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"
26 #include "qemu/units.h"
28 #include "hw/i386/pc.h"
29 #include "hw/char/serial.h"
30 #include "hw/char/parallel.h"
31 #include "hw/i386/apic.h"
32 #include "hw/i386/topology.h"
33 #include "sysemu/cpus.h"
34 #include "hw/block/fdc.h"
36 #include "hw/pci/pci.h"
37 #include "hw/pci/pci_bus.h"
38 #include "hw/nvram/fw_cfg.h"
39 #include "hw/timer/hpet.h"
40 #include "hw/firmware/smbios.h"
41 #include "hw/loader.h"
43 #include "multiboot.h"
44 #include "hw/timer/mc146818rtc.h"
45 #include "hw/dma/i8257.h"
46 #include "hw/timer/i8254.h"
47 #include "hw/input/i8042.h"
48 #include "hw/audio/pcspk.h"
49 #include "hw/pci/msi.h"
50 #include "hw/sysbus.h"
51 #include "sysemu/sysemu.h"
52 #include "sysemu/numa.h"
53 #include "sysemu/kvm.h"
54 #include "sysemu/qtest.h"
56 #include "hw/xen/xen.h"
57 #include "ui/qemu-spice.h"
58 #include "exec/memory.h"
59 #include "exec/address-spaces.h"
60 #include "sysemu/arch_init.h"
61 #include "qemu/bitmap.h"
62 #include "qemu/config-file.h"
63 #include "qemu/error-report.h"
64 #include "qemu/option.h"
65 #include "hw/acpi/acpi.h"
66 #include "hw/acpi/cpu_hotplug.h"
67 #include "hw/boards.h"
68 #include "acpi-build.h"
69 #include "hw/mem/pc-dimm.h"
70 #include "qapi/error.h"
71 #include "qapi/qapi-visit-common.h"
72 #include "qapi/visitor.h"
75 #include "hw/i386/intel_iommu.h"
76 #include "hw/net/ne2000-isa.h"
78 /* debug PC/ISA interrupts */
82 #define DPRINTF(fmt, ...) \
83 do { printf("CPUIRQ: " fmt , ## __VA_ARGS__); } while (0)
85 #define DPRINTF(fmt, ...)
88 #define FW_CFG_ACPI_TABLES (FW_CFG_ARCH_LOCAL + 0)
89 #define FW_CFG_SMBIOS_ENTRIES (FW_CFG_ARCH_LOCAL + 1)
90 #define FW_CFG_IRQ0_OVERRIDE (FW_CFG_ARCH_LOCAL + 2)
91 #define FW_CFG_E820_TABLE (FW_CFG_ARCH_LOCAL + 3)
92 #define FW_CFG_HPET (FW_CFG_ARCH_LOCAL + 4)
94 #define E820_NR_ENTRIES 16
100 } QEMU_PACKED
__attribute((__aligned__(4)));
104 struct e820_entry entry
[E820_NR_ENTRIES
];
105 } QEMU_PACKED
__attribute((__aligned__(4)));
107 static struct e820_table e820_reserve
;
108 static struct e820_entry
*e820_table
;
109 static unsigned e820_entries
;
110 struct hpet_fw_config hpet_cfg
= {.count
= UINT8_MAX
};
112 GlobalProperty pc_compat_3_1
[] = {
114 .driver
= "intel-iommu",
115 .property
= "dma-drain",
119 const size_t pc_compat_3_1_len
= G_N_ELEMENTS(pc_compat_3_1
);
121 void gsi_handler(void *opaque
, int n
, int level
)
123 GSIState
*s
= opaque
;
125 DPRINTF("pc: %s GSI %d\n", level
? "raising" : "lowering", n
);
126 if (n
< ISA_NUM_IRQS
) {
127 qemu_set_irq(s
->i8259_irq
[n
], level
);
129 qemu_set_irq(s
->ioapic_irq
[n
], level
);
132 static void ioport80_write(void *opaque
, hwaddr addr
, uint64_t data
,
137 static uint64_t ioport80_read(void *opaque
, hwaddr addr
, unsigned size
)
139 return 0xffffffffffffffffULL
;
142 /* MSDOS compatibility mode FPU exception support */
143 static qemu_irq ferr_irq
;
145 void pc_register_ferr_irq(qemu_irq irq
)
150 /* XXX: add IGNNE support */
151 void cpu_set_ferr(CPUX86State
*s
)
153 qemu_irq_raise(ferr_irq
);
156 static void ioportF0_write(void *opaque
, hwaddr addr
, uint64_t data
,
159 qemu_irq_lower(ferr_irq
);
162 static uint64_t ioportF0_read(void *opaque
, hwaddr addr
, unsigned size
)
164 return 0xffffffffffffffffULL
;
168 uint64_t cpu_get_tsc(CPUX86State
*env
)
170 return cpu_get_ticks();
174 int cpu_get_pic_interrupt(CPUX86State
*env
)
176 X86CPU
*cpu
= x86_env_get_cpu(env
);
179 if (!kvm_irqchip_in_kernel()) {
180 intno
= apic_get_interrupt(cpu
->apic_state
);
184 /* read the irq from the PIC */
185 if (!apic_accept_pic_intr(cpu
->apic_state
)) {
190 intno
= pic_read_irq(isa_pic
);
194 static void pic_irq_request(void *opaque
, int irq
, int level
)
196 CPUState
*cs
= first_cpu
;
197 X86CPU
*cpu
= X86_CPU(cs
);
199 DPRINTF("pic_irqs: %s irq %d\n", level
? "raise" : "lower", irq
);
200 if (cpu
->apic_state
&& !kvm_irqchip_in_kernel()) {
203 if (apic_accept_pic_intr(cpu
->apic_state
)) {
204 apic_deliver_pic_intr(cpu
->apic_state
, level
);
209 cpu_interrupt(cs
, CPU_INTERRUPT_HARD
);
211 cpu_reset_interrupt(cs
, CPU_INTERRUPT_HARD
);
216 /* PC cmos mappings */
218 #define REG_EQUIPMENT_BYTE 0x14
220 int cmos_get_fd_drive_type(FloppyDriveType fd0
)
225 case FLOPPY_DRIVE_TYPE_144
:
226 /* 1.44 Mb 3"5 drive */
229 case FLOPPY_DRIVE_TYPE_288
:
230 /* 2.88 Mb 3"5 drive */
233 case FLOPPY_DRIVE_TYPE_120
:
234 /* 1.2 Mb 5"5 drive */
237 case FLOPPY_DRIVE_TYPE_NONE
:
245 static void cmos_init_hd(ISADevice
*s
, int type_ofs
, int info_ofs
,
246 int16_t cylinders
, int8_t heads
, int8_t sectors
)
248 rtc_set_memory(s
, type_ofs
, 47);
249 rtc_set_memory(s
, info_ofs
, cylinders
);
250 rtc_set_memory(s
, info_ofs
+ 1, cylinders
>> 8);
251 rtc_set_memory(s
, info_ofs
+ 2, heads
);
252 rtc_set_memory(s
, info_ofs
+ 3, 0xff);
253 rtc_set_memory(s
, info_ofs
+ 4, 0xff);
254 rtc_set_memory(s
, info_ofs
+ 5, 0xc0 | ((heads
> 8) << 3));
255 rtc_set_memory(s
, info_ofs
+ 6, cylinders
);
256 rtc_set_memory(s
, info_ofs
+ 7, cylinders
>> 8);
257 rtc_set_memory(s
, info_ofs
+ 8, sectors
);
260 /* convert boot_device letter to something recognizable by the bios */
261 static int boot_device2nibble(char boot_device
)
263 switch(boot_device
) {
266 return 0x01; /* floppy boot */
268 return 0x02; /* hard drive boot */
270 return 0x03; /* CD-ROM boot */
272 return 0x04; /* Network boot */
277 static void set_boot_dev(ISADevice
*s
, const char *boot_device
, Error
**errp
)
279 #define PC_MAX_BOOT_DEVICES 3
280 int nbds
, bds
[3] = { 0, };
283 nbds
= strlen(boot_device
);
284 if (nbds
> PC_MAX_BOOT_DEVICES
) {
285 error_setg(errp
, "Too many boot devices for PC");
288 for (i
= 0; i
< nbds
; i
++) {
289 bds
[i
] = boot_device2nibble(boot_device
[i
]);
291 error_setg(errp
, "Invalid boot device for PC: '%c'",
296 rtc_set_memory(s
, 0x3d, (bds
[1] << 4) | bds
[0]);
297 rtc_set_memory(s
, 0x38, (bds
[2] << 4) | (fd_bootchk
? 0x0 : 0x1));
300 static void pc_boot_set(void *opaque
, const char *boot_device
, Error
**errp
)
302 set_boot_dev(opaque
, boot_device
, errp
);
305 static void pc_cmos_init_floppy(ISADevice
*rtc_state
, ISADevice
*floppy
)
308 FloppyDriveType fd_type
[2] = { FLOPPY_DRIVE_TYPE_NONE
,
309 FLOPPY_DRIVE_TYPE_NONE
};
313 for (i
= 0; i
< 2; i
++) {
314 fd_type
[i
] = isa_fdc_get_drive_type(floppy
, i
);
317 val
= (cmos_get_fd_drive_type(fd_type
[0]) << 4) |
318 cmos_get_fd_drive_type(fd_type
[1]);
319 rtc_set_memory(rtc_state
, 0x10, val
);
321 val
= rtc_get_memory(rtc_state
, REG_EQUIPMENT_BYTE
);
323 if (fd_type
[0] != FLOPPY_DRIVE_TYPE_NONE
) {
326 if (fd_type
[1] != FLOPPY_DRIVE_TYPE_NONE
) {
333 val
|= 0x01; /* 1 drive, ready for boot */
336 val
|= 0x41; /* 2 drives, ready for boot */
339 rtc_set_memory(rtc_state
, REG_EQUIPMENT_BYTE
, val
);
342 typedef struct pc_cmos_init_late_arg
{
343 ISADevice
*rtc_state
;
345 } pc_cmos_init_late_arg
;
347 typedef struct check_fdc_state
{
352 static int check_fdc(Object
*obj
, void *opaque
)
354 CheckFdcState
*state
= opaque
;
357 Error
*local_err
= NULL
;
359 fdc
= object_dynamic_cast(obj
, TYPE_ISA_FDC
);
364 iobase
= object_property_get_uint(obj
, "iobase", &local_err
);
365 if (local_err
|| iobase
!= 0x3f0) {
366 error_free(local_err
);
371 state
->multiple
= true;
373 state
->floppy
= ISA_DEVICE(obj
);
378 static const char * const fdc_container_path
[] = {
379 "/unattached", "/peripheral", "/peripheral-anon"
383 * Locate the FDC at IO address 0x3f0, in order to configure the CMOS registers
386 ISADevice
*pc_find_fdc0(void)
390 CheckFdcState state
= { 0 };
392 for (i
= 0; i
< ARRAY_SIZE(fdc_container_path
); i
++) {
393 container
= container_get(qdev_get_machine(), fdc_container_path
[i
]);
394 object_child_foreach(container
, check_fdc
, &state
);
397 if (state
.multiple
) {
398 warn_report("multiple floppy disk controllers with "
399 "iobase=0x3f0 have been found");
400 error_printf("the one being picked for CMOS setup might not reflect "
407 static void pc_cmos_init_late(void *opaque
)
409 pc_cmos_init_late_arg
*arg
= opaque
;
410 ISADevice
*s
= arg
->rtc_state
;
412 int8_t heads
, sectors
;
417 if (arg
->idebus
[0] && ide_get_geometry(arg
->idebus
[0], 0,
418 &cylinders
, &heads
, §ors
) >= 0) {
419 cmos_init_hd(s
, 0x19, 0x1b, cylinders
, heads
, sectors
);
422 if (arg
->idebus
[0] && ide_get_geometry(arg
->idebus
[0], 1,
423 &cylinders
, &heads
, §ors
) >= 0) {
424 cmos_init_hd(s
, 0x1a, 0x24, cylinders
, heads
, sectors
);
427 rtc_set_memory(s
, 0x12, val
);
430 for (i
= 0; i
< 4; i
++) {
431 /* NOTE: ide_get_geometry() returns the physical
432 geometry. It is always such that: 1 <= sects <= 63, 1
433 <= heads <= 16, 1 <= cylinders <= 16383. The BIOS
434 geometry can be different if a translation is done. */
435 if (arg
->idebus
[i
/ 2] &&
436 ide_get_geometry(arg
->idebus
[i
/ 2], i
% 2,
437 &cylinders
, &heads
, §ors
) >= 0) {
438 trans
= ide_get_bios_chs_trans(arg
->idebus
[i
/ 2], i
% 2) - 1;
439 assert((trans
& ~3) == 0);
440 val
|= trans
<< (i
* 2);
443 rtc_set_memory(s
, 0x39, val
);
445 pc_cmos_init_floppy(s
, pc_find_fdc0());
447 qemu_unregister_reset(pc_cmos_init_late
, opaque
);
450 void pc_cmos_init(PCMachineState
*pcms
,
451 BusState
*idebus0
, BusState
*idebus1
,
455 static pc_cmos_init_late_arg arg
;
457 /* various important CMOS locations needed by PC/Bochs bios */
460 /* base memory (first MiB) */
461 val
= MIN(pcms
->below_4g_mem_size
/ KiB
, 640);
462 rtc_set_memory(s
, 0x15, val
);
463 rtc_set_memory(s
, 0x16, val
>> 8);
464 /* extended memory (next 64MiB) */
465 if (pcms
->below_4g_mem_size
> 1 * MiB
) {
466 val
= (pcms
->below_4g_mem_size
- 1 * MiB
) / KiB
;
472 rtc_set_memory(s
, 0x17, val
);
473 rtc_set_memory(s
, 0x18, val
>> 8);
474 rtc_set_memory(s
, 0x30, val
);
475 rtc_set_memory(s
, 0x31, val
>> 8);
476 /* memory between 16MiB and 4GiB */
477 if (pcms
->below_4g_mem_size
> 16 * MiB
) {
478 val
= (pcms
->below_4g_mem_size
- 16 * MiB
) / (64 * KiB
);
484 rtc_set_memory(s
, 0x34, val
);
485 rtc_set_memory(s
, 0x35, val
>> 8);
486 /* memory above 4GiB */
487 val
= pcms
->above_4g_mem_size
/ 65536;
488 rtc_set_memory(s
, 0x5b, val
);
489 rtc_set_memory(s
, 0x5c, val
>> 8);
490 rtc_set_memory(s
, 0x5d, val
>> 16);
492 object_property_add_link(OBJECT(pcms
), "rtc_state",
494 (Object
**)&pcms
->rtc
,
495 object_property_allow_set_link
,
496 OBJ_PROP_LINK_STRONG
, &error_abort
);
497 object_property_set_link(OBJECT(pcms
), OBJECT(s
),
498 "rtc_state", &error_abort
);
500 set_boot_dev(s
, MACHINE(pcms
)->boot_order
, &error_fatal
);
503 val
|= 0x02; /* FPU is there */
504 val
|= 0x04; /* PS/2 mouse installed */
505 rtc_set_memory(s
, REG_EQUIPMENT_BYTE
, val
);
507 /* hard drives and FDC */
509 arg
.idebus
[0] = idebus0
;
510 arg
.idebus
[1] = idebus1
;
511 qemu_register_reset(pc_cmos_init_late
, &arg
);
514 #define TYPE_PORT92 "port92"
515 #define PORT92(obj) OBJECT_CHECK(Port92State, (obj), TYPE_PORT92)
517 /* port 92 stuff: could be split off */
518 typedef struct Port92State
{
519 ISADevice parent_obj
;
526 static void port92_write(void *opaque
, hwaddr addr
, uint64_t val
,
529 Port92State
*s
= opaque
;
530 int oldval
= s
->outport
;
532 DPRINTF("port92: write 0x%02" PRIx64
"\n", val
);
534 qemu_set_irq(s
->a20_out
, (val
>> 1) & 1);
535 if ((val
& 1) && !(oldval
& 1)) {
536 qemu_system_reset_request(SHUTDOWN_CAUSE_GUEST_RESET
);
540 static uint64_t port92_read(void *opaque
, hwaddr addr
,
543 Port92State
*s
= opaque
;
547 DPRINTF("port92: read 0x%02x\n", ret
);
551 static void port92_init(ISADevice
*dev
, qemu_irq a20_out
)
553 qdev_connect_gpio_out_named(DEVICE(dev
), PORT92_A20_LINE
, 0, a20_out
);
556 static const VMStateDescription vmstate_port92_isa
= {
559 .minimum_version_id
= 1,
560 .fields
= (VMStateField
[]) {
561 VMSTATE_UINT8(outport
, Port92State
),
562 VMSTATE_END_OF_LIST()
566 static void port92_reset(DeviceState
*d
)
568 Port92State
*s
= PORT92(d
);
573 static const MemoryRegionOps port92_ops
= {
575 .write
= port92_write
,
577 .min_access_size
= 1,
578 .max_access_size
= 1,
580 .endianness
= DEVICE_LITTLE_ENDIAN
,
583 static void port92_initfn(Object
*obj
)
585 Port92State
*s
= PORT92(obj
);
587 memory_region_init_io(&s
->io
, OBJECT(s
), &port92_ops
, s
, "port92", 1);
591 qdev_init_gpio_out_named(DEVICE(obj
), &s
->a20_out
, PORT92_A20_LINE
, 1);
594 static void port92_realizefn(DeviceState
*dev
, Error
**errp
)
596 ISADevice
*isadev
= ISA_DEVICE(dev
);
597 Port92State
*s
= PORT92(dev
);
599 isa_register_ioport(isadev
, &s
->io
, 0x92);
602 static void port92_class_initfn(ObjectClass
*klass
, void *data
)
604 DeviceClass
*dc
= DEVICE_CLASS(klass
);
606 dc
->realize
= port92_realizefn
;
607 dc
->reset
= port92_reset
;
608 dc
->vmsd
= &vmstate_port92_isa
;
610 * Reason: unlike ordinary ISA devices, this one needs additional
611 * wiring: its A20 output line needs to be wired up by
614 dc
->user_creatable
= false;
617 static const TypeInfo port92_info
= {
619 .parent
= TYPE_ISA_DEVICE
,
620 .instance_size
= sizeof(Port92State
),
621 .instance_init
= port92_initfn
,
622 .class_init
= port92_class_initfn
,
625 static void port92_register_types(void)
627 type_register_static(&port92_info
);
630 type_init(port92_register_types
)
632 static void handle_a20_line_change(void *opaque
, int irq
, int level
)
634 X86CPU
*cpu
= opaque
;
636 /* XXX: send to all CPUs ? */
637 /* XXX: add logic to handle multiple A20 line sources */
638 x86_cpu_set_a20(cpu
, level
);
641 int e820_add_entry(uint64_t address
, uint64_t length
, uint32_t type
)
643 int index
= le32_to_cpu(e820_reserve
.count
);
644 struct e820_entry
*entry
;
646 if (type
!= E820_RAM
) {
647 /* old FW_CFG_E820_TABLE entry -- reservations only */
648 if (index
>= E820_NR_ENTRIES
) {
651 entry
= &e820_reserve
.entry
[index
++];
653 entry
->address
= cpu_to_le64(address
);
654 entry
->length
= cpu_to_le64(length
);
655 entry
->type
= cpu_to_le32(type
);
657 e820_reserve
.count
= cpu_to_le32(index
);
660 /* new "etc/e820" file -- include ram too */
661 e820_table
= g_renew(struct e820_entry
, e820_table
, e820_entries
+ 1);
662 e820_table
[e820_entries
].address
= cpu_to_le64(address
);
663 e820_table
[e820_entries
].length
= cpu_to_le64(length
);
664 e820_table
[e820_entries
].type
= cpu_to_le32(type
);
670 int e820_get_num_entries(void)
675 bool e820_get_entry(int idx
, uint32_t type
, uint64_t *address
, uint64_t *length
)
677 if (idx
< e820_entries
&& e820_table
[idx
].type
== cpu_to_le32(type
)) {
678 *address
= le64_to_cpu(e820_table
[idx
].address
);
679 *length
= le64_to_cpu(e820_table
[idx
].length
);
685 /* Enables contiguous-apic-ID mode, for compatibility */
686 static bool compat_apic_id_mode
;
688 void enable_compat_apic_id_mode(void)
690 compat_apic_id_mode
= true;
693 /* Calculates initial APIC ID for a specific CPU index
695 * Currently we need to be able to calculate the APIC ID from the CPU index
696 * alone (without requiring a CPU object), as the QEMU<->Seabios interfaces have
697 * no concept of "CPU index", and the NUMA tables on fw_cfg need the APIC ID of
698 * all CPUs up to max_cpus.
700 static uint32_t x86_cpu_apic_id_from_index(unsigned int cpu_index
)
705 correct_id
= x86_apicid_from_cpu_idx(smp_cores
, smp_threads
, cpu_index
);
706 if (compat_apic_id_mode
) {
707 if (cpu_index
!= correct_id
&& !warned
&& !qtest_enabled()) {
708 error_report("APIC IDs set in compatibility mode, "
709 "CPU topology won't match the configuration");
718 static void pc_build_smbios(PCMachineState
*pcms
)
720 uint8_t *smbios_tables
, *smbios_anchor
;
721 size_t smbios_tables_len
, smbios_anchor_len
;
722 struct smbios_phys_mem_area
*mem_array
;
723 unsigned i
, array_count
;
724 MachineState
*ms
= MACHINE(pcms
);
725 X86CPU
*cpu
= X86_CPU(ms
->possible_cpus
->cpus
[0].cpu
);
727 /* tell smbios about cpuid version and features */
728 smbios_set_cpuid(cpu
->env
.cpuid_version
, cpu
->env
.features
[FEAT_1_EDX
]);
730 smbios_tables
= smbios_get_table_legacy(&smbios_tables_len
);
732 fw_cfg_add_bytes(pcms
->fw_cfg
, FW_CFG_SMBIOS_ENTRIES
,
733 smbios_tables
, smbios_tables_len
);
736 /* build the array of physical mem area from e820 table */
737 mem_array
= g_malloc0(sizeof(*mem_array
) * e820_get_num_entries());
738 for (i
= 0, array_count
= 0; i
< e820_get_num_entries(); i
++) {
741 if (e820_get_entry(i
, E820_RAM
, &addr
, &len
)) {
742 mem_array
[array_count
].address
= addr
;
743 mem_array
[array_count
].length
= len
;
747 smbios_get_tables(mem_array
, array_count
,
748 &smbios_tables
, &smbios_tables_len
,
749 &smbios_anchor
, &smbios_anchor_len
);
753 fw_cfg_add_file(pcms
->fw_cfg
, "etc/smbios/smbios-tables",
754 smbios_tables
, smbios_tables_len
);
755 fw_cfg_add_file(pcms
->fw_cfg
, "etc/smbios/smbios-anchor",
756 smbios_anchor
, smbios_anchor_len
);
760 static FWCfgState
*bochs_bios_init(AddressSpace
*as
, PCMachineState
*pcms
)
763 uint64_t *numa_fw_cfg
;
765 const CPUArchIdList
*cpus
;
766 MachineClass
*mc
= MACHINE_GET_CLASS(pcms
);
768 fw_cfg
= fw_cfg_init_io_dma(FW_CFG_IO_BASE
, FW_CFG_IO_BASE
+ 4, as
);
769 fw_cfg_add_i16(fw_cfg
, FW_CFG_NB_CPUS
, pcms
->boot_cpus
);
771 /* FW_CFG_MAX_CPUS is a bit confusing/problematic on x86:
773 * For machine types prior to 1.8, SeaBIOS needs FW_CFG_MAX_CPUS for
774 * building MPTable, ACPI MADT, ACPI CPU hotplug and ACPI SRAT table,
775 * that tables are based on xAPIC ID and QEMU<->SeaBIOS interface
776 * for CPU hotplug also uses APIC ID and not "CPU index".
777 * This means that FW_CFG_MAX_CPUS is not the "maximum number of CPUs",
778 * but the "limit to the APIC ID values SeaBIOS may see".
780 * So for compatibility reasons with old BIOSes we are stuck with
781 * "etc/max-cpus" actually being apic_id_limit
783 fw_cfg_add_i16(fw_cfg
, FW_CFG_MAX_CPUS
, (uint16_t)pcms
->apic_id_limit
);
784 fw_cfg_add_i64(fw_cfg
, FW_CFG_RAM_SIZE
, (uint64_t)ram_size
);
785 fw_cfg_add_bytes(fw_cfg
, FW_CFG_ACPI_TABLES
,
786 acpi_tables
, acpi_tables_len
);
787 fw_cfg_add_i32(fw_cfg
, FW_CFG_IRQ0_OVERRIDE
, kvm_allows_irq0_override());
789 fw_cfg_add_bytes(fw_cfg
, FW_CFG_E820_TABLE
,
790 &e820_reserve
, sizeof(e820_reserve
));
791 fw_cfg_add_file(fw_cfg
, "etc/e820", e820_table
,
792 sizeof(struct e820_entry
) * e820_entries
);
794 fw_cfg_add_bytes(fw_cfg
, FW_CFG_HPET
, &hpet_cfg
, sizeof(hpet_cfg
));
795 /* allocate memory for the NUMA channel: one (64bit) word for the number
796 * of nodes, one word for each VCPU->node and one word for each node to
797 * hold the amount of memory.
799 numa_fw_cfg
= g_new0(uint64_t, 1 + pcms
->apic_id_limit
+ nb_numa_nodes
);
800 numa_fw_cfg
[0] = cpu_to_le64(nb_numa_nodes
);
801 cpus
= mc
->possible_cpu_arch_ids(MACHINE(pcms
));
802 for (i
= 0; i
< cpus
->len
; i
++) {
803 unsigned int apic_id
= cpus
->cpus
[i
].arch_id
;
804 assert(apic_id
< pcms
->apic_id_limit
);
805 numa_fw_cfg
[apic_id
+ 1] = cpu_to_le64(cpus
->cpus
[i
].props
.node_id
);
807 for (i
= 0; i
< nb_numa_nodes
; i
++) {
808 numa_fw_cfg
[pcms
->apic_id_limit
+ 1 + i
] =
809 cpu_to_le64(numa_info
[i
].node_mem
);
811 fw_cfg_add_bytes(fw_cfg
, FW_CFG_NUMA
, numa_fw_cfg
,
812 (1 + pcms
->apic_id_limit
+ nb_numa_nodes
) *
813 sizeof(*numa_fw_cfg
));
818 static long get_file_size(FILE *f
)
822 /* XXX: on Unix systems, using fstat() probably makes more sense */
825 fseek(f
, 0, SEEK_END
);
827 fseek(f
, where
, SEEK_SET
);
832 /* setup_data types */
834 #define SETUP_E820_EXT 1
844 } __attribute__((packed
));
846 static void load_linux(PCMachineState
*pcms
,
850 int setup_size
, kernel_size
, cmdline_size
;
851 int dtb_size
, setup_data_offset
;
853 uint8_t header
[8192], *setup
, *kernel
;
854 hwaddr real_addr
, prot_addr
, cmdline_addr
, initrd_addr
= 0;
857 MachineState
*machine
= MACHINE(pcms
);
858 PCMachineClass
*pcmc
= PC_MACHINE_GET_CLASS(pcms
);
859 struct setup_data
*setup_data
;
860 const char *kernel_filename
= machine
->kernel_filename
;
861 const char *initrd_filename
= machine
->initrd_filename
;
862 const char *dtb_filename
= machine
->dtb
;
863 const char *kernel_cmdline
= machine
->kernel_cmdline
;
865 /* Align to 16 bytes as a paranoia measure */
866 cmdline_size
= (strlen(kernel_cmdline
)+16) & ~15;
868 /* load the kernel header */
869 f
= fopen(kernel_filename
, "rb");
870 if (!f
|| !(kernel_size
= get_file_size(f
)) ||
871 fread(header
, 1, MIN(ARRAY_SIZE(header
), kernel_size
), f
) !=
872 MIN(ARRAY_SIZE(header
), kernel_size
)) {
873 fprintf(stderr
, "qemu: could not load kernel '%s': %s\n",
874 kernel_filename
, strerror(errno
));
878 /* kernel protocol version */
880 fprintf(stderr
, "header magic: %#x\n", ldl_p(header
+0x202));
882 if (ldl_p(header
+0x202) == 0x53726448) {
883 protocol
= lduw_p(header
+0x206);
885 /* This looks like a multiboot kernel. If it is, let's stop
886 treating it like a Linux kernel. */
887 if (load_multiboot(fw_cfg
, f
, kernel_filename
, initrd_filename
,
888 kernel_cmdline
, kernel_size
, header
)) {
894 if (protocol
< 0x200 || !(header
[0x211] & 0x01)) {
897 cmdline_addr
= 0x9a000 - cmdline_size
;
899 } else if (protocol
< 0x202) {
900 /* High but ancient kernel */
902 cmdline_addr
= 0x9a000 - cmdline_size
;
903 prot_addr
= 0x100000;
905 /* High and recent kernel */
907 cmdline_addr
= 0x20000;
908 prot_addr
= 0x100000;
913 "qemu: real_addr = 0x" TARGET_FMT_plx
"\n"
914 "qemu: cmdline_addr = 0x" TARGET_FMT_plx
"\n"
915 "qemu: prot_addr = 0x" TARGET_FMT_plx
"\n",
921 /* highest address for loading the initrd */
922 if (protocol
>= 0x203) {
923 initrd_max
= ldl_p(header
+0x22c);
925 initrd_max
= 0x37ffffff;
928 if (initrd_max
>= pcms
->below_4g_mem_size
- pcmc
->acpi_data_size
) {
929 initrd_max
= pcms
->below_4g_mem_size
- pcmc
->acpi_data_size
- 1;
932 fw_cfg_add_i32(fw_cfg
, FW_CFG_CMDLINE_ADDR
, cmdline_addr
);
933 fw_cfg_add_i32(fw_cfg
, FW_CFG_CMDLINE_SIZE
, strlen(kernel_cmdline
)+1);
934 fw_cfg_add_string(fw_cfg
, FW_CFG_CMDLINE_DATA
, kernel_cmdline
);
936 if (protocol
>= 0x202) {
937 stl_p(header
+0x228, cmdline_addr
);
939 stw_p(header
+0x20, 0xA33F);
940 stw_p(header
+0x22, cmdline_addr
-real_addr
);
943 /* handle vga= parameter */
944 vmode
= strstr(kernel_cmdline
, "vga=");
946 unsigned int video_mode
;
949 if (!strncmp(vmode
, "normal", 6)) {
951 } else if (!strncmp(vmode
, "ext", 3)) {
953 } else if (!strncmp(vmode
, "ask", 3)) {
956 video_mode
= strtol(vmode
, NULL
, 0);
958 stw_p(header
+0x1fa, video_mode
);
962 /* High nybble = B reserved for QEMU; low nybble is revision number.
963 If this code is substantially changed, you may want to consider
964 incrementing the revision. */
965 if (protocol
>= 0x200) {
966 header
[0x210] = 0xB0;
969 if (protocol
>= 0x201) {
970 header
[0x211] |= 0x80; /* CAN_USE_HEAP */
971 stw_p(header
+0x224, cmdline_addr
-real_addr
-0x200);
975 if (initrd_filename
) {
980 if (protocol
< 0x200) {
981 fprintf(stderr
, "qemu: linux kernel too old to load a ram disk\n");
985 if (!g_file_get_contents(initrd_filename
, &initrd_data
,
986 &initrd_size
, &gerr
)) {
987 fprintf(stderr
, "qemu: error reading initrd %s: %s\n",
988 initrd_filename
, gerr
->message
);
991 if (initrd_size
>= initrd_max
) {
992 fprintf(stderr
, "qemu: initrd is too large, cannot support."
993 "(max: %"PRIu32
", need %"PRId64
")\n",
994 initrd_max
, (uint64_t)initrd_size
);
998 initrd_addr
= (initrd_max
-initrd_size
) & ~4095;
1000 fw_cfg_add_i32(fw_cfg
, FW_CFG_INITRD_ADDR
, initrd_addr
);
1001 fw_cfg_add_i32(fw_cfg
, FW_CFG_INITRD_SIZE
, initrd_size
);
1002 fw_cfg_add_bytes(fw_cfg
, FW_CFG_INITRD_DATA
, initrd_data
, initrd_size
);
1004 stl_p(header
+0x218, initrd_addr
);
1005 stl_p(header
+0x21c, initrd_size
);
1008 /* load kernel and setup */
1009 setup_size
= header
[0x1f1];
1010 if (setup_size
== 0) {
1013 setup_size
= (setup_size
+1)*512;
1014 if (setup_size
> kernel_size
) {
1015 fprintf(stderr
, "qemu: invalid kernel header\n");
1018 kernel_size
-= setup_size
;
1020 setup
= g_malloc(setup_size
);
1021 kernel
= g_malloc(kernel_size
);
1022 fseek(f
, 0, SEEK_SET
);
1023 if (fread(setup
, 1, setup_size
, f
) != setup_size
) {
1024 fprintf(stderr
, "fread() failed\n");
1027 if (fread(kernel
, 1, kernel_size
, f
) != kernel_size
) {
1028 fprintf(stderr
, "fread() failed\n");
1033 /* append dtb to kernel */
1035 if (protocol
< 0x209) {
1036 fprintf(stderr
, "qemu: Linux kernel too old to load a dtb\n");
1040 dtb_size
= get_image_size(dtb_filename
);
1041 if (dtb_size
<= 0) {
1042 fprintf(stderr
, "qemu: error reading dtb %s: %s\n",
1043 dtb_filename
, strerror(errno
));
1047 setup_data_offset
= QEMU_ALIGN_UP(kernel_size
, 16);
1048 kernel_size
= setup_data_offset
+ sizeof(struct setup_data
) + dtb_size
;
1049 kernel
= g_realloc(kernel
, kernel_size
);
1051 stq_p(header
+0x250, prot_addr
+ setup_data_offset
);
1053 setup_data
= (struct setup_data
*)(kernel
+ setup_data_offset
);
1054 setup_data
->next
= 0;
1055 setup_data
->type
= cpu_to_le32(SETUP_DTB
);
1056 setup_data
->len
= cpu_to_le32(dtb_size
);
1058 load_image_size(dtb_filename
, setup_data
->data
, dtb_size
);
1061 memcpy(setup
, header
, MIN(sizeof(header
), setup_size
));
1063 fw_cfg_add_i32(fw_cfg
, FW_CFG_KERNEL_ADDR
, prot_addr
);
1064 fw_cfg_add_i32(fw_cfg
, FW_CFG_KERNEL_SIZE
, kernel_size
);
1065 fw_cfg_add_bytes(fw_cfg
, FW_CFG_KERNEL_DATA
, kernel
, kernel_size
);
1067 fw_cfg_add_i32(fw_cfg
, FW_CFG_SETUP_ADDR
, real_addr
);
1068 fw_cfg_add_i32(fw_cfg
, FW_CFG_SETUP_SIZE
, setup_size
);
1069 fw_cfg_add_bytes(fw_cfg
, FW_CFG_SETUP_DATA
, setup
, setup_size
);
1071 option_rom
[nb_option_roms
].bootindex
= 0;
1072 option_rom
[nb_option_roms
].name
= "linuxboot.bin";
1073 if (pcmc
->linuxboot_dma_enabled
&& fw_cfg_dma_enabled(fw_cfg
)) {
1074 option_rom
[nb_option_roms
].name
= "linuxboot_dma.bin";
1079 #define NE2000_NB_MAX 6
1081 static const int ne2000_io
[NE2000_NB_MAX
] = { 0x300, 0x320, 0x340, 0x360,
1083 static const int ne2000_irq
[NE2000_NB_MAX
] = { 9, 10, 11, 3, 4, 5 };
1085 void pc_init_ne2k_isa(ISABus
*bus
, NICInfo
*nd
)
1087 static int nb_ne2k
= 0;
1089 if (nb_ne2k
== NE2000_NB_MAX
)
1091 isa_ne2000_init(bus
, ne2000_io
[nb_ne2k
],
1092 ne2000_irq
[nb_ne2k
], nd
);
1096 DeviceState
*cpu_get_current_apic(void)
1099 X86CPU
*cpu
= X86_CPU(current_cpu
);
1100 return cpu
->apic_state
;
1106 void pc_acpi_smi_interrupt(void *opaque
, int irq
, int level
)
1108 X86CPU
*cpu
= opaque
;
1111 cpu_interrupt(CPU(cpu
), CPU_INTERRUPT_SMI
);
1115 static void pc_new_cpu(const char *typename
, int64_t apic_id
, Error
**errp
)
1118 Error
*local_err
= NULL
;
1120 cpu
= object_new(typename
);
1122 object_property_set_uint(cpu
, apic_id
, "apic-id", &local_err
);
1123 object_property_set_bool(cpu
, true, "realized", &local_err
);
1126 error_propagate(errp
, local_err
);
1129 void pc_hot_add_cpu(const int64_t id
, Error
**errp
)
1131 MachineState
*ms
= MACHINE(qdev_get_machine());
1132 int64_t apic_id
= x86_cpu_apic_id_from_index(id
);
1133 Error
*local_err
= NULL
;
1136 error_setg(errp
, "Invalid CPU id: %" PRIi64
, id
);
1140 if (apic_id
>= ACPI_CPU_HOTPLUG_ID_LIMIT
) {
1141 error_setg(errp
, "Unable to add CPU: %" PRIi64
1142 ", resulting APIC ID (%" PRIi64
") is too large",
1147 pc_new_cpu(ms
->cpu_type
, apic_id
, &local_err
);
1149 error_propagate(errp
, local_err
);
1154 void pc_cpus_init(PCMachineState
*pcms
)
1157 const CPUArchIdList
*possible_cpus
;
1158 MachineState
*ms
= MACHINE(pcms
);
1159 MachineClass
*mc
= MACHINE_GET_CLASS(pcms
);
1161 /* Calculates the limit to CPU APIC ID values
1163 * Limit for the APIC ID value, so that all
1164 * CPU APIC IDs are < pcms->apic_id_limit.
1166 * This is used for FW_CFG_MAX_CPUS. See comments on bochs_bios_init().
1168 pcms
->apic_id_limit
= x86_cpu_apic_id_from_index(max_cpus
- 1) + 1;
1169 possible_cpus
= mc
->possible_cpu_arch_ids(ms
);
1170 for (i
= 0; i
< smp_cpus
; i
++) {
1171 pc_new_cpu(possible_cpus
->cpus
[i
].type
, possible_cpus
->cpus
[i
].arch_id
,
1176 static void pc_build_feature_control_file(PCMachineState
*pcms
)
1178 MachineState
*ms
= MACHINE(pcms
);
1179 X86CPU
*cpu
= X86_CPU(ms
->possible_cpus
->cpus
[0].cpu
);
1180 CPUX86State
*env
= &cpu
->env
;
1181 uint32_t unused
, ecx
, edx
;
1182 uint64_t feature_control_bits
= 0;
1185 cpu_x86_cpuid(env
, 1, 0, &unused
, &unused
, &ecx
, &edx
);
1186 if (ecx
& CPUID_EXT_VMX
) {
1187 feature_control_bits
|= FEATURE_CONTROL_VMXON_ENABLED_OUTSIDE_SMX
;
1190 if ((edx
& (CPUID_EXT2_MCE
| CPUID_EXT2_MCA
)) ==
1191 (CPUID_EXT2_MCE
| CPUID_EXT2_MCA
) &&
1192 (env
->mcg_cap
& MCG_LMCE_P
)) {
1193 feature_control_bits
|= FEATURE_CONTROL_LMCE
;
1196 if (!feature_control_bits
) {
1200 val
= g_malloc(sizeof(*val
));
1201 *val
= cpu_to_le64(feature_control_bits
| FEATURE_CONTROL_LOCKED
);
1202 fw_cfg_add_file(pcms
->fw_cfg
, "etc/msr_feature_control", val
, sizeof(*val
));
1205 static void rtc_set_cpus_count(ISADevice
*rtc
, uint16_t cpus_count
)
1207 if (cpus_count
> 0xff) {
1208 /* If the number of CPUs can't be represented in 8 bits, the
1209 * BIOS must use "FW_CFG_NB_CPUS". Set RTC field to 0 just
1210 * to make old BIOSes fail more predictably.
1212 rtc_set_memory(rtc
, 0x5f, 0);
1214 rtc_set_memory(rtc
, 0x5f, cpus_count
- 1);
1219 void pc_machine_done(Notifier
*notifier
, void *data
)
1221 PCMachineState
*pcms
= container_of(notifier
,
1222 PCMachineState
, machine_done
);
1223 PCIBus
*bus
= pcms
->bus
;
1225 /* set the number of CPUs */
1226 rtc_set_cpus_count(pcms
->rtc
, pcms
->boot_cpus
);
1229 int extra_hosts
= 0;
1231 QLIST_FOREACH(bus
, &bus
->child
, sibling
) {
1232 /* look for expander root buses */
1233 if (pci_bus_is_root(bus
)) {
1237 if (extra_hosts
&& pcms
->fw_cfg
) {
1238 uint64_t *val
= g_malloc(sizeof(*val
));
1239 *val
= cpu_to_le64(extra_hosts
);
1240 fw_cfg_add_file(pcms
->fw_cfg
,
1241 "etc/extra-pci-roots", val
, sizeof(*val
));
1247 pc_build_smbios(pcms
);
1248 pc_build_feature_control_file(pcms
);
1249 /* update FW_CFG_NB_CPUS to account for -device added CPUs */
1250 fw_cfg_modify_i16(pcms
->fw_cfg
, FW_CFG_NB_CPUS
, pcms
->boot_cpus
);
1253 if (pcms
->apic_id_limit
> 255 && !xen_enabled()) {
1254 IntelIOMMUState
*iommu
= INTEL_IOMMU_DEVICE(x86_iommu_get_default());
1256 if (!iommu
|| !x86_iommu_ir_supported(X86_IOMMU_DEVICE(iommu
)) ||
1257 iommu
->intr_eim
!= ON_OFF_AUTO_ON
) {
1258 error_report("current -smp configuration requires "
1259 "Extended Interrupt Mode enabled. "
1260 "You can add an IOMMU using: "
1261 "-device intel-iommu,intremap=on,eim=on");
1267 void pc_guest_info_init(PCMachineState
*pcms
)
1271 pcms
->apic_xrupt_override
= kvm_allows_irq0_override();
1272 pcms
->numa_nodes
= nb_numa_nodes
;
1273 pcms
->node_mem
= g_malloc0(pcms
->numa_nodes
*
1274 sizeof *pcms
->node_mem
);
1275 for (i
= 0; i
< nb_numa_nodes
; i
++) {
1276 pcms
->node_mem
[i
] = numa_info
[i
].node_mem
;
1279 pcms
->machine_done
.notify
= pc_machine_done
;
1280 qemu_add_machine_init_done_notifier(&pcms
->machine_done
);
1283 /* setup pci memory address space mapping into system address space */
1284 void pc_pci_as_mapping_init(Object
*owner
, MemoryRegion
*system_memory
,
1285 MemoryRegion
*pci_address_space
)
1287 /* Set to lower priority than RAM */
1288 memory_region_add_subregion_overlap(system_memory
, 0x0,
1289 pci_address_space
, -1);
1292 void pc_acpi_init(const char *default_dsdt
)
1296 if (acpi_tables
!= NULL
) {
1297 /* manually set via -acpitable, leave it alone */
1301 filename
= qemu_find_file(QEMU_FILE_TYPE_BIOS
, default_dsdt
);
1302 if (filename
== NULL
) {
1303 warn_report("failed to find %s", default_dsdt
);
1305 QemuOpts
*opts
= qemu_opts_create(qemu_find_opts("acpi"), NULL
, 0,
1309 qemu_opt_set(opts
, "file", filename
, &error_abort
);
1311 acpi_table_add_builtin(opts
, &err
);
1313 warn_reportf_err(err
, "failed to load %s: ", filename
);
1319 void xen_load_linux(PCMachineState
*pcms
)
1324 assert(MACHINE(pcms
)->kernel_filename
!= NULL
);
1326 fw_cfg
= fw_cfg_init_io(FW_CFG_IO_BASE
);
1327 fw_cfg_add_i16(fw_cfg
, FW_CFG_NB_CPUS
, pcms
->boot_cpus
);
1330 load_linux(pcms
, fw_cfg
);
1331 for (i
= 0; i
< nb_option_roms
; i
++) {
1332 assert(!strcmp(option_rom
[i
].name
, "linuxboot.bin") ||
1333 !strcmp(option_rom
[i
].name
, "linuxboot_dma.bin") ||
1334 !strcmp(option_rom
[i
].name
, "multiboot.bin"));
1335 rom_add_option(option_rom
[i
].name
, option_rom
[i
].bootindex
);
1337 pcms
->fw_cfg
= fw_cfg
;
1340 void pc_memory_init(PCMachineState
*pcms
,
1341 MemoryRegion
*system_memory
,
1342 MemoryRegion
*rom_memory
,
1343 MemoryRegion
**ram_memory
)
1346 MemoryRegion
*ram
, *option_rom_mr
;
1347 MemoryRegion
*ram_below_4g
, *ram_above_4g
;
1349 MachineState
*machine
= MACHINE(pcms
);
1350 PCMachineClass
*pcmc
= PC_MACHINE_GET_CLASS(pcms
);
1352 assert(machine
->ram_size
== pcms
->below_4g_mem_size
+
1353 pcms
->above_4g_mem_size
);
1355 linux_boot
= (machine
->kernel_filename
!= NULL
);
1357 /* Allocate RAM. We allocate it as a single memory region and use
1358 * aliases to address portions of it, mostly for backwards compatibility
1359 * with older qemus that used qemu_ram_alloc().
1361 ram
= g_malloc(sizeof(*ram
));
1362 memory_region_allocate_system_memory(ram
, NULL
, "pc.ram",
1365 ram_below_4g
= g_malloc(sizeof(*ram_below_4g
));
1366 memory_region_init_alias(ram_below_4g
, NULL
, "ram-below-4g", ram
,
1367 0, pcms
->below_4g_mem_size
);
1368 memory_region_add_subregion(system_memory
, 0, ram_below_4g
);
1369 e820_add_entry(0, pcms
->below_4g_mem_size
, E820_RAM
);
1370 if (pcms
->above_4g_mem_size
> 0) {
1371 ram_above_4g
= g_malloc(sizeof(*ram_above_4g
));
1372 memory_region_init_alias(ram_above_4g
, NULL
, "ram-above-4g", ram
,
1373 pcms
->below_4g_mem_size
,
1374 pcms
->above_4g_mem_size
);
1375 memory_region_add_subregion(system_memory
, 0x100000000ULL
,
1377 e820_add_entry(0x100000000ULL
, pcms
->above_4g_mem_size
, E820_RAM
);
1380 if (!pcmc
->has_reserved_memory
&&
1381 (machine
->ram_slots
||
1382 (machine
->maxram_size
> machine
->ram_size
))) {
1383 MachineClass
*mc
= MACHINE_GET_CLASS(machine
);
1385 error_report("\"-memory 'slots|maxmem'\" is not supported by: %s",
1390 /* always allocate the device memory information */
1391 machine
->device_memory
= g_malloc0(sizeof(*machine
->device_memory
));
1393 /* initialize device memory address space */
1394 if (pcmc
->has_reserved_memory
&&
1395 (machine
->ram_size
< machine
->maxram_size
)) {
1396 ram_addr_t device_mem_size
= machine
->maxram_size
- machine
->ram_size
;
1398 if (machine
->ram_slots
> ACPI_MAX_RAM_SLOTS
) {
1399 error_report("unsupported amount of memory slots: %"PRIu64
,
1400 machine
->ram_slots
);
1404 if (QEMU_ALIGN_UP(machine
->maxram_size
,
1405 TARGET_PAGE_SIZE
) != machine
->maxram_size
) {
1406 error_report("maximum memory size must by aligned to multiple of "
1407 "%d bytes", TARGET_PAGE_SIZE
);
1411 machine
->device_memory
->base
=
1412 ROUND_UP(0x100000000ULL
+ pcms
->above_4g_mem_size
, 1 * GiB
);
1414 if (pcmc
->enforce_aligned_dimm
) {
1415 /* size device region assuming 1G page max alignment per slot */
1416 device_mem_size
+= (1 * GiB
) * machine
->ram_slots
;
1419 if ((machine
->device_memory
->base
+ device_mem_size
) <
1421 error_report("unsupported amount of maximum memory: " RAM_ADDR_FMT
,
1422 machine
->maxram_size
);
1426 memory_region_init(&machine
->device_memory
->mr
, OBJECT(pcms
),
1427 "device-memory", device_mem_size
);
1428 memory_region_add_subregion(system_memory
, machine
->device_memory
->base
,
1429 &machine
->device_memory
->mr
);
1432 /* Initialize PC system firmware */
1433 pc_system_firmware_init(rom_memory
, !pcmc
->pci_enabled
);
1435 option_rom_mr
= g_malloc(sizeof(*option_rom_mr
));
1436 memory_region_init_ram(option_rom_mr
, NULL
, "pc.rom", PC_ROM_SIZE
,
1438 if (pcmc
->pci_enabled
) {
1439 memory_region_set_readonly(option_rom_mr
, true);
1441 memory_region_add_subregion_overlap(rom_memory
,
1446 fw_cfg
= bochs_bios_init(&address_space_memory
, pcms
);
1450 if (pcmc
->has_reserved_memory
&& machine
->device_memory
->base
) {
1451 uint64_t *val
= g_malloc(sizeof(*val
));
1452 PCMachineClass
*pcmc
= PC_MACHINE_GET_CLASS(pcms
);
1453 uint64_t res_mem_end
= machine
->device_memory
->base
;
1455 if (!pcmc
->broken_reserved_end
) {
1456 res_mem_end
+= memory_region_size(&machine
->device_memory
->mr
);
1458 *val
= cpu_to_le64(ROUND_UP(res_mem_end
, 1 * GiB
));
1459 fw_cfg_add_file(fw_cfg
, "etc/reserved-memory-end", val
, sizeof(*val
));
1463 load_linux(pcms
, fw_cfg
);
1466 for (i
= 0; i
< nb_option_roms
; i
++) {
1467 rom_add_option(option_rom
[i
].name
, option_rom
[i
].bootindex
);
1469 pcms
->fw_cfg
= fw_cfg
;
1471 /* Init default IOAPIC address space */
1472 pcms
->ioapic_as
= &address_space_memory
;
1476 * The 64bit pci hole starts after "above 4G RAM" and
1477 * potentially the space reserved for memory hotplug.
1479 uint64_t pc_pci_hole64_start(void)
1481 PCMachineState
*pcms
= PC_MACHINE(qdev_get_machine());
1482 PCMachineClass
*pcmc
= PC_MACHINE_GET_CLASS(pcms
);
1483 MachineState
*ms
= MACHINE(pcms
);
1484 uint64_t hole64_start
= 0;
1486 if (pcmc
->has_reserved_memory
&& ms
->device_memory
->base
) {
1487 hole64_start
= ms
->device_memory
->base
;
1488 if (!pcmc
->broken_reserved_end
) {
1489 hole64_start
+= memory_region_size(&ms
->device_memory
->mr
);
1492 hole64_start
= 0x100000000ULL
+ pcms
->above_4g_mem_size
;
1495 return ROUND_UP(hole64_start
, 1 * GiB
);
1498 qemu_irq
pc_allocate_cpu_irq(void)
1500 return qemu_allocate_irq(pic_irq_request
, NULL
, 0);
1503 DeviceState
*pc_vga_init(ISABus
*isa_bus
, PCIBus
*pci_bus
)
1505 DeviceState
*dev
= NULL
;
1507 rom_set_order_override(FW_CFG_ORDER_OVERRIDE_VGA
);
1509 PCIDevice
*pcidev
= pci_vga_init(pci_bus
);
1510 dev
= pcidev
? &pcidev
->qdev
: NULL
;
1511 } else if (isa_bus
) {
1512 ISADevice
*isadev
= isa_vga_init(isa_bus
);
1513 dev
= isadev
? DEVICE(isadev
) : NULL
;
1515 rom_reset_order_override();
1519 static const MemoryRegionOps ioport80_io_ops
= {
1520 .write
= ioport80_write
,
1521 .read
= ioport80_read
,
1522 .endianness
= DEVICE_NATIVE_ENDIAN
,
1524 .min_access_size
= 1,
1525 .max_access_size
= 1,
1529 static const MemoryRegionOps ioportF0_io_ops
= {
1530 .write
= ioportF0_write
,
1531 .read
= ioportF0_read
,
1532 .endianness
= DEVICE_NATIVE_ENDIAN
,
1534 .min_access_size
= 1,
1535 .max_access_size
= 1,
1539 static void pc_superio_init(ISABus
*isa_bus
, bool create_fdctrl
, bool no_vmport
)
1542 DriveInfo
*fd
[MAX_FD
];
1544 ISADevice
*i8042
, *port92
, *vmmouse
;
1546 serial_hds_isa_init(isa_bus
, 0, MAX_ISA_SERIAL_PORTS
);
1547 parallel_hds_isa_init(isa_bus
, MAX_PARALLEL_PORTS
);
1549 for (i
= 0; i
< MAX_FD
; i
++) {
1550 fd
[i
] = drive_get(IF_FLOPPY
, 0, i
);
1551 create_fdctrl
|= !!fd
[i
];
1553 if (create_fdctrl
) {
1554 fdctrl_init_isa(isa_bus
, fd
);
1557 i8042
= isa_create_simple(isa_bus
, "i8042");
1559 vmport_init(isa_bus
);
1560 vmmouse
= isa_try_create(isa_bus
, "vmmouse");
1565 DeviceState
*dev
= DEVICE(vmmouse
);
1566 qdev_prop_set_ptr(dev
, "ps2_mouse", i8042
);
1567 qdev_init_nofail(dev
);
1569 port92
= isa_create_simple(isa_bus
, "port92");
1571 a20_line
= qemu_allocate_irqs(handle_a20_line_change
, first_cpu
, 2);
1572 i8042_setup_a20_line(i8042
, a20_line
[0]);
1573 port92_init(port92
, a20_line
[1]);
1577 void pc_basic_device_init(ISABus
*isa_bus
, qemu_irq
*gsi
,
1578 ISADevice
**rtc_state
,
1585 DeviceState
*hpet
= NULL
;
1586 int pit_isa_irq
= 0;
1587 qemu_irq pit_alt_irq
= NULL
;
1588 qemu_irq rtc_irq
= NULL
;
1589 ISADevice
*pit
= NULL
;
1590 MemoryRegion
*ioport80_io
= g_new(MemoryRegion
, 1);
1591 MemoryRegion
*ioportF0_io
= g_new(MemoryRegion
, 1);
1593 memory_region_init_io(ioport80_io
, NULL
, &ioport80_io_ops
, NULL
, "ioport80", 1);
1594 memory_region_add_subregion(isa_bus
->address_space_io
, 0x80, ioport80_io
);
1596 memory_region_init_io(ioportF0_io
, NULL
, &ioportF0_io_ops
, NULL
, "ioportF0", 1);
1597 memory_region_add_subregion(isa_bus
->address_space_io
, 0xf0, ioportF0_io
);
1600 * Check if an HPET shall be created.
1602 * Without KVM_CAP_PIT_STATE2, we cannot switch off the in-kernel PIT
1603 * when the HPET wants to take over. Thus we have to disable the latter.
1605 if (!no_hpet
&& (!kvm_irqchip_in_kernel() || kvm_has_pit_state2())) {
1606 /* In order to set property, here not using sysbus_try_create_simple */
1607 hpet
= qdev_try_create(NULL
, TYPE_HPET
);
1609 /* For pc-piix-*, hpet's intcap is always IRQ2. For pc-q35-1.7
1610 * and earlier, use IRQ2 for compat. Otherwise, use IRQ16~23,
1613 uint8_t compat
= object_property_get_uint(OBJECT(hpet
),
1616 qdev_prop_set_uint32(hpet
, HPET_INTCAP
, hpet_irqs
);
1618 qdev_init_nofail(hpet
);
1619 sysbus_mmio_map(SYS_BUS_DEVICE(hpet
), 0, HPET_BASE
);
1621 for (i
= 0; i
< GSI_NUM_PINS
; i
++) {
1622 sysbus_connect_irq(SYS_BUS_DEVICE(hpet
), i
, gsi
[i
]);
1625 pit_alt_irq
= qdev_get_gpio_in(hpet
, HPET_LEGACY_PIT_INT
);
1626 rtc_irq
= qdev_get_gpio_in(hpet
, HPET_LEGACY_RTC_INT
);
1629 *rtc_state
= mc146818_rtc_init(isa_bus
, 2000, rtc_irq
);
1631 qemu_register_boot_set(pc_boot_set
, *rtc_state
);
1633 if (!xen_enabled() && has_pit
) {
1634 if (kvm_pit_in_kernel()) {
1635 pit
= kvm_pit_init(isa_bus
, 0x40);
1637 pit
= i8254_pit_init(isa_bus
, 0x40, pit_isa_irq
, pit_alt_irq
);
1640 /* connect PIT to output control line of the HPET */
1641 qdev_connect_gpio_out(hpet
, 0, qdev_get_gpio_in(DEVICE(pit
), 0));
1643 pcspk_init(isa_bus
, pit
);
1646 i8257_dma_init(isa_bus
, 0);
1649 pc_superio_init(isa_bus
, create_fdctrl
, no_vmport
);
1652 void pc_nic_init(PCMachineClass
*pcmc
, ISABus
*isa_bus
, PCIBus
*pci_bus
)
1656 rom_set_order_override(FW_CFG_ORDER_OVERRIDE_NIC
);
1657 for (i
= 0; i
< nb_nics
; i
++) {
1658 NICInfo
*nd
= &nd_table
[i
];
1659 const char *model
= nd
->model
? nd
->model
: pcmc
->default_nic_model
;
1661 if (g_str_equal(model
, "ne2k_isa")) {
1662 pc_init_ne2k_isa(isa_bus
, nd
);
1664 pci_nic_init_nofail(nd
, pci_bus
, model
, NULL
);
1667 rom_reset_order_override();
1670 void ioapic_init_gsi(GSIState
*gsi_state
, const char *parent_name
)
1676 if (kvm_ioapic_in_kernel()) {
1677 dev
= qdev_create(NULL
, "kvm-ioapic");
1679 dev
= qdev_create(NULL
, "ioapic");
1682 object_property_add_child(object_resolve_path(parent_name
, NULL
),
1683 "ioapic", OBJECT(dev
), NULL
);
1685 qdev_init_nofail(dev
);
1686 d
= SYS_BUS_DEVICE(dev
);
1687 sysbus_mmio_map(d
, 0, IO_APIC_DEFAULT_ADDRESS
);
1689 for (i
= 0; i
< IOAPIC_NUM_PINS
; i
++) {
1690 gsi_state
->ioapic_irq
[i
] = qdev_get_gpio_in(dev
, i
);
1694 static void pc_memory_pre_plug(HotplugHandler
*hotplug_dev
, DeviceState
*dev
,
1697 const PCMachineState
*pcms
= PC_MACHINE(hotplug_dev
);
1698 const PCMachineClass
*pcmc
= PC_MACHINE_GET_CLASS(pcms
);
1699 const bool is_nvdimm
= object_dynamic_cast(OBJECT(dev
), TYPE_NVDIMM
);
1700 const uint64_t legacy_align
= TARGET_PAGE_SIZE
;
1703 * When -no-acpi is used with Q35 machine type, no ACPI is built,
1704 * but pcms->acpi_dev is still created. Check !acpi_enabled in
1705 * addition to cover this case.
1707 if (!pcms
->acpi_dev
|| !acpi_enabled
) {
1709 "memory hotplug is not enabled: missing acpi device or acpi disabled");
1713 if (is_nvdimm
&& !pcms
->acpi_nvdimm_state
.is_enabled
) {
1714 error_setg(errp
, "nvdimm is not enabled: missing 'nvdimm' in '-M'");
1718 pc_dimm_pre_plug(PC_DIMM(dev
), MACHINE(hotplug_dev
),
1719 pcmc
->enforce_aligned_dimm
? NULL
: &legacy_align
, errp
);
1722 static void pc_memory_plug(HotplugHandler
*hotplug_dev
,
1723 DeviceState
*dev
, Error
**errp
)
1725 HotplugHandlerClass
*hhc
;
1726 Error
*local_err
= NULL
;
1727 PCMachineState
*pcms
= PC_MACHINE(hotplug_dev
);
1728 bool is_nvdimm
= object_dynamic_cast(OBJECT(dev
), TYPE_NVDIMM
);
1730 pc_dimm_plug(PC_DIMM(dev
), MACHINE(pcms
), &local_err
);
1736 nvdimm_plug(&pcms
->acpi_nvdimm_state
);
1739 hhc
= HOTPLUG_HANDLER_GET_CLASS(pcms
->acpi_dev
);
1740 hhc
->plug(HOTPLUG_HANDLER(pcms
->acpi_dev
), dev
, &error_abort
);
1742 error_propagate(errp
, local_err
);
1745 static void pc_memory_unplug_request(HotplugHandler
*hotplug_dev
,
1746 DeviceState
*dev
, Error
**errp
)
1748 HotplugHandlerClass
*hhc
;
1749 Error
*local_err
= NULL
;
1750 PCMachineState
*pcms
= PC_MACHINE(hotplug_dev
);
1753 * When -no-acpi is used with Q35 machine type, no ACPI is built,
1754 * but pcms->acpi_dev is still created. Check !acpi_enabled in
1755 * addition to cover this case.
1757 if (!pcms
->acpi_dev
|| !acpi_enabled
) {
1758 error_setg(&local_err
,
1759 "memory hotplug is not enabled: missing acpi device or acpi disabled");
1763 if (object_dynamic_cast(OBJECT(dev
), TYPE_NVDIMM
)) {
1764 error_setg(&local_err
,
1765 "nvdimm device hot unplug is not supported yet.");
1769 hhc
= HOTPLUG_HANDLER_GET_CLASS(pcms
->acpi_dev
);
1770 hhc
->unplug_request(HOTPLUG_HANDLER(pcms
->acpi_dev
), dev
, &local_err
);
1773 error_propagate(errp
, local_err
);
1776 static void pc_memory_unplug(HotplugHandler
*hotplug_dev
,
1777 DeviceState
*dev
, Error
**errp
)
1779 PCMachineState
*pcms
= PC_MACHINE(hotplug_dev
);
1780 HotplugHandlerClass
*hhc
;
1781 Error
*local_err
= NULL
;
1783 hhc
= HOTPLUG_HANDLER_GET_CLASS(pcms
->acpi_dev
);
1784 hhc
->unplug(HOTPLUG_HANDLER(pcms
->acpi_dev
), dev
, &local_err
);
1790 pc_dimm_unplug(PC_DIMM(dev
), MACHINE(pcms
));
1791 object_unparent(OBJECT(dev
));
1794 error_propagate(errp
, local_err
);
1797 static int pc_apic_cmp(const void *a
, const void *b
)
1799 CPUArchId
*apic_a
= (CPUArchId
*)a
;
1800 CPUArchId
*apic_b
= (CPUArchId
*)b
;
1802 return apic_a
->arch_id
- apic_b
->arch_id
;
1805 /* returns pointer to CPUArchId descriptor that matches CPU's apic_id
1806 * in ms->possible_cpus->cpus, if ms->possible_cpus->cpus has no
1807 * entry corresponding to CPU's apic_id returns NULL.
1809 static CPUArchId
*pc_find_cpu_slot(MachineState
*ms
, uint32_t id
, int *idx
)
1811 CPUArchId apic_id
, *found_cpu
;
1813 apic_id
.arch_id
= id
;
1814 found_cpu
= bsearch(&apic_id
, ms
->possible_cpus
->cpus
,
1815 ms
->possible_cpus
->len
, sizeof(*ms
->possible_cpus
->cpus
),
1817 if (found_cpu
&& idx
) {
1818 *idx
= found_cpu
- ms
->possible_cpus
->cpus
;
1823 static void pc_cpu_plug(HotplugHandler
*hotplug_dev
,
1824 DeviceState
*dev
, Error
**errp
)
1826 CPUArchId
*found_cpu
;
1827 HotplugHandlerClass
*hhc
;
1828 Error
*local_err
= NULL
;
1829 X86CPU
*cpu
= X86_CPU(dev
);
1830 PCMachineState
*pcms
= PC_MACHINE(hotplug_dev
);
1832 if (pcms
->acpi_dev
) {
1833 hhc
= HOTPLUG_HANDLER_GET_CLASS(pcms
->acpi_dev
);
1834 hhc
->plug(HOTPLUG_HANDLER(pcms
->acpi_dev
), dev
, &local_err
);
1840 /* increment the number of CPUs */
1843 rtc_set_cpus_count(pcms
->rtc
, pcms
->boot_cpus
);
1846 fw_cfg_modify_i16(pcms
->fw_cfg
, FW_CFG_NB_CPUS
, pcms
->boot_cpus
);
1849 found_cpu
= pc_find_cpu_slot(MACHINE(pcms
), cpu
->apic_id
, NULL
);
1850 found_cpu
->cpu
= OBJECT(dev
);
1852 error_propagate(errp
, local_err
);
1854 static void pc_cpu_unplug_request_cb(HotplugHandler
*hotplug_dev
,
1855 DeviceState
*dev
, Error
**errp
)
1858 HotplugHandlerClass
*hhc
;
1859 Error
*local_err
= NULL
;
1860 X86CPU
*cpu
= X86_CPU(dev
);
1861 PCMachineState
*pcms
= PC_MACHINE(hotplug_dev
);
1863 if (!pcms
->acpi_dev
) {
1864 error_setg(&local_err
, "CPU hot unplug not supported without ACPI");
1868 pc_find_cpu_slot(MACHINE(pcms
), cpu
->apic_id
, &idx
);
1871 error_setg(&local_err
, "Boot CPU is unpluggable");
1875 hhc
= HOTPLUG_HANDLER_GET_CLASS(pcms
->acpi_dev
);
1876 hhc
->unplug_request(HOTPLUG_HANDLER(pcms
->acpi_dev
), dev
, &local_err
);
1883 error_propagate(errp
, local_err
);
1887 static void pc_cpu_unplug_cb(HotplugHandler
*hotplug_dev
,
1888 DeviceState
*dev
, Error
**errp
)
1890 CPUArchId
*found_cpu
;
1891 HotplugHandlerClass
*hhc
;
1892 Error
*local_err
= NULL
;
1893 X86CPU
*cpu
= X86_CPU(dev
);
1894 PCMachineState
*pcms
= PC_MACHINE(hotplug_dev
);
1896 hhc
= HOTPLUG_HANDLER_GET_CLASS(pcms
->acpi_dev
);
1897 hhc
->unplug(HOTPLUG_HANDLER(pcms
->acpi_dev
), dev
, &local_err
);
1903 found_cpu
= pc_find_cpu_slot(MACHINE(pcms
), cpu
->apic_id
, NULL
);
1904 found_cpu
->cpu
= NULL
;
1905 object_unparent(OBJECT(dev
));
1907 /* decrement the number of CPUs */
1909 /* Update the number of CPUs in CMOS */
1910 rtc_set_cpus_count(pcms
->rtc
, pcms
->boot_cpus
);
1911 fw_cfg_modify_i16(pcms
->fw_cfg
, FW_CFG_NB_CPUS
, pcms
->boot_cpus
);
1913 error_propagate(errp
, local_err
);
1916 static void pc_cpu_pre_plug(HotplugHandler
*hotplug_dev
,
1917 DeviceState
*dev
, Error
**errp
)
1921 CPUArchId
*cpu_slot
;
1922 X86CPUTopoInfo topo
;
1923 X86CPU
*cpu
= X86_CPU(dev
);
1924 MachineState
*ms
= MACHINE(hotplug_dev
);
1925 PCMachineState
*pcms
= PC_MACHINE(hotplug_dev
);
1927 if(!object_dynamic_cast(OBJECT(cpu
), ms
->cpu_type
)) {
1928 error_setg(errp
, "Invalid CPU type, expected cpu type: '%s'",
1933 /* if APIC ID is not set, set it based on socket/core/thread properties */
1934 if (cpu
->apic_id
== UNASSIGNED_APIC_ID
) {
1935 int max_socket
= (max_cpus
- 1) / smp_threads
/ smp_cores
;
1937 if (cpu
->socket_id
< 0) {
1938 error_setg(errp
, "CPU socket-id is not set");
1940 } else if (cpu
->socket_id
> max_socket
) {
1941 error_setg(errp
, "Invalid CPU socket-id: %u must be in range 0:%u",
1942 cpu
->socket_id
, max_socket
);
1945 if (cpu
->core_id
< 0) {
1946 error_setg(errp
, "CPU core-id is not set");
1948 } else if (cpu
->core_id
> (smp_cores
- 1)) {
1949 error_setg(errp
, "Invalid CPU core-id: %u must be in range 0:%u",
1950 cpu
->core_id
, smp_cores
- 1);
1953 if (cpu
->thread_id
< 0) {
1954 error_setg(errp
, "CPU thread-id is not set");
1956 } else if (cpu
->thread_id
> (smp_threads
- 1)) {
1957 error_setg(errp
, "Invalid CPU thread-id: %u must be in range 0:%u",
1958 cpu
->thread_id
, smp_threads
- 1);
1962 topo
.pkg_id
= cpu
->socket_id
;
1963 topo
.core_id
= cpu
->core_id
;
1964 topo
.smt_id
= cpu
->thread_id
;
1965 cpu
->apic_id
= apicid_from_topo_ids(smp_cores
, smp_threads
, &topo
);
1968 cpu_slot
= pc_find_cpu_slot(MACHINE(pcms
), cpu
->apic_id
, &idx
);
1970 MachineState
*ms
= MACHINE(pcms
);
1972 x86_topo_ids_from_apicid(cpu
->apic_id
, smp_cores
, smp_threads
, &topo
);
1973 error_setg(errp
, "Invalid CPU [socket: %u, core: %u, thread: %u] with"
1974 " APIC ID %" PRIu32
", valid index range 0:%d",
1975 topo
.pkg_id
, topo
.core_id
, topo
.smt_id
, cpu
->apic_id
,
1976 ms
->possible_cpus
->len
- 1);
1980 if (cpu_slot
->cpu
) {
1981 error_setg(errp
, "CPU[%d] with APIC ID %" PRIu32
" exists",
1986 /* if 'address' properties socket-id/core-id/thread-id are not set, set them
1987 * so that machine_query_hotpluggable_cpus would show correct values
1989 /* TODO: move socket_id/core_id/thread_id checks into x86_cpu_realizefn()
1990 * once -smp refactoring is complete and there will be CPU private
1991 * CPUState::nr_cores and CPUState::nr_threads fields instead of globals */
1992 x86_topo_ids_from_apicid(cpu
->apic_id
, smp_cores
, smp_threads
, &topo
);
1993 if (cpu
->socket_id
!= -1 && cpu
->socket_id
!= topo
.pkg_id
) {
1994 error_setg(errp
, "property socket-id: %u doesn't match set apic-id:"
1995 " 0x%x (socket-id: %u)", cpu
->socket_id
, cpu
->apic_id
, topo
.pkg_id
);
1998 cpu
->socket_id
= topo
.pkg_id
;
2000 if (cpu
->core_id
!= -1 && cpu
->core_id
!= topo
.core_id
) {
2001 error_setg(errp
, "property core-id: %u doesn't match set apic-id:"
2002 " 0x%x (core-id: %u)", cpu
->core_id
, cpu
->apic_id
, topo
.core_id
);
2005 cpu
->core_id
= topo
.core_id
;
2007 if (cpu
->thread_id
!= -1 && cpu
->thread_id
!= topo
.smt_id
) {
2008 error_setg(errp
, "property thread-id: %u doesn't match set apic-id:"
2009 " 0x%x (thread-id: %u)", cpu
->thread_id
, cpu
->apic_id
, topo
.smt_id
);
2012 cpu
->thread_id
= topo
.smt_id
;
2014 if (cpu
->hyperv_vpindex
&& !kvm_hv_vpindex_settable()) {
2015 error_setg(errp
, "kernel doesn't allow setting HyperV VP_INDEX");
2020 cs
->cpu_index
= idx
;
2022 numa_cpu_pre_plug(cpu_slot
, dev
, errp
);
2025 static void pc_machine_device_pre_plug_cb(HotplugHandler
*hotplug_dev
,
2026 DeviceState
*dev
, Error
**errp
)
2028 if (object_dynamic_cast(OBJECT(dev
), TYPE_PC_DIMM
)) {
2029 pc_memory_pre_plug(hotplug_dev
, dev
, errp
);
2030 } else if (object_dynamic_cast(OBJECT(dev
), TYPE_CPU
)) {
2031 pc_cpu_pre_plug(hotplug_dev
, dev
, errp
);
2035 static void pc_machine_device_plug_cb(HotplugHandler
*hotplug_dev
,
2036 DeviceState
*dev
, Error
**errp
)
2038 if (object_dynamic_cast(OBJECT(dev
), TYPE_PC_DIMM
)) {
2039 pc_memory_plug(hotplug_dev
, dev
, errp
);
2040 } else if (object_dynamic_cast(OBJECT(dev
), TYPE_CPU
)) {
2041 pc_cpu_plug(hotplug_dev
, dev
, errp
);
2045 static void pc_machine_device_unplug_request_cb(HotplugHandler
*hotplug_dev
,
2046 DeviceState
*dev
, Error
**errp
)
2048 if (object_dynamic_cast(OBJECT(dev
), TYPE_PC_DIMM
)) {
2049 pc_memory_unplug_request(hotplug_dev
, dev
, errp
);
2050 } else if (object_dynamic_cast(OBJECT(dev
), TYPE_CPU
)) {
2051 pc_cpu_unplug_request_cb(hotplug_dev
, dev
, errp
);
2053 error_setg(errp
, "acpi: device unplug request for not supported device"
2054 " type: %s", object_get_typename(OBJECT(dev
)));
2058 static void pc_machine_device_unplug_cb(HotplugHandler
*hotplug_dev
,
2059 DeviceState
*dev
, Error
**errp
)
2061 if (object_dynamic_cast(OBJECT(dev
), TYPE_PC_DIMM
)) {
2062 pc_memory_unplug(hotplug_dev
, dev
, errp
);
2063 } else if (object_dynamic_cast(OBJECT(dev
), TYPE_CPU
)) {
2064 pc_cpu_unplug_cb(hotplug_dev
, dev
, errp
);
2066 error_setg(errp
, "acpi: device unplug for not supported device"
2067 " type: %s", object_get_typename(OBJECT(dev
)));
2071 static HotplugHandler
*pc_get_hotpug_handler(MachineState
*machine
,
2074 if (object_dynamic_cast(OBJECT(dev
), TYPE_PC_DIMM
) ||
2075 object_dynamic_cast(OBJECT(dev
), TYPE_CPU
)) {
2076 return HOTPLUG_HANDLER(machine
);
2083 pc_machine_get_device_memory_region_size(Object
*obj
, Visitor
*v
,
2084 const char *name
, void *opaque
,
2087 MachineState
*ms
= MACHINE(obj
);
2088 int64_t value
= memory_region_size(&ms
->device_memory
->mr
);
2090 visit_type_int(v
, name
, &value
, errp
);
2093 static void pc_machine_get_max_ram_below_4g(Object
*obj
, Visitor
*v
,
2094 const char *name
, void *opaque
,
2097 PCMachineState
*pcms
= PC_MACHINE(obj
);
2098 uint64_t value
= pcms
->max_ram_below_4g
;
2100 visit_type_size(v
, name
, &value
, errp
);
2103 static void pc_machine_set_max_ram_below_4g(Object
*obj
, Visitor
*v
,
2104 const char *name
, void *opaque
,
2107 PCMachineState
*pcms
= PC_MACHINE(obj
);
2108 Error
*error
= NULL
;
2111 visit_type_size(v
, name
, &value
, &error
);
2113 error_propagate(errp
, error
);
2116 if (value
> 4 * GiB
) {
2118 "Machine option 'max-ram-below-4g=%"PRIu64
2119 "' expects size less than or equal to 4G", value
);
2120 error_propagate(errp
, error
);
2124 if (value
< 1 * MiB
) {
2125 warn_report("Only %" PRIu64
" bytes of RAM below the 4GiB boundary,"
2126 "BIOS may not work with less than 1MiB", value
);
2129 pcms
->max_ram_below_4g
= value
;
2132 static void pc_machine_get_vmport(Object
*obj
, Visitor
*v
, const char *name
,
2133 void *opaque
, Error
**errp
)
2135 PCMachineState
*pcms
= PC_MACHINE(obj
);
2136 OnOffAuto vmport
= pcms
->vmport
;
2138 visit_type_OnOffAuto(v
, name
, &vmport
, errp
);
2141 static void pc_machine_set_vmport(Object
*obj
, Visitor
*v
, const char *name
,
2142 void *opaque
, Error
**errp
)
2144 PCMachineState
*pcms
= PC_MACHINE(obj
);
2146 visit_type_OnOffAuto(v
, name
, &pcms
->vmport
, errp
);
2149 bool pc_machine_is_smm_enabled(PCMachineState
*pcms
)
2151 bool smm_available
= false;
2153 if (pcms
->smm
== ON_OFF_AUTO_OFF
) {
2157 if (tcg_enabled() || qtest_enabled()) {
2158 smm_available
= true;
2159 } else if (kvm_enabled()) {
2160 smm_available
= kvm_has_smm();
2163 if (smm_available
) {
2167 if (pcms
->smm
== ON_OFF_AUTO_ON
) {
2168 error_report("System Management Mode not supported by this hypervisor.");
2174 static void pc_machine_get_smm(Object
*obj
, Visitor
*v
, const char *name
,
2175 void *opaque
, Error
**errp
)
2177 PCMachineState
*pcms
= PC_MACHINE(obj
);
2178 OnOffAuto smm
= pcms
->smm
;
2180 visit_type_OnOffAuto(v
, name
, &smm
, errp
);
2183 static void pc_machine_set_smm(Object
*obj
, Visitor
*v
, const char *name
,
2184 void *opaque
, Error
**errp
)
2186 PCMachineState
*pcms
= PC_MACHINE(obj
);
2188 visit_type_OnOffAuto(v
, name
, &pcms
->smm
, errp
);
2191 static bool pc_machine_get_nvdimm(Object
*obj
, Error
**errp
)
2193 PCMachineState
*pcms
= PC_MACHINE(obj
);
2195 return pcms
->acpi_nvdimm_state
.is_enabled
;
2198 static void pc_machine_set_nvdimm(Object
*obj
, bool value
, Error
**errp
)
2200 PCMachineState
*pcms
= PC_MACHINE(obj
);
2202 pcms
->acpi_nvdimm_state
.is_enabled
= value
;
2205 static char *pc_machine_get_nvdimm_persistence(Object
*obj
, Error
**errp
)
2207 PCMachineState
*pcms
= PC_MACHINE(obj
);
2209 return g_strdup(pcms
->acpi_nvdimm_state
.persistence_string
);
2212 static void pc_machine_set_nvdimm_persistence(Object
*obj
, const char *value
,
2215 PCMachineState
*pcms
= PC_MACHINE(obj
);
2216 AcpiNVDIMMState
*nvdimm_state
= &pcms
->acpi_nvdimm_state
;
2218 if (strcmp(value
, "cpu") == 0)
2219 nvdimm_state
->persistence
= 3;
2220 else if (strcmp(value
, "mem-ctrl") == 0)
2221 nvdimm_state
->persistence
= 2;
2223 error_setg(errp
, "-machine nvdimm-persistence=%s: unsupported option",
2228 g_free(nvdimm_state
->persistence_string
);
2229 nvdimm_state
->persistence_string
= g_strdup(value
);
2232 static bool pc_machine_get_smbus(Object
*obj
, Error
**errp
)
2234 PCMachineState
*pcms
= PC_MACHINE(obj
);
2236 return pcms
->smbus_enabled
;
2239 static void pc_machine_set_smbus(Object
*obj
, bool value
, Error
**errp
)
2241 PCMachineState
*pcms
= PC_MACHINE(obj
);
2243 pcms
->smbus_enabled
= value
;
2246 static bool pc_machine_get_sata(Object
*obj
, Error
**errp
)
2248 PCMachineState
*pcms
= PC_MACHINE(obj
);
2250 return pcms
->sata_enabled
;
2253 static void pc_machine_set_sata(Object
*obj
, bool value
, Error
**errp
)
2255 PCMachineState
*pcms
= PC_MACHINE(obj
);
2257 pcms
->sata_enabled
= value
;
2260 static bool pc_machine_get_pit(Object
*obj
, Error
**errp
)
2262 PCMachineState
*pcms
= PC_MACHINE(obj
);
2264 return pcms
->pit_enabled
;
2267 static void pc_machine_set_pit(Object
*obj
, bool value
, Error
**errp
)
2269 PCMachineState
*pcms
= PC_MACHINE(obj
);
2271 pcms
->pit_enabled
= value
;
2274 static void pc_machine_initfn(Object
*obj
)
2276 PCMachineState
*pcms
= PC_MACHINE(obj
);
2278 pcms
->max_ram_below_4g
= 0; /* use default */
2279 pcms
->smm
= ON_OFF_AUTO_AUTO
;
2280 pcms
->vmport
= ON_OFF_AUTO_AUTO
;
2281 /* nvdimm is disabled on default. */
2282 pcms
->acpi_nvdimm_state
.is_enabled
= false;
2283 /* acpi build is enabled by default if machine supports it */
2284 pcms
->acpi_build_enabled
= PC_MACHINE_GET_CLASS(pcms
)->has_acpi_build
;
2285 pcms
->smbus_enabled
= true;
2286 pcms
->sata_enabled
= true;
2287 pcms
->pit_enabled
= true;
2290 static void pc_machine_reset(void)
2295 qemu_devices_reset();
2297 /* Reset APIC after devices have been reset to cancel
2298 * any changes that qemu_devices_reset() might have done.
2303 if (cpu
->apic_state
) {
2304 device_reset(cpu
->apic_state
);
2309 static CpuInstanceProperties
2310 pc_cpu_index_to_props(MachineState
*ms
, unsigned cpu_index
)
2312 MachineClass
*mc
= MACHINE_GET_CLASS(ms
);
2313 const CPUArchIdList
*possible_cpus
= mc
->possible_cpu_arch_ids(ms
);
2315 assert(cpu_index
< possible_cpus
->len
);
2316 return possible_cpus
->cpus
[cpu_index
].props
;
2319 static int64_t pc_get_default_cpu_node_id(const MachineState
*ms
, int idx
)
2321 X86CPUTopoInfo topo
;
2323 assert(idx
< ms
->possible_cpus
->len
);
2324 x86_topo_ids_from_apicid(ms
->possible_cpus
->cpus
[idx
].arch_id
,
2325 smp_cores
, smp_threads
, &topo
);
2326 return topo
.pkg_id
% nb_numa_nodes
;
2329 static const CPUArchIdList
*pc_possible_cpu_arch_ids(MachineState
*ms
)
2333 if (ms
->possible_cpus
) {
2335 * make sure that max_cpus hasn't changed since the first use, i.e.
2336 * -smp hasn't been parsed after it
2338 assert(ms
->possible_cpus
->len
== max_cpus
);
2339 return ms
->possible_cpus
;
2342 ms
->possible_cpus
= g_malloc0(sizeof(CPUArchIdList
) +
2343 sizeof(CPUArchId
) * max_cpus
);
2344 ms
->possible_cpus
->len
= max_cpus
;
2345 for (i
= 0; i
< ms
->possible_cpus
->len
; i
++) {
2346 X86CPUTopoInfo topo
;
2348 ms
->possible_cpus
->cpus
[i
].type
= ms
->cpu_type
;
2349 ms
->possible_cpus
->cpus
[i
].vcpus_count
= 1;
2350 ms
->possible_cpus
->cpus
[i
].arch_id
= x86_cpu_apic_id_from_index(i
);
2351 x86_topo_ids_from_apicid(ms
->possible_cpus
->cpus
[i
].arch_id
,
2352 smp_cores
, smp_threads
, &topo
);
2353 ms
->possible_cpus
->cpus
[i
].props
.has_socket_id
= true;
2354 ms
->possible_cpus
->cpus
[i
].props
.socket_id
= topo
.pkg_id
;
2355 ms
->possible_cpus
->cpus
[i
].props
.has_core_id
= true;
2356 ms
->possible_cpus
->cpus
[i
].props
.core_id
= topo
.core_id
;
2357 ms
->possible_cpus
->cpus
[i
].props
.has_thread_id
= true;
2358 ms
->possible_cpus
->cpus
[i
].props
.thread_id
= topo
.smt_id
;
2360 return ms
->possible_cpus
;
2363 static void x86_nmi(NMIState
*n
, int cpu_index
, Error
**errp
)
2365 /* cpu index isn't used */
2369 X86CPU
*cpu
= X86_CPU(cs
);
2371 if (!cpu
->apic_state
) {
2372 cpu_interrupt(cs
, CPU_INTERRUPT_NMI
);
2374 apic_deliver_nmi(cpu
->apic_state
);
2379 static void pc_machine_class_init(ObjectClass
*oc
, void *data
)
2381 MachineClass
*mc
= MACHINE_CLASS(oc
);
2382 PCMachineClass
*pcmc
= PC_MACHINE_CLASS(oc
);
2383 HotplugHandlerClass
*hc
= HOTPLUG_HANDLER_CLASS(oc
);
2384 NMIClass
*nc
= NMI_CLASS(oc
);
2386 pcmc
->pci_enabled
= true;
2387 pcmc
->has_acpi_build
= true;
2388 pcmc
->rsdp_in_ram
= true;
2389 pcmc
->smbios_defaults
= true;
2390 pcmc
->smbios_uuid_encoded
= true;
2391 pcmc
->gigabyte_align
= true;
2392 pcmc
->has_reserved_memory
= true;
2393 pcmc
->kvmclock_enabled
= true;
2394 pcmc
->enforce_aligned_dimm
= true;
2395 /* BIOS ACPI tables: 128K. Other BIOS datastructures: less than 4K reported
2396 * to be used at the moment, 32K should be enough for a while. */
2397 pcmc
->acpi_data_size
= 0x20000 + 0x8000;
2398 pcmc
->save_tsc_khz
= true;
2399 pcmc
->linuxboot_dma_enabled
= true;
2400 assert(!mc
->get_hotplug_handler
);
2401 mc
->get_hotplug_handler
= pc_get_hotpug_handler
;
2402 mc
->cpu_index_to_instance_props
= pc_cpu_index_to_props
;
2403 mc
->get_default_cpu_node_id
= pc_get_default_cpu_node_id
;
2404 mc
->possible_cpu_arch_ids
= pc_possible_cpu_arch_ids
;
2405 mc
->auto_enable_numa_with_memhp
= true;
2406 mc
->has_hotpluggable_cpus
= true;
2407 mc
->default_boot_order
= "cad";
2408 mc
->hot_add_cpu
= pc_hot_add_cpu
;
2409 mc
->block_default_type
= IF_IDE
;
2411 mc
->reset
= pc_machine_reset
;
2412 hc
->pre_plug
= pc_machine_device_pre_plug_cb
;
2413 hc
->plug
= pc_machine_device_plug_cb
;
2414 hc
->unplug_request
= pc_machine_device_unplug_request_cb
;
2415 hc
->unplug
= pc_machine_device_unplug_cb
;
2416 nc
->nmi_monitor_handler
= x86_nmi
;
2417 mc
->default_cpu_type
= TARGET_DEFAULT_CPU_TYPE
;
2419 object_class_property_add(oc
, PC_MACHINE_DEVMEM_REGION_SIZE
, "int",
2420 pc_machine_get_device_memory_region_size
, NULL
,
2421 NULL
, NULL
, &error_abort
);
2423 object_class_property_add(oc
, PC_MACHINE_MAX_RAM_BELOW_4G
, "size",
2424 pc_machine_get_max_ram_below_4g
, pc_machine_set_max_ram_below_4g
,
2425 NULL
, NULL
, &error_abort
);
2427 object_class_property_set_description(oc
, PC_MACHINE_MAX_RAM_BELOW_4G
,
2428 "Maximum ram below the 4G boundary (32bit boundary)", &error_abort
);
2430 object_class_property_add(oc
, PC_MACHINE_SMM
, "OnOffAuto",
2431 pc_machine_get_smm
, pc_machine_set_smm
,
2432 NULL
, NULL
, &error_abort
);
2433 object_class_property_set_description(oc
, PC_MACHINE_SMM
,
2434 "Enable SMM (pc & q35)", &error_abort
);
2436 object_class_property_add(oc
, PC_MACHINE_VMPORT
, "OnOffAuto",
2437 pc_machine_get_vmport
, pc_machine_set_vmport
,
2438 NULL
, NULL
, &error_abort
);
2439 object_class_property_set_description(oc
, PC_MACHINE_VMPORT
,
2440 "Enable vmport (pc & q35)", &error_abort
);
2442 object_class_property_add_bool(oc
, PC_MACHINE_NVDIMM
,
2443 pc_machine_get_nvdimm
, pc_machine_set_nvdimm
, &error_abort
);
2445 object_class_property_add_str(oc
, PC_MACHINE_NVDIMM_PERSIST
,
2446 pc_machine_get_nvdimm_persistence
,
2447 pc_machine_set_nvdimm_persistence
, &error_abort
);
2449 object_class_property_add_bool(oc
, PC_MACHINE_SMBUS
,
2450 pc_machine_get_smbus
, pc_machine_set_smbus
, &error_abort
);
2452 object_class_property_add_bool(oc
, PC_MACHINE_SATA
,
2453 pc_machine_get_sata
, pc_machine_set_sata
, &error_abort
);
2455 object_class_property_add_bool(oc
, PC_MACHINE_PIT
,
2456 pc_machine_get_pit
, pc_machine_set_pit
, &error_abort
);
2459 static const TypeInfo pc_machine_info
= {
2460 .name
= TYPE_PC_MACHINE
,
2461 .parent
= TYPE_MACHINE
,
2463 .instance_size
= sizeof(PCMachineState
),
2464 .instance_init
= pc_machine_initfn
,
2465 .class_size
= sizeof(PCMachineClass
),
2466 .class_init
= pc_machine_class_init
,
2467 .interfaces
= (InterfaceInfo
[]) {
2468 { TYPE_HOTPLUG_HANDLER
},
2474 static void pc_machine_register_types(void)
2476 type_register_static(&pc_machine_info
);
2479 type_init(pc_machine_register_types
)