unplug_request(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &local_err);
1854 if (local_err) {
1855 goto out;
1858 out:
1859 error_propagate(errp, local_err);
1863 static void pc_cpu_unplug_cb(HotplugHandler *hotplug_dev,
1864 DeviceState *dev, Error **errp)
1866 CPUArchId *found_cpu;
1867 HotplugHandlerClass *hhc;
1868 Error *local_err = NULL;
1869 X86CPU *cpu = X86_CPU(dev);
1870 PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1872 hhc = HOTPLUG_HANDLER_GET_CLASS(pcms->acpi_dev);
1873 hhc->unplug(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &local_err);
1875 if (local_err) {
1876 goto out;
1879 found_cpu = pc_find_cpu_slot(MACHINE(pcms), cpu->apic_id, NULL);
1880 found_cpu->cpu = NULL;
1881 object_unparent(OBJECT(dev));
1883 /* decrement the number of CPUs */
1884 pcms->boot_cpus--;
1885 /* Update the number of CPUs in CMOS */
1886 rtc_set_cpus_count(pcms->rtc, pcms->boot_cpus);
1887 fw_cfg_modify_i16(pcms->fw_cfg, FW_CFG_NB_CPUS, pcms->boot_cpus);
1888 out:
1889 error_propagate(errp, local_err);
1892 static void pc_cpu_pre_plug(HotplugHandler *hotplug_dev,
1893 DeviceState *dev, Error **errp)
1895 int idx;
1896 int node_id;
1897 CPUState *cs;
1898 CPUArchId *cpu_slot;
1899 X86CPUTopoInfo topo;
1900 X86CPU *cpu = X86_CPU(dev);
1901 PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1903 /* if APIC ID is not set, set it based on socket/core/thread properties */
1904 if (cpu->apic_id == UNASSIGNED_APIC_ID) {
1905 int max_socket = (max_cpus - 1) / smp_threads / smp_cores;
1907 if (cpu->socket_id < 0) {
1908 error_setg(errp, "CPU socket-id is not set");
1909 return;
1910 } else if (cpu->socket_id > max_socket) {
1911 error_setg(errp, "Invalid CPU socket-id: %u must be in range 0:%u",
1912 cpu->socket_id, max_socket);
1913 return;
1915 if (cpu->core_id < 0) {
1916 error_setg(errp, "CPU core-id is not set");
1917 return;
1918 } else if (cpu->core_id > (smp_cores - 1)) {
1919 error_setg(errp, "Invalid CPU core-id: %u must be in range 0:%u",
1920 cpu->core_id, smp_cores - 1);
1921 return;
1923 if (cpu->thread_id < 0) {
1924 error_setg(errp, "CPU thread-id is not set");
1925 return;
1926 } else if (cpu->thread_id > (smp_threads - 1)) {
1927 error_setg(errp, "Invalid CPU thread-id: %u must be in range 0:%u",
1928 cpu->thread_id, smp_threads - 1);
1929 return;
1932 topo.pkg_id = cpu->socket_id;
1933 topo.core_id = cpu->core_id;
1934 topo.smt_id = cpu->thread_id;
1935 cpu->apic_id = apicid_from_topo_ids(smp_cores, smp_threads, &topo);
1938 cpu_slot = pc_find_cpu_slot(MACHINE(pcms), cpu->apic_id, &idx);
1939 if (!cpu_slot) {
1940 MachineState *ms = MACHINE(pcms);
1942 x86_topo_ids_from_apicid(cpu->apic_id, smp_cores, smp_threads, &topo);
1943 error_setg(errp, "Invalid CPU [socket: %u, core: %u, thread: %u] with"
1944 " APIC ID %" PRIu32 ", valid index range 0:%d",
1945 topo.pkg_id, topo.core_id, topo.smt_id, cpu->apic_id,
1946 ms->possible_cpus->len - 1);
1947 return;
1950 if (cpu_slot->cpu) {
1951 error_setg(errp, "CPU[%d] with APIC ID %" PRIu32 " exists",
1952 idx, cpu->apic_id);
1953 return;
1956 /* if 'address' properties socket-id/core-id/thread-id are not set, set them
1957 * so that machine_query_hotpluggable_cpus would show correct values
1959 /* TODO: move socket_id/core_id/thread_id checks into x86_cpu_realizefn()
1960 * once -smp refactoring is complete and there will be CPU private
1961 * CPUState::nr_cores and CPUState::nr_threads fields instead of globals */
1962 x86_topo_ids_from_apicid(cpu->apic_id, smp_cores, smp_threads, &topo);
1963 if (cpu->socket_id != -1 && cpu->socket_id != topo.pkg_id) {
1964 error_setg(errp, "property socket-id: %u doesn't match set apic-id:"
1965 " 0x%x (socket-id: %u)", cpu->socket_id, cpu->apic_id, topo.pkg_id);
1966 return;
1968 cpu->socket_id = topo.pkg_id;
1970 if (cpu->core_id != -1 && cpu->core_id != topo.core_id) {
1971 error_setg(errp, "property core-id: %u doesn't match set apic-id:"
1972 " 0x%x (core-id: %u)", cpu->core_id, cpu->apic_id, topo.core_id);
1973 return;
1975 cpu->core_id = topo.core_id;
1977 if (cpu->thread_id != -1 && cpu->thread_id != topo.smt_id) {
1978 error_setg(errp, "property thread-id: %u doesn't match set apic-id:"
1979 " 0x%x (thread-id: %u)", cpu->thread_id, cpu->apic_id, topo.smt_id);
1980 return;
1982 cpu->thread_id = topo.smt_id;
1984 cs = CPU(cpu);
1985 cs->cpu_index = idx;
1987 node_id = cpu_slot->props.node_id;
1988 if (!cpu_slot->props.has_node_id) {
1989 /* by default CPUState::numa_node was 0 if it's not set via CLI
1990 * keep it this way for now but in future we probably should
1991 * refuse to start up with incomplete numa mapping */
1992 node_id = 0;
1994 if (cs->numa_node == CPU_UNSET_NUMA_NODE_ID) {
1995 cs->numa_node = node_id;
1996 } else if (cs->numa_node != node_id) {
1997 error_setg(errp, "node-id %d must match numa node specified"
1998 "with -numa option for cpu-index %d",
1999 cs->numa_node, cs->cpu_index);
2000 return;
2004 static void pc_machine_device_pre_plug_cb(HotplugHandler *hotplug_dev,
2005 DeviceState *dev, Error **errp)
2007 if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
2008 pc_cpu_pre_plug(hotplug_dev, dev, errp);
2012 static void pc_machine_device_plug_cb(HotplugHandler *hotplug_dev,
2013 DeviceState *dev, Error **errp)
2015 if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2016 pc_dimm_plug(hotplug_dev, dev, errp);
2017 } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
2018 pc_cpu_plug(hotplug_dev, dev, errp);
2022 static void pc_machine_device_unplug_request_cb(HotplugHandler *hotplug_dev,
2023 DeviceState *dev, Error **errp)
2025 if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2026 pc_dimm_unplug_request(hotplug_dev, dev, errp);
2027 } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
2028 pc_cpu_unplug_request_cb(hotplug_dev, dev, errp);
2029 } else {
2030 error_setg(errp, "acpi: device unplug request for not supported device"
2031 " type: %s", object_get_typename(OBJECT(dev)));
2035 static void pc_machine_device_unplug_cb(HotplugHandler *hotplug_dev,
2036 DeviceState *dev, Error **errp)
2038 if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2039 pc_dimm_unplug(hotplug_dev, dev, errp);
2040 } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
2041 pc_cpu_unplug_cb(hotplug_dev, dev, errp);
2042 } else {
2043 error_setg(errp, "acpi: device unplug for not supported device"
2044 " type: %s", object_get_typename(OBJECT(dev)));
2048 static HotplugHandler *pc_get_hotpug_handler(MachineState *machine,
2049 DeviceState *dev)
2051 PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(machine);
2053 if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM) ||
2054 object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
2055 return HOTPLUG_HANDLER(machine);
2058 return pcmc->get_hotplug_handler ?
2059 pcmc->get_hotplug_handler(machine, dev) : NULL;
2062 static void
2063 pc_machine_get_hotplug_memory_region_size(Object *obj, Visitor *v,
2064 const char *name, void *opaque,
2065 Error **errp)
2067 PCMachineState *pcms = PC_MACHINE(obj);
2068 int64_t value = memory_region_size(&pcms->hotplug_memory.mr);
2070 visit_type_int(v, name, &value, errp);
2073 static void pc_machine_get_max_ram_below_4g(Object *obj, Visitor *v,
2074 const char *name, void *opaque,
2075 Error **errp)
2077 PCMachineState *pcms = PC_MACHINE(obj);
2078 uint64_t value = pcms->max_ram_below_4g;
2080 visit_type_size(v, name, &value, errp);
2083 static void pc_machine_set_max_ram_below_4g(Object *obj, Visitor *v,
2084 const char *name, void *opaque,
2085 Error **errp)
2087 PCMachineState *pcms = PC_MACHINE(obj);
2088 Error *error = NULL;
2089 uint64_t value;
2091 visit_type_size(v, name, &value, &error);
2092 if (error) {
2093 error_propagate(errp, error);
2094 return;
2096 if (value > (1ULL << 32)) {
2097 error_setg(&error,
2098 "Machine option 'max-ram-below-4g=%"PRIu64
2099 "' expects size less than or equal to 4G", value);
2100 error_propagate(errp, error);
2101 return;
2104 if (value < (1ULL << 20)) {
2105 error_report("Warning: small max_ram_below_4g(%"PRIu64
2106 ") less than 1M. BIOS may not work..",
2107 value);
2110 pcms->max_ram_below_4g = value;
2113 static void pc_machine_get_vmport(Object *obj, Visitor *v, const char *name,
2114 void *opaque, Error **errp)
2116 PCMachineState *pcms = PC_MACHINE(obj);
2117 OnOffAuto vmport = pcms->vmport;
2119 visit_type_OnOffAuto(v, name, &vmport, errp);
2122 static void pc_machine_set_vmport(Object *obj, Visitor *v, const char *name,
2123 void *opaque, Error **errp)
2125 PCMachineState *pcms = PC_MACHINE(obj);
2127 visit_type_OnOffAuto(v, name, &pcms->vmport, errp);
2130 bool pc_machine_is_smm_enabled(PCMachineState *pcms)
2132 bool smm_available = false;
2134 if (pcms->smm == ON_OFF_AUTO_OFF) {
2135 return false;
2138 if (tcg_enabled() || qtest_enabled()) {
2139 smm_available = true;
2140 } else if (kvm_enabled()) {
2141 smm_available = kvm_has_smm();
2144 if (smm_available) {
2145 return true;
2148 if (pcms->smm == ON_OFF_AUTO_ON) {
2149 error_report("System Management Mode not supported by this hypervisor.");
2150 exit(1);
2152 return false;
2155 static void pc_machine_get_smm(Object *obj, Visitor *v, const char *name,
2156 void *opaque, Error **errp)
2158 PCMachineState *pcms = PC_MACHINE(obj);
2159 OnOffAuto smm = pcms->smm;
2161 visit_type_OnOffAuto(v, name, &smm, errp);
2164 static void pc_machine_set_smm(Object *obj, Visitor *v, const char *name,
2165 void *opaque, Error **errp)
2167 PCMachineState *pcms = PC_MACHINE(obj);
2169 visit_type_OnOffAuto(v, name, &pcms->smm, errp);
2172 static bool pc_machine_get_nvdimm(Object *obj, Error **errp)
2174 PCMachineState *pcms = PC_MACHINE(obj);
2176 return pcms->acpi_nvdimm_state.is_enabled;
2179 static void pc_machine_set_nvdimm(Object *obj, bool value, Error **errp)
2181 PCMachineState *pcms = PC_MACHINE(obj);
2183 pcms->acpi_nvdimm_state.is_enabled = value;
2186 static bool pc_machine_get_smbus(Object *obj, Error **errp)
2188 PCMachineState *pcms = PC_MACHINE(obj);
2190 return pcms->smbus;
2193 static void pc_machine_set_smbus(Object *obj, bool value, Error **errp)
2195 PCMachineState *pcms = PC_MACHINE(obj);
2197 pcms->smbus = value;
2200 static bool pc_machine_get_sata(Object *obj, Error **errp)
2202 PCMachineState *pcms = PC_MACHINE(obj);
2204 return pcms->sata;
2207 static void pc_machine_set_sata(Object *obj, bool value, Error **errp)
2209 PCMachineState *pcms = PC_MACHINE(obj);
2211 pcms->sata = value;
2214 static bool pc_machine_get_pit(Object *obj, Error **errp)
2216 PCMachineState *pcms = PC_MACHINE(obj);
2218 return pcms->pit;
2221 static void pc_machine_set_pit(Object *obj, bool value, Error **errp)
2223 PCMachineState *pcms = PC_MACHINE(obj);
2225 pcms->pit = value;
2228 static void pc_machine_initfn(Object *obj)
2230 PCMachineState *pcms = PC_MACHINE(obj);
2232 pcms->max_ram_below_4g = 0; /* use default */
2233 pcms->smm = ON_OFF_AUTO_AUTO;
2234 pcms->vmport = ON_OFF_AUTO_AUTO;
2235 /* nvdimm is disabled on default. */
2236 pcms->acpi_nvdimm_state.is_enabled = false;
2237 /* acpi build is enabled by default if machine supports it */
2238 pcms->acpi_build_enabled = PC_MACHINE_GET_CLASS(pcms)->has_acpi_build;
2239 pcms->smbus = true;
2240 pcms->sata = true;
2241 pcms->pit = true;
2244 static void pc_machine_reset(void)
2246 CPUState *cs;
2247 X86CPU *cpu;
2249 qemu_devices_reset();
2251 /* Reset APIC after devices have been reset to cancel
2252 * any changes that qemu_devices_reset() might have done.
2254 CPU_FOREACH(cs) {
2255 cpu = X86_CPU(cs);
2257 if (cpu->apic_state) {
2258 device_reset(cpu->apic_state);
2263 static CpuInstanceProperties
2264 pc_cpu_index_to_props(MachineState *ms, unsigned cpu_index)
2266 MachineClass *mc = MACHINE_GET_CLASS(ms);
2267 const CPUArchIdList *possible_cpus = mc->possible_cpu_arch_ids(ms);
2269 assert(cpu_index < possible_cpus->len);
2270 return possible_cpus->cpus[cpu_index].props;
2273 static const CPUArchIdList *pc_possible_cpu_arch_ids(MachineState *ms)
2275 int i;
2277 if (ms->possible_cpus) {
2279 * make sure that max_cpus hasn't changed since the first use, i.e.
2280 * -smp hasn't been parsed after it
2282 assert(ms->possible_cpus->len == max_cpus);
2283 return ms->possible_cpus;
2286 ms->possible_cpus = g_malloc0(sizeof(CPUArchIdList) +
2287 sizeof(CPUArchId) * max_cpus);
2288 ms->possible_cpus->len = max_cpus;
2289 for (i = 0; i < ms->possible_cpus->len; i++) {
2290 X86CPUTopoInfo topo;
2292 ms->possible_cpus->cpus[i].vcpus_count = 1;
2293 ms->possible_cpus->cpus[i].arch_id = x86_cpu_apic_id_from_index(i);
2294 x86_topo_ids_from_apicid(ms->possible_cpus->cpus[i].arch_id,
2295 smp_cores, smp_threads, &topo);
2296 ms->possible_cpus->cpus[i].props.has_socket_id = true;
2297 ms->possible_cpus->cpus[i].props.socket_id = topo.pkg_id;
2298 ms->possible_cpus->cpus[i].props.has_core_id = true;
2299 ms->possible_cpus->cpus[i].props.core_id = topo.core_id;
2300 ms->possible_cpus->cpus[i].props.has_thread_id = true;
2301 ms->possible_cpus->cpus[i].props.thread_id = topo.smt_id;
2303 /* default distribution of CPUs over NUMA nodes */
2304 if (nb_numa_nodes) {
2305 /* preset values but do not enable them i.e. 'has_node_id = false',
2306 * numa init code will enable them later if manual mapping wasn't
2307 * present on CLI */
2308 ms->possible_cpus->cpus[i].props.node_id =
2309 topo.pkg_id % nb_numa_nodes;
2312 return ms->possible_cpus;
2315 static void x86_nmi(NMIState *n, int cpu_index, Error **errp)
2317 /* cpu index isn't used */
2318 CPUState *cs;
2320 CPU_FOREACH(cs) {
2321 X86CPU *cpu = X86_CPU(cs);
2323 if (!cpu->apic_state) {
2324 cpu_interrupt(cs, CPU_INTERRUPT_NMI);
2325 } else {
2326 apic_deliver_nmi(cpu->apic_state);
2331 static void pc_machine_class_init(ObjectClass *oc, void *data)
2333 MachineClass *mc = MACHINE_CLASS(oc);
2334 PCMachineClass *pcmc = PC_MACHINE_CLASS(oc);
2335 HotplugHandlerClass *hc = HOTPLUG_HANDLER_CLASS(oc);
2336 NMIClass *nc = NMI_CLASS(oc);
2338 pcmc->get_hotplug_handler = mc->get_hotplug_handler;
2339 pcmc->pci_enabled = true;
2340 pcmc->has_acpi_build = true;
2341 pcmc->rsdp_in_ram = true;
2342 pcmc->smbios_defaults = true;
2343 pcmc->smbios_uuid_encoded = true;
2344 pcmc->gigabyte_align = true;
2345 pcmc->has_reserved_memory = true;
2346 pcmc->kvmclock_enabled = true;
2347 pcmc->enforce_aligned_dimm = true;
2348 /* BIOS ACPI tables: 128K. Other BIOS datastructures: less than 4K reported
2349 * to be used at the moment, 32K should be enough for a while. */
2350 pcmc->acpi_data_size = 0x20000 + 0x8000;
2351 pcmc->save_tsc_khz = true;
2352 pcmc->linuxboot_dma_enabled = true;
2353 mc->get_hotplug_handler = pc_get_hotpug_handler;
2354 mc->cpu_index_to_instance_props = pc_cpu_index_to_props;
2355 mc->possible_cpu_arch_ids = pc_possible_cpu_arch_ids;
2356 mc->has_hotpluggable_cpus = true;
2357 mc->default_boot_order = "cad";
2358 mc->hot_add_cpu = pc_hot_add_cpu;
2359 mc->block_default_type = IF_IDE;
2360 mc->max_cpus = 255;
2361 mc->reset = pc_machine_reset;
2362 hc->pre_plug = pc_machine_device_pre_plug_cb;
2363 hc->plug = pc_machine_device_plug_cb;
2364 hc->unplug_request = pc_machine_device_unplug_request_cb;
2365 hc->unplug = pc_machine_device_unplug_cb;
2366 nc->nmi_monitor_handler = x86_nmi;
2368 object_class_property_add(oc, PC_MACHINE_MEMHP_REGION_SIZE, "int",
2369 pc_machine_get_hotplug_memory_region_size, NULL,
2370 NULL, NULL, &error_abort);
2372 object_class_property_add(oc, PC_MACHINE_MAX_RAM_BELOW_4G, "size",
2373 pc_machine_get_max_ram_below_4g, pc_machine_set_max_ram_below_4g,
2374 NULL, NULL, &error_abort);
2376 object_class_property_set_description(oc, PC_MACHINE_MAX_RAM_BELOW_4G,
2377 "Maximum ram below the 4G boundary (32bit boundary)", &error_abort);
2379 object_class_property_add(oc, PC_MACHINE_SMM, "OnOffAuto",
2380 pc_machine_get_smm, pc_machine_set_smm,
2381 NULL, NULL, &error_abort);
2382 object_class_property_set_description(oc, PC_MACHINE_SMM,
2383 "Enable SMM (pc & q35)", &error_abort);
2385 object_class_property_add(oc, PC_MACHINE_VMPORT, "OnOffAuto",
2386 pc_machine_get_vmport, pc_machine_set_vmport,
2387 NULL, NULL, &error_abort);
2388 object_class_property_set_description(oc, PC_MACHINE_VMPORT,
2389 "Enable vmport (pc & q35)", &error_abort);
2391 object_class_property_add_bool(oc, PC_MACHINE_NVDIMM,
2392 pc_machine_get_nvdimm, pc_machine_set_nvdimm, &error_abort);
2394 object_class_property_add_bool(oc, PC_MACHINE_SMBUS,
2395 pc_machine_get_smbus, pc_machine_set_smbus, &error_abort);
2397 object_class_property_add_bool(oc, PC_MACHINE_SATA,
2398 pc_machine_get_sata, pc_machine_set_sata, &error_abort);
2400 object_class_property_add_bool(oc, PC_MACHINE_PIT,
2401 pc_machine_get_pit, pc_machine_set_pit, &error_abort);
2404 static const TypeInfo pc_machine_info = {
2405 .name = TYPE_PC_MACHINE,
2406 .parent = TYPE_MACHINE,
2407 .abstract = true,
2408 .instance_size = sizeof(PCMachineState),
2409 .instance_init = pc_machine_initfn,
2410 .class_size = sizeof(PCMachineClass),
2411 .class_init = pc_machine_class_init,
2412 .interfaces = (InterfaceInfo[]) {
2413 { TYPE_HOTPLUG_HANDLER },
2414 { TYPE_NMI },
2419 static void pc_machine_register_types(void)
2421 type_register_static(&pc_machine_info);
2424 type_init(pc_machine_register_types)