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
24 #include "qemu/osdep.h"
26 #include "hw/i386/pc.h"
27 #include "hw/char/serial.h"
28 #include "hw/i386/apic.h"
29 #include "hw/i386/topology.h"
30 #include "sysemu/cpus.h"
31 #include "hw/block/fdc.h"
33 #include "hw/pci/pci.h"
34 #include "hw/pci/pci_bus.h"
35 #include "hw/nvram/fw_cfg.h"
36 #include "hw/timer/hpet.h"
37 #include "hw/smbios/smbios.h"
38 #include "hw/loader.h"
40 #include "multiboot.h"
41 #include "hw/timer/mc146818rtc.h"
42 #include "hw/timer/i8254.h"
43 #include "hw/audio/pcspk.h"
44 #include "hw/pci/msi.h"
45 #include "hw/sysbus.h"
46 #include "sysemu/sysemu.h"
47 #include "sysemu/numa.h"
48 #include "sysemu/kvm.h"
49 #include "sysemu/qtest.h"
51 #include "hw/xen/xen.h"
52 #include "sysemu/block-backend.h"
53 #include "hw/block/block.h"
54 #include "ui/qemu-spice.h"
55 #include "exec/memory.h"
56 #include "exec/address-spaces.h"
57 #include "sysemu/arch_init.h"
58 #include "qemu/bitmap.h"
59 #include "qemu/config-file.h"
60 #include "qemu/error-report.h"
61 #include "hw/acpi/acpi.h"
62 #include "hw/acpi/cpu_hotplug.h"
63 #include "hw/boards.h"
64 #include "hw/pci/pci_host.h"
65 #include "acpi-build.h"
66 #include "hw/mem/pc-dimm.h"
67 #include "qapi/visitor.h"
68 #include "qapi-visit.h"
71 #include "hw/i386/intel_iommu.h"
72 #include "hw/net/ne2000-isa.h"
74 /* debug PC/ISA interrupts */
78 #define DPRINTF(fmt, ...) \
79 do { printf("CPUIRQ: " fmt , ## __VA_ARGS__); } while (0)
81 #define DPRINTF(fmt, ...)
84 #define FW_CFG_ACPI_TABLES (FW_CFG_ARCH_LOCAL + 0)
85 #define FW_CFG_SMBIOS_ENTRIES (FW_CFG_ARCH_LOCAL + 1)
86 #define FW_CFG_IRQ0_OVERRIDE (FW_CFG_ARCH_LOCAL + 2)
87 #define FW_CFG_E820_TABLE (FW_CFG_ARCH_LOCAL + 3)
88 #define FW_CFG_HPET (FW_CFG_ARCH_LOCAL + 4)
90 #define E820_NR_ENTRIES 16
96 } QEMU_PACKED
__attribute((__aligned__(4)));
100 struct e820_entry entry
[E820_NR_ENTRIES
];
101 } QEMU_PACKED
__attribute((__aligned__(4)));
103 static struct e820_table e820_reserve
;
104 static struct e820_entry
*e820_table
;
105 static unsigned e820_entries
;
106 struct hpet_fw_config hpet_cfg
= {.count
= UINT8_MAX
};
108 void gsi_handler(void *opaque
, int n
, int level
)
110 GSIState
*s
= opaque
;
112 DPRINTF("pc: %s GSI %d\n", level
? "raising" : "lowering", n
);
113 if (n
< ISA_NUM_IRQS
) {
114 qemu_set_irq(s
->i8259_irq
[n
], level
);
116 qemu_set_irq(s
->ioapic_irq
[n
], level
);
119 static void ioport80_write(void *opaque
, hwaddr addr
, uint64_t data
,
124 static uint64_t ioport80_read(void *opaque
, hwaddr addr
, unsigned size
)
126 return 0xffffffffffffffffULL
;
129 /* MSDOS compatibility mode FPU exception support */
130 static qemu_irq ferr_irq
;
132 void pc_register_ferr_irq(qemu_irq irq
)
137 /* XXX: add IGNNE support */
138 void cpu_set_ferr(CPUX86State
*s
)
140 qemu_irq_raise(ferr_irq
);
143 static void ioportF0_write(void *opaque
, hwaddr addr
, uint64_t data
,
146 qemu_irq_lower(ferr_irq
);
149 static uint64_t ioportF0_read(void *opaque
, hwaddr addr
, unsigned size
)
151 return 0xffffffffffffffffULL
;
155 uint64_t cpu_get_tsc(CPUX86State
*env
)
157 return cpu_get_ticks();
161 int cpu_get_pic_interrupt(CPUX86State
*env
)
163 X86CPU
*cpu
= x86_env_get_cpu(env
);
166 if (!kvm_irqchip_in_kernel()) {
167 intno
= apic_get_interrupt(cpu
->apic_state
);
171 /* read the irq from the PIC */
172 if (!apic_accept_pic_intr(cpu
->apic_state
)) {
177 intno
= pic_read_irq(isa_pic
);
181 static void pic_irq_request(void *opaque
, int irq
, int level
)
183 CPUState
*cs
= first_cpu
;
184 X86CPU
*cpu
= X86_CPU(cs
);
186 DPRINTF("pic_irqs: %s irq %d\n", level
? "raise" : "lower", irq
);
187 if (cpu
->apic_state
&& !kvm_irqchip_in_kernel()) {
190 if (apic_accept_pic_intr(cpu
->apic_state
)) {
191 apic_deliver_pic_intr(cpu
->apic_state
, level
);
196 cpu_interrupt(cs
, CPU_INTERRUPT_HARD
);
198 cpu_reset_interrupt(cs
, CPU_INTERRUPT_HARD
);
203 /* PC cmos mappings */
205 #define REG_EQUIPMENT_BYTE 0x14
207 int cmos_get_fd_drive_type(FloppyDriveType fd0
)
212 case FLOPPY_DRIVE_TYPE_144
:
213 /* 1.44 Mb 3"5 drive */
216 case FLOPPY_DRIVE_TYPE_288
:
217 /* 2.88 Mb 3"5 drive */
220 case FLOPPY_DRIVE_TYPE_120
:
221 /* 1.2 Mb 5"5 drive */
224 case FLOPPY_DRIVE_TYPE_NONE
:
232 static void cmos_init_hd(ISADevice
*s
, int type_ofs
, int info_ofs
,
233 int16_t cylinders
, int8_t heads
, int8_t sectors
)
235 rtc_set_memory(s
, type_ofs
, 47);
236 rtc_set_memory(s
, info_ofs
, cylinders
);
237 rtc_set_memory(s
, info_ofs
+ 1, cylinders
>> 8);
238 rtc_set_memory(s
, info_ofs
+ 2, heads
);
239 rtc_set_memory(s
, info_ofs
+ 3, 0xff);
240 rtc_set_memory(s
, info_ofs
+ 4, 0xff);
241 rtc_set_memory(s
, info_ofs
+ 5, 0xc0 | ((heads
> 8) << 3));
242 rtc_set_memory(s
, info_ofs
+ 6, cylinders
);
243 rtc_set_memory(s
, info_ofs
+ 7, cylinders
>> 8);
244 rtc_set_memory(s
, info_ofs
+ 8, sectors
);
247 /* convert boot_device letter to something recognizable by the bios */
248 static int boot_device2nibble(char boot_device
)
250 switch(boot_device
) {
253 return 0x01; /* floppy boot */
255 return 0x02; /* hard drive boot */
257 return 0x03; /* CD-ROM boot */
259 return 0x04; /* Network boot */
264 static void set_boot_dev(ISADevice
*s
, const char *boot_device
, Error
**errp
)
266 #define PC_MAX_BOOT_DEVICES 3
267 int nbds
, bds
[3] = { 0, };
270 nbds
= strlen(boot_device
);
271 if (nbds
> PC_MAX_BOOT_DEVICES
) {
272 error_setg(errp
, "Too many boot devices for PC");
275 for (i
= 0; i
< nbds
; i
++) {
276 bds
[i
] = boot_device2nibble(boot_device
[i
]);
278 error_setg(errp
, "Invalid boot device for PC: '%c'",
283 rtc_set_memory(s
, 0x3d, (bds
[1] << 4) | bds
[0]);
284 rtc_set_memory(s
, 0x38, (bds
[2] << 4) | (fd_bootchk
? 0x0 : 0x1));
287 static void pc_boot_set(void *opaque
, const char *boot_device
, Error
**errp
)
289 set_boot_dev(opaque
, boot_device
, errp
);
292 static void pc_cmos_init_floppy(ISADevice
*rtc_state
, ISADevice
*floppy
)
295 FloppyDriveType fd_type
[2] = { FLOPPY_DRIVE_TYPE_NONE
,
296 FLOPPY_DRIVE_TYPE_NONE
};
300 for (i
= 0; i
< 2; i
++) {
301 fd_type
[i
] = isa_fdc_get_drive_type(floppy
, i
);
304 val
= (cmos_get_fd_drive_type(fd_type
[0]) << 4) |
305 cmos_get_fd_drive_type(fd_type
[1]);
306 rtc_set_memory(rtc_state
, 0x10, val
);
308 val
= rtc_get_memory(rtc_state
, REG_EQUIPMENT_BYTE
);
310 if (fd_type
[0] != FLOPPY_DRIVE_TYPE_NONE
) {
313 if (fd_type
[1] != FLOPPY_DRIVE_TYPE_NONE
) {
320 val
|= 0x01; /* 1 drive, ready for boot */
323 val
|= 0x41; /* 2 drives, ready for boot */
326 rtc_set_memory(rtc_state
, REG_EQUIPMENT_BYTE
, val
);
329 typedef struct pc_cmos_init_late_arg
{
330 ISADevice
*rtc_state
;
332 } pc_cmos_init_late_arg
;
334 typedef struct check_fdc_state
{
339 static int check_fdc(Object
*obj
, void *opaque
)
341 CheckFdcState
*state
= opaque
;
344 Error
*local_err
= NULL
;
346 fdc
= object_dynamic_cast(obj
, TYPE_ISA_FDC
);
351 iobase
= object_property_get_uint(obj
, "iobase", &local_err
);
352 if (local_err
|| iobase
!= 0x3f0) {
353 error_free(local_err
);
358 state
->multiple
= true;
360 state
->floppy
= ISA_DEVICE(obj
);
365 static const char * const fdc_container_path
[] = {
366 "/unattached", "/peripheral", "/peripheral-anon"
370 * Locate the FDC at IO address 0x3f0, in order to configure the CMOS registers
373 ISADevice
*pc_find_fdc0(void)
377 CheckFdcState state
= { 0 };
379 for (i
= 0; i
< ARRAY_SIZE(fdc_container_path
); i
++) {
380 container
= container_get(qdev_get_machine(), fdc_container_path
[i
]);
381 object_child_foreach(container
, check_fdc
, &state
);
384 if (state
.multiple
) {
385 warn_report("multiple floppy disk controllers with "
386 "iobase=0x3f0 have been found");
387 error_printf("the one being picked for CMOS setup might not reflect "
394 static void pc_cmos_init_late(void *opaque
)
396 pc_cmos_init_late_arg
*arg
= opaque
;
397 ISADevice
*s
= arg
->rtc_state
;
399 int8_t heads
, sectors
;
404 if (arg
->idebus
[0] && ide_get_geometry(arg
->idebus
[0], 0,
405 &cylinders
, &heads
, §ors
) >= 0) {
406 cmos_init_hd(s
, 0x19, 0x1b, cylinders
, heads
, sectors
);
409 if (arg
->idebus
[0] && ide_get_geometry(arg
->idebus
[0], 1,
410 &cylinders
, &heads
, §ors
) >= 0) {
411 cmos_init_hd(s
, 0x1a, 0x24, cylinders
, heads
, sectors
);
414 rtc_set_memory(s
, 0x12, val
);
417 for (i
= 0; i
< 4; i
++) {
418 /* NOTE: ide_get_geometry() returns the physical
419 geometry. It is always such that: 1 <= sects <= 63, 1
420 <= heads <= 16, 1 <= cylinders <= 16383. The BIOS
421 geometry can be different if a translation is done. */
422 if (arg
->idebus
[i
/ 2] &&
423 ide_get_geometry(arg
->idebus
[i
/ 2], i
% 2,
424 &cylinders
, &heads
, §ors
) >= 0) {
425 trans
= ide_get_bios_chs_trans(arg
->idebus
[i
/ 2], i
% 2) - 1;
426 assert((trans
& ~3) == 0);
427 val
|= trans
<< (i
* 2);
430 rtc_set_memory(s
, 0x39, val
);
432 pc_cmos_init_floppy(s
, pc_find_fdc0());
434 qemu_unregister_reset(pc_cmos_init_late
, opaque
);
437 void pc_cmos_init(PCMachineState
*pcms
,
438 BusState
*idebus0
, BusState
*idebus1
,
442 static pc_cmos_init_late_arg arg
;
444 /* various important CMOS locations needed by PC/Bochs bios */
447 /* base memory (first MiB) */
448 val
= MIN(pcms
->below_4g_mem_size
/ 1024, 640);
449 rtc_set_memory(s
, 0x15, val
);
450 rtc_set_memory(s
, 0x16, val
>> 8);
451 /* extended memory (next 64MiB) */
452 if (pcms
->below_4g_mem_size
> 1024 * 1024) {
453 val
= (pcms
->below_4g_mem_size
- 1024 * 1024) / 1024;
459 rtc_set_memory(s
, 0x17, val
);
460 rtc_set_memory(s
, 0x18, val
>> 8);
461 rtc_set_memory(s
, 0x30, val
);
462 rtc_set_memory(s
, 0x31, val
>> 8);
463 /* memory between 16MiB and 4GiB */
464 if (pcms
->below_4g_mem_size
> 16 * 1024 * 1024) {
465 val
= (pcms
->below_4g_mem_size
- 16 * 1024 * 1024) / 65536;
471 rtc_set_memory(s
, 0x34, val
);
472 rtc_set_memory(s
, 0x35, val
>> 8);
473 /* memory above 4GiB */
474 val
= pcms
->above_4g_mem_size
/ 65536;
475 rtc_set_memory(s
, 0x5b, val
);
476 rtc_set_memory(s
, 0x5c, val
>> 8);
477 rtc_set_memory(s
, 0x5d, val
>> 16);
479 object_property_add_link(OBJECT(pcms
), "rtc_state",
481 (Object
**)&pcms
->rtc
,
482 object_property_allow_set_link
,
483 OBJ_PROP_LINK_UNREF_ON_RELEASE
, &error_abort
);
484 object_property_set_link(OBJECT(pcms
), OBJECT(s
),
485 "rtc_state", &error_abort
);
487 set_boot_dev(s
, MACHINE(pcms
)->boot_order
, &error_fatal
);
490 val
|= 0x02; /* FPU is there */
491 val
|= 0x04; /* PS/2 mouse installed */
492 rtc_set_memory(s
, REG_EQUIPMENT_BYTE
, val
);
494 /* hard drives and FDC */
496 arg
.idebus
[0] = idebus0
;
497 arg
.idebus
[1] = idebus1
;
498 qemu_register_reset(pc_cmos_init_late
, &arg
);
501 #define TYPE_PORT92 "port92"
502 #define PORT92(obj) OBJECT_CHECK(Port92State, (obj), TYPE_PORT92)
504 /* port 92 stuff: could be split off */
505 typedef struct Port92State
{
506 ISADevice parent_obj
;
513 static void port92_write(void *opaque
, hwaddr addr
, uint64_t val
,
516 Port92State
*s
= opaque
;
517 int oldval
= s
->outport
;
519 DPRINTF("port92: write 0x%02" PRIx64
"\n", val
);
521 qemu_set_irq(s
->a20_out
, (val
>> 1) & 1);
522 if ((val
& 1) && !(oldval
& 1)) {
523 qemu_system_reset_request(SHUTDOWN_CAUSE_GUEST_RESET
);
527 static uint64_t port92_read(void *opaque
, hwaddr addr
,
530 Port92State
*s
= opaque
;
534 DPRINTF("port92: read 0x%02x\n", ret
);
538 static void port92_init(ISADevice
*dev
, qemu_irq a20_out
)
540 qdev_connect_gpio_out_named(DEVICE(dev
), PORT92_A20_LINE
, 0, a20_out
);
543 static const VMStateDescription vmstate_port92_isa
= {
546 .minimum_version_id
= 1,
547 .fields
= (VMStateField
[]) {
548 VMSTATE_UINT8(outport
, Port92State
),
549 VMSTATE_END_OF_LIST()
553 static void port92_reset(DeviceState
*d
)
555 Port92State
*s
= PORT92(d
);
560 static const MemoryRegionOps port92_ops
= {
562 .write
= port92_write
,
564 .min_access_size
= 1,
565 .max_access_size
= 1,
567 .endianness
= DEVICE_LITTLE_ENDIAN
,
570 static void port92_initfn(Object
*obj
)
572 Port92State
*s
= PORT92(obj
);
574 memory_region_init_io(&s
->io
, OBJECT(s
), &port92_ops
, s
, "port92", 1);
578 qdev_init_gpio_out_named(DEVICE(obj
), &s
->a20_out
, PORT92_A20_LINE
, 1);
581 static void port92_realizefn(DeviceState
*dev
, Error
**errp
)
583 ISADevice
*isadev
= ISA_DEVICE(dev
);
584 Port92State
*s
= PORT92(dev
);
586 isa_register_ioport(isadev
, &s
->io
, 0x92);
589 static void port92_class_initfn(ObjectClass
*klass
, void *data
)
591 DeviceClass
*dc
= DEVICE_CLASS(klass
);
593 dc
->realize
= port92_realizefn
;
594 dc
->reset
= port92_reset
;
595 dc
->vmsd
= &vmstate_port92_isa
;
597 * Reason: unlike ordinary ISA devices, this one needs additional
598 * wiring: its A20 output line needs to be wired up by
601 dc
->user_creatable
= false;
604 static const TypeInfo port92_info
= {
606 .parent
= TYPE_ISA_DEVICE
,
607 .instance_size
= sizeof(Port92State
),
608 .instance_init
= port92_initfn
,
609 .class_init
= port92_class_initfn
,
612 static void port92_register_types(void)
614 type_register_static(&port92_info
);
617 type_init(port92_register_types
)
619 static void handle_a20_line_change(void *opaque
, int irq
, int level
)
621 X86CPU
*cpu
= opaque
;
623 /* XXX: send to all CPUs ? */
624 /* XXX: add logic to handle multiple A20 line sources */
625 x86_cpu_set_a20(cpu
, level
);
628 int e820_add_entry(uint64_t address
, uint64_t length
, uint32_t type
)
630 int index
= le32_to_cpu(e820_reserve
.count
);
631 struct e820_entry
*entry
;
633 if (type
!= E820_RAM
) {
634 /* old FW_CFG_E820_TABLE entry -- reservations only */
635 if (index
>= E820_NR_ENTRIES
) {
638 entry
= &e820_reserve
.entry
[index
++];
640 entry
->address
= cpu_to_le64(address
);
641 entry
->length
= cpu_to_le64(length
);
642 entry
->type
= cpu_to_le32(type
);
644 e820_reserve
.count
= cpu_to_le32(index
);
647 /* new "etc/e820" file -- include ram too */
648 e820_table
= g_renew(struct e820_entry
, e820_table
, e820_entries
+ 1);
649 e820_table
[e820_entries
].address
= cpu_to_le64(address
);
650 e820_table
[e820_entries
].length
= cpu_to_le64(length
);
651 e820_table
[e820_entries
].type
= cpu_to_le32(type
);
657 int e820_get_num_entries(void)
662 bool e820_get_entry(int idx
, uint32_t type
, uint64_t *address
, uint64_t *length
)
664 if (idx
< e820_entries
&& e820_table
[idx
].type
== cpu_to_le32(type
)) {
665 *address
= le64_to_cpu(e820_table
[idx
].address
);
666 *length
= le64_to_cpu(e820_table
[idx
].length
);
672 /* Enables contiguous-apic-ID mode, for compatibility */
673 static bool compat_apic_id_mode
;
675 void enable_compat_apic_id_mode(void)
677 compat_apic_id_mode
= true;
680 /* Calculates initial APIC ID for a specific CPU index
682 * Currently we need to be able to calculate the APIC ID from the CPU index
683 * alone (without requiring a CPU object), as the QEMU<->Seabios interfaces have
684 * no concept of "CPU index", and the NUMA tables on fw_cfg need the APIC ID of
685 * all CPUs up to max_cpus.
687 static uint32_t x86_cpu_apic_id_from_index(unsigned int cpu_index
)
692 correct_id
= x86_apicid_from_cpu_idx(smp_cores
, smp_threads
, cpu_index
);
693 if (compat_apic_id_mode
) {
694 if (cpu_index
!= correct_id
&& !warned
&& !qtest_enabled()) {
695 error_report("APIC IDs set in compatibility mode, "
696 "CPU topology won't match the configuration");
705 static void pc_build_smbios(PCMachineState
*pcms
)
707 uint8_t *smbios_tables
, *smbios_anchor
;
708 size_t smbios_tables_len
, smbios_anchor_len
;
709 struct smbios_phys_mem_area
*mem_array
;
710 unsigned i
, array_count
;
711 MachineState
*ms
= MACHINE(pcms
);
712 X86CPU
*cpu
= X86_CPU(ms
->possible_cpus
->cpus
[0].cpu
);
714 /* tell smbios about cpuid version and features */
715 smbios_set_cpuid(cpu
->env
.cpuid_version
, cpu
->env
.features
[FEAT_1_EDX
]);
717 smbios_tables
= smbios_get_table_legacy(&smbios_tables_len
);
719 fw_cfg_add_bytes(pcms
->fw_cfg
, FW_CFG_SMBIOS_ENTRIES
,
720 smbios_tables
, smbios_tables_len
);
723 /* build the array of physical mem area from e820 table */
724 mem_array
= g_malloc0(sizeof(*mem_array
) * e820_get_num_entries());
725 for (i
= 0, array_count
= 0; i
< e820_get_num_entries(); i
++) {
728 if (e820_get_entry(i
, E820_RAM
, &addr
, &len
)) {
729 mem_array
[array_count
].address
= addr
;
730 mem_array
[array_count
].length
= len
;
734 smbios_get_tables(mem_array
, array_count
,
735 &smbios_tables
, &smbios_tables_len
,
736 &smbios_anchor
, &smbios_anchor_len
);
740 fw_cfg_add_file(pcms
->fw_cfg
, "etc/smbios/smbios-tables",
741 smbios_tables
, smbios_tables_len
);
742 fw_cfg_add_file(pcms
->fw_cfg
, "etc/smbios/smbios-anchor",
743 smbios_anchor
, smbios_anchor_len
);
747 static FWCfgState
*bochs_bios_init(AddressSpace
*as
, PCMachineState
*pcms
)
750 uint64_t *numa_fw_cfg
;
752 const CPUArchIdList
*cpus
;
753 MachineClass
*mc
= MACHINE_GET_CLASS(pcms
);
755 fw_cfg
= fw_cfg_init_io_dma(FW_CFG_IO_BASE
, FW_CFG_IO_BASE
+ 4, as
);
756 fw_cfg_add_i16(fw_cfg
, FW_CFG_NB_CPUS
, pcms
->boot_cpus
);
758 /* FW_CFG_MAX_CPUS is a bit confusing/problematic on x86:
760 * For machine types prior to 1.8, SeaBIOS needs FW_CFG_MAX_CPUS for
761 * building MPTable, ACPI MADT, ACPI CPU hotplug and ACPI SRAT table,
762 * that tables are based on xAPIC ID and QEMU<->SeaBIOS interface
763 * for CPU hotplug also uses APIC ID and not "CPU index".
764 * This means that FW_CFG_MAX_CPUS is not the "maximum number of CPUs",
765 * but the "limit to the APIC ID values SeaBIOS may see".
767 * So for compatibility reasons with old BIOSes we are stuck with
768 * "etc/max-cpus" actually being apic_id_limit
770 fw_cfg_add_i16(fw_cfg
, FW_CFG_MAX_CPUS
, (uint16_t)pcms
->apic_id_limit
);
771 fw_cfg_add_i64(fw_cfg
, FW_CFG_RAM_SIZE
, (uint64_t)ram_size
);
772 fw_cfg_add_bytes(fw_cfg
, FW_CFG_ACPI_TABLES
,
773 acpi_tables
, acpi_tables_len
);
774 fw_cfg_add_i32(fw_cfg
, FW_CFG_IRQ0_OVERRIDE
, kvm_allows_irq0_override());
776 fw_cfg_add_bytes(fw_cfg
, FW_CFG_E820_TABLE
,
777 &e820_reserve
, sizeof(e820_reserve
));
778 fw_cfg_add_file(fw_cfg
, "etc/e820", e820_table
,
779 sizeof(struct e820_entry
) * e820_entries
);
781 fw_cfg_add_bytes(fw_cfg
, FW_CFG_HPET
, &hpet_cfg
, sizeof(hpet_cfg
));
782 /* allocate memory for the NUMA channel: one (64bit) word for the number
783 * of nodes, one word for each VCPU->node and one word for each node to
784 * hold the amount of memory.
786 numa_fw_cfg
= g_new0(uint64_t, 1 + pcms
->apic_id_limit
+ nb_numa_nodes
);
787 numa_fw_cfg
[0] = cpu_to_le64(nb_numa_nodes
);
788 cpus
= mc
->possible_cpu_arch_ids(MACHINE(pcms
));
789 for (i
= 0; i
< cpus
->len
; i
++) {
790 unsigned int apic_id
= cpus
->cpus
[i
].arch_id
;
791 assert(apic_id
< pcms
->apic_id_limit
);
792 numa_fw_cfg
[apic_id
+ 1] = cpu_to_le64(cpus
->cpus
[i
].props
.node_id
);
794 for (i
= 0; i
< nb_numa_nodes
; i
++) {
795 numa_fw_cfg
[pcms
->apic_id_limit
+ 1 + i
] =
796 cpu_to_le64(numa_info
[i
].node_mem
);
798 fw_cfg_add_bytes(fw_cfg
, FW_CFG_NUMA
, numa_fw_cfg
,
799 (1 + pcms
->apic_id_limit
+ nb_numa_nodes
) *
800 sizeof(*numa_fw_cfg
));
805 static long get_file_size(FILE *f
)
809 /* XXX: on Unix systems, using fstat() probably makes more sense */
812 fseek(f
, 0, SEEK_END
);
814 fseek(f
, where
, SEEK_SET
);
819 /* setup_data types */
821 #define SETUP_E820_EXT 1
831 } __attribute__((packed
));
833 static void load_linux(PCMachineState
*pcms
,
837 int setup_size
, kernel_size
, initrd_size
= 0, cmdline_size
;
838 int dtb_size
, setup_data_offset
;
840 uint8_t header
[8192], *setup
, *kernel
, *initrd_data
;
841 hwaddr real_addr
, prot_addr
, cmdline_addr
, initrd_addr
= 0;
844 MachineState
*machine
= MACHINE(pcms
);
845 PCMachineClass
*pcmc
= PC_MACHINE_GET_CLASS(pcms
);
846 struct setup_data
*setup_data
;
847 const char *kernel_filename
= machine
->kernel_filename
;
848 const char *initrd_filename
= machine
->initrd_filename
;
849 const char *dtb_filename
= machine
->dtb
;
850 const char *kernel_cmdline
= machine
->kernel_cmdline
;
852 /* Align to 16 bytes as a paranoia measure */
853 cmdline_size
= (strlen(kernel_cmdline
)+16) & ~15;
855 /* load the kernel header */
856 f
= fopen(kernel_filename
, "rb");
857 if (!f
|| !(kernel_size
= get_file_size(f
)) ||
858 fread(header
, 1, MIN(ARRAY_SIZE(header
), kernel_size
), f
) !=
859 MIN(ARRAY_SIZE(header
), kernel_size
)) {
860 fprintf(stderr
, "qemu: could not load kernel '%s': %s\n",
861 kernel_filename
, strerror(errno
));
865 /* kernel protocol version */
867 fprintf(stderr
, "header magic: %#x\n", ldl_p(header
+0x202));
869 if (ldl_p(header
+0x202) == 0x53726448) {
870 protocol
= lduw_p(header
+0x206);
872 /* This looks like a multiboot kernel. If it is, let's stop
873 treating it like a Linux kernel. */
874 if (load_multiboot(fw_cfg
, f
, kernel_filename
, initrd_filename
,
875 kernel_cmdline
, kernel_size
, header
)) {
881 if (protocol
< 0x200 || !(header
[0x211] & 0x01)) {
884 cmdline_addr
= 0x9a000 - cmdline_size
;
886 } else if (protocol
< 0x202) {
887 /* High but ancient kernel */
889 cmdline_addr
= 0x9a000 - cmdline_size
;
890 prot_addr
= 0x100000;
892 /* High and recent kernel */
894 cmdline_addr
= 0x20000;
895 prot_addr
= 0x100000;
900 "qemu: real_addr = 0x" TARGET_FMT_plx
"\n"
901 "qemu: cmdline_addr = 0x" TARGET_FMT_plx
"\n"
902 "qemu: prot_addr = 0x" TARGET_FMT_plx
"\n",
908 /* highest address for loading the initrd */
909 if (protocol
>= 0x203) {
910 initrd_max
= ldl_p(header
+0x22c);
912 initrd_max
= 0x37ffffff;
915 if (initrd_max
>= pcms
->below_4g_mem_size
- pcmc
->acpi_data_size
) {
916 initrd_max
= pcms
->below_4g_mem_size
- pcmc
->acpi_data_size
- 1;
919 fw_cfg_add_i32(fw_cfg
, FW_CFG_CMDLINE_ADDR
, cmdline_addr
);
920 fw_cfg_add_i32(fw_cfg
, FW_CFG_CMDLINE_SIZE
, strlen(kernel_cmdline
)+1);
921 fw_cfg_add_string(fw_cfg
, FW_CFG_CMDLINE_DATA
, kernel_cmdline
);
923 if (protocol
>= 0x202) {
924 stl_p(header
+0x228, cmdline_addr
);
926 stw_p(header
+0x20, 0xA33F);
927 stw_p(header
+0x22, cmdline_addr
-real_addr
);
930 /* handle vga= parameter */
931 vmode
= strstr(kernel_cmdline
, "vga=");
933 unsigned int video_mode
;
936 if (!strncmp(vmode
, "normal", 6)) {
938 } else if (!strncmp(vmode
, "ext", 3)) {
940 } else if (!strncmp(vmode
, "ask", 3)) {
943 video_mode
= strtol(vmode
, NULL
, 0);
945 stw_p(header
+0x1fa, video_mode
);
949 /* High nybble = B reserved for QEMU; low nybble is revision number.
950 If this code is substantially changed, you may want to consider
951 incrementing the revision. */
952 if (protocol
>= 0x200) {
953 header
[0x210] = 0xB0;
956 if (protocol
>= 0x201) {
957 header
[0x211] |= 0x80; /* CAN_USE_HEAP */
958 stw_p(header
+0x224, cmdline_addr
-real_addr
-0x200);
962 if (initrd_filename
) {
963 if (protocol
< 0x200) {
964 fprintf(stderr
, "qemu: linux kernel too old to load a ram disk\n");
968 initrd_size
= get_image_size(initrd_filename
);
969 if (initrd_size
< 0) {
970 fprintf(stderr
, "qemu: error reading initrd %s: %s\n",
971 initrd_filename
, strerror(errno
));
975 initrd_addr
= (initrd_max
-initrd_size
) & ~4095;
977 initrd_data
= g_malloc(initrd_size
);
978 load_image(initrd_filename
, initrd_data
);
980 fw_cfg_add_i32(fw_cfg
, FW_CFG_INITRD_ADDR
, initrd_addr
);
981 fw_cfg_add_i32(fw_cfg
, FW_CFG_INITRD_SIZE
, initrd_size
);
982 fw_cfg_add_bytes(fw_cfg
, FW_CFG_INITRD_DATA
, initrd_data
, initrd_size
);
984 stl_p(header
+0x218, initrd_addr
);
985 stl_p(header
+0x21c, initrd_size
);
988 /* load kernel and setup */
989 setup_size
= header
[0x1f1];
990 if (setup_size
== 0) {
993 setup_size
= (setup_size
+1)*512;
994 if (setup_size
> kernel_size
) {
995 fprintf(stderr
, "qemu: invalid kernel header\n");
998 kernel_size
-= setup_size
;
1000 setup
= g_malloc(setup_size
);
1001 kernel
= g_malloc(kernel_size
);
1002 fseek(f
, 0, SEEK_SET
);
1003 if (fread(setup
, 1, setup_size
, f
) != setup_size
) {
1004 fprintf(stderr
, "fread() failed\n");
1007 if (fread(kernel
, 1, kernel_size
, f
) != kernel_size
) {
1008 fprintf(stderr
, "fread() failed\n");
1013 /* append dtb to kernel */
1015 if (protocol
< 0x209) {
1016 fprintf(stderr
, "qemu: Linux kernel too old to load a dtb\n");
1020 dtb_size
= get_image_size(dtb_filename
);
1021 if (dtb_size
<= 0) {
1022 fprintf(stderr
, "qemu: error reading dtb %s: %s\n",
1023 dtb_filename
, strerror(errno
));
1027 setup_data_offset
= QEMU_ALIGN_UP(kernel_size
, 16);
1028 kernel_size
= setup_data_offset
+ sizeof(struct setup_data
) + dtb_size
;
1029 kernel
= g_realloc(kernel
, kernel_size
);
1031 stq_p(header
+0x250, prot_addr
+ setup_data_offset
);
1033 setup_data
= (struct setup_data
*)(kernel
+ setup_data_offset
);
1034 setup_data
->next
= 0;
1035 setup_data
->type
= cpu_to_le32(SETUP_DTB
);
1036 setup_data
->len
= cpu_to_le32(dtb_size
);
1038 load_image_size(dtb_filename
, setup_data
->data
, dtb_size
);
1041 memcpy(setup
, header
, MIN(sizeof(header
), setup_size
));
1043 fw_cfg_add_i32(fw_cfg
, FW_CFG_KERNEL_ADDR
, prot_addr
);
1044 fw_cfg_add_i32(fw_cfg
, FW_CFG_KERNEL_SIZE
, kernel_size
);
1045 fw_cfg_add_bytes(fw_cfg
, FW_CFG_KERNEL_DATA
, kernel
, kernel_size
);
1047 fw_cfg_add_i32(fw_cfg
, FW_CFG_SETUP_ADDR
, real_addr
);
1048 fw_cfg_add_i32(fw_cfg
, FW_CFG_SETUP_SIZE
, setup_size
);
1049 fw_cfg_add_bytes(fw_cfg
, FW_CFG_SETUP_DATA
, setup
, setup_size
);
1051 option_rom
[nb_option_roms
].bootindex
= 0;
1052 option_rom
[nb_option_roms
].name
= "linuxboot.bin";
1053 if (pcmc
->linuxboot_dma_enabled
&& fw_cfg_dma_enabled(fw_cfg
)) {
1054 option_rom
[nb_option_roms
].name
= "linuxboot_dma.bin";
1059 #define NE2000_NB_MAX 6
1061 static const int ne2000_io
[NE2000_NB_MAX
] = { 0x300, 0x320, 0x340, 0x360,
1063 static const int ne2000_irq
[NE2000_NB_MAX
] = { 9, 10, 11, 3, 4, 5 };
1065 void pc_init_ne2k_isa(ISABus
*bus
, NICInfo
*nd
)
1067 static int nb_ne2k
= 0;
1069 if (nb_ne2k
== NE2000_NB_MAX
)
1071 isa_ne2000_init(bus
, ne2000_io
[nb_ne2k
],
1072 ne2000_irq
[nb_ne2k
], nd
);
1076 DeviceState
*cpu_get_current_apic(void)
1079 X86CPU
*cpu
= X86_CPU(current_cpu
);
1080 return cpu
->apic_state
;
1086 void pc_acpi_smi_interrupt(void *opaque
, int irq
, int level
)
1088 X86CPU
*cpu
= opaque
;
1091 cpu_interrupt(CPU(cpu
), CPU_INTERRUPT_SMI
);
1095 static void pc_new_cpu(const char *typename
, int64_t apic_id
, Error
**errp
)
1098 Error
*local_err
= NULL
;
1100 cpu
= object_new(typename
);
1102 object_property_set_uint(cpu
, apic_id
, "apic-id", &local_err
);
1103 object_property_set_bool(cpu
, true, "realized", &local_err
);
1106 error_propagate(errp
, local_err
);
1109 void pc_hot_add_cpu(const int64_t id
, Error
**errp
)
1111 MachineState
*ms
= MACHINE(qdev_get_machine());
1112 int64_t apic_id
= x86_cpu_apic_id_from_index(id
);
1113 Error
*local_err
= NULL
;
1116 error_setg(errp
, "Invalid CPU id: %" PRIi64
, id
);
1120 if (apic_id
>= ACPI_CPU_HOTPLUG_ID_LIMIT
) {
1121 error_setg(errp
, "Unable to add CPU: %" PRIi64
1122 ", resulting APIC ID (%" PRIi64
") is too large",
1127 pc_new_cpu(ms
->cpu_type
, apic_id
, &local_err
);
1129 error_propagate(errp
, local_err
);
1134 void pc_cpus_init(PCMachineState
*pcms
)
1137 const CPUArchIdList
*possible_cpus
;
1138 MachineState
*ms
= MACHINE(pcms
);
1139 MachineClass
*mc
= MACHINE_GET_CLASS(pcms
);
1141 /* Calculates the limit to CPU APIC ID values
1143 * Limit for the APIC ID value, so that all
1144 * CPU APIC IDs are < pcms->apic_id_limit.
1146 * This is used for FW_CFG_MAX_CPUS. See comments on bochs_bios_init().
1148 pcms
->apic_id_limit
= x86_cpu_apic_id_from_index(max_cpus
- 1) + 1;
1149 possible_cpus
= mc
->possible_cpu_arch_ids(ms
);
1150 for (i
= 0; i
< smp_cpus
; i
++) {
1151 pc_new_cpu(ms
->cpu_type
, possible_cpus
->cpus
[i
].arch_id
, &error_fatal
);
1155 static void pc_build_feature_control_file(PCMachineState
*pcms
)
1157 MachineState
*ms
= MACHINE(pcms
);
1158 X86CPU
*cpu
= X86_CPU(ms
->possible_cpus
->cpus
[0].cpu
);
1159 CPUX86State
*env
= &cpu
->env
;
1160 uint32_t unused
, ecx
, edx
;
1161 uint64_t feature_control_bits
= 0;
1164 cpu_x86_cpuid(env
, 1, 0, &unused
, &unused
, &ecx
, &edx
);
1165 if (ecx
& CPUID_EXT_VMX
) {
1166 feature_control_bits
|= FEATURE_CONTROL_VMXON_ENABLED_OUTSIDE_SMX
;
1169 if ((edx
& (CPUID_EXT2_MCE
| CPUID_EXT2_MCA
)) ==
1170 (CPUID_EXT2_MCE
| CPUID_EXT2_MCA
) &&
1171 (env
->mcg_cap
& MCG_LMCE_P
)) {
1172 feature_control_bits
|= FEATURE_CONTROL_LMCE
;
1175 if (!feature_control_bits
) {
1179 val
= g_malloc(sizeof(*val
));
1180 *val
= cpu_to_le64(feature_control_bits
| FEATURE_CONTROL_LOCKED
);
1181 fw_cfg_add_file(pcms
->fw_cfg
, "etc/msr_feature_control", val
, sizeof(*val
));
1184 static void rtc_set_cpus_count(ISADevice
*rtc
, uint16_t cpus_count
)
1186 if (cpus_count
> 0xff) {
1187 /* If the number of CPUs can't be represented in 8 bits, the
1188 * BIOS must use "FW_CFG_NB_CPUS". Set RTC field to 0 just
1189 * to make old BIOSes fail more predictably.
1191 rtc_set_memory(rtc
, 0x5f, 0);
1193 rtc_set_memory(rtc
, 0x5f, cpus_count
- 1);
1198 void pc_machine_done(Notifier
*notifier
, void *data
)
1200 PCMachineState
*pcms
= container_of(notifier
,
1201 PCMachineState
, machine_done
);
1202 PCIBus
*bus
= pcms
->bus
;
1204 /* set the number of CPUs */
1205 rtc_set_cpus_count(pcms
->rtc
, pcms
->boot_cpus
);
1208 int extra_hosts
= 0;
1210 QLIST_FOREACH(bus
, &bus
->child
, sibling
) {
1211 /* look for expander root buses */
1212 if (pci_bus_is_root(bus
)) {
1216 if (extra_hosts
&& pcms
->fw_cfg
) {
1217 uint64_t *val
= g_malloc(sizeof(*val
));
1218 *val
= cpu_to_le64(extra_hosts
);
1219 fw_cfg_add_file(pcms
->fw_cfg
,
1220 "etc/extra-pci-roots", val
, sizeof(*val
));
1226 pc_build_smbios(pcms
);
1227 pc_build_feature_control_file(pcms
);
1228 /* update FW_CFG_NB_CPUS to account for -device added CPUs */
1229 fw_cfg_modify_i16(pcms
->fw_cfg
, FW_CFG_NB_CPUS
, pcms
->boot_cpus
);
1232 if (pcms
->apic_id_limit
> 255 && !xen_enabled()) {
1233 IntelIOMMUState
*iommu
= INTEL_IOMMU_DEVICE(x86_iommu_get_default());
1235 if (!iommu
|| !iommu
->x86_iommu
.intr_supported
||
1236 iommu
->intr_eim
!= ON_OFF_AUTO_ON
) {
1237 error_report("current -smp configuration requires "
1238 "Extended Interrupt Mode enabled. "
1239 "You can add an IOMMU using: "
1240 "-device intel-iommu,intremap=on,eim=on");
1246 void pc_guest_info_init(PCMachineState
*pcms
)
1250 pcms
->apic_xrupt_override
= kvm_allows_irq0_override();
1251 pcms
->numa_nodes
= nb_numa_nodes
;
1252 pcms
->node_mem
= g_malloc0(pcms
->numa_nodes
*
1253 sizeof *pcms
->node_mem
);
1254 for (i
= 0; i
< nb_numa_nodes
; i
++) {
1255 pcms
->node_mem
[i
] = numa_info
[i
].node_mem
;
1258 pcms
->machine_done
.notify
= pc_machine_done
;
1259 qemu_add_machine_init_done_notifier(&pcms
->machine_done
);
1262 /* setup pci memory address space mapping into system address space */
1263 void pc_pci_as_mapping_init(Object
*owner
, MemoryRegion
*system_memory
,
1264 MemoryRegion
*pci_address_space
)
1266 /* Set to lower priority than RAM */
1267 memory_region_add_subregion_overlap(system_memory
, 0x0,
1268 pci_address_space
, -1);
1271 void pc_acpi_init(const char *default_dsdt
)
1275 if (acpi_tables
!= NULL
) {
1276 /* manually set via -acpitable, leave it alone */
1280 filename
= qemu_find_file(QEMU_FILE_TYPE_BIOS
, default_dsdt
);
1281 if (filename
== NULL
) {
1282 warn_report("failed to find %s", default_dsdt
);
1284 QemuOpts
*opts
= qemu_opts_create(qemu_find_opts("acpi"), NULL
, 0,
1288 qemu_opt_set(opts
, "file", filename
, &error_abort
);
1290 acpi_table_add_builtin(opts
, &err
);
1292 warn_reportf_err(err
, "failed to load %s: ", filename
);
1298 void xen_load_linux(PCMachineState
*pcms
)
1303 assert(MACHINE(pcms
)->kernel_filename
!= NULL
);
1305 fw_cfg
= fw_cfg_init_io(FW_CFG_IO_BASE
);
1306 fw_cfg_add_i16(fw_cfg
, FW_CFG_NB_CPUS
, pcms
->boot_cpus
);
1309 load_linux(pcms
, fw_cfg
);
1310 for (i
= 0; i
< nb_option_roms
; i
++) {
1311 assert(!strcmp(option_rom
[i
].name
, "linuxboot.bin") ||
1312 !strcmp(option_rom
[i
].name
, "linuxboot_dma.bin") ||
1313 !strcmp(option_rom
[i
].name
, "multiboot.bin"));
1314 rom_add_option(option_rom
[i
].name
, option_rom
[i
].bootindex
);
1316 pcms
->fw_cfg
= fw_cfg
;
1319 void pc_memory_init(PCMachineState
*pcms
,
1320 MemoryRegion
*system_memory
,
1321 MemoryRegion
*rom_memory
,
1322 MemoryRegion
**ram_memory
)
1325 MemoryRegion
*ram
, *option_rom_mr
;
1326 MemoryRegion
*ram_below_4g
, *ram_above_4g
;
1328 MachineState
*machine
= MACHINE(pcms
);
1329 PCMachineClass
*pcmc
= PC_MACHINE_GET_CLASS(pcms
);
1331 assert(machine
->ram_size
== pcms
->below_4g_mem_size
+
1332 pcms
->above_4g_mem_size
);
1334 linux_boot
= (machine
->kernel_filename
!= NULL
);
1336 /* Allocate RAM. We allocate it as a single memory region and use
1337 * aliases to address portions of it, mostly for backwards compatibility
1338 * with older qemus that used qemu_ram_alloc().
1340 ram
= g_malloc(sizeof(*ram
));
1341 memory_region_allocate_system_memory(ram
, NULL
, "pc.ram",
1344 ram_below_4g
= g_malloc(sizeof(*ram_below_4g
));
1345 memory_region_init_alias(ram_below_4g
, NULL
, "ram-below-4g", ram
,
1346 0, pcms
->below_4g_mem_size
);
1347 memory_region_add_subregion(system_memory
, 0, ram_below_4g
);
1348 e820_add_entry(0, pcms
->below_4g_mem_size
, E820_RAM
);
1349 if (pcms
->above_4g_mem_size
> 0) {
1350 ram_above_4g
= g_malloc(sizeof(*ram_above_4g
));
1351 memory_region_init_alias(ram_above_4g
, NULL
, "ram-above-4g", ram
,
1352 pcms
->below_4g_mem_size
,
1353 pcms
->above_4g_mem_size
);
1354 memory_region_add_subregion(system_memory
, 0x100000000ULL
,
1356 e820_add_entry(0x100000000ULL
, pcms
->above_4g_mem_size
, E820_RAM
);
1359 if (!pcmc
->has_reserved_memory
&&
1360 (machine
->ram_slots
||
1361 (machine
->maxram_size
> machine
->ram_size
))) {
1362 MachineClass
*mc
= MACHINE_GET_CLASS(machine
);
1364 error_report("\"-memory 'slots|maxmem'\" is not supported by: %s",
1369 /* initialize hotplug memory address space */
1370 if (pcmc
->has_reserved_memory
&&
1371 (machine
->ram_size
< machine
->maxram_size
)) {
1372 ram_addr_t hotplug_mem_size
=
1373 machine
->maxram_size
- machine
->ram_size
;
1375 if (machine
->ram_slots
> ACPI_MAX_RAM_SLOTS
) {
1376 error_report("unsupported amount of memory slots: %"PRIu64
,
1377 machine
->ram_slots
);
1381 if (QEMU_ALIGN_UP(machine
->maxram_size
,
1382 TARGET_PAGE_SIZE
) != machine
->maxram_size
) {
1383 error_report("maximum memory size must by aligned to multiple of "
1384 "%d bytes", TARGET_PAGE_SIZE
);
1388 pcms
->hotplug_memory
.base
=
1389 ROUND_UP(0x100000000ULL
+ pcms
->above_4g_mem_size
, 1ULL << 30);
1391 if (pcmc
->enforce_aligned_dimm
) {
1392 /* size hotplug region assuming 1G page max alignment per slot */
1393 hotplug_mem_size
+= (1ULL << 30) * machine
->ram_slots
;
1396 if ((pcms
->hotplug_memory
.base
+ hotplug_mem_size
) <
1398 error_report("unsupported amount of maximum memory: " RAM_ADDR_FMT
,
1399 machine
->maxram_size
);
1403 memory_region_init(&pcms
->hotplug_memory
.mr
, OBJECT(pcms
),
1404 "hotplug-memory", hotplug_mem_size
);
1405 memory_region_add_subregion(system_memory
, pcms
->hotplug_memory
.base
,
1406 &pcms
->hotplug_memory
.mr
);
1409 /* Initialize PC system firmware */
1410 pc_system_firmware_init(rom_memory
, !pcmc
->pci_enabled
);
1412 option_rom_mr
= g_malloc(sizeof(*option_rom_mr
));
1413 memory_region_init_ram(option_rom_mr
, NULL
, "pc.rom", PC_ROM_SIZE
,
1415 if (pcmc
->pci_enabled
) {
1416 memory_region_set_readonly(option_rom_mr
, true);
1418 memory_region_add_subregion_overlap(rom_memory
,
1423 fw_cfg
= bochs_bios_init(&address_space_memory
, pcms
);
1427 if (pcmc
->has_reserved_memory
&& pcms
->hotplug_memory
.base
) {
1428 uint64_t *val
= g_malloc(sizeof(*val
));
1429 PCMachineClass
*pcmc
= PC_MACHINE_GET_CLASS(pcms
);
1430 uint64_t res_mem_end
= pcms
->hotplug_memory
.base
;
1432 if (!pcmc
->broken_reserved_end
) {
1433 res_mem_end
+= memory_region_size(&pcms
->hotplug_memory
.mr
);
1435 *val
= cpu_to_le64(ROUND_UP(res_mem_end
, 0x1ULL
<< 30));
1436 fw_cfg_add_file(fw_cfg
, "etc/reserved-memory-end", val
, sizeof(*val
));
1440 load_linux(pcms
, fw_cfg
);
1443 for (i
= 0; i
< nb_option_roms
; i
++) {
1444 rom_add_option(option_rom
[i
].name
, option_rom
[i
].bootindex
);
1446 pcms
->fw_cfg
= fw_cfg
;
1448 /* Init default IOAPIC address space */
1449 pcms
->ioapic_as
= &address_space_memory
;
1453 * The 64bit pci hole starts after "above 4G RAM" and
1454 * potentially the space reserved for memory hotplug.
1456 uint64_t pc_pci_hole64_start(void)
1458 PCMachineState
*pcms
= PC_MACHINE(qdev_get_machine());
1459 PCMachineClass
*pcmc
= PC_MACHINE_GET_CLASS(pcms
);
1460 uint64_t hole64_start
= 0;
1462 if (pcmc
->has_reserved_memory
&& pcms
->hotplug_memory
.base
) {
1463 hole64_start
= pcms
->hotplug_memory
.base
;
1464 if (!pcmc
->broken_reserved_end
) {
1465 hole64_start
+= memory_region_size(&pcms
->hotplug_memory
.mr
);
1468 hole64_start
= 0x100000000ULL
+ pcms
->above_4g_mem_size
;
1471 return ROUND_UP(hole64_start
, 1ULL << 30);
1474 qemu_irq
pc_allocate_cpu_irq(void)
1476 return qemu_allocate_irq(pic_irq_request
, NULL
, 0);
1479 DeviceState
*pc_vga_init(ISABus
*isa_bus
, PCIBus
*pci_bus
)
1481 DeviceState
*dev
= NULL
;
1483 rom_set_order_override(FW_CFG_ORDER_OVERRIDE_VGA
);
1485 PCIDevice
*pcidev
= pci_vga_init(pci_bus
);
1486 dev
= pcidev
? &pcidev
->qdev
: NULL
;
1487 } else if (isa_bus
) {
1488 ISADevice
*isadev
= isa_vga_init(isa_bus
);
1489 dev
= isadev
? DEVICE(isadev
) : NULL
;
1491 rom_reset_order_override();
1495 static const MemoryRegionOps ioport80_io_ops
= {
1496 .write
= ioport80_write
,
1497 .read
= ioport80_read
,
1498 .endianness
= DEVICE_NATIVE_ENDIAN
,
1500 .min_access_size
= 1,
1501 .max_access_size
= 1,
1505 static const MemoryRegionOps ioportF0_io_ops
= {
1506 .write
= ioportF0_write
,
1507 .read
= ioportF0_read
,
1508 .endianness
= DEVICE_NATIVE_ENDIAN
,
1510 .min_access_size
= 1,
1511 .max_access_size
= 1,
1515 void pc_basic_device_init(ISABus
*isa_bus
, qemu_irq
*gsi
,
1516 ISADevice
**rtc_state
,
1523 DriveInfo
*fd
[MAX_FD
];
1524 DeviceState
*hpet
= NULL
;
1525 int pit_isa_irq
= 0;
1526 qemu_irq pit_alt_irq
= NULL
;
1527 qemu_irq rtc_irq
= NULL
;
1529 ISADevice
*i8042
, *port92
, *vmmouse
, *pit
= NULL
;
1530 MemoryRegion
*ioport80_io
= g_new(MemoryRegion
, 1);
1531 MemoryRegion
*ioportF0_io
= g_new(MemoryRegion
, 1);
1533 memory_region_init_io(ioport80_io
, NULL
, &ioport80_io_ops
, NULL
, "ioport80", 1);
1534 memory_region_add_subregion(isa_bus
->address_space_io
, 0x80, ioport80_io
);
1536 memory_region_init_io(ioportF0_io
, NULL
, &ioportF0_io_ops
, NULL
, "ioportF0", 1);
1537 memory_region_add_subregion(isa_bus
->address_space_io
, 0xf0, ioportF0_io
);
1540 * Check if an HPET shall be created.
1542 * Without KVM_CAP_PIT_STATE2, we cannot switch off the in-kernel PIT
1543 * when the HPET wants to take over. Thus we have to disable the latter.
1545 if (!no_hpet
&& (!kvm_irqchip_in_kernel() || kvm_has_pit_state2())) {
1546 /* In order to set property, here not using sysbus_try_create_simple */
1547 hpet
= qdev_try_create(NULL
, TYPE_HPET
);
1549 /* For pc-piix-*, hpet's intcap is always IRQ2. For pc-q35-1.7
1550 * and earlier, use IRQ2 for compat. Otherwise, use IRQ16~23,
1553 uint8_t compat
= object_property_get_uint(OBJECT(hpet
),
1556 qdev_prop_set_uint32(hpet
, HPET_INTCAP
, hpet_irqs
);
1558 qdev_init_nofail(hpet
);
1559 sysbus_mmio_map(SYS_BUS_DEVICE(hpet
), 0, HPET_BASE
);
1561 for (i
= 0; i
< GSI_NUM_PINS
; i
++) {
1562 sysbus_connect_irq(SYS_BUS_DEVICE(hpet
), i
, gsi
[i
]);
1565 pit_alt_irq
= qdev_get_gpio_in(hpet
, HPET_LEGACY_PIT_INT
);
1566 rtc_irq
= qdev_get_gpio_in(hpet
, HPET_LEGACY_RTC_INT
);
1569 *rtc_state
= mc146818_rtc_init(isa_bus
, 2000, rtc_irq
);
1571 qemu_register_boot_set(pc_boot_set
, *rtc_state
);
1573 if (!xen_enabled() && has_pit
) {
1574 if (kvm_pit_in_kernel()) {
1575 pit
= kvm_pit_init(isa_bus
, 0x40);
1577 pit
= i8254_pit_init(isa_bus
, 0x40, pit_isa_irq
, pit_alt_irq
);
1580 /* connect PIT to output control line of the HPET */
1581 qdev_connect_gpio_out(hpet
, 0, qdev_get_gpio_in(DEVICE(pit
), 0));
1583 pcspk_init(isa_bus
, pit
);
1586 serial_hds_isa_init(isa_bus
, 0, MAX_SERIAL_PORTS
);
1587 parallel_hds_isa_init(isa_bus
, MAX_PARALLEL_PORTS
);
1589 a20_line
= qemu_allocate_irqs(handle_a20_line_change
, first_cpu
, 2);
1590 i8042
= isa_create_simple(isa_bus
, "i8042");
1591 i8042_setup_a20_line(i8042
, a20_line
[0]);
1593 vmport_init(isa_bus
);
1594 vmmouse
= isa_try_create(isa_bus
, "vmmouse");
1599 DeviceState
*dev
= DEVICE(vmmouse
);
1600 qdev_prop_set_ptr(dev
, "ps2_mouse", i8042
);
1601 qdev_init_nofail(dev
);
1603 port92
= isa_create_simple(isa_bus
, "port92");
1604 port92_init(port92
, a20_line
[1]);
1607 DMA_init(isa_bus
, 0);
1609 for(i
= 0; i
< MAX_FD
; i
++) {
1610 fd
[i
] = drive_get(IF_FLOPPY
, 0, i
);
1611 create_fdctrl
|= !!fd
[i
];
1613 if (create_fdctrl
) {
1614 fdctrl_init_isa(isa_bus
, fd
);
1618 void pc_nic_init(ISABus
*isa_bus
, PCIBus
*pci_bus
)
1622 rom_set_order_override(FW_CFG_ORDER_OVERRIDE_NIC
);
1623 for (i
= 0; i
< nb_nics
; i
++) {
1624 NICInfo
*nd
= &nd_table
[i
];
1626 if (!pci_bus
|| (nd
->model
&& strcmp(nd
->model
, "ne2k_isa") == 0)) {
1627 pc_init_ne2k_isa(isa_bus
, nd
);
1629 pci_nic_init_nofail(nd
, pci_bus
, "e1000", NULL
);
1632 rom_reset_order_override();
1635 void pc_pci_device_init(PCIBus
*pci_bus
)
1640 /* Note: if=scsi is deprecated with PC machine types */
1641 max_bus
= drive_get_max_bus(IF_SCSI
);
1642 for (bus
= 0; bus
<= max_bus
; bus
++) {
1643 pci_create_simple(pci_bus
, -1, "lsi53c895a");
1645 * By not creating frontends here, we make
1646 * scsi_legacy_handle_cmdline() create them, and warn that
1647 * this usage is deprecated.
1652 void ioapic_init_gsi(GSIState
*gsi_state
, const char *parent_name
)
1658 if (kvm_ioapic_in_kernel()) {
1659 dev
= qdev_create(NULL
, "kvm-ioapic");
1661 dev
= qdev_create(NULL
, "ioapic");
1664 object_property_add_child(object_resolve_path(parent_name
, NULL
),
1665 "ioapic", OBJECT(dev
), NULL
);
1667 qdev_init_nofail(dev
);
1668 d
= SYS_BUS_DEVICE(dev
);
1669 sysbus_mmio_map(d
, 0, IO_APIC_DEFAULT_ADDRESS
);
1671 for (i
= 0; i
< IOAPIC_NUM_PINS
; i
++) {
1672 gsi_state
->ioapic_irq
[i
] = qdev_get_gpio_in(dev
, i
);
1676 static void pc_dimm_plug(HotplugHandler
*hotplug_dev
,
1677 DeviceState
*dev
, Error
**errp
)
1679 HotplugHandlerClass
*hhc
;
1680 Error
*local_err
= NULL
;
1681 PCMachineState
*pcms
= PC_MACHINE(hotplug_dev
);
1682 PCMachineClass
*pcmc
= PC_MACHINE_GET_CLASS(pcms
);
1683 PCDIMMDevice
*dimm
= PC_DIMM(dev
);
1684 PCDIMMDeviceClass
*ddc
= PC_DIMM_GET_CLASS(dimm
);
1686 uint64_t align
= TARGET_PAGE_SIZE
;
1687 bool is_nvdimm
= object_dynamic_cast(OBJECT(dev
), TYPE_NVDIMM
);
1689 mr
= ddc
->get_memory_region(dimm
, &local_err
);
1694 if (memory_region_get_alignment(mr
) && pcmc
->enforce_aligned_dimm
) {
1695 align
= memory_region_get_alignment(mr
);
1699 * When -no-acpi is used with Q35 machine type, no ACPI is built,
1700 * but pcms->acpi_dev is still created. Check !acpi_enabled in
1701 * addition to cover this case.
1703 if (!pcms
->acpi_dev
|| !acpi_enabled
) {
1704 error_setg(&local_err
,
1705 "memory hotplug is not enabled: missing acpi device or acpi disabled");
1709 if (is_nvdimm
&& !pcms
->acpi_nvdimm_state
.is_enabled
) {
1710 error_setg(&local_err
,
1711 "nvdimm is not enabled: missing 'nvdimm' in '-M'");
1715 pc_dimm_memory_plug(dev
, &pcms
->hotplug_memory
, mr
, align
, &local_err
);
1721 nvdimm_plug(&pcms
->acpi_nvdimm_state
);
1724 hhc
= HOTPLUG_HANDLER_GET_CLASS(pcms
->acpi_dev
);
1725 hhc
->plug(HOTPLUG_HANDLER(pcms
->acpi_dev
), dev
, &error_abort
);
1727 error_propagate(errp
, local_err
);
1730 static void pc_dimm_unplug_request(HotplugHandler
*hotplug_dev
,
1731 DeviceState
*dev
, Error
**errp
)
1733 HotplugHandlerClass
*hhc
;
1734 Error
*local_err
= NULL
;
1735 PCMachineState
*pcms
= PC_MACHINE(hotplug_dev
);
1738 * When -no-acpi is used with Q35 machine type, no ACPI is built,
1739 * but pcms->acpi_dev is still created. Check !acpi_enabled in
1740 * addition to cover this case.
1742 if (!pcms
->acpi_dev
|| !acpi_enabled
) {
1743 error_setg(&local_err
,
1744 "memory hotplug is not enabled: missing acpi device or acpi disabled");
1748 if (object_dynamic_cast(OBJECT(dev
), TYPE_NVDIMM
)) {
1749 error_setg(&local_err
,
1750 "nvdimm device hot unplug is not supported yet.");
1754 hhc
= HOTPLUG_HANDLER_GET_CLASS(pcms
->acpi_dev
);
1755 hhc
->unplug_request(HOTPLUG_HANDLER(pcms
->acpi_dev
), dev
, &local_err
);
1758 error_propagate(errp
, local_err
);
1761 static void pc_dimm_unplug(HotplugHandler
*hotplug_dev
,
1762 DeviceState
*dev
, Error
**errp
)
1764 PCMachineState
*pcms
= PC_MACHINE(hotplug_dev
);
1765 PCDIMMDevice
*dimm
= PC_DIMM(dev
);
1766 PCDIMMDeviceClass
*ddc
= PC_DIMM_GET_CLASS(dimm
);
1768 HotplugHandlerClass
*hhc
;
1769 Error
*local_err
= NULL
;
1771 mr
= ddc
->get_memory_region(dimm
, &local_err
);
1776 hhc
= HOTPLUG_HANDLER_GET_CLASS(pcms
->acpi_dev
);
1777 hhc
->unplug(HOTPLUG_HANDLER(pcms
->acpi_dev
), dev
, &local_err
);
1783 pc_dimm_memory_unplug(dev
, &pcms
->hotplug_memory
, mr
);
1784 object_unparent(OBJECT(dev
));
1787 error_propagate(errp
, local_err
);
1790 static int pc_apic_cmp(const void *a
, const void *b
)
1792 CPUArchId
*apic_a
= (CPUArchId
*)a
;
1793 CPUArchId
*apic_b
= (CPUArchId
*)b
;
1795 return apic_a
->arch_id
- apic_b
->arch_id
;
1798 /* returns pointer to CPUArchId descriptor that matches CPU's apic_id
1799 * in ms->possible_cpus->cpus, if ms->possible_cpus->cpus has no
1800 * entry corresponding to CPU's apic_id returns NULL.
1802 static CPUArchId
*pc_find_cpu_slot(MachineState
*ms
, uint32_t id
, int *idx
)
1804 CPUArchId apic_id
, *found_cpu
;
1806 apic_id
.arch_id
= id
;
1807 found_cpu
= bsearch(&apic_id
, ms
->possible_cpus
->cpus
,
1808 ms
->possible_cpus
->len
, sizeof(*ms
->possible_cpus
->cpus
),
1810 if (found_cpu
&& idx
) {
1811 *idx
= found_cpu
- ms
->possible_cpus
->cpus
;
1816 static void pc_cpu_plug(HotplugHandler
*hotplug_dev
,
1817 DeviceState
*dev
, Error
**errp
)
1819 CPUArchId
*found_cpu
;
1820 HotplugHandlerClass
*hhc
;
1821 Error
*local_err
= NULL
;
1822 X86CPU
*cpu
= X86_CPU(dev
);
1823 PCMachineState
*pcms
= PC_MACHINE(hotplug_dev
);
1825 if (pcms
->acpi_dev
) {
1826 hhc
= HOTPLUG_HANDLER_GET_CLASS(pcms
->acpi_dev
);
1827 hhc
->plug(HOTPLUG_HANDLER(pcms
->acpi_dev
), dev
, &local_err
);
1833 /* increment the number of CPUs */
1836 rtc_set_cpus_count(pcms
->rtc
, pcms
->boot_cpus
);
1839 fw_cfg_modify_i16(pcms
->fw_cfg
, FW_CFG_NB_CPUS
, pcms
->boot_cpus
);
1842 found_cpu
= pc_find_cpu_slot(MACHINE(pcms
), cpu
->apic_id
, NULL
);
1843 found_cpu
->cpu
= OBJECT(dev
);
1845 error_propagate(errp
, local_err
);
1847 static void pc_cpu_unplug_request_cb(HotplugHandler
*hotplug_dev
,
1848 DeviceState
*dev
, Error
**errp
)
1851 HotplugHandlerClass
*hhc
;
1852 Error
*local_err
= NULL
;
1853 X86CPU
*cpu
= X86_CPU(dev
);
1854 PCMachineState
*pcms
= PC_MACHINE(hotplug_dev
);
1856 if (!pcms
->acpi_dev
) {
1857 error_setg(&local_err
, "CPU hot unplug not supported without ACPI");
1861 pc_find_cpu_slot(MACHINE(pcms
), cpu
->apic_id
, &idx
);
1864 error_setg(&local_err
, "Boot CPU is unpluggable");
1868 hhc
= HOTPLUG_HANDLER_GET_CLASS(pcms
->acpi_dev
);
1869 hhc
->unplug_request(HOTPLUG_HANDLER(pcms
->acpi_dev
), dev
, &local_err
);
1876 error_propagate(errp
, local_err
);
1880 static void pc_cpu_unplug_cb(HotplugHandler
*hotplug_dev
,
1881 DeviceState
*dev
, Error
**errp
)
1883 CPUArchId
*found_cpu
;
1884 HotplugHandlerClass
*hhc
;
1885 Error
*local_err
= NULL
;
1886 X86CPU
*cpu
= X86_CPU(dev
);
1887 PCMachineState
*pcms
= PC_MACHINE(hotplug_dev
);
1889 hhc
= HOTPLUG_HANDLER_GET_CLASS(pcms
->acpi_dev
);
1890 hhc
->unplug(HOTPLUG_HANDLER(pcms
->acpi_dev
), dev
, &local_err
);
1896 found_cpu
= pc_find_cpu_slot(MACHINE(pcms
), cpu
->apic_id
, NULL
);
1897 found_cpu
->cpu
= NULL
;
1898 object_unparent(OBJECT(dev
));
1900 /* decrement the number of CPUs */
1902 /* Update the number of CPUs in CMOS */
1903 rtc_set_cpus_count(pcms
->rtc
, pcms
->boot_cpus
);
1904 fw_cfg_modify_i16(pcms
->fw_cfg
, FW_CFG_NB_CPUS
, pcms
->boot_cpus
);
1906 error_propagate(errp
, local_err
);
1909 static void pc_cpu_pre_plug(HotplugHandler
*hotplug_dev
,
1910 DeviceState
*dev
, Error
**errp
)
1914 CPUArchId
*cpu_slot
;
1915 X86CPUTopoInfo topo
;
1916 X86CPU
*cpu
= X86_CPU(dev
);
1917 MachineState
*ms
= MACHINE(hotplug_dev
);
1918 PCMachineState
*pcms
= PC_MACHINE(hotplug_dev
);
1920 if(!object_dynamic_cast(OBJECT(cpu
), ms
->cpu_type
)) {
1921 error_setg(errp
, "Invalid CPU type, expected cpu type: '%s'",
1926 /* if APIC ID is not set, set it based on socket/core/thread properties */
1927 if (cpu
->apic_id
== UNASSIGNED_APIC_ID
) {
1928 int max_socket
= (max_cpus
- 1) / smp_threads
/ smp_cores
;
1930 if (cpu
->socket_id
< 0) {
1931 error_setg(errp
, "CPU socket-id is not set");
1933 } else if (cpu
->socket_id
> max_socket
) {
1934 error_setg(errp
, "Invalid CPU socket-id: %u must be in range 0:%u",
1935 cpu
->socket_id
, max_socket
);
1938 if (cpu
->core_id
< 0) {
1939 error_setg(errp
, "CPU core-id is not set");
1941 } else if (cpu
->core_id
> (smp_cores
- 1)) {
1942 error_setg(errp
, "Invalid CPU core-id: %u must be in range 0:%u",
1943 cpu
->core_id
, smp_cores
- 1);
1946 if (cpu
->thread_id
< 0) {
1947 error_setg(errp
, "CPU thread-id is not set");
1949 } else if (cpu
->thread_id
> (smp_threads
- 1)) {
1950 error_setg(errp
, "Invalid CPU thread-id: %u must be in range 0:%u",
1951 cpu
->thread_id
, smp_threads
- 1);
1955 topo
.pkg_id
= cpu
->socket_id
;
1956 topo
.core_id
= cpu
->core_id
;
1957 topo
.smt_id
= cpu
->thread_id
;
1958 cpu
->apic_id
= apicid_from_topo_ids(smp_cores
, smp_threads
, &topo
);
1961 cpu_slot
= pc_find_cpu_slot(MACHINE(pcms
), cpu
->apic_id
, &idx
);
1963 MachineState
*ms
= MACHINE(pcms
);
1965 x86_topo_ids_from_apicid(cpu
->apic_id
, smp_cores
, smp_threads
, &topo
);
1966 error_setg(errp
, "Invalid CPU [socket: %u, core: %u, thread: %u] with"
1967 " APIC ID %" PRIu32
", valid index range 0:%d",
1968 topo
.pkg_id
, topo
.core_id
, topo
.smt_id
, cpu
->apic_id
,
1969 ms
->possible_cpus
->len
- 1);
1973 if (cpu_slot
->cpu
) {
1974 error_setg(errp
, "CPU[%d] with APIC ID %" PRIu32
" exists",
1979 /* if 'address' properties socket-id/core-id/thread-id are not set, set them
1980 * so that machine_query_hotpluggable_cpus would show correct values
1982 /* TODO: move socket_id/core_id/thread_id checks into x86_cpu_realizefn()
1983 * once -smp refactoring is complete and there will be CPU private
1984 * CPUState::nr_cores and CPUState::nr_threads fields instead of globals */
1985 x86_topo_ids_from_apicid(cpu
->apic_id
, smp_cores
, smp_threads
, &topo
);
1986 if (cpu
->socket_id
!= -1 && cpu
->socket_id
!= topo
.pkg_id
) {
1987 error_setg(errp
, "property socket-id: %u doesn't match set apic-id:"
1988 " 0x%x (socket-id: %u)", cpu
->socket_id
, cpu
->apic_id
, topo
.pkg_id
);
1991 cpu
->socket_id
= topo
.pkg_id
;
1993 if (cpu
->core_id
!= -1 && cpu
->core_id
!= topo
.core_id
) {
1994 error_setg(errp
, "property core-id: %u doesn't match set apic-id:"
1995 " 0x%x (core-id: %u)", cpu
->core_id
, cpu
->apic_id
, topo
.core_id
);
1998 cpu
->core_id
= topo
.core_id
;
2000 if (cpu
->thread_id
!= -1 && cpu
->thread_id
!= topo
.smt_id
) {
2001 error_setg(errp
, "property thread-id: %u doesn't match set apic-id:"
2002 " 0x%x (thread-id: %u)", cpu
->thread_id
, cpu
->apic_id
, topo
.smt_id
);
2005 cpu
->thread_id
= topo
.smt_id
;
2008 cs
->cpu_index
= idx
;
2010 numa_cpu_pre_plug(cpu_slot
, dev
, errp
);
2013 static void pc_machine_device_pre_plug_cb(HotplugHandler
*hotplug_dev
,
2014 DeviceState
*dev
, Error
**errp
)
2016 if (object_dynamic_cast(OBJECT(dev
), TYPE_CPU
)) {
2017 pc_cpu_pre_plug(hotplug_dev
, dev
, errp
);
2021 static void pc_machine_device_plug_cb(HotplugHandler
*hotplug_dev
,
2022 DeviceState
*dev
, Error
**errp
)
2024 if (object_dynamic_cast(OBJECT(dev
), TYPE_PC_DIMM
)) {
2025 pc_dimm_plug(hotplug_dev
, dev
, errp
);
2026 } else if (object_dynamic_cast(OBJECT(dev
), TYPE_CPU
)) {
2027 pc_cpu_plug(hotplug_dev
, dev
, errp
);
2031 static void pc_machine_device_unplug_request_cb(HotplugHandler
*hotplug_dev
,
2032 DeviceState
*dev
, Error
**errp
)
2034 if (object_dynamic_cast(OBJECT(dev
), TYPE_PC_DIMM
)) {
2035 pc_dimm_unplug_request(hotplug_dev
, dev
, errp
);
2036 } else if (object_dynamic_cast(OBJECT(dev
), TYPE_CPU
)) {
2037 pc_cpu_unplug_request_cb(hotplug_dev
, dev
, errp
);
2039 error_setg(errp
, "acpi: device unplug request for not supported device"
2040 " type: %s", object_get_typename(OBJECT(dev
)));
2044 static void pc_machine_device_unplug_cb(HotplugHandler
*hotplug_dev
,
2045 DeviceState
*dev
, Error
**errp
)
2047 if (object_dynamic_cast(OBJECT(dev
), TYPE_PC_DIMM
)) {
2048 pc_dimm_unplug(hotplug_dev
, dev
, errp
);
2049 } else if (object_dynamic_cast(OBJECT(dev
), TYPE_CPU
)) {
2050 pc_cpu_unplug_cb(hotplug_dev
, dev
, errp
);
2052 error_setg(errp
, "acpi: device unplug for not supported device"
2053 " type: %s", object_get_typename(OBJECT(dev
)));
2057 static HotplugHandler
*pc_get_hotpug_handler(MachineState
*machine
,
2060 PCMachineClass
*pcmc
= PC_MACHINE_GET_CLASS(machine
);
2062 if (object_dynamic_cast(OBJECT(dev
), TYPE_PC_DIMM
) ||
2063 object_dynamic_cast(OBJECT(dev
), TYPE_CPU
)) {
2064 return HOTPLUG_HANDLER(machine
);
2067 return pcmc
->get_hotplug_handler
?
2068 pcmc
->get_hotplug_handler(machine
, dev
) : NULL
;
2072 pc_machine_get_hotplug_memory_region_size(Object
*obj
, Visitor
*v
,
2073 const char *name
, void *opaque
,
2076 PCMachineState
*pcms
= PC_MACHINE(obj
);
2077 int64_t value
= memory_region_size(&pcms
->hotplug_memory
.mr
);
2079 visit_type_int(v
, name
, &value
, errp
);
2082 static void pc_machine_get_max_ram_below_4g(Object
*obj
, Visitor
*v
,
2083 const char *name
, void *opaque
,
2086 PCMachineState
*pcms
= PC_MACHINE(obj
);
2087 uint64_t value
= pcms
->max_ram_below_4g
;
2089 visit_type_size(v
, name
, &value
, errp
);
2092 static void pc_machine_set_max_ram_below_4g(Object
*obj
, Visitor
*v
,
2093 const char *name
, void *opaque
,
2096 PCMachineState
*pcms
= PC_MACHINE(obj
);
2097 Error
*error
= NULL
;
2100 visit_type_size(v
, name
, &value
, &error
);
2102 error_propagate(errp
, error
);
2105 if (value
> (1ULL << 32)) {
2107 "Machine option 'max-ram-below-4g=%"PRIu64
2108 "' expects size less than or equal to 4G", value
);
2109 error_propagate(errp
, error
);
2113 if (value
< (1ULL << 20)) {
2114 warn_report("Only %" PRIu64
" bytes of RAM below the 4GiB boundary,"
2115 "BIOS may not work with less than 1MiB", value
);
2118 pcms
->max_ram_below_4g
= value
;
2121 static void pc_machine_get_vmport(Object
*obj
, Visitor
*v
, const char *name
,
2122 void *opaque
, Error
**errp
)
2124 PCMachineState
*pcms
= PC_MACHINE(obj
);
2125 OnOffAuto vmport
= pcms
->vmport
;
2127 visit_type_OnOffAuto(v
, name
, &vmport
, errp
);
2130 static void pc_machine_set_vmport(Object
*obj
, Visitor
*v
, const char *name
,
2131 void *opaque
, Error
**errp
)
2133 PCMachineState
*pcms
= PC_MACHINE(obj
);
2135 visit_type_OnOffAuto(v
, name
, &pcms
->vmport
, errp
);
2138 bool pc_machine_is_smm_enabled(PCMachineState
*pcms
)
2140 bool smm_available
= false;
2142 if (pcms
->smm
== ON_OFF_AUTO_OFF
) {
2146 if (tcg_enabled() || qtest_enabled()) {
2147 smm_available
= true;
2148 } else if (kvm_enabled()) {
2149 smm_available
= kvm_has_smm();
2152 if (smm_available
) {
2156 if (pcms
->smm
== ON_OFF_AUTO_ON
) {
2157 error_report("System Management Mode not supported by this hypervisor.");
2163 static void pc_machine_get_smm(Object
*obj
, Visitor
*v
, const char *name
,
2164 void *opaque
, Error
**errp
)
2166 PCMachineState
*pcms
= PC_MACHINE(obj
);
2167 OnOffAuto smm
= pcms
->smm
;
2169 visit_type_OnOffAuto(v
, name
, &smm
, errp
);
2172 static void pc_machine_set_smm(Object
*obj
, Visitor
*v
, const char *name
,
2173 void *opaque
, Error
**errp
)
2175 PCMachineState
*pcms
= PC_MACHINE(obj
);
2177 visit_type_OnOffAuto(v
, name
, &pcms
->smm
, errp
);
2180 static bool pc_machine_get_nvdimm(Object
*obj
, Error
**errp
)
2182 PCMachineState
*pcms
= PC_MACHINE(obj
);
2184 return pcms
->acpi_nvdimm_state
.is_enabled
;
2187 static void pc_machine_set_nvdimm(Object
*obj
, bool value
, Error
**errp
)
2189 PCMachineState
*pcms
= PC_MACHINE(obj
);
2191 pcms
->acpi_nvdimm_state
.is_enabled
= value
;
2194 static bool pc_machine_get_smbus(Object
*obj
, Error
**errp
)
2196 PCMachineState
*pcms
= PC_MACHINE(obj
);
2201 static void pc_machine_set_smbus(Object
*obj
, bool value
, Error
**errp
)
2203 PCMachineState
*pcms
= PC_MACHINE(obj
);
2205 pcms
->smbus
= value
;
2208 static bool pc_machine_get_sata(Object
*obj
, Error
**errp
)
2210 PCMachineState
*pcms
= PC_MACHINE(obj
);
2215 static void pc_machine_set_sata(Object
*obj
, bool value
, Error
**errp
)
2217 PCMachineState
*pcms
= PC_MACHINE(obj
);
2222 static bool pc_machine_get_pit(Object
*obj
, Error
**errp
)
2224 PCMachineState
*pcms
= PC_MACHINE(obj
);
2229 static void pc_machine_set_pit(Object
*obj
, bool value
, Error
**errp
)
2231 PCMachineState
*pcms
= PC_MACHINE(obj
);
2236 static void pc_machine_initfn(Object
*obj
)
2238 PCMachineState
*pcms
= PC_MACHINE(obj
);
2240 pcms
->max_ram_below_4g
= 0; /* use default */
2241 pcms
->smm
= ON_OFF_AUTO_AUTO
;
2242 pcms
->vmport
= ON_OFF_AUTO_AUTO
;
2243 /* nvdimm is disabled on default. */
2244 pcms
->acpi_nvdimm_state
.is_enabled
= false;
2245 /* acpi build is enabled by default if machine supports it */
2246 pcms
->acpi_build_enabled
= PC_MACHINE_GET_CLASS(pcms
)->has_acpi_build
;
2252 static void pc_machine_reset(void)
2257 qemu_devices_reset();
2259 /* Reset APIC after devices have been reset to cancel
2260 * any changes that qemu_devices_reset() might have done.
2265 if (cpu
->apic_state
) {
2266 device_reset(cpu
->apic_state
);
2271 static CpuInstanceProperties
2272 pc_cpu_index_to_props(MachineState
*ms
, unsigned cpu_index
)
2274 MachineClass
*mc
= MACHINE_GET_CLASS(ms
);
2275 const CPUArchIdList
*possible_cpus
= mc
->possible_cpu_arch_ids(ms
);
2277 assert(cpu_index
< possible_cpus
->len
);
2278 return possible_cpus
->cpus
[cpu_index
].props
;
2281 static int64_t pc_get_default_cpu_node_id(const MachineState
*ms
, int idx
)
2283 X86CPUTopoInfo topo
;
2285 assert(idx
< ms
->possible_cpus
->len
);
2286 x86_topo_ids_from_apicid(ms
->possible_cpus
->cpus
[idx
].arch_id
,
2287 smp_cores
, smp_threads
, &topo
);
2288 return topo
.pkg_id
% nb_numa_nodes
;
2291 static const CPUArchIdList
*pc_possible_cpu_arch_ids(MachineState
*ms
)
2295 if (ms
->possible_cpus
) {
2297 * make sure that max_cpus hasn't changed since the first use, i.e.
2298 * -smp hasn't been parsed after it
2300 assert(ms
->possible_cpus
->len
== max_cpus
);
2301 return ms
->possible_cpus
;
2304 ms
->possible_cpus
= g_malloc0(sizeof(CPUArchIdList
) +
2305 sizeof(CPUArchId
) * max_cpus
);
2306 ms
->possible_cpus
->len
= max_cpus
;
2307 for (i
= 0; i
< ms
->possible_cpus
->len
; i
++) {
2308 X86CPUTopoInfo topo
;
2310 ms
->possible_cpus
->cpus
[i
].vcpus_count
= 1;
2311 ms
->possible_cpus
->cpus
[i
].arch_id
= x86_cpu_apic_id_from_index(i
);
2312 x86_topo_ids_from_apicid(ms
->possible_cpus
->cpus
[i
].arch_id
,
2313 smp_cores
, smp_threads
, &topo
);
2314 ms
->possible_cpus
->cpus
[i
].props
.has_socket_id
= true;
2315 ms
->possible_cpus
->cpus
[i
].props
.socket_id
= topo
.pkg_id
;
2316 ms
->possible_cpus
->cpus
[i
].props
.has_core_id
= true;
2317 ms
->possible_cpus
->cpus
[i
].props
.core_id
= topo
.core_id
;
2318 ms
->possible_cpus
->cpus
[i
].props
.has_thread_id
= true;
2319 ms
->possible_cpus
->cpus
[i
].props
.thread_id
= topo
.smt_id
;
2321 return ms
->possible_cpus
;
2324 static void x86_nmi(NMIState
*n
, int cpu_index
, Error
**errp
)
2326 /* cpu index isn't used */
2330 X86CPU
*cpu
= X86_CPU(cs
);
2332 if (!cpu
->apic_state
) {
2333 cpu_interrupt(cs
, CPU_INTERRUPT_NMI
);
2335 apic_deliver_nmi(cpu
->apic_state
);
2340 static void pc_machine_class_init(ObjectClass
*oc
, void *data
)
2342 MachineClass
*mc
= MACHINE_CLASS(oc
);
2343 PCMachineClass
*pcmc
= PC_MACHINE_CLASS(oc
);
2344 HotplugHandlerClass
*hc
= HOTPLUG_HANDLER_CLASS(oc
);
2345 NMIClass
*nc
= NMI_CLASS(oc
);
2347 pcmc
->get_hotplug_handler
= mc
->get_hotplug_handler
;
2348 pcmc
->pci_enabled
= true;
2349 pcmc
->has_acpi_build
= true;
2350 pcmc
->rsdp_in_ram
= true;
2351 pcmc
->smbios_defaults
= true;
2352 pcmc
->smbios_uuid_encoded
= true;
2353 pcmc
->gigabyte_align
= true;
2354 pcmc
->has_reserved_memory
= true;
2355 pcmc
->kvmclock_enabled
= true;
2356 pcmc
->enforce_aligned_dimm
= true;
2357 /* BIOS ACPI tables: 128K. Other BIOS datastructures: less than 4K reported
2358 * to be used at the moment, 32K should be enough for a while. */
2359 pcmc
->acpi_data_size
= 0x20000 + 0x8000;
2360 pcmc
->save_tsc_khz
= true;
2361 pcmc
->linuxboot_dma_enabled
= true;
2362 mc
->get_hotplug_handler
= pc_get_hotpug_handler
;
2363 mc
->cpu_index_to_instance_props
= pc_cpu_index_to_props
;
2364 mc
->get_default_cpu_node_id
= pc_get_default_cpu_node_id
;
2365 mc
->possible_cpu_arch_ids
= pc_possible_cpu_arch_ids
;
2366 mc
->auto_enable_numa_with_memhp
= true;
2367 mc
->has_hotpluggable_cpus
= true;
2368 mc
->default_boot_order
= "cad";
2369 mc
->hot_add_cpu
= pc_hot_add_cpu
;
2370 mc
->block_default_type
= IF_IDE
;
2372 mc
->reset
= pc_machine_reset
;
2373 hc
->pre_plug
= pc_machine_device_pre_plug_cb
;
2374 hc
->plug
= pc_machine_device_plug_cb
;
2375 hc
->unplug_request
= pc_machine_device_unplug_request_cb
;
2376 hc
->unplug
= pc_machine_device_unplug_cb
;
2377 nc
->nmi_monitor_handler
= x86_nmi
;
2378 mc
->default_cpu_type
= TARGET_DEFAULT_CPU_TYPE
;
2380 object_class_property_add(oc
, PC_MACHINE_MEMHP_REGION_SIZE
, "int",
2381 pc_machine_get_hotplug_memory_region_size
, NULL
,
2382 NULL
, NULL
, &error_abort
);
2384 object_class_property_add(oc
, PC_MACHINE_MAX_RAM_BELOW_4G
, "size",
2385 pc_machine_get_max_ram_below_4g
, pc_machine_set_max_ram_below_4g
,
2386 NULL
, NULL
, &error_abort
);
2388 object_class_property_set_description(oc
, PC_MACHINE_MAX_RAM_BELOW_4G
,
2389 "Maximum ram below the 4G boundary (32bit boundary)", &error_abort
);
2391 object_class_property_add(oc
, PC_MACHINE_SMM
, "OnOffAuto",
2392 pc_machine_get_smm
, pc_machine_set_smm
,
2393 NULL
, NULL
, &error_abort
);
2394 object_class_property_set_description(oc
, PC_MACHINE_SMM
,
2395 "Enable SMM (pc & q35)", &error_abort
);
2397 object_class_property_add(oc
, PC_MACHINE_VMPORT
, "OnOffAuto",
2398 pc_machine_get_vmport
, pc_machine_set_vmport
,
2399 NULL
, NULL
, &error_abort
);
2400 object_class_property_set_description(oc
, PC_MACHINE_VMPORT
,
2401 "Enable vmport (pc & q35)", &error_abort
);
2403 object_class_property_add_bool(oc
, PC_MACHINE_NVDIMM
,
2404 pc_machine_get_nvdimm
, pc_machine_set_nvdimm
, &error_abort
);
2406 object_class_property_add_bool(oc
, PC_MACHINE_SMBUS
,
2407 pc_machine_get_smbus
, pc_machine_set_smbus
, &error_abort
);
2409 object_class_property_add_bool(oc
, PC_MACHINE_SATA
,
2410 pc_machine_get_sata
, pc_machine_set_sata
, &error_abort
);
2412 object_class_property_add_bool(oc
, PC_MACHINE_PIT
,
2413 pc_machine_get_pit
, pc_machine_set_pit
, &error_abort
);
2416 static const TypeInfo pc_machine_info
= {
2417 .name
= TYPE_PC_MACHINE
,
2418 .parent
= TYPE_MACHINE
,
2420 .instance_size
= sizeof(PCMachineState
),
2421 .instance_init
= pc_machine_initfn
,
2422 .class_size
= sizeof(PCMachineClass
),
2423 .class_init
= pc_machine_class_init
,
2424 .interfaces
= (InterfaceInfo
[]) {
2425 { TYPE_HOTPLUG_HANDLER
},
2431 static void pc_machine_register_types(void)
2433 type_register_static(&pc_machine_info
);
2436 type_init(pc_machine_register_types
)