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
25 #include "qemu/osdep.h"
27 #include "hw/i386/pc.h"
28 #include "hw/char/serial.h"
29 #include "hw/char/parallel.h"
30 #include "hw/i386/apic.h"
31 #include "hw/i386/topology.h"
32 #include "sysemu/cpus.h"
33 #include "hw/block/fdc.h"
35 #include "hw/pci/pci.h"
36 #include "hw/pci/pci_bus.h"
37 #include "hw/nvram/fw_cfg.h"
38 #include "hw/timer/hpet.h"
39 #include "hw/smbios/smbios.h"
40 #include "hw/loader.h"
42 #include "multiboot.h"
43 #include "hw/timer/mc146818rtc.h"
44 #include "hw/dma/i8257.h"
45 #include "hw/timer/i8254.h"
46 #include "hw/input/i8042.h"
47 #include "hw/audio/pcspk.h"
48 #include "hw/pci/msi.h"
49 #include "hw/sysbus.h"
50 #include "sysemu/sysemu.h"
51 #include "sysemu/numa.h"
52 #include "sysemu/kvm.h"
53 #include "sysemu/qtest.h"
55 #include "hw/xen/xen.h"
56 #include "ui/qemu-spice.h"
57 #include "exec/memory.h"
58 #include "exec/address-spaces.h"
59 #include "sysemu/arch_init.h"
60 #include "qemu/bitmap.h"
61 #include "qemu/config-file.h"
62 #include "qemu/error-report.h"
63 #include "qemu/option.h"
64 #include "hw/acpi/acpi.h"
65 #include "hw/acpi/cpu_hotplug.h"
66 #include "hw/boards.h"
67 #include "acpi-build.h"
68 #include "hw/mem/pc-dimm.h"
69 #include "qapi/error.h"
70 #include "qapi/qapi-visit-common.h"
71 #include "qapi/visitor.h"
74 #include "hw/i386/intel_iommu.h"
75 #include "hw/net/ne2000-isa.h"
77 /* debug PC/ISA interrupts */
81 #define DPRINTF(fmt, ...) \
82 do { printf("CPUIRQ: " fmt , ## __VA_ARGS__); } while (0)
84 #define DPRINTF(fmt, ...)
87 #define FW_CFG_ACPI_TABLES (FW_CFG_ARCH_LOCAL + 0)
88 #define FW_CFG_SMBIOS_ENTRIES (FW_CFG_ARCH_LOCAL + 1)
89 #define FW_CFG_IRQ0_OVERRIDE (FW_CFG_ARCH_LOCAL + 2)
90 #define FW_CFG_E820_TABLE (FW_CFG_ARCH_LOCAL + 3)
91 #define FW_CFG_HPET (FW_CFG_ARCH_LOCAL + 4)
93 #define E820_NR_ENTRIES 16
99 } QEMU_PACKED __attribute((__aligned__(4)));
103 struct e820_entry entry[E820_NR_ENTRIES];
104 } QEMU_PACKED __attribute((__aligned__(4)));
106 static struct e820_table e820_reserve;
107 static struct e820_entry *e820_table;
108 static unsigned e820_entries;
109 struct hpet_fw_config hpet_cfg = {.count = UINT8_MAX};
111 void gsi_handler(void *opaque, int n, int level)
113 GSIState *s = opaque;
115 DPRINTF("pc: %s GSI %d\n", level ? "raising" : "lowering", n);
116 if (n < ISA_NUM_IRQS) {
117 qemu_set_irq(s->i8259_irq[n], level);
119 qemu_set_irq(s->ioapic_irq[n], level);
122 static void ioport80_write(void *opaque, hwaddr addr, uint64_t data,
127 static uint64_t ioport80_read(void *opaque, hwaddr addr, unsigned size)
129 return 0xffffffffffffffffULL;
132 /* MSDOS compatibility mode FPU exception support */
133 static qemu_irq ferr_irq;
135 void pc_register_ferr_irq(qemu_irq irq)
140 /* XXX: add IGNNE support */
141 void cpu_set_ferr(CPUX86State *s)
143 qemu_irq_raise(ferr_irq);
146 static void ioportF0_write(void *opaque, hwaddr addr, uint64_t data,
149 qemu_irq_lower(ferr_irq);
152 static uint64_t ioportF0_read(void *opaque, hwaddr addr, unsigned size)
154 return 0xffffffffffffffffULL;
158 uint64_t cpu_get_tsc(CPUX86State *env)
160 return cpu_get_ticks();
164 int cpu_get_pic_interrupt(CPUX86State *env)
166 X86CPU *cpu = x86_env_get_cpu(env);
169 if (!kvm_irqchip_in_kernel()) {
170 intno = apic_get_interrupt(cpu->apic_state);
174 /* read the irq from the PIC */
175 if (!apic_accept_pic_intr(cpu->apic_state)) {
180 intno = pic_read_irq(isa_pic);
184 static void pic_irq_request(void *opaque, int irq, int level)
186 CPUState *cs = first_cpu;
187 X86CPU *cpu = X86_CPU(cs);
189 DPRINTF("pic_irqs: %s irq %d\n", level? "raise" : "lower", irq);
190 if (cpu->apic_state && !kvm_irqchip_in_kernel()) {
193 if (apic_accept_pic_intr(cpu->apic_state)) {
194 apic_deliver_pic_intr(cpu->apic_state, level);
199 cpu_interrupt(cs, CPU_INTERRUPT_HARD);
201 cpu_reset_interrupt(cs, CPU_INTERRUPT_HARD);
206 /* PC cmos mappings */
208 #define REG_EQUIPMENT_BYTE 0x14
210 int cmos_get_fd_drive_type(FloppyDriveType fd0)
215 case FLOPPY_DRIVE_TYPE_144:
216 /* 1.44 Mb 3"5 drive */
219 case FLOPPY_DRIVE_TYPE_288:
220 /* 2.88 Mb 3"5 drive */
223 case FLOPPY_DRIVE_TYPE_120:
224 /* 1.2 Mb 5"5 drive */
227 case FLOPPY_DRIVE_TYPE_NONE:
235 static void cmos_init_hd(ISADevice *s, int type_ofs, int info_ofs,
236 int16_t cylinders, int8_t heads, int8_t sectors)
238 rtc_set_memory(s, type_ofs, 47);
239 rtc_set_memory(s, info_ofs, cylinders);
240 rtc_set_memory(s, info_ofs + 1, cylinders >> 8);
241 rtc_set_memory(s, info_ofs + 2, heads);
242 rtc_set_memory(s, info_ofs + 3, 0xff);
243 rtc_set_memory(s, info_ofs + 4, 0xff);
244 rtc_set_memory(s, info_ofs + 5, 0xc0 | ((heads > 8) << 3));
245 rtc_set_memory(s, info_ofs + 6, cylinders);
246 rtc_set_memory(s, info_ofs + 7, cylinders >> 8);
247 rtc_set_memory(s, info_ofs + 8, sectors);
250 /* convert boot_device letter to something recognizable by the bios */
251 static int boot_device2nibble(char boot_device)
253 switch(boot_device) {
256 return 0x01; /* floppy boot */
258 return 0x02; /* hard drive boot */
260 return 0x03; /* CD-ROM boot */
262 return 0x04; /* Network boot */
267 static void set_boot_dev(ISADevice *s, const char *boot_device, Error **errp)
269 #define PC_MAX_BOOT_DEVICES 3
270 int nbds, bds[3] = { 0, };
273 nbds = strlen(boot_device);
274 if (nbds > PC_MAX_BOOT_DEVICES) {
275 error_setg(errp, "Too many boot devices for PC");
278 for (i = 0; i < nbds; i++) {
279 bds[i] = boot_device2nibble(boot_device[i]);
281 error_setg(errp, "Invalid boot device for PC: '%c'",
286 rtc_set_memory(s, 0x3d, (bds[1] << 4) | bds[0]);
287 rtc_set_memory(s, 0x38, (bds[2] << 4) | (fd_bootchk ? 0x0 : 0x1));
290 static void pc_boot_set(void *opaque, const char *boot_device, Error **errp)
292 set_boot_dev(opaque, boot_device, errp);
295 static void pc_cmos_init_floppy(ISADevice *rtc_state, ISADevice *floppy)
298 FloppyDriveType fd_type[2] = { FLOPPY_DRIVE_TYPE_NONE,
299 FLOPPY_DRIVE_TYPE_NONE };
303 for (i = 0; i < 2; i++) {
304 fd_type[i] = isa_fdc_get_drive_type(floppy, i);
307 val = (cmos_get_fd_drive_type(fd_type[0]) << 4) |
308 cmos_get_fd_drive_type(fd_type[1]);
309 rtc_set_memory(rtc_state, 0x10, val);
311 val = rtc_get_memory(rtc_state, REG_EQUIPMENT_BYTE);
313 if (fd_type[0] != FLOPPY_DRIVE_TYPE_NONE) {
316 if (fd_type[1] != FLOPPY_DRIVE_TYPE_NONE) {
323 val |= 0x01; /* 1 drive, ready for boot */
326 val |= 0x41; /* 2 drives, ready for boot */
329 rtc_set_memory(rtc_state, REG_EQUIPMENT_BYTE, val);
332 typedef struct pc_cmos_init_late_arg {
333 ISADevice *rtc_state;
335 } pc_cmos_init_late_arg;
337 typedef struct check_fdc_state {
342 static int check_fdc(Object *obj, void *opaque)
344 CheckFdcState *state = opaque;
347 Error *local_err = NULL;
349 fdc = object_dynamic_cast(obj, TYPE_ISA_FDC);
354 iobase = object_property_get_uint(obj, "iobase", &local_err);
355 if (local_err || iobase != 0x3f0) {
356 error_free(local_err);
361 state->multiple = true;
363 state->floppy = ISA_DEVICE(obj);
368 static const char * const fdc_container_path[] = {
369 "/unattached", "/peripheral", "/peripheral-anon"
373 * Locate the FDC at IO address 0x3f0, in order to configure the CMOS registers
376 ISADevice *pc_find_fdc0(void)
380 CheckFdcState state = { 0 };
382 for (i = 0; i < ARRAY_SIZE(fdc_container_path); i++) {
383 container = container_get(qdev_get_machine(), fdc_container_path[i]);
384 object_child_foreach(container, check_fdc, &state);
387 if (state.multiple) {
388 warn_report("multiple floppy disk controllers with "
389 "iobase=0x3f0 have been found");
390 error_printf("the one being picked for CMOS setup might not reflect "
397 static void pc_cmos_init_late(void *opaque)
399 pc_cmos_init_late_arg *arg = opaque;
400 ISADevice *s = arg->rtc_state;
402 int8_t heads, sectors;
407 if (arg->idebus[0] && ide_get_geometry(arg->idebus[0], 0,
408 &cylinders, &heads, §ors) >= 0) {
409 cmos_init_hd(s, 0x19, 0x1b, cylinders, heads, sectors);
412 if (arg->idebus[0] && ide_get_geometry(arg->idebus[0], 1,
413 &cylinders, &heads, §ors) >= 0) {
414 cmos_init_hd(s, 0x1a, 0x24, cylinders, heads, sectors);
417 rtc_set_memory(s, 0x12, val);
420 for (i = 0; i < 4; i++) {
421 /* NOTE: ide_get_geometry() returns the physical
422 geometry. It is always such that: 1 <= sects <= 63, 1
423 <= heads <= 16, 1 <= cylinders <= 16383. The BIOS
424 geometry can be different if a translation is done. */
425 if (arg->idebus[i / 2] &&
426 ide_get_geometry(arg->idebus[i / 2], i % 2,
427 &cylinders, &heads, §ors) >= 0) {
428 trans = ide_get_bios_chs_trans(arg->idebus[i / 2], i % 2) - 1;
429 assert((trans & ~3) == 0);
430 val |= trans << (i * 2);
433 rtc_set_memory(s, 0x39, val);
435 pc_cmos_init_floppy(s, pc_find_fdc0());
437 qemu_unregister_reset(pc_cmos_init_late, opaque);
440 void pc_cmos_init(PCMachineState *pcms,
441 BusState *idebus0, BusState *idebus1,
445 static pc_cmos_init_late_arg arg;
447 /* various important CMOS locations needed by PC/Bochs bios */
450 /* base memory (first MiB) */
451 val = MIN(pcms->below_4g_mem_size / 1024, 640);
452 rtc_set_memory(s, 0x15, val);
453 rtc_set_memory(s, 0x16, val >> 8);
454 /* extended memory (next 64MiB) */
455 if (pcms->below_4g_mem_size > 1024 * 1024) {
456 val = (pcms->below_4g_mem_size - 1024 * 1024) / 1024;
462 rtc_set_memory(s, 0x17, val);
463 rtc_set_memory(s, 0x18, val >> 8);
464 rtc_set_memory(s, 0x30, val);
465 rtc_set_memory(s, 0x31, val >> 8);
466 /* memory between 16MiB and 4GiB */
467 if (pcms->below_4g_mem_size > 16 * 1024 * 1024) {
468 val = (pcms->below_4g_mem_size - 16 * 1024 * 1024) / 65536;
474 rtc_set_memory(s, 0x34, val);
475 rtc_set_memory(s, 0x35, val >> 8);
476 /* memory above 4GiB */
477 val = pcms->above_4g_mem_size / 65536;
478 rtc_set_memory(s, 0x5b, val);
479 rtc_set_memory(s, 0x5c, val >> 8);
480 rtc_set_memory(s, 0x5d, val >> 16);
482 object_property_add_link(OBJECT(pcms), "rtc_state",
484 (Object **)&pcms->rtc,
485 object_property_allow_set_link,
486 OBJ_PROP_LINK_STRONG, &error_abort);
487 object_property_set_link(OBJECT(pcms), OBJECT(s),
488 "rtc_state", &error_abort);
490 set_boot_dev(s, MACHINE(pcms)->boot_order, &error_fatal);
493 val |= 0x02; /* FPU is there */
494 val |= 0x04; /* PS/2 mouse installed */
495 rtc_set_memory(s, REG_EQUIPMENT_BYTE, val);
497 /* hard drives and FDC */
499 arg.idebus[0] = idebus0;
500 arg.idebus[1] = idebus1;
501 qemu_register_reset(pc_cmos_init_late, &arg);
504 #define TYPE_PORT92 "port92"
505 #define PORT92(obj) OBJECT_CHECK(Port92State, (obj), TYPE_PORT92)
507 /* port 92 stuff: could be split off */
508 typedef struct Port92State {
509 ISADevice parent_obj;
516 static void port92_write(void *opaque, hwaddr addr, uint64_t val,
519 Port92State *s = opaque;
520 int oldval = s->outport;
522 DPRINTF("port92: write 0x%02" PRIx64 "\n", val);
524 qemu_set_irq(s->a20_out, (val >> 1) & 1);
525 if ((val & 1) && !(oldval & 1)) {
526 qemu_system_reset_request(SHUTDOWN_CAUSE_GUEST_RESET);
530 static uint64_t port92_read(void *opaque, hwaddr addr,
533 Port92State *s = opaque;
537 DPRINTF("port92: read 0x%02x\n", ret);
541 static void port92_init(ISADevice *dev, qemu_irq a20_out)
543 qdev_connect_gpio_out_named(DEVICE(dev), PORT92_A20_LINE, 0, a20_out);
546 static const VMStateDescription vmstate_port92_isa = {
549 .minimum_version_id = 1,
550 .fields = (VMStateField[]) {
551 VMSTATE_UINT8(outport, Port92State),
552 VMSTATE_END_OF_LIST()
556 static void port92_reset(DeviceState *d)
558 Port92State *s = PORT92(d);
563 static const MemoryRegionOps port92_ops = {
565 .write = port92_write,
567 .min_access_size = 1,
568 .max_access_size = 1,
570 .endianness = DEVICE_LITTLE_ENDIAN,
573 static void port92_initfn(Object *obj)
575 Port92State *s = PORT92(obj);
577 memory_region_init_io(&s->io, OBJECT(s), &port92_ops, s, "port92", 1);
581 qdev_init_gpio_out_named(DEVICE(obj), &s->a20_out, PORT92_A20_LINE, 1);
584 static void port92_realizefn(DeviceState *dev, Error **errp)
586 ISADevice *isadev = ISA_DEVICE(dev);
587 Port92State *s = PORT92(dev);
589 isa_register_ioport(isadev, &s->io, 0x92);
592 static void port92_class_initfn(ObjectClass *klass, void *data)
594 DeviceClass *dc = DEVICE_CLASS(klass);
596 dc->realize = port92_realizefn;
597 dc->reset = port92_reset;
598 dc->vmsd = &vmstate_port92_isa;
600 * Reason: unlike ordinary ISA devices, this one needs additional
601 * wiring: its A20 output line needs to be wired up by
604 dc->user_creatable = false;
607 static const TypeInfo port92_info = {
609 .parent = TYPE_ISA_DEVICE,
610 .instance_size = sizeof(Port92State),
611 .instance_init = port92_initfn,
612 .class_init = port92_class_initfn,
615 static void port92_register_types(void)
617 type_register_static(&port92_info);
620 type_init(port92_register_types)
622 static void handle_a20_line_change(void *opaque, int irq, int level)
624 X86CPU *cpu = opaque;
626 /* XXX: send to all CPUs ? */
627 /* XXX: add logic to handle multiple A20 line sources */
628 x86_cpu_set_a20(cpu, level);
631 int e820_add_entry(uint64_t address, uint64_t length, uint32_t type)
633 int index = le32_to_cpu(e820_reserve.count);
634 struct e820_entry *entry;
636 if (type != E820_RAM) {
637 /* old FW_CFG_E820_TABLE entry -- reservations only */
638 if (index >= E820_NR_ENTRIES) {
641 entry = &e820_reserve.entry[index++];
643 entry->address = cpu_to_le64(address);
644 entry->length = cpu_to_le64(length);
645 entry->type = cpu_to_le32(type);
647 e820_reserve.count = cpu_to_le32(index);
650 /* new "etc/e820" file -- include ram too */
651 e820_table = g_renew(struct e820_entry, e820_table, e820_entries + 1);
652 e820_table[e820_entries].address = cpu_to_le64(address);
653 e820_table[e820_entries].length = cpu_to_le64(length);
654 e820_table[e820_entries].type = cpu_to_le32(type);
660 int e820_get_num_entries(void)
665 bool e820_get_entry(int idx, uint32_t type, uint64_t *address, uint64_t *length)
667 if (idx < e820_entries && e820_table[idx].type == cpu_to_le32(type)) {
668 *address = le64_to_cpu(e820_table[idx].address);
669 *length = le64_to_cpu(e820_table[idx].length);
675 /* Enables contiguous-apic-ID mode, for compatibility */
676 static bool compat_apic_id_mode;
678 void enable_compat_apic_id_mode(void)
680 compat_apic_id_mode = true;
683 /* Calculates initial APIC ID for a specific CPU index
685 * Currently we need to be able to calculate the APIC ID from the CPU index
686 * alone (without requiring a CPU object), as the QEMU<->Seabios interfaces have
687 * no concept of "CPU index", and the NUMA tables on fw_cfg need the APIC ID of
688 * all CPUs up to max_cpus.
690 static uint32_t x86_cpu_apic_id_from_index(unsigned int cpu_index)
695 correct_id = x86_apicid_from_cpu_idx(smp_cores, smp_threads, cpu_index);
696 if (compat_apic_id_mode) {
697 if (cpu_index != correct_id && !warned && !qtest_enabled()) {
698 error_report("APIC IDs set in compatibility mode, "
699 "CPU topology won't match the configuration");
708 static void pc_build_smbios(PCMachineState *pcms)
710 uint8_t *smbios_tables, *smbios_anchor;
711 size_t smbios_tables_len, smbios_anchor_len;
712 struct smbios_phys_mem_area *mem_array;
713 unsigned i, array_count;
714 MachineState *ms = MACHINE(pcms);
715 X86CPU *cpu = X86_CPU(ms->possible_cpus->cpus[0].cpu);
717 /* tell smbios about cpuid version and features */
718 smbios_set_cpuid(cpu->env.cpuid_version, cpu->env.features[FEAT_1_EDX]);
720 smbios_tables = smbios_get_table_legacy(&smbios_tables_len);
722 fw_cfg_add_bytes(pcms->fw_cfg, FW_CFG_SMBIOS_ENTRIES,
723 smbios_tables, smbios_tables_len);
726 /* build the array of physical mem area from e820 table */
727 mem_array = g_malloc0(sizeof(*mem_array) * e820_get_num_entries());
728 for (i = 0, array_count = 0; i < e820_get_num_entries(); i++) {
731 if (e820_get_entry(i, E820_RAM, &addr, &len)) {
732 mem_array[array_count].address = addr;
733 mem_array[array_count].length = len;
737 smbios_get_tables(mem_array, array_count,
738 &smbios_tables, &smbios_tables_len,
739 &smbios_anchor, &smbios_anchor_len);
743 fw_cfg_add_file(pcms->fw_cfg, "etc/smbios/smbios-tables",
744 smbios_tables, smbios_tables_len);
745 fw_cfg_add_file(pcms->fw_cfg, "etc/smbios/smbios-anchor",
746 smbios_anchor, smbios_anchor_len);
750 static FWCfgState *bochs_bios_init(AddressSpace *as, PCMachineState *pcms)
753 uint64_t *numa_fw_cfg;
755 const CPUArchIdList *cpus;
756 MachineClass *mc = MACHINE_GET_CLASS(pcms);
758 fw_cfg = fw_cfg_init_io_dma(FW_CFG_IO_BASE, FW_CFG_IO_BASE + 4, as);
759 fw_cfg_add_i16(fw_cfg, FW_CFG_NB_CPUS, pcms->boot_cpus);
761 /* FW_CFG_MAX_CPUS is a bit confusing/problematic on x86:
763 * For machine types prior to 1.8, SeaBIOS needs FW_CFG_MAX_CPUS for
764 * building MPTable, ACPI MADT, ACPI CPU hotplug and ACPI SRAT table,
765 * that tables are based on xAPIC ID and QEMU<->SeaBIOS interface
766 * for CPU hotplug also uses APIC ID and not "CPU index".
767 * This means that FW_CFG_MAX_CPUS is not the "maximum number of CPUs",
768 * but the "limit to the APIC ID values SeaBIOS may see".
770 * So for compatibility reasons with old BIOSes we are stuck with
771 * "etc/max-cpus" actually being apic_id_limit
773 fw_cfg_add_i16(fw_cfg, FW_CFG_MAX_CPUS, (uint16_t)pcms->apic_id_limit);
774 fw_cfg_add_i64(fw_cfg, FW_CFG_RAM_SIZE, (uint64_t)ram_size);
775 fw_cfg_add_bytes(fw_cfg, FW_CFG_ACPI_TABLES,
776 acpi_tables, acpi_tables_len);
777 fw_cfg_add_i32(fw_cfg, FW_CFG_IRQ0_OVERRIDE, kvm_allows_irq0_override());
779 fw_cfg_add_bytes(fw_cfg, FW_CFG_E820_TABLE,
780 &e820_reserve, sizeof(e820_reserve));
781 fw_cfg_add_file(fw_cfg, "etc/e820", e820_table,
782 sizeof(struct e820_entry) * e820_entries);
784 fw_cfg_add_bytes(fw_cfg, FW_CFG_HPET, &hpet_cfg, sizeof(hpet_cfg));
785 /* allocate memory for the NUMA channel: one (64bit) word for the number
786 * of nodes, one word for each VCPU->node and one word for each node to
787 * hold the amount of memory.
789 numa_fw_cfg = g_new0(uint64_t, 1 + pcms->apic_id_limit + nb_numa_nodes);
790 numa_fw_cfg[0] = cpu_to_le64(nb_numa_nodes);
791 cpus = mc->possible_cpu_arch_ids(MACHINE(pcms));
792 for (i = 0; i < cpus->len; i++) {
793 unsigned int apic_id = cpus->cpus[i].arch_id;
794 assert(apic_id < pcms->apic_id_limit);
795 numa_fw_cfg[apic_id + 1] = cpu_to_le64(cpus->cpus[i].props.node_id);
797 for (i = 0; i < nb_numa_nodes; i++) {
798 numa_fw_cfg[pcms->apic_id_limit + 1 + i] =
799 cpu_to_le64(numa_info[i].node_mem);
801 fw_cfg_add_bytes(fw_cfg, FW_CFG_NUMA, numa_fw_cfg,
802 (1 + pcms->apic_id_limit + nb_numa_nodes) *
803 sizeof(*numa_fw_cfg));
808 static long get_file_size(FILE *f)
812 /* XXX: on Unix systems, using fstat() probably makes more sense */
815 fseek(f, 0, SEEK_END);
817 fseek(f, where, SEEK_SET);
822 /* setup_data types */
824 #define SETUP_E820_EXT 1
834 } __attribute__((packed));
836 static void load_linux(PCMachineState *pcms,
840 int setup_size, kernel_size, initrd_size = 0, cmdline_size;
841 int dtb_size, setup_data_offset;
843 uint8_t header[8192], *setup, *kernel, *initrd_data;
844 hwaddr real_addr, prot_addr, cmdline_addr, initrd_addr = 0;
847 MachineState *machine = MACHINE(pcms);
848 PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
849 struct setup_data *setup_data;
850 const char *kernel_filename = machine->kernel_filename;
851 const char *initrd_filename = machine->initrd_filename;
852 const char *dtb_filename = machine->dtb;
853 const char *kernel_cmdline = machine->kernel_cmdline;
855 /* Align to 16 bytes as a paranoia measure */
856 cmdline_size = (strlen(kernel_cmdline)+16) & ~15;
858 /* load the kernel header */
859 f = fopen(kernel_filename, "rb");
860 if (!f || !(kernel_size = get_file_size(f)) ||
861 fread(header, 1, MIN(ARRAY_SIZE(header), kernel_size), f) !=
862 MIN(ARRAY_SIZE(header), kernel_size)) {
863 fprintf(stderr, "qemu: could not load kernel '%s': %s\n",
864 kernel_filename, strerror(errno));
868 /* kernel protocol version */
870 fprintf(stderr, "header magic: %#x\n", ldl_p(header+0x202));
872 if (ldl_p(header+0x202) == 0x53726448) {
873 protocol = lduw_p(header+0x206);
875 /* This looks like a multiboot kernel. If it is, let's stop
876 treating it like a Linux kernel. */
877 if (load_multiboot(fw_cfg, f, kernel_filename, initrd_filename,
878 kernel_cmdline, kernel_size, header)) {
884 if (protocol < 0x200 || !(header[0x211] & 0x01)) {
887 cmdline_addr = 0x9a000 - cmdline_size;
889 } else if (protocol < 0x202) {
890 /* High but ancient kernel */
892 cmdline_addr = 0x9a000 - cmdline_size;
893 prot_addr = 0x100000;
895 /* High and recent kernel */
897 cmdline_addr = 0x20000;
898 prot_addr = 0x100000;
903 "qemu: real_addr = 0x" TARGET_FMT_plx "\n"
904 "qemu: cmdline_addr = 0x" TARGET_FMT_plx "\n"
905 "qemu: prot_addr = 0x" TARGET_FMT_plx "\n",
911 /* highest address for loading the initrd */
912 if (protocol >= 0x203) {
913 initrd_max = ldl_p(header+0x22c);
915 initrd_max = 0x37ffffff;
918 if (initrd_max >= pcms->below_4g_mem_size - pcmc->acpi_data_size) {
919 initrd_max = pcms->below_4g_mem_size - pcmc->acpi_data_size - 1;
922 fw_cfg_add_i32(fw_cfg, FW_CFG_CMDLINE_ADDR, cmdline_addr);
923 fw_cfg_add_i32(fw_cfg, FW_CFG_CMDLINE_SIZE, strlen(kernel_cmdline)+1);
924 fw_cfg_add_string(fw_cfg, FW_CFG_CMDLINE_DATA, kernel_cmdline);
926 if (protocol >= 0x202) {
927 stl_p(header+0x228, cmdline_addr);
929 stw_p(header+0x20, 0xA33F);
930 stw_p(header+0x22, cmdline_addr-real_addr);
933 /* handle vga= parameter */
934 vmode = strstr(kernel_cmdline, "vga=");
936 unsigned int video_mode;
939 if (!strncmp(vmode, "normal", 6)) {
941 } else if (!strncmp(vmode, "ext", 3)) {
943 } else if (!strncmp(vmode, "ask", 3)) {
946 video_mode = strtol(vmode, NULL, 0);
948 stw_p(header+0x1fa, video_mode);
952 /* High nybble = B reserved for QEMU; low nybble is revision number.
953 If this code is substantially changed, you may want to consider
954 incrementing the revision. */
955 if (protocol >= 0x200) {
956 header[0x210] = 0xB0;
959 if (protocol >= 0x201) {
960 header[0x211] |= 0x80; /* CAN_USE_HEAP */
961 stw_p(header+0x224, cmdline_addr-real_addr-0x200);
965 if (initrd_filename) {
966 if (protocol < 0x200) {
967 fprintf(stderr, "qemu: linux kernel too old to load a ram disk\n");
971 initrd_size = get_image_size(initrd_filename);
972 if (initrd_size < 0) {
973 fprintf(stderr, "qemu: error reading initrd %s: %s\n",
974 initrd_filename, strerror(errno));
978 initrd_addr = (initrd_max-initrd_size) & ~4095;
980 initrd_data = g_malloc(initrd_size);
981 load_image(initrd_filename, initrd_data);
983 fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_ADDR, initrd_addr);
984 fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_SIZE, initrd_size);
985 fw_cfg_add_bytes(fw_cfg, FW_CFG_INITRD_DATA, initrd_data, initrd_size);
987 stl_p(header+0x218, initrd_addr);
988 stl_p(header+0x21c, initrd_size);
991 /* load kernel and setup */
992 setup_size = header[0x1f1];
993 if (setup_size == 0) {
996 setup_size = (setup_size+1)*512;
997 if (setup_size > kernel_size) {
998 fprintf(stderr, "qemu: invalid kernel header\n");
1001 kernel_size -= setup_size;
1003 setup = g_malloc(setup_size);
1004 kernel = g_malloc(kernel_size);
1005 fseek(f, 0, SEEK_SET);
1006 if (fread(setup, 1, setup_size, f) != setup_size) {
1007 fprintf(stderr, "fread() failed\n");
1010 if (fread(kernel, 1, kernel_size, f) != kernel_size) {
1011 fprintf(stderr, "fread() failed\n");
1016 /* append dtb to kernel */
1018 if (protocol < 0x209) {
1019 fprintf(stderr, "qemu: Linux kernel too old to load a dtb\n");
1023 dtb_size = get_image_size(dtb_filename);
1024 if (dtb_size <= 0) {
1025 fprintf(stderr, "qemu: error reading dtb %s: %s\n",
1026 dtb_filename, strerror(errno));
1030 setup_data_offset = QEMU_ALIGN_UP(kernel_size, 16);
1031 kernel_size = setup_data_offset + sizeof(struct setup_data) + dtb_size;
1032 kernel = g_realloc(kernel, kernel_size);
1034 stq_p(header+0x250, prot_addr + setup_data_offset);
1036 setup_data = (struct setup_data *)(kernel + setup_data_offset);
1037 setup_data->next = 0;
1038 setup_data->type = cpu_to_le32(SETUP_DTB);
1039 setup_data->len = cpu_to_le32(dtb_size);
1041 load_image_size(dtb_filename, setup_data->data, dtb_size);
1044 memcpy(setup, header, MIN(sizeof(header), setup_size));
1046 fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_ADDR, prot_addr);
1047 fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_SIZE, kernel_size);
1048 fw_cfg_add_bytes(fw_cfg, FW_CFG_KERNEL_DATA, kernel, kernel_size);
1050 fw_cfg_add_i32(fw_cfg, FW_CFG_SETUP_ADDR, real_addr);
1051 fw_cfg_add_i32(fw_cfg, FW_CFG_SETUP_SIZE, setup_size);
1052 fw_cfg_add_bytes(fw_cfg, FW_CFG_SETUP_DATA, setup, setup_size);
1054 option_rom[nb_option_roms].bootindex = 0;
1055 option_rom[nb_option_roms].name = "linuxboot.bin";
1056 if (pcmc->linuxboot_dma_enabled && fw_cfg_dma_enabled(fw_cfg)) {
1057 option_rom[nb_option_roms].name = "linuxboot_dma.bin";
1062 #define NE2000_NB_MAX 6
1064 static const int ne2000_io[NE2000_NB_MAX] = { 0x300, 0x320, 0x340, 0x360,
1066 static const int ne2000_irq[NE2000_NB_MAX] = { 9, 10, 11, 3, 4, 5 };
1068 void pc_init_ne2k_isa(ISABus *bus, NICInfo *nd)
1070 static int nb_ne2k = 0;
1072 if (nb_ne2k == NE2000_NB_MAX)
1074 isa_ne2000_init(bus, ne2000_io[nb_ne2k],
1075 ne2000_irq[nb_ne2k], nd);
1079 DeviceState *cpu_get_current_apic(void)
1082 X86CPU *cpu = X86_CPU(current_cpu);
1083 return cpu->apic_state;
1089 void pc_acpi_smi_interrupt(void *opaque, int irq, int level)
1091 X86CPU *cpu = opaque;
1094 cpu_interrupt(CPU(cpu), CPU_INTERRUPT_SMI);
1098 static void pc_new_cpu(const char *typename, int64_t apic_id, Error **errp)
1101 Error *local_err = NULL;
1103 cpu = object_new(typename);
1105 object_property_set_uint(cpu, apic_id, "apic-id", &local_err);
1106 object_property_set_bool(cpu, true, "realized", &local_err);
1109 error_propagate(errp, local_err);
1112 void pc_hot_add_cpu(const int64_t id, Error **errp)
1114 MachineState *ms = MACHINE(qdev_get_machine());
1115 int64_t apic_id = x86_cpu_apic_id_from_index(id);
1116 Error *local_err = NULL;
1119 error_setg(errp, "Invalid CPU id: %" PRIi64, id);
1123 if (apic_id >= ACPI_CPU_HOTPLUG_ID_LIMIT) {
1124 error_setg(errp, "Unable to add CPU: %" PRIi64
1125 ", resulting APIC ID (%" PRIi64 ") is too large",
1130 pc_new_cpu(ms->cpu_type, apic_id, &local_err);
1132 error_propagate(errp, local_err);
1137 void pc_cpus_init(PCMachineState *pcms)
1140 const CPUArchIdList *possible_cpus;
1141 MachineState *ms = MACHINE(pcms);
1142 MachineClass *mc = MACHINE_GET_CLASS(pcms);
1144 /* Calculates the limit to CPU APIC ID values
1146 * Limit for the APIC ID value, so that all
1147 * CPU APIC IDs are < pcms->apic_id_limit.
1149 * This is used for FW_CFG_MAX_CPUS. See comments on bochs_bios_init().
1151 pcms->apic_id_limit = x86_cpu_apic_id_from_index(max_cpus - 1) + 1;
1152 possible_cpus = mc->possible_cpu_arch_ids(ms);
1153 for (i = 0; i < smp_cpus; i++) {
1154 pc_new_cpu(possible_cpus->cpus[i].type, possible_cpus->cpus[i].arch_id,
1159 static void pc_build_feature_control_file(PCMachineState *pcms)
1161 MachineState *ms = MACHINE(pcms);
1162 X86CPU *cpu = X86_CPU(ms->possible_cpus->cpus[0].cpu);
1163 CPUX86State *env = &cpu->env;
1164 uint32_t unused, ecx, edx;
1165 uint64_t feature_control_bits = 0;
1168 cpu_x86_cpuid(env, 1, 0, &unused, &unused, &ecx, &edx);
1169 if (ecx & CPUID_EXT_VMX) {
1170 feature_control_bits |= FEATURE_CONTROL_VMXON_ENABLED_OUTSIDE_SMX;
1173 if ((edx & (CPUID_EXT2_MCE | CPUID_EXT2_MCA)) ==
1174 (CPUID_EXT2_MCE | CPUID_EXT2_MCA) &&
1175 (env->mcg_cap & MCG_LMCE_P)) {
1176 feature_control_bits |= FEATURE_CONTROL_LMCE;
1179 if (!feature_control_bits) {
1183 val = g_malloc(sizeof(*val));
1184 *val = cpu_to_le64(feature_control_bits | FEATURE_CONTROL_LOCKED);
1185 fw_cfg_add_file(pcms->fw_cfg, "etc/msr_feature_control", val, sizeof(*val));
1188 static void rtc_set_cpus_count(ISADevice *rtc, uint16_t cpus_count)
1190 if (cpus_count > 0xff) {
1191 /* If the number of CPUs can't be represented in 8 bits, the
1192 * BIOS must use "FW_CFG_NB_CPUS". Set RTC field to 0 just
1193 * to make old BIOSes fail more predictably.
1195 rtc_set_memory(rtc, 0x5f, 0);
1197 rtc_set_memory(rtc, 0x5f, cpus_count - 1);
1202 void pc_machine_done(Notifier *notifier, void *data)
1204 PCMachineState *pcms = container_of(notifier,
1205 PCMachineState, machine_done);
1206 PCIBus *bus = pcms->bus;
1208 /* set the number of CPUs */
1209 rtc_set_cpus_count(pcms->rtc, pcms->boot_cpus);
1212 int extra_hosts = 0;
1214 QLIST_FOREACH(bus, &bus->child, sibling) {
1215 /* look for expander root buses */
1216 if (pci_bus_is_root(bus)) {
1220 if (extra_hosts && pcms->fw_cfg) {
1221 uint64_t *val = g_malloc(sizeof(*val));
1222 *val = cpu_to_le64(extra_hosts);
1223 fw_cfg_add_file(pcms->fw_cfg,
1224 "etc/extra-pci-roots", val, sizeof(*val));
1230 pc_build_smbios(pcms);
1231 pc_build_feature_control_file(pcms);
1232 /* update FW_CFG_NB_CPUS to account for -device added CPUs */
1233 fw_cfg_modify_i16(pcms->fw_cfg, FW_CFG_NB_CPUS, pcms->boot_cpus);
1236 if (pcms->apic_id_limit > 255 && !xen_enabled()) {
1237 IntelIOMMUState *iommu = INTEL_IOMMU_DEVICE(x86_iommu_get_default());
1239 if (!iommu || !iommu->x86_iommu.intr_supported ||
1240 iommu->intr_eim != ON_OFF_AUTO_ON) {
1241 error_report("current -smp configuration requires "
1242 "Extended Interrupt Mode enabled. "
1243 "You can add an IOMMU using: "
1244 "-device intel-iommu,intremap=on,eim=on");
1250 void pc_guest_info_init(PCMachineState *pcms)
1254 pcms->apic_xrupt_override = kvm_allows_irq0_override();
1255 pcms->numa_nodes = nb_numa_nodes;
1256 pcms->node_mem = g_malloc0(pcms->numa_nodes *
1257 sizeof *pcms->node_mem);
1258 for (i = 0; i < nb_numa_nodes; i++) {
1259 pcms->node_mem[i] = numa_info[i].node_mem;
1262 pcms->machine_done.notify = pc_machine_done;
1263 qemu_add_machine_init_done_notifier(&pcms->machine_done);
1266 /* setup pci memory address space mapping into system address space */
1267 void pc_pci_as_mapping_init(Object *owner, MemoryRegion *system_memory,
1268 MemoryRegion *pci_address_space)
1270 /* Set to lower priority than RAM */
1271 memory_region_add_subregion_overlap(system_memory, 0x0,
1272 pci_address_space, -1);
1275 void pc_acpi_init(const char *default_dsdt)
1279 if (acpi_tables != NULL) {
1280 /* manually set via -acpitable, leave it alone */
1284 filename = qemu_find_file(QEMU_FILE_TYPE_BIOS, default_dsdt);
1285 if (filename == NULL) {
1286 warn_report("failed to find %s", default_dsdt);
1288 QemuOpts *opts = qemu_opts_create(qemu_find_opts("acpi"), NULL, 0,
1292 qemu_opt_set(opts, "file", filename, &error_abort);
1294 acpi_table_add_builtin(opts, &err);
1296 warn_reportf_err(err, "failed to load %s: ", filename);
1302 void xen_load_linux(PCMachineState *pcms)
1307 assert(MACHINE(pcms)->kernel_filename != NULL);
1309 fw_cfg = fw_cfg_init_io(FW_CFG_IO_BASE);
1310 fw_cfg_add_i16(fw_cfg, FW_CFG_NB_CPUS, pcms->boot_cpus);
1313 load_linux(pcms, fw_cfg);
1314 for (i = 0; i < nb_option_roms; i++) {
1315 assert(!strcmp(option_rom[i].name, "linuxboot.bin") ||
1316 !strcmp(option_rom[i].name, "linuxboot_dma.bin") ||
1317 !strcmp(option_rom[i].name, "multiboot.bin"));
1318 rom_add_option(option_rom[i].name, option_rom[i].bootindex);
1320 pcms->fw_cfg = fw_cfg;
1323 void pc_memory_init(PCMachineState *pcms,
1324 MemoryRegion *system_memory,
1325 MemoryRegion *rom_memory,
1326 MemoryRegion **ram_memory)
1329 MemoryRegion *ram, *option_rom_mr;
1330 MemoryRegion *ram_below_4g, *ram_above_4g;
1332 MachineState *machine = MACHINE(pcms);
1333 PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
1335 assert(machine->ram_size == pcms->below_4g_mem_size +
1336 pcms->above_4g_mem_size);
1338 linux_boot = (machine->kernel_filename != NULL);
1340 /* Allocate RAM. We allocate it as a single memory region and use
1341 * aliases to address portions of it, mostly for backwards compatibility
1342 * with older qemus that used qemu_ram_alloc().
1344 ram = g_malloc(sizeof(*ram));
1345 memory_region_allocate_system_memory(ram, NULL, "pc.ram",
1348 ram_below_4g = g_malloc(sizeof(*ram_below_4g));
1349 memory_region_init_alias(ram_below_4g, NULL, "ram-below-4g", ram,
1350 0, pcms->below_4g_mem_size);
1351 memory_region_add_subregion(system_memory, 0, ram_below_4g);
1352 e820_add_entry(0, pcms->below_4g_mem_size, E820_RAM);
1353 if (pcms->above_4g_mem_size > 0) {
1354 ram_above_4g = g_malloc(sizeof(*ram_above_4g));
1355 memory_region_init_alias(ram_above_4g, NULL, "ram-above-4g", ram,
1356 pcms->below_4g_mem_size,
1357 pcms->above_4g_mem_size);
1358 memory_region_add_subregion(system_memory, 0x100000000ULL,
1360 e820_add_entry(0x100000000ULL, pcms->above_4g_mem_size, E820_RAM);
1363 if (!pcmc->has_reserved_memory &&
1364 (machine->ram_slots ||
1365 (machine->maxram_size > machine->ram_size))) {
1366 MachineClass *mc = MACHINE_GET_CLASS(machine);
1368 error_report("\"-memory 'slots|maxmem'\" is not supported by: %s",
1373 /* always allocate the device memory information */
1374 machine->device_memory = g_malloc0(sizeof(*machine->device_memory));
1376 /* initialize device memory address space */
1377 if (pcmc->has_reserved_memory &&
1378 (machine->ram_size < machine->maxram_size)) {
1379 ram_addr_t device_mem_size = machine->maxram_size - machine->ram_size;
1381 if (machine->ram_slots > ACPI_MAX_RAM_SLOTS) {
1382 error_report("unsupported amount of memory slots: %"PRIu64,
1383 machine->ram_slots);
1387 if (QEMU_ALIGN_UP(machine->maxram_size,
1388 TARGET_PAGE_SIZE) != machine->maxram_size) {
1389 error_report("maximum memory size must by aligned to multiple of "
1390 "%d bytes", TARGET_PAGE_SIZE);
1394 machine->device_memory->base =
1395 ROUND_UP(0x100000000ULL + pcms->above_4g_mem_size, 1ULL << 30);
1397 if (pcmc->enforce_aligned_dimm) {
1398 /* size device region assuming 1G page max alignment per slot */
1399 device_mem_size += (1ULL << 30) * machine->ram_slots;
1402 if ((machine->device_memory->base + device_mem_size) <
1404 error_report("unsupported amount of maximum memory: " RAM_ADDR_FMT,
1405 machine->maxram_size);
1409 memory_region_init(&machine->device_memory->mr, OBJECT(pcms),
1410 "device-memory", device_mem_size);
1411 memory_region_add_subregion(system_memory, machine->device_memory->base,
1412 &machine->device_memory->mr);
1415 /* Initialize PC system firmware */
1416 pc_system_firmware_init(rom_memory, !pcmc->pci_enabled);
1418 option_rom_mr = g_malloc(sizeof(*option_rom_mr));
1419 memory_region_init_ram(option_rom_mr, NULL, "pc.rom", PC_ROM_SIZE,
1421 if (pcmc->pci_enabled) {
1422 memory_region_set_readonly(option_rom_mr, true);
1424 memory_region_add_subregion_overlap(rom_memory,
1429 fw_cfg = bochs_bios_init(&address_space_memory, pcms);
1433 if (pcmc->has_reserved_memory && machine->device_memory->base) {
1434 uint64_t *val = g_malloc(sizeof(*val));
1435 PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
1436 uint64_t res_mem_end = machine->device_memory->base;
1438 if (!pcmc->broken_reserved_end) {
1439 res_mem_end += memory_region_size(&machine->device_memory->mr);
1441 *val = cpu_to_le64(ROUND_UP(res_mem_end, 0x1ULL << 30));
1442 fw_cfg_add_file(fw_cfg, "etc/reserved-memory-end", val, sizeof(*val));
1446 load_linux(pcms, fw_cfg);
1449 for (i = 0; i < nb_option_roms; i++) {
1450 rom_add_option(option_rom[i].name, option_rom[i].bootindex);
1452 pcms->fw_cfg = fw_cfg;
1454 /* Init default IOAPIC address space */
1455 pcms->ioapic_as = &address_space_memory;
1459 * The 64bit pci hole starts after "above 4G RAM" and
1460 * potentially the space reserved for memory hotplug.
1462 uint64_t pc_pci_hole64_start(void)
1464 PCMachineState *pcms = PC_MACHINE(qdev_get_machine());
1465 PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
1466 MachineState *ms = MACHINE(pcms);
1467 uint64_t hole64_start = 0;
1469 if (pcmc->has_reserved_memory && ms->device_memory->base) {
1470 hole64_start = ms->device_memory->base;
1471 if (!pcmc->broken_reserved_end) {
1472 hole64_start += memory_region_size(&ms->device_memory->mr);
1475 hole64_start = 0x100000000ULL + pcms->above_4g_mem_size;
1478 return ROUND_UP(hole64_start, 1ULL << 30);
1481 qemu_irq pc_allocate_cpu_irq(void)
1483 return qemu_allocate_irq(pic_irq_request, NULL, 0);
1486 DeviceState *pc_vga_init(ISABus *isa_bus, PCIBus *pci_bus)
1488 DeviceState *dev = NULL;
1490 rom_set_order_override(FW_CFG_ORDER_OVERRIDE_VGA);
1492 PCIDevice *pcidev = pci_vga_init(pci_bus);
1493 dev = pcidev ? &pcidev->qdev : NULL;
1494 } else if (isa_bus) {
1495 ISADevice *isadev = isa_vga_init(isa_bus);
1496 dev = isadev ? DEVICE(isadev) : NULL;
1498 rom_reset_order_override();
1502 static const MemoryRegionOps ioport80_io_ops = {
1503 .write = ioport80_write,
1504 .read = ioport80_read,
1505 .endianness = DEVICE_NATIVE_ENDIAN,
1507 .min_access_size = 1,
1508 .max_access_size = 1,
1512 static const MemoryRegionOps ioportF0_io_ops = {
1513 .write = ioportF0_write,
1514 .read = ioportF0_read,
1515 .endianness = DEVICE_NATIVE_ENDIAN,
1517 .min_access_size = 1,
1518 .max_access_size = 1,
1522 static void pc_superio_init(ISABus *isa_bus, bool create_fdctrl, bool no_vmport)
1525 DriveInfo *fd[MAX_FD];
1527 ISADevice *i8042, *port92, *vmmouse;
1529 serial_hds_isa_init(isa_bus, 0, MAX_ISA_SERIAL_PORTS);
1530 parallel_hds_isa_init(isa_bus, MAX_PARALLEL_PORTS);
1532 for (i = 0; i < MAX_FD; i++) {
1533 fd[i] = drive_get(IF_FLOPPY, 0, i);
1534 create_fdctrl |= !!fd[i];
1536 if (create_fdctrl) {
1537 fdctrl_init_isa(isa_bus, fd);
1540 i8042 = isa_create_simple(isa_bus, "i8042");
1542 vmport_init(isa_bus);
1543 vmmouse = isa_try_create(isa_bus, "vmmouse");
1548 DeviceState *dev = DEVICE(vmmouse);
1549 qdev_prop_set_ptr(dev, "ps2_mouse", i8042);
1550 qdev_init_nofail(dev);
1552 port92 = isa_create_simple(isa_bus, "port92");
1554 a20_line = qemu_allocate_irqs(handle_a20_line_change, first_cpu, 2);
1555 i8042_setup_a20_line(i8042, a20_line[0]);
1556 port92_init(port92, a20_line[1]);
1560 void pc_basic_device_init(ISABus *isa_bus, qemu_irq *gsi,
1561 ISADevice **rtc_state,
1568 DeviceState *hpet = NULL;
1569 int pit_isa_irq = 0;
1570 qemu_irq pit_alt_irq = NULL;
1571 qemu_irq rtc_irq = NULL;
1572 ISADevice *pit = NULL;
1573 MemoryRegion *ioport80_io = g_new(MemoryRegion, 1);
1574 MemoryRegion *ioportF0_io = g_new(MemoryRegion, 1);
1576 memory_region_init_io(ioport80_io, NULL, &ioport80_io_ops, NULL, "ioport80", 1);
1577 memory_region_add_subregion(isa_bus->address_space_io, 0x80, ioport80_io);
1579 memory_region_init_io(ioportF0_io, NULL, &ioportF0_io_ops, NULL, "ioportF0", 1);
1580 memory_region_add_subregion(isa_bus->address_space_io, 0xf0, ioportF0_io);
1583 * Check if an HPET shall be created.
1585 * Without KVM_CAP_PIT_STATE2, we cannot switch off the in-kernel PIT
1586 * when the HPET wants to take over. Thus we have to disable the latter.
1588 if (!no_hpet && (!kvm_irqchip_in_kernel() || kvm_has_pit_state2())) {
1589 /* In order to set property, here not using sysbus_try_create_simple */
1590 hpet = qdev_try_create(NULL, TYPE_HPET);
1592 /* For pc-piix-*, hpet's intcap is always IRQ2. For pc-q35-1.7
1593 * and earlier, use IRQ2 for compat. Otherwise, use IRQ16~23,
1596 uint8_t compat = object_property_get_uint(OBJECT(hpet),
1599 qdev_prop_set_uint32(hpet, HPET_INTCAP, hpet_irqs);
1601 qdev_init_nofail(hpet);
1602 sysbus_mmio_map(SYS_BUS_DEVICE(hpet), 0, HPET_BASE);
1604 for (i = 0; i < GSI_NUM_PINS; i++) {
1605 sysbus_connect_irq(SYS_BUS_DEVICE(hpet), i, gsi[i]);
1608 pit_alt_irq = qdev_get_gpio_in(hpet, HPET_LEGACY_PIT_INT);
1609 rtc_irq = qdev_get_gpio_in(hpet, HPET_LEGACY_RTC_INT);
1612 *rtc_state = mc146818_rtc_init(isa_bus, 2000, rtc_irq);
1614 qemu_register_boot_set(pc_boot_set, *rtc_state);
1616 if (!xen_enabled() && has_pit) {
1617 if (kvm_pit_in_kernel()) {
1618 pit = kvm_pit_init(isa_bus, 0x40);
1620 pit = i8254_pit_init(isa_bus, 0x40, pit_isa_irq, pit_alt_irq);
1623 /* connect PIT to output control line of the HPET */
1624 qdev_connect_gpio_out(hpet, 0, qdev_get_gpio_in(DEVICE(pit), 0));
1626 pcspk_init(isa_bus, pit);
1629 i8257_dma_init(isa_bus, 0);
1632 pc_superio_init(isa_bus, create_fdctrl, no_vmport);
1635 void pc_nic_init(PCMachineClass *pcmc, ISABus *isa_bus, PCIBus *pci_bus)
1639 rom_set_order_override(FW_CFG_ORDER_OVERRIDE_NIC);
1640 for (i = 0; i < nb_nics; i++) {
1641 NICInfo *nd = &nd_table[i];
1642 const char *model = nd->model ? nd->model : pcmc->default_nic_model;
1644 if (g_str_equal(model, "ne2k_isa")) {
1645 pc_init_ne2k_isa(isa_bus, nd);
1647 pci_nic_init_nofail(nd, pci_bus, model, NULL);
1650 rom_reset_order_override();
1653 void ioapic_init_gsi(GSIState *gsi_state, const char *parent_name)
1659 if (kvm_ioapic_in_kernel()) {
1660 dev = qdev_create(NULL, "kvm-ioapic");
1662 dev = qdev_create(NULL, "ioapic");
1665 object_property_add_child(object_resolve_path(parent_name, NULL),
1666 "ioapic", OBJECT(dev), NULL);
1668 qdev_init_nofail(dev);
1669 d = SYS_BUS_DEVICE(dev);
1670 sysbus_mmio_map(d, 0, IO_APIC_DEFAULT_ADDRESS);
1672 for (i = 0; i < IOAPIC_NUM_PINS; i++) {
1673 gsi_state->ioapic_irq[i] = qdev_get_gpio_in(dev, i);
1677 static void pc_memory_plug(HotplugHandler *hotplug_dev,
1678 DeviceState *dev, Error **errp)
1680 HotplugHandlerClass *hhc;
1681 Error *local_err = NULL;
1682 PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1683 PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
1684 PCDIMMDevice *dimm = PC_DIMM(dev);
1685 PCDIMMDeviceClass *ddc = PC_DIMM_GET_CLASS(dimm);
1687 uint64_t align = TARGET_PAGE_SIZE;
1688 bool is_nvdimm = object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM);
1690 mr = ddc->get_memory_region(dimm, &local_err);
1695 if (memory_region_get_alignment(mr) && pcmc->enforce_aligned_dimm) {
1696 align = memory_region_get_alignment(mr);
1700 * When -no-acpi is used with Q35 machine type, no ACPI is built,
1701 * but pcms->acpi_dev is still created. Check !acpi_enabled in
1702 * addition to cover this case.
1704 if (!pcms->acpi_dev || !acpi_enabled) {
1705 error_setg(&local_err,
1706 "memory hotplug is not enabled: missing acpi device or acpi disabled");
1710 if (is_nvdimm && !pcms->acpi_nvdimm_state.is_enabled) {
1711 error_setg(&local_err,
1712 "nvdimm is not enabled: missing 'nvdimm' in '-M'");
1716 pc_dimm_memory_plug(dev, MACHINE(pcms), align, &local_err);
1722 nvdimm_plug(&pcms->acpi_nvdimm_state);
1725 hhc = HOTPLUG_HANDLER_GET_CLASS(pcms->acpi_dev);
1726 hhc->plug(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &error_abort);
1728 error_propagate(errp, local_err);
1731 static void pc_memory_unplug_request(HotplugHandler *hotplug_dev,
1732 DeviceState *dev, Error **errp)
1734 HotplugHandlerClass *hhc;
1735 Error *local_err = NULL;
1736 PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1739 * When -no-acpi is used with Q35 machine type, no ACPI is built,
1740 * but pcms->acpi_dev is still created. Check !acpi_enabled in
1741 * addition to cover this case.
1743 if (!pcms->acpi_dev || !acpi_enabled) {
1744 error_setg(&local_err,
1745 "memory hotplug is not enabled: missing acpi device or acpi disabled");
1749 if (object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM)) {
1750 error_setg(&local_err,
1751 "nvdimm device hot unplug is not supported yet.");
1755 hhc = HOTPLUG_HANDLER_GET_CLASS(pcms->acpi_dev);
1756 hhc->unplug_request(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &local_err);
1759 error_propagate(errp, local_err);
1762 static void pc_memory_unplug(HotplugHandler *hotplug_dev,
1763 DeviceState *dev, Error **errp)
1765 PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1766 HotplugHandlerClass *hhc;
1767 Error *local_err = NULL;
1769 hhc = HOTPLUG_HANDLER_GET_CLASS(pcms->acpi_dev);
1770 hhc->unplug(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &local_err);
1776 pc_dimm_memory_unplug(dev, MACHINE(pcms));
1777 object_unparent(OBJECT(dev));
1780 error_propagate(errp, local_err);
1783 static int pc_apic_cmp(const void *a, const void *b)
1785 CPUArchId *apic_a = (CPUArchId *)a;
1786 CPUArchId *apic_b = (CPUArchId *)b;
1788 return apic_a->arch_id - apic_b->arch_id;
1791 /* returns pointer to CPUArchId descriptor that matches CPU's apic_id
1792 * in ms->possible_cpus->cpus, if ms->possible_cpus->cpus has no
1793 * entry corresponding to CPU's apic_id returns NULL.
1795 static CPUArchId *pc_find_cpu_slot(MachineState *ms, uint32_t id, int *idx)
1797 CPUArchId apic_id, *found_cpu;
1799 apic_id.arch_id = id;
1800 found_cpu = bsearch(&apic_id, ms->possible_cpus->cpus,
1801 ms->possible_cpus->len, sizeof(*ms->possible_cpus->cpus),
1803 if (found_cpu && idx) {
1804 *idx = found_cpu - ms->possible_cpus->cpus;
1809 static void pc_cpu_plug(HotplugHandler *hotplug_dev,
1810 DeviceState *dev, Error **errp)
1812 CPUArchId *found_cpu;
1813 HotplugHandlerClass *hhc;
1814 Error *local_err = NULL;
1815 X86CPU *cpu = X86_CPU(dev);
1816 PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1818 if (pcms->acpi_dev) {
1819 hhc = HOTPLUG_HANDLER_GET_CLASS(pcms->acpi_dev);
1820 hhc->plug(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &local_err);
1826 /* increment the number of CPUs */
1829 rtc_set_cpus_count(pcms->rtc, pcms->boot_cpus);
1832 fw_cfg_modify_i16(pcms->fw_cfg, FW_CFG_NB_CPUS, pcms->boot_cpus);
1835 found_cpu = pc_find_cpu_slot(MACHINE(pcms), cpu->apic_id, NULL);
1836 found_cpu->cpu = OBJECT(dev);
1838 error_propagate(errp, local_err);
1840 static void pc_cpu_unplug_request_cb(HotplugHandler *hotplug_dev,
1841 DeviceState *dev, Error **errp)
1844 HotplugHandlerClass *hhc;
1845 Error *local_err = NULL;
1846 X86CPU *cpu = X86_CPU(dev);
1847 PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1849 if (!pcms->acpi_dev) {
1850 error_setg(&local_err, "CPU hot unplug not supported without ACPI");
1854 pc_find_cpu_slot(MACHINE(pcms), cpu->apic_id, &idx);
1857 error_setg(&local_err, "Boot CPU is unpluggable");
1861 hhc = HOTPLUG_HANDLER_GET_CLASS(pcms->acpi_dev);
1862 hhc->unplug_request(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &local_err);
1869 error_propagate(errp, local_err);
1873 static void pc_cpu_unplug_cb(HotplugHandler *hotplug_dev,
1874 DeviceState *dev, Error **errp)
1876 CPUArchId *found_cpu;
1877 HotplugHandlerClass *hhc;
1878 Error *local_err = NULL;
1879 X86CPU *cpu = X86_CPU(dev);
1880 PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1882 hhc = HOTPLUG_HANDLER_GET_CLASS(pcms->acpi_dev);
1883 hhc->unplug(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &local_err);
1889 found_cpu = pc_find_cpu_slot(MACHINE(pcms), cpu->apic_id, NULL);
1890 found_cpu->cpu = NULL;
1891 object_unparent(OBJECT(dev));
1893 /* decrement the number of CPUs */
1895 /* Update the number of CPUs in CMOS */
1896 rtc_set_cpus_count(pcms->rtc, pcms->boot_cpus);
1897 fw_cfg_modify_i16(pcms->fw_cfg, FW_CFG_NB_CPUS, pcms->boot_cpus);
1899 error_propagate(errp, local_err);
1902 static void pc_cpu_pre_plug(HotplugHandler *hotplug_dev,
1903 DeviceState *dev, Error **errp)
1907 CPUArchId *cpu_slot;
1908 X86CPUTopoInfo topo;
1909 X86CPU *cpu = X86_CPU(dev);
1910 MachineState *ms = MACHINE(hotplug_dev);
1911 PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1913 if(!object_dynamic_cast(OBJECT(cpu), ms->cpu_type)) {
1914 error_setg(errp, "Invalid CPU type, expected cpu type: '%s'",
1919 /* if APIC ID is not set, set it based on socket/core/thread properties */
1920 if (cpu->apic_id == UNASSIGNED_APIC_ID) {
1921 int max_socket = (max_cpus - 1) / smp_threads / smp_cores;
1923 if (cpu->socket_id < 0) {
1924 error_setg(errp, "CPU socket-id is not set");
1926 } else if (cpu->socket_id > max_socket) {
1927 error_setg(errp, "Invalid CPU socket-id: %u must be in range 0:%u",
1928 cpu->socket_id, max_socket);
1931 if (cpu->core_id < 0) {
1932 error_setg(errp, "CPU core-id is not set");
1934 } else if (cpu->core_id > (smp_cores - 1)) {
1935 error_setg(errp, "Invalid CPU core-id: %u must be in range 0:%u",
1936 cpu->core_id, smp_cores - 1);
1939 if (cpu->thread_id < 0) {
1940 error_setg(errp, "CPU thread-id is not set");
1942 } else if (cpu->thread_id > (smp_threads - 1)) {
1943 error_setg(errp, "Invalid CPU thread-id: %u must be in range 0:%u",
1944 cpu->thread_id, smp_threads - 1);
1948 topo.pkg_id = cpu->socket_id;
1949 topo.core_id = cpu->core_id;
1950 topo.smt_id = cpu->thread_id;
1951 cpu->apic_id = apicid_from_topo_ids(smp_cores, smp_threads, &topo);
1954 cpu_slot = pc_find_cpu_slot(MACHINE(pcms), cpu->apic_id, &idx);
1956 MachineState *ms = MACHINE(pcms);
1958 x86_topo_ids_from_apicid(cpu->apic_id, smp_cores, smp_threads, &topo);
1959 error_setg(errp, "Invalid CPU [socket: %u, core: %u, thread: %u] with"
1960 " APIC ID %" PRIu32 ", valid index range 0:%d",
1961 topo.pkg_id, topo.core_id, topo.smt_id, cpu->apic_id,
1962 ms->possible_cpus->len - 1);
1966 if (cpu_slot->cpu) {
1967 error_setg(errp, "CPU[%d] with APIC ID %" PRIu32 " exists",
1972 /* if 'address' properties socket-id/core-id/thread-id are not set, set them
1973 * so that machine_query_hotpluggable_cpus would show correct values
1975 /* TODO: move socket_id/core_id/thread_id checks into x86_cpu_realizefn()
1976 * once -smp refactoring is complete and there will be CPU private
1977 * CPUState::nr_cores and CPUState::nr_threads fields instead of globals */
1978 x86_topo_ids_from_apicid(cpu->apic_id, smp_cores, smp_threads, &topo);
1979 if (cpu->socket_id != -1 && cpu->socket_id != topo.pkg_id) {
1980 error_setg(errp, "property socket-id: %u doesn't match set apic-id:"
1981 " 0x%x (socket-id: %u)", cpu->socket_id, cpu->apic_id, topo.pkg_id);
1984 cpu->socket_id = topo.pkg_id;
1986 if (cpu->core_id != -1 && cpu->core_id != topo.core_id) {
1987 error_setg(errp, "property core-id: %u doesn't match set apic-id:"
1988 " 0x%x (core-id: %u)", cpu->core_id, cpu->apic_id, topo.core_id);
1991 cpu->core_id = topo.core_id;
1993 if (cpu->thread_id != -1 && cpu->thread_id != topo.smt_id) {
1994 error_setg(errp, "property thread-id: %u doesn't match set apic-id:"
1995 " 0x%x (thread-id: %u)", cpu->thread_id, cpu->apic_id, topo.smt_id);
1998 cpu->thread_id = topo.smt_id;
2001 cs->cpu_index = idx;
2003 numa_cpu_pre_plug(cpu_slot, dev, errp);
2006 static void pc_machine_device_pre_plug_cb(HotplugHandler *hotplug_dev,
2007 DeviceState *dev, Error **errp)
2009 if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
2010 pc_cpu_pre_plug(hotplug_dev, dev, errp);
2014 static void pc_machine_device_plug_cb(HotplugHandler *hotplug_dev,
2015 DeviceState *dev, Error **errp)
2017 if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2018 pc_memory_plug(hotplug_dev, dev, errp);
2019 } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
2020 pc_cpu_plug(hotplug_dev, dev, errp);
2024 static void pc_machine_device_unplug_request_cb(HotplugHandler *hotplug_dev,
2025 DeviceState *dev, Error **errp)
2027 if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2028 pc_memory_unplug_request(hotplug_dev, dev, errp);
2029 } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
2030 pc_cpu_unplug_request_cb(hotplug_dev, dev, errp);
2032 error_setg(errp, "acpi: device unplug request for not supported device"
2033 " type: %s", object_get_typename(OBJECT(dev)));
2037 static void pc_machine_device_unplug_cb(HotplugHandler *hotplug_dev,
2038 DeviceState *dev, Error **errp)
2040 if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2041 pc_memory_unplug(hotplug_dev, dev, errp);
2042 } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
2043 pc_cpu_unplug_cb(hotplug_dev, dev, errp);
2045 error_setg(errp, "acpi: device unplug for not supported device"
2046 " type: %s", object_get_typename(OBJECT(dev)));
2050 static HotplugHandler *pc_get_hotpug_handler(MachineState *machine,
2053 if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM) ||
2054 object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
2055 return HOTPLUG_HANDLER(machine);
2062 pc_machine_get_device_memory_region_size(Object *obj, Visitor *v,
2063 const char *name, void *opaque,
2066 MachineState *ms = MACHINE(obj);
2067 int64_t value = memory_region_size(&ms->device_memory->mr);
2069 visit_type_int(v, name, &value, errp);
2072 static void pc_machine_get_max_ram_below_4g(Object *obj, Visitor *v,
2073 const char *name, void *opaque,
2076 PCMachineState *pcms = PC_MACHINE(obj);
2077 uint64_t value = pcms->max_ram_below_4g;
2079 visit_type_size(v, name, &value, errp);
2082 static void pc_machine_set_max_ram_below_4g(Object *obj, Visitor *v,
2083 const char *name, void *opaque,
2086 PCMachineState *pcms = PC_MACHINE(obj);
2087 Error *error = NULL;
2090 visit_type_size(v, name, &value, &error);
2092 error_propagate(errp, error);
2095 if (value > (1ULL << 32)) {
2097 "Machine option 'max-ram-below-4g=%"PRIu64
2098 "' expects size less than or equal to 4G", value);
2099 error_propagate(errp, error);
2103 if (value < (1ULL << 20)) {
2104 warn_report("Only %" PRIu64 " bytes of RAM below the 4GiB boundary,"
2105 "BIOS may not work with less than 1MiB", value);
2108 pcms->max_ram_below_4g = value;
2111 static void pc_machine_get_vmport(Object *obj, Visitor *v, const char *name,
2112 void *opaque, Error **errp)
2114 PCMachineState *pcms = PC_MACHINE(obj);
2115 OnOffAuto vmport = pcms->vmport;
2117 visit_type_OnOffAuto(v, name, &vmport, errp);
2120 static void pc_machine_set_vmport(Object *obj, Visitor *v, const char *name,
2121 void *opaque, Error **errp)
2123 PCMachineState *pcms = PC_MACHINE(obj);
2125 visit_type_OnOffAuto(v, name, &pcms->vmport, errp);
2128 bool pc_machine_is_smm_enabled(PCMachineState *pcms)
2130 bool smm_available = false;
2132 if (pcms->smm == ON_OFF_AUTO_OFF) {
2136 if (tcg_enabled() || qtest_enabled()) {
2137 smm_available = true;
2138 } else if (kvm_enabled()) {
2139 smm_available = kvm_has_smm();
2142 if (smm_available) {
2146 if (pcms->smm == ON_OFF_AUTO_ON) {
2147 error_report("System Management Mode not supported by this hypervisor.");
2153 static void pc_machine_get_smm(Object *obj, Visitor *v, const char *name,
2154 void *opaque, Error **errp)
2156 PCMachineState *pcms = PC_MACHINE(obj);
2157 OnOffAuto smm = pcms->smm;
2159 visit_type_OnOffAuto(v, name, &smm, errp);
2162 static void pc_machine_set_smm(Object *obj, Visitor *v, const char *name,
2163 void *opaque, Error **errp)
2165 PCMachineState *pcms = PC_MACHINE(obj);
2167 visit_type_OnOffAuto(v, name, &pcms->smm, errp);
2170 static bool pc_machine_get_nvdimm(Object *obj, Error **errp)
2172 PCMachineState *pcms = PC_MACHINE(obj);
2174 return pcms->acpi_nvdimm_state.is_enabled;
2177 static void pc_machine_set_nvdimm(Object *obj, bool value, Error **errp)
2179 PCMachineState *pcms = PC_MACHINE(obj);
2181 pcms->acpi_nvdimm_state.is_enabled = value;
2184 static char *pc_machine_get_nvdimm_persistence(Object *obj, Error **errp)
2186 PCMachineState *pcms = PC_MACHINE(obj);
2188 return g_strdup(pcms->acpi_nvdimm_state.persistence_string);
2191 static void pc_machine_set_nvdimm_persistence(Object *obj, const char *value,
2194 PCMachineState *pcms = PC_MACHINE(obj);
2195 AcpiNVDIMMState *nvdimm_state = &pcms->acpi_nvdimm_state;
2197 if (strcmp(value, "cpu") == 0)
2198 nvdimm_state->persistence = 3;
2199 else if (strcmp(value, "mem-ctrl") == 0)
2200 nvdimm_state->persistence = 2;
2202 error_report("-machine nvdimm-persistence=%s: unsupported option", value);
2206 g_free(nvdimm_state->persistence_string);
2207 nvdimm_state->persistence_string = g_strdup(value);
2210 static bool pc_machine_get_smbus(Object *obj, Error **errp)
2212 PCMachineState *pcms = PC_MACHINE(obj);
2217 static void pc_machine_set_smbus(Object *obj, bool value, Error **errp)
2219 PCMachineState *pcms = PC_MACHINE(obj);
2221 pcms->smbus = value;
2224 static bool pc_machine_get_sata(Object *obj, Error **errp)
2226 PCMachineState *pcms = PC_MACHINE(obj);
2231 static void pc_machine_set_sata(Object *obj, bool value, Error **errp)
2233 PCMachineState *pcms = PC_MACHINE(obj);
2238 static bool pc_machine_get_pit(Object *obj, Error **errp)
2240 PCMachineState *pcms = PC_MACHINE(obj);
2245 static void pc_machine_set_pit(Object *obj, bool value, Error **errp)
2247 PCMachineState *pcms = PC_MACHINE(obj);
2252 static void pc_machine_initfn(Object *obj)
2254 PCMachineState *pcms = PC_MACHINE(obj);
2256 pcms->max_ram_below_4g = 0; /* use default */
2257 pcms->smm = ON_OFF_AUTO_AUTO;
2258 pcms->vmport = ON_OFF_AUTO_AUTO;
2259 /* nvdimm is disabled on default. */
2260 pcms->acpi_nvdimm_state.is_enabled = false;
2261 /* acpi build is enabled by default if machine supports it */
2262 pcms->acpi_build_enabled = PC_MACHINE_GET_CLASS(pcms)->has_acpi_build;
2268 static void pc_machine_reset(void)
2273 qemu_devices_reset();
2275 /* Reset APIC after devices have been reset to cancel
2276 * any changes that qemu_devices_reset() might have done.
2281 if (cpu->apic_state) {
2282 device_reset(cpu->apic_state);
2287 static CpuInstanceProperties
2288 pc_cpu_index_to_props(MachineState *ms, unsigned cpu_index)
2290 MachineClass *mc = MACHINE_GET_CLASS(ms);
2291 const CPUArchIdList *possible_cpus = mc->possible_cpu_arch_ids(ms);
2293 assert(cpu_index < possible_cpus->len);
2294 return possible_cpus->cpus[cpu_index].props;
2297 static int64_t pc_get_default_cpu_node_id(const MachineState *ms, int idx)
2299 X86CPUTopoInfo topo;
2301 assert(idx < ms->possible_cpus->len);
2302 x86_topo_ids_from_apicid(ms->possible_cpus->cpus[idx].arch_id,
2303 smp_cores, smp_threads, &topo);
2304 return topo.pkg_id % nb_numa_nodes;
2307 static const CPUArchIdList *pc_possible_cpu_arch_ids(MachineState *ms)
2311 if (ms->possible_cpus) {
2313 * make sure that max_cpus hasn't changed since the first use, i.e.
2314 * -smp hasn't been parsed after it
2316 assert(ms->possible_cpus->len == max_cpus);
2317 return ms->possible_cpus;
2320 ms->possible_cpus = g_malloc0(sizeof(CPUArchIdList) +
2321 sizeof(CPUArchId) * max_cpus);
2322 ms->possible_cpus->len = max_cpus;
2323 for (i = 0; i < ms->possible_cpus->len; i++) {
2324 X86CPUTopoInfo topo;
2326 ms->possible_cpus->cpus[i].type = ms->cpu_type;
2327 ms->possible_cpus->cpus[i].vcpus_count = 1;
2328 ms->possible_cpus->cpus[i].arch_id = x86_cpu_apic_id_from_index(i);
2329 x86_topo_ids_from_apicid(ms->possible_cpus->cpus[i].arch_id,
2330 smp_cores, smp_threads, &topo);
2331 ms->possible_cpus->cpus[i].props.has_socket_id = true;
2332 ms->possible_cpus->cpus[i].props.socket_id = topo.pkg_id;
2333 ms->possible_cpus->cpus[i].props.has_core_id = true;
2334 ms->possible_cpus->cpus[i].props.core_id = topo.core_id;
2335 ms->possible_cpus->cpus[i].props.has_thread_id = true;
2336 ms->possible_cpus->cpus[i].props.thread_id = topo.smt_id;
2338 return ms->possible_cpus;
2341 static void x86_nmi(NMIState *n, int cpu_index, Error **errp)
2343 /* cpu index isn't used */
2347 X86CPU *cpu = X86_CPU(cs);
2349 if (!cpu->apic_state) {
2350 cpu_interrupt(cs, CPU_INTERRUPT_NMI);
2352 apic_deliver_nmi(cpu->apic_state);
2357 static void pc_machine_class_init(ObjectClass *oc, void *data)
2359 MachineClass *mc = MACHINE_CLASS(oc);
2360 PCMachineClass *pcmc = PC_MACHINE_CLASS(oc);
2361 HotplugHandlerClass *hc = HOTPLUG_HANDLER_CLASS(oc);
2362 NMIClass *nc = NMI_CLASS(oc);
2364 pcmc->pci_enabled = true;
2365 pcmc->has_acpi_build = true;
2366 pcmc->rsdp_in_ram = true;
2367 pcmc->smbios_defaults = true;
2368 pcmc->smbios_uuid_encoded = true;
2369 pcmc->gigabyte_align = true;
2370 pcmc->has_reserved_memory = true;
2371 pcmc->kvmclock_enabled = true;
2372 pcmc->enforce_aligned_dimm = true;
2373 /* BIOS ACPI tables: 128K. Other BIOS datastructures: less than 4K reported
2374 * to be used at the moment, 32K should be enough for a while. */
2375 pcmc->acpi_data_size = 0x20000 + 0x8000;
2376 pcmc->save_tsc_khz = true;
2377 pcmc->linuxboot_dma_enabled = true;
2378 assert(!mc->get_hotplug_handler);
2379 mc->get_hotplug_handler = pc_get_hotpug_handler;
2380 mc->cpu_index_to_instance_props = pc_cpu_index_to_props;
2381 mc->get_default_cpu_node_id = pc_get_default_cpu_node_id;
2382 mc->possible_cpu_arch_ids = pc_possible_cpu_arch_ids;
2383 mc->auto_enable_numa_with_memhp = true;
2384 mc->has_hotpluggable_cpus = true;
2385 mc->default_boot_order = "cad";
2386 mc->hot_add_cpu = pc_hot_add_cpu;
2387 mc->block_default_type = IF_IDE;
2389 mc->reset = pc_machine_reset;
2390 hc->pre_plug = pc_machine_device_pre_plug_cb;
2391 hc->plug = pc_machine_device_plug_cb;
2392 hc->unplug_request = pc_machine_device_unplug_request_cb;
2393 hc->unplug = pc_machine_device_unplug_cb;
2394 nc->nmi_monitor_handler = x86_nmi;
2395 mc->default_cpu_type = TARGET_DEFAULT_CPU_TYPE;
2397 object_class_property_add(oc, PC_MACHINE_DEVMEM_REGION_SIZE, "int",
2398 pc_machine_get_device_memory_region_size, NULL,
2399 NULL, NULL, &error_abort);
2401 object_class_property_add(oc, PC_MACHINE_MAX_RAM_BELOW_4G, "size",
2402 pc_machine_get_max_ram_below_4g, pc_machine_set_max_ram_below_4g,
2403 NULL, NULL, &error_abort);
2405 object_class_property_set_description(oc, PC_MACHINE_MAX_RAM_BELOW_4G,
2406 "Maximum ram below the 4G boundary (32bit boundary)", &error_abort);
2408 object_class_property_add(oc, PC_MACHINE_SMM, "OnOffAuto",
2409 pc_machine_get_smm, pc_machine_set_smm,
2410 NULL, NULL, &error_abort);
2411 object_class_property_set_description(oc, PC_MACHINE_SMM,
2412 "Enable SMM (pc & q35)", &error_abort);
2414 object_class_property_add(oc, PC_MACHINE_VMPORT, "OnOffAuto",
2415 pc_machine_get_vmport, pc_machine_set_vmport,
2416 NULL, NULL, &error_abort);
2417 object_class_property_set_description(oc, PC_MACHINE_VMPORT,
2418 "Enable vmport (pc & q35)", &error_abort);
2420 object_class_property_add_bool(oc, PC_MACHINE_NVDIMM,
2421 pc_machine_get_nvdimm, pc_machine_set_nvdimm, &error_abort);
2423 object_class_property_add_str(oc, PC_MACHINE_NVDIMM_PERSIST,
2424 pc_machine_get_nvdimm_persistence,
2425 pc_machine_set_nvdimm_persistence, &error_abort);
2427 object_class_property_add_bool(oc, PC_MACHINE_SMBUS,
2428 pc_machine_get_smbus, pc_machine_set_smbus, &error_abort);
2430 object_class_property_add_bool(oc, PC_MACHINE_SATA,
2431 pc_machine_get_sata, pc_machine_set_sata, &error_abort);
2433 object_class_property_add_bool(oc, PC_MACHINE_PIT,
2434 pc_machine_get_pit, pc_machine_set_pit, &error_abort);
2437 static const TypeInfo pc_machine_info = {
2438 .name = TYPE_PC_MACHINE,
2439 .parent = TYPE_MACHINE,
2441 .instance_size = sizeof(PCMachineState),
2442 .instance_init = pc_machine_initfn,
2443 .class_size = sizeof(PCMachineClass),
2444 .class_init = pc_machine_class_init,
2445 .interfaces = (InterfaceInfo[]) {
2446 { TYPE_HOTPLUG_HANDLER },
2452 static void pc_machine_register_types(void)
2454 type_register_static(&pc_machine_info);
2457 type_init(pc_machine_register_types)