s390x: upgrade status of KVM cores to "supported"
[qemu/ar7.git] / hw / i386 / pc.c
blob3889eccdc3225f2944b7a3b9f6679b24ac11110e
1 /*
2 * QEMU PC System Emulator
4 * Copyright (c) 2003-2004 Fabrice Bellard
6 * Permission is hereby granted, free of charge, to any person obtaining a copy
7 * of this software and associated documentation files (the "Software"), to deal
8 * in the Software without restriction, including without limitation the rights
9 * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
10 * copies of the Software, and to permit persons to whom the Software is
11 * furnished to do so, subject to the following conditions:
13 * The above copyright notice and this permission notice shall be included in
14 * all copies or substantial portions of the Software.
16 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
17 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
18 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
19 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
20 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
21 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
22 * THE SOFTWARE.
25 #include "qemu/osdep.h"
26 #include "qemu/units.h"
27 #include "hw/hw.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"
35 #include "hw/ide.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"
42 #include "elf.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"
55 #include "kvm_i386.h"
56 #include "hw/xen/xen.h"
57 #include "hw/xen/start_info.h"
58 #include "ui/qemu-spice.h"
59 #include "exec/memory.h"
60 #include "exec/address-spaces.h"
61 #include "sysemu/arch_init.h"
62 #include "qemu/bitmap.h"
63 #include "qemu/config-file.h"
64 #include "qemu/error-report.h"
65 #include "qemu/option.h"
66 #include "hw/acpi/acpi.h"
67 #include "hw/acpi/cpu_hotplug.h"
68 #include "hw/boards.h"
69 #include "acpi-build.h"
70 #include "hw/mem/pc-dimm.h"
71 #include "qapi/error.h"
72 #include "qapi/qapi-visit-common.h"
73 #include "qapi/visitor.h"
74 #include "qom/cpu.h"
75 #include "hw/nmi.h"
76 #include "hw/usb.h"
77 #include "hw/i386/intel_iommu.h"
78 #include "hw/net/ne2000-isa.h"
79 #include "standard-headers/asm-x86/bootparam.h"
81 /* debug PC/ISA interrupts */
82 //#define DEBUG_IRQ
84 #ifdef DEBUG_IRQ
85 #define DPRINTF(fmt, ...) \
86 do { printf("CPUIRQ: " fmt , ## __VA_ARGS__); } while (0)
87 #else
88 #define DPRINTF(fmt, ...)
89 #endif
91 #define FW_CFG_ACPI_TABLES (FW_CFG_ARCH_LOCAL + 0)
92 #define FW_CFG_SMBIOS_ENTRIES (FW_CFG_ARCH_LOCAL + 1)
93 #define FW_CFG_IRQ0_OVERRIDE (FW_CFG_ARCH_LOCAL + 2)
94 #define FW_CFG_E820_TABLE (FW_CFG_ARCH_LOCAL + 3)
95 #define FW_CFG_HPET (FW_CFG_ARCH_LOCAL + 4)
97 #define E820_NR_ENTRIES 16
99 struct e820_entry {
100 uint64_t address;
101 uint64_t length;
102 uint32_t type;
103 } QEMU_PACKED __attribute((__aligned__(4)));
105 struct e820_table {
106 uint32_t count;
107 struct e820_entry entry[E820_NR_ENTRIES];
108 } QEMU_PACKED __attribute((__aligned__(4)));
110 static struct e820_table e820_reserve;
111 static struct e820_entry *e820_table;
112 static unsigned e820_entries;
113 struct hpet_fw_config hpet_cfg = {.count = UINT8_MAX};
115 /* Physical Address of PVH entry point read from kernel ELF NOTE */
116 static size_t pvh_start_addr;
118 GlobalProperty pc_compat_3_1[] = {
119 { "intel-iommu", "dma-drain", "off" },
120 { "Opteron_G3" "-" TYPE_X86_CPU, "rdtscp", "off" },
121 { "Opteron_G4" "-" TYPE_X86_CPU, "rdtscp", "off" },
122 { "Opteron_G4" "-" TYPE_X86_CPU, "npt", "off" },
123 { "Opteron_G4" "-" TYPE_X86_CPU, "nrip-save", "off" },
124 { "Opteron_G5" "-" TYPE_X86_CPU, "rdtscp", "off" },
125 { "Opteron_G5" "-" TYPE_X86_CPU, "npt", "off" },
126 { "Opteron_G5" "-" TYPE_X86_CPU, "nrip-save", "off" },
127 { "EPYC" "-" TYPE_X86_CPU, "npt", "off" },
128 { "EPYC" "-" TYPE_X86_CPU, "nrip-save", "off" },
129 { "EPYC-IBPB" "-" TYPE_X86_CPU, "npt", "off" },
130 { "EPYC-IBPB" "-" TYPE_X86_CPU, "nrip-save", "off" },
131 { "Skylake-Client" "-" TYPE_X86_CPU, "mpx", "on" },
132 { "Skylake-Client-IBRS" "-" TYPE_X86_CPU, "mpx", "on" },
133 { "Skylake-Server" "-" TYPE_X86_CPU, "mpx", "on" },
134 { "Skylake-Server-IBRS" "-" TYPE_X86_CPU, "mpx", "on" },
135 { "Cascadelake-Server" "-" TYPE_X86_CPU, "mpx", "on" },
136 { "Icelake-Client" "-" TYPE_X86_CPU, "mpx", "on" },
137 { "Icelake-Server" "-" TYPE_X86_CPU, "mpx", "on" },
138 { "Cascadelake-Server" "-" TYPE_X86_CPU, "stepping", "5" },
140 const size_t pc_compat_3_1_len = G_N_ELEMENTS(pc_compat_3_1);
142 GlobalProperty pc_compat_3_0[] = {
143 { TYPE_X86_CPU, "x-hv-synic-kvm-only", "on" },
144 { "Skylake-Server" "-" TYPE_X86_CPU, "pku", "off" },
145 { "Skylake-Server-IBRS" "-" TYPE_X86_CPU, "pku", "off" },
147 const size_t pc_compat_3_0_len = G_N_ELEMENTS(pc_compat_3_0);
149 GlobalProperty pc_compat_2_12[] = {
150 { TYPE_X86_CPU, "legacy-cache", "on" },
151 { TYPE_X86_CPU, "topoext", "off" },
152 { "EPYC-" TYPE_X86_CPU, "xlevel", "0x8000000a" },
153 { "EPYC-IBPB-" TYPE_X86_CPU, "xlevel", "0x8000000a" },
155 const size_t pc_compat_2_12_len = G_N_ELEMENTS(pc_compat_2_12);
157 GlobalProperty pc_compat_2_11[] = {
158 { TYPE_X86_CPU, "x-migrate-smi-count", "off" },
159 { "Skylake-Server" "-" TYPE_X86_CPU, "clflushopt", "off" },
161 const size_t pc_compat_2_11_len = G_N_ELEMENTS(pc_compat_2_11);
163 GlobalProperty pc_compat_2_10[] = {
164 { TYPE_X86_CPU, "x-hv-max-vps", "0x40" },
165 { "i440FX-pcihost", "x-pci-hole64-fix", "off" },
166 { "q35-pcihost", "x-pci-hole64-fix", "off" },
168 const size_t pc_compat_2_10_len = G_N_ELEMENTS(pc_compat_2_10);
170 GlobalProperty pc_compat_2_9[] = {
171 { "mch", "extended-tseg-mbytes", "0" },
173 const size_t pc_compat_2_9_len = G_N_ELEMENTS(pc_compat_2_9);
175 GlobalProperty pc_compat_2_8[] = {
176 { TYPE_X86_CPU, "tcg-cpuid", "off" },
177 { "kvmclock", "x-mach-use-reliable-get-clock", "off" },
178 { "ICH9-LPC", "x-smi-broadcast", "off" },
179 { TYPE_X86_CPU, "vmware-cpuid-freq", "off" },
180 { "Haswell-" TYPE_X86_CPU, "stepping", "1" },
182 const size_t pc_compat_2_8_len = G_N_ELEMENTS(pc_compat_2_8);
184 GlobalProperty pc_compat_2_7[] = {
185 { TYPE_X86_CPU, "l3-cache", "off" },
186 { TYPE_X86_CPU, "full-cpuid-auto-level", "off" },
187 { "Opteron_G3" "-" TYPE_X86_CPU, "family", "15" },
188 { "Opteron_G3" "-" TYPE_X86_CPU, "model", "6" },
189 { "Opteron_G3" "-" TYPE_X86_CPU, "stepping", "1" },
190 { "isa-pcspk", "migrate", "off" },
192 const size_t pc_compat_2_7_len = G_N_ELEMENTS(pc_compat_2_7);
194 GlobalProperty pc_compat_2_6[] = {
195 { TYPE_X86_CPU, "cpuid-0xb", "off" },
196 { "vmxnet3", "romfile", "" },
197 { TYPE_X86_CPU, "fill-mtrr-mask", "off" },
198 { "apic-common", "legacy-instance-id", "on", }
200 const size_t pc_compat_2_6_len = G_N_ELEMENTS(pc_compat_2_6);
202 GlobalProperty pc_compat_2_5[] = {};
203 const size_t pc_compat_2_5_len = G_N_ELEMENTS(pc_compat_2_5);
205 GlobalProperty pc_compat_2_4[] = {
206 PC_CPU_MODEL_IDS("2.4.0")
207 { "Haswell-" TYPE_X86_CPU, "abm", "off" },
208 { "Haswell-noTSX-" TYPE_X86_CPU, "abm", "off" },
209 { "Broadwell-" TYPE_X86_CPU, "abm", "off" },
210 { "Broadwell-noTSX-" TYPE_X86_CPU, "abm", "off" },
211 { "host" "-" TYPE_X86_CPU, "host-cache-info", "on" },
212 { TYPE_X86_CPU, "check", "off" },
213 { "qemu64" "-" TYPE_X86_CPU, "sse4a", "on" },
214 { "qemu64" "-" TYPE_X86_CPU, "abm", "on" },
215 { "qemu64" "-" TYPE_X86_CPU, "popcnt", "on" },
216 { "qemu32" "-" TYPE_X86_CPU, "popcnt", "on" },
217 { "Opteron_G2" "-" TYPE_X86_CPU, "rdtscp", "on" },
218 { "Opteron_G3" "-" TYPE_X86_CPU, "rdtscp", "on" },
219 { "Opteron_G4" "-" TYPE_X86_CPU, "rdtscp", "on" },
220 { "Opteron_G5" "-" TYPE_X86_CPU, "rdtscp", "on", }
222 const size_t pc_compat_2_4_len = G_N_ELEMENTS(pc_compat_2_4);
224 GlobalProperty pc_compat_2_3[] = {
225 PC_CPU_MODEL_IDS("2.3.0")
226 { TYPE_X86_CPU, "arat", "off" },
227 { "qemu64" "-" TYPE_X86_CPU, "min-level", "4" },
228 { "kvm64" "-" TYPE_X86_CPU, "min-level", "5" },
229 { "pentium3" "-" TYPE_X86_CPU, "min-level", "2" },
230 { "n270" "-" TYPE_X86_CPU, "min-level", "5" },
231 { "Conroe" "-" TYPE_X86_CPU, "min-level", "4" },
232 { "Penryn" "-" TYPE_X86_CPU, "min-level", "4" },
233 { "Nehalem" "-" TYPE_X86_CPU, "min-level", "4" },
234 { "n270" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
235 { "Penryn" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
236 { "Conroe" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
237 { "Nehalem" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
238 { "Westmere" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
239 { "SandyBridge" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
240 { "IvyBridge" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
241 { "Haswell" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
242 { "Haswell-noTSX" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
243 { "Broadwell" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
244 { "Broadwell-noTSX" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
245 { TYPE_X86_CPU, "kvm-no-smi-migration", "on" },
247 const size_t pc_compat_2_3_len = G_N_ELEMENTS(pc_compat_2_3);
249 GlobalProperty pc_compat_2_2[] = {
250 PC_CPU_MODEL_IDS("2.2.0")
251 { "kvm64" "-" TYPE_X86_CPU, "vme", "off" },
252 { "kvm32" "-" TYPE_X86_CPU, "vme", "off" },
253 { "Conroe" "-" TYPE_X86_CPU, "vme", "off" },
254 { "Penryn" "-" TYPE_X86_CPU, "vme", "off" },
255 { "Nehalem" "-" TYPE_X86_CPU, "vme", "off" },
256 { "Westmere" "-" TYPE_X86_CPU, "vme", "off" },
257 { "SandyBridge" "-" TYPE_X86_CPU, "vme", "off" },
258 { "Haswell" "-" TYPE_X86_CPU, "vme", "off" },
259 { "Broadwell" "-" TYPE_X86_CPU, "vme", "off" },
260 { "Opteron_G1" "-" TYPE_X86_CPU, "vme", "off" },
261 { "Opteron_G2" "-" TYPE_X86_CPU, "vme", "off" },
262 { "Opteron_G3" "-" TYPE_X86_CPU, "vme", "off" },
263 { "Opteron_G4" "-" TYPE_X86_CPU, "vme", "off" },
264 { "Opteron_G5" "-" TYPE_X86_CPU, "vme", "off" },
265 { "Haswell" "-" TYPE_X86_CPU, "f16c", "off" },
266 { "Haswell" "-" TYPE_X86_CPU, "rdrand", "off" },
267 { "Broadwell" "-" TYPE_X86_CPU, "f16c", "off" },
268 { "Broadwell" "-" TYPE_X86_CPU, "rdrand", "off" },
270 const size_t pc_compat_2_2_len = G_N_ELEMENTS(pc_compat_2_2);
272 GlobalProperty pc_compat_2_1[] = {
273 PC_CPU_MODEL_IDS("2.1.0")
274 { "coreduo" "-" TYPE_X86_CPU, "vmx", "on" },
275 { "core2duo" "-" TYPE_X86_CPU, "vmx", "on" },
277 const size_t pc_compat_2_1_len = G_N_ELEMENTS(pc_compat_2_1);
279 GlobalProperty pc_compat_2_0[] = {
280 PC_CPU_MODEL_IDS("2.0.0")
281 { "virtio-scsi-pci", "any_layout", "off" },
282 { "PIIX4_PM", "memory-hotplug-support", "off" },
283 { "apic", "version", "0x11" },
284 { "nec-usb-xhci", "superspeed-ports-first", "off" },
285 { "nec-usb-xhci", "force-pcie-endcap", "on" },
286 { "pci-serial", "prog_if", "0" },
287 { "pci-serial-2x", "prog_if", "0" },
288 { "pci-serial-4x", "prog_if", "0" },
289 { "virtio-net-pci", "guest_announce", "off" },
290 { "ICH9-LPC", "memory-hotplug-support", "off" },
291 { "xio3130-downstream", COMPAT_PROP_PCP, "off" },
292 { "ioh3420", COMPAT_PROP_PCP, "off" },
294 const size_t pc_compat_2_0_len = G_N_ELEMENTS(pc_compat_2_0);
296 GlobalProperty pc_compat_1_7[] = {
297 PC_CPU_MODEL_IDS("1.7.0")
298 { TYPE_USB_DEVICE, "msos-desc", "no" },
299 { "PIIX4_PM", "acpi-pci-hotplug-with-bridge-support", "off" },
300 { "hpet", HPET_INTCAP, "4" },
302 const size_t pc_compat_1_7_len = G_N_ELEMENTS(pc_compat_1_7);
304 GlobalProperty pc_compat_1_6[] = {
305 PC_CPU_MODEL_IDS("1.6.0")
306 { "e1000", "mitigation", "off" },
307 { "qemu64-" TYPE_X86_CPU, "model", "2" },
308 { "qemu32-" TYPE_X86_CPU, "model", "3" },
309 { "i440FX-pcihost", "short_root_bus", "1" },
310 { "q35-pcihost", "short_root_bus", "1" },
312 const size_t pc_compat_1_6_len = G_N_ELEMENTS(pc_compat_1_6);
314 GlobalProperty pc_compat_1_5[] = {
315 PC_CPU_MODEL_IDS("1.5.0")
316 { "Conroe-" TYPE_X86_CPU, "model", "2" },
317 { "Conroe-" TYPE_X86_CPU, "min-level", "2" },
318 { "Penryn-" TYPE_X86_CPU, "model", "2" },
319 { "Penryn-" TYPE_X86_CPU, "min-level", "2" },
320 { "Nehalem-" TYPE_X86_CPU, "model", "2" },
321 { "Nehalem-" TYPE_X86_CPU, "min-level", "2" },
322 { "virtio-net-pci", "any_layout", "off" },
323 { TYPE_X86_CPU, "pmu", "on" },
324 { "i440FX-pcihost", "short_root_bus", "0" },
325 { "q35-pcihost", "short_root_bus", "0" },
327 const size_t pc_compat_1_5_len = G_N_ELEMENTS(pc_compat_1_5);
329 GlobalProperty pc_compat_1_4[] = {
330 PC_CPU_MODEL_IDS("1.4.0")
331 { "scsi-hd", "discard_granularity", "0" },
332 { "scsi-cd", "discard_granularity", "0" },
333 { "scsi-disk", "discard_granularity", "0" },
334 { "ide-hd", "discard_granularity", "0" },
335 { "ide-cd", "discard_granularity", "0" },
336 { "ide-drive", "discard_granularity", "0" },
337 { "virtio-blk-pci", "discard_granularity", "0" },
338 /* DEV_NVECTORS_UNSPECIFIED as a uint32_t string: */
339 { "virtio-serial-pci", "vectors", "0xFFFFFFFF" },
340 { "virtio-net-pci", "ctrl_guest_offloads", "off" },
341 { "e1000", "romfile", "pxe-e1000.rom" },
342 { "ne2k_pci", "romfile", "pxe-ne2k_pci.rom" },
343 { "pcnet", "romfile", "pxe-pcnet.rom" },
344 { "rtl8139", "romfile", "pxe-rtl8139.rom" },
345 { "virtio-net-pci", "romfile", "pxe-virtio.rom" },
346 { "486-" TYPE_X86_CPU, "model", "0" },
347 { "n270" "-" TYPE_X86_CPU, "movbe", "off" },
348 { "Westmere" "-" TYPE_X86_CPU, "pclmulqdq", "off" },
350 const size_t pc_compat_1_4_len = G_N_ELEMENTS(pc_compat_1_4);
352 void gsi_handler(void *opaque, int n, int level)
354 GSIState *s = opaque;
356 DPRINTF("pc: %s GSI %d\n", level ? "raising" : "lowering", n);
357 if (n < ISA_NUM_IRQS) {
358 qemu_set_irq(s->i8259_irq[n], level);
360 qemu_set_irq(s->ioapic_irq[n], level);
363 static void ioport80_write(void *opaque, hwaddr addr, uint64_t data,
364 unsigned size)
368 static uint64_t ioport80_read(void *opaque, hwaddr addr, unsigned size)
370 return 0xffffffffffffffffULL;
373 /* MSDOS compatibility mode FPU exception support */
374 static qemu_irq ferr_irq;
376 void pc_register_ferr_irq(qemu_irq irq)
378 ferr_irq = irq;
381 /* XXX: add IGNNE support */
382 void cpu_set_ferr(CPUX86State *s)
384 qemu_irq_raise(ferr_irq);
387 static void ioportF0_write(void *opaque, hwaddr addr, uint64_t data,
388 unsigned size)
390 qemu_irq_lower(ferr_irq);
393 static uint64_t ioportF0_read(void *opaque, hwaddr addr, unsigned size)
395 return 0xffffffffffffffffULL;
398 /* TSC handling */
399 uint64_t cpu_get_tsc(CPUX86State *env)
401 return cpu_get_ticks();
404 /* IRQ handling */
405 int cpu_get_pic_interrupt(CPUX86State *env)
407 X86CPU *cpu = x86_env_get_cpu(env);
408 int intno;
410 if (!kvm_irqchip_in_kernel()) {
411 intno = apic_get_interrupt(cpu->apic_state);
412 if (intno >= 0) {
413 return intno;
415 /* read the irq from the PIC */
416 if (!apic_accept_pic_intr(cpu->apic_state)) {
417 return -1;
421 intno = pic_read_irq(isa_pic);
422 return intno;
425 static void pic_irq_request(void *opaque, int irq, int level)
427 CPUState *cs = first_cpu;
428 X86CPU *cpu = X86_CPU(cs);
430 DPRINTF("pic_irqs: %s irq %d\n", level? "raise" : "lower", irq);
431 if (cpu->apic_state && !kvm_irqchip_in_kernel()) {
432 CPU_FOREACH(cs) {
433 cpu = X86_CPU(cs);
434 if (apic_accept_pic_intr(cpu->apic_state)) {
435 apic_deliver_pic_intr(cpu->apic_state, level);
438 } else {
439 if (level) {
440 cpu_interrupt(cs, CPU_INTERRUPT_HARD);
441 } else {
442 cpu_reset_interrupt(cs, CPU_INTERRUPT_HARD);
447 /* PC cmos mappings */
449 #define REG_EQUIPMENT_BYTE 0x14
451 int cmos_get_fd_drive_type(FloppyDriveType fd0)
453 int val;
455 switch (fd0) {
456 case FLOPPY_DRIVE_TYPE_144:
457 /* 1.44 Mb 3"5 drive */
458 val = 4;
459 break;
460 case FLOPPY_DRIVE_TYPE_288:
461 /* 2.88 Mb 3"5 drive */
462 val = 5;
463 break;
464 case FLOPPY_DRIVE_TYPE_120:
465 /* 1.2 Mb 5"5 drive */
466 val = 2;
467 break;
468 case FLOPPY_DRIVE_TYPE_NONE:
469 default:
470 val = 0;
471 break;
473 return val;
476 static void cmos_init_hd(ISADevice *s, int type_ofs, int info_ofs,
477 int16_t cylinders, int8_t heads, int8_t sectors)
479 rtc_set_memory(s, type_ofs, 47);
480 rtc_set_memory(s, info_ofs, cylinders);
481 rtc_set_memory(s, info_ofs + 1, cylinders >> 8);
482 rtc_set_memory(s, info_ofs + 2, heads);
483 rtc_set_memory(s, info_ofs + 3, 0xff);
484 rtc_set_memory(s, info_ofs + 4, 0xff);
485 rtc_set_memory(s, info_ofs + 5, 0xc0 | ((heads > 8) << 3));
486 rtc_set_memory(s, info_ofs + 6, cylinders);
487 rtc_set_memory(s, info_ofs + 7, cylinders >> 8);
488 rtc_set_memory(s, info_ofs + 8, sectors);
491 /* convert boot_device letter to something recognizable by the bios */
492 static int boot_device2nibble(char boot_device)
494 switch(boot_device) {
495 case 'a':
496 case 'b':
497 return 0x01; /* floppy boot */
498 case 'c':
499 return 0x02; /* hard drive boot */
500 case 'd':
501 return 0x03; /* CD-ROM boot */
502 case 'n':
503 return 0x04; /* Network boot */
505 return 0;
508 static void set_boot_dev(ISADevice *s, const char *boot_device, Error **errp)
510 #define PC_MAX_BOOT_DEVICES 3
511 int nbds, bds[3] = { 0, };
512 int i;
514 nbds = strlen(boot_device);
515 if (nbds > PC_MAX_BOOT_DEVICES) {
516 error_setg(errp, "Too many boot devices for PC");
517 return;
519 for (i = 0; i < nbds; i++) {
520 bds[i] = boot_device2nibble(boot_device[i]);
521 if (bds[i] == 0) {
522 error_setg(errp, "Invalid boot device for PC: '%c'",
523 boot_device[i]);
524 return;
527 rtc_set_memory(s, 0x3d, (bds[1] << 4) | bds[0]);
528 rtc_set_memory(s, 0x38, (bds[2] << 4) | (fd_bootchk ? 0x0 : 0x1));
531 static void pc_boot_set(void *opaque, const char *boot_device, Error **errp)
533 set_boot_dev(opaque, boot_device, errp);
536 static void pc_cmos_init_floppy(ISADevice *rtc_state, ISADevice *floppy)
538 int val, nb, i;
539 FloppyDriveType fd_type[2] = { FLOPPY_DRIVE_TYPE_NONE,
540 FLOPPY_DRIVE_TYPE_NONE };
542 /* floppy type */
543 if (floppy) {
544 for (i = 0; i < 2; i++) {
545 fd_type[i] = isa_fdc_get_drive_type(floppy, i);
548 val = (cmos_get_fd_drive_type(fd_type[0]) << 4) |
549 cmos_get_fd_drive_type(fd_type[1]);
550 rtc_set_memory(rtc_state, 0x10, val);
552 val = rtc_get_memory(rtc_state, REG_EQUIPMENT_BYTE);
553 nb = 0;
554 if (fd_type[0] != FLOPPY_DRIVE_TYPE_NONE) {
555 nb++;
557 if (fd_type[1] != FLOPPY_DRIVE_TYPE_NONE) {
558 nb++;
560 switch (nb) {
561 case 0:
562 break;
563 case 1:
564 val |= 0x01; /* 1 drive, ready for boot */
565 break;
566 case 2:
567 val |= 0x41; /* 2 drives, ready for boot */
568 break;
570 rtc_set_memory(rtc_state, REG_EQUIPMENT_BYTE, val);
573 typedef struct pc_cmos_init_late_arg {
574 ISADevice *rtc_state;
575 BusState *idebus[2];
576 } pc_cmos_init_late_arg;
578 typedef struct check_fdc_state {
579 ISADevice *floppy;
580 bool multiple;
581 } CheckFdcState;
583 static int check_fdc(Object *obj, void *opaque)
585 CheckFdcState *state = opaque;
586 Object *fdc;
587 uint32_t iobase;
588 Error *local_err = NULL;
590 fdc = object_dynamic_cast(obj, TYPE_ISA_FDC);
591 if (!fdc) {
592 return 0;
595 iobase = object_property_get_uint(obj, "iobase", &local_err);
596 if (local_err || iobase != 0x3f0) {
597 error_free(local_err);
598 return 0;
601 if (state->floppy) {
602 state->multiple = true;
603 } else {
604 state->floppy = ISA_DEVICE(obj);
606 return 0;
609 static const char * const fdc_container_path[] = {
610 "/unattached", "/peripheral", "/peripheral-anon"
614 * Locate the FDC at IO address 0x3f0, in order to configure the CMOS registers
615 * and ACPI objects.
617 ISADevice *pc_find_fdc0(void)
619 int i;
620 Object *container;
621 CheckFdcState state = { 0 };
623 for (i = 0; i < ARRAY_SIZE(fdc_container_path); i++) {
624 container = container_get(qdev_get_machine(), fdc_container_path[i]);
625 object_child_foreach(container, check_fdc, &state);
628 if (state.multiple) {
629 warn_report("multiple floppy disk controllers with "
630 "iobase=0x3f0 have been found");
631 error_printf("the one being picked for CMOS setup might not reflect "
632 "your intent");
635 return state.floppy;
638 static void pc_cmos_init_late(void *opaque)
640 pc_cmos_init_late_arg *arg = opaque;
641 ISADevice *s = arg->rtc_state;
642 int16_t cylinders;
643 int8_t heads, sectors;
644 int val;
645 int i, trans;
647 val = 0;
648 if (arg->idebus[0] && ide_get_geometry(arg->idebus[0], 0,
649 &cylinders, &heads, &sectors) >= 0) {
650 cmos_init_hd(s, 0x19, 0x1b, cylinders, heads, sectors);
651 val |= 0xf0;
653 if (arg->idebus[0] && ide_get_geometry(arg->idebus[0], 1,
654 &cylinders, &heads, &sectors) >= 0) {
655 cmos_init_hd(s, 0x1a, 0x24, cylinders, heads, sectors);
656 val |= 0x0f;
658 rtc_set_memory(s, 0x12, val);
660 val = 0;
661 for (i = 0; i < 4; i++) {
662 /* NOTE: ide_get_geometry() returns the physical
663 geometry. It is always such that: 1 <= sects <= 63, 1
664 <= heads <= 16, 1 <= cylinders <= 16383. The BIOS
665 geometry can be different if a translation is done. */
666 if (arg->idebus[i / 2] &&
667 ide_get_geometry(arg->idebus[i / 2], i % 2,
668 &cylinders, &heads, &sectors) >= 0) {
669 trans = ide_get_bios_chs_trans(arg->idebus[i / 2], i % 2) - 1;
670 assert((trans & ~3) == 0);
671 val |= trans << (i * 2);
674 rtc_set_memory(s, 0x39, val);
676 pc_cmos_init_floppy(s, pc_find_fdc0());
678 qemu_unregister_reset(pc_cmos_init_late, opaque);
681 void pc_cmos_init(PCMachineState *pcms,
682 BusState *idebus0, BusState *idebus1,
683 ISADevice *s)
685 int val;
686 static pc_cmos_init_late_arg arg;
688 /* various important CMOS locations needed by PC/Bochs bios */
690 /* memory size */
691 /* base memory (first MiB) */
692 val = MIN(pcms->below_4g_mem_size / KiB, 640);
693 rtc_set_memory(s, 0x15, val);
694 rtc_set_memory(s, 0x16, val >> 8);
695 /* extended memory (next 64MiB) */
696 if (pcms->below_4g_mem_size > 1 * MiB) {
697 val = (pcms->below_4g_mem_size - 1 * MiB) / KiB;
698 } else {
699 val = 0;
701 if (val > 65535)
702 val = 65535;
703 rtc_set_memory(s, 0x17, val);
704 rtc_set_memory(s, 0x18, val >> 8);
705 rtc_set_memory(s, 0x30, val);
706 rtc_set_memory(s, 0x31, val >> 8);
707 /* memory between 16MiB and 4GiB */
708 if (pcms->below_4g_mem_size > 16 * MiB) {
709 val = (pcms->below_4g_mem_size - 16 * MiB) / (64 * KiB);
710 } else {
711 val = 0;
713 if (val > 65535)
714 val = 65535;
715 rtc_set_memory(s, 0x34, val);
716 rtc_set_memory(s, 0x35, val >> 8);
717 /* memory above 4GiB */
718 val = pcms->above_4g_mem_size / 65536;
719 rtc_set_memory(s, 0x5b, val);
720 rtc_set_memory(s, 0x5c, val >> 8);
721 rtc_set_memory(s, 0x5d, val >> 16);
723 object_property_add_link(OBJECT(pcms), "rtc_state",
724 TYPE_ISA_DEVICE,
725 (Object **)&pcms->rtc,
726 object_property_allow_set_link,
727 OBJ_PROP_LINK_STRONG, &error_abort);
728 object_property_set_link(OBJECT(pcms), OBJECT(s),
729 "rtc_state", &error_abort);
731 set_boot_dev(s, MACHINE(pcms)->boot_order, &error_fatal);
733 val = 0;
734 val |= 0x02; /* FPU is there */
735 val |= 0x04; /* PS/2 mouse installed */
736 rtc_set_memory(s, REG_EQUIPMENT_BYTE, val);
738 /* hard drives and FDC */
739 arg.rtc_state = s;
740 arg.idebus[0] = idebus0;
741 arg.idebus[1] = idebus1;
742 qemu_register_reset(pc_cmos_init_late, &arg);
745 #define TYPE_PORT92 "port92"
746 #define PORT92(obj) OBJECT_CHECK(Port92State, (obj), TYPE_PORT92)
748 /* port 92 stuff: could be split off */
749 typedef struct Port92State {
750 ISADevice parent_obj;
752 MemoryRegion io;
753 uint8_t outport;
754 qemu_irq a20_out;
755 } Port92State;
757 static void port92_write(void *opaque, hwaddr addr, uint64_t val,
758 unsigned size)
760 Port92State *s = opaque;
761 int oldval = s->outport;
763 DPRINTF("port92: write 0x%02" PRIx64 "\n", val);
764 s->outport = val;
765 qemu_set_irq(s->a20_out, (val >> 1) & 1);
766 if ((val & 1) && !(oldval & 1)) {
767 qemu_system_reset_request(SHUTDOWN_CAUSE_GUEST_RESET);
771 static uint64_t port92_read(void *opaque, hwaddr addr,
772 unsigned size)
774 Port92State *s = opaque;
775 uint32_t ret;
777 ret = s->outport;
778 DPRINTF("port92: read 0x%02x\n", ret);
779 return ret;
782 static void port92_init(ISADevice *dev, qemu_irq a20_out)
784 qdev_connect_gpio_out_named(DEVICE(dev), PORT92_A20_LINE, 0, a20_out);
787 static const VMStateDescription vmstate_port92_isa = {
788 .name = "port92",
789 .version_id = 1,
790 .minimum_version_id = 1,
791 .fields = (VMStateField[]) {
792 VMSTATE_UINT8(outport, Port92State),
793 VMSTATE_END_OF_LIST()
797 static void port92_reset(DeviceState *d)
799 Port92State *s = PORT92(d);
801 s->outport &= ~1;
804 static const MemoryRegionOps port92_ops = {
805 .read = port92_read,
806 .write = port92_write,
807 .impl = {
808 .min_access_size = 1,
809 .max_access_size = 1,
811 .endianness = DEVICE_LITTLE_ENDIAN,
814 static void port92_initfn(Object *obj)
816 Port92State *s = PORT92(obj);
818 memory_region_init_io(&s->io, OBJECT(s), &port92_ops, s, "port92", 1);
820 s->outport = 0;
822 qdev_init_gpio_out_named(DEVICE(obj), &s->a20_out, PORT92_A20_LINE, 1);
825 static void port92_realizefn(DeviceState *dev, Error **errp)
827 ISADevice *isadev = ISA_DEVICE(dev);
828 Port92State *s = PORT92(dev);
830 isa_register_ioport(isadev, &s->io, 0x92);
833 static void port92_class_initfn(ObjectClass *klass, void *data)
835 DeviceClass *dc = DEVICE_CLASS(klass);
837 dc->realize = port92_realizefn;
838 dc->reset = port92_reset;
839 dc->vmsd = &vmstate_port92_isa;
841 * Reason: unlike ordinary ISA devices, this one needs additional
842 * wiring: its A20 output line needs to be wired up by
843 * port92_init().
845 dc->user_creatable = false;
848 static const TypeInfo port92_info = {
849 .name = TYPE_PORT92,
850 .parent = TYPE_ISA_DEVICE,
851 .instance_size = sizeof(Port92State),
852 .instance_init = port92_initfn,
853 .class_init = port92_class_initfn,
856 static void port92_register_types(void)
858 type_register_static(&port92_info);
861 type_init(port92_register_types)
863 static void handle_a20_line_change(void *opaque, int irq, int level)
865 X86CPU *cpu = opaque;
867 /* XXX: send to all CPUs ? */
868 /* XXX: add logic to handle multiple A20 line sources */
869 x86_cpu_set_a20(cpu, level);
872 int e820_add_entry(uint64_t address, uint64_t length, uint32_t type)
874 int index = le32_to_cpu(e820_reserve.count);
875 struct e820_entry *entry;
877 if (type != E820_RAM) {
878 /* old FW_CFG_E820_TABLE entry -- reservations only */
879 if (index >= E820_NR_ENTRIES) {
880 return -EBUSY;
882 entry = &e820_reserve.entry[index++];
884 entry->address = cpu_to_le64(address);
885 entry->length = cpu_to_le64(length);
886 entry->type = cpu_to_le32(type);
888 e820_reserve.count = cpu_to_le32(index);
891 /* new "etc/e820" file -- include ram too */
892 e820_table = g_renew(struct e820_entry, e820_table, e820_entries + 1);
893 e820_table[e820_entries].address = cpu_to_le64(address);
894 e820_table[e820_entries].length = cpu_to_le64(length);
895 e820_table[e820_entries].type = cpu_to_le32(type);
896 e820_entries++;
898 return e820_entries;
901 int e820_get_num_entries(void)
903 return e820_entries;
906 bool e820_get_entry(int idx, uint32_t type, uint64_t *address, uint64_t *length)
908 if (idx < e820_entries && e820_table[idx].type == cpu_to_le32(type)) {
909 *address = le64_to_cpu(e820_table[idx].address);
910 *length = le64_to_cpu(e820_table[idx].length);
911 return true;
913 return false;
916 /* Enables contiguous-apic-ID mode, for compatibility */
917 static bool compat_apic_id_mode;
919 void enable_compat_apic_id_mode(void)
921 compat_apic_id_mode = true;
924 /* Calculates initial APIC ID for a specific CPU index
926 * Currently we need to be able to calculate the APIC ID from the CPU index
927 * alone (without requiring a CPU object), as the QEMU<->Seabios interfaces have
928 * no concept of "CPU index", and the NUMA tables on fw_cfg need the APIC ID of
929 * all CPUs up to max_cpus.
931 static uint32_t x86_cpu_apic_id_from_index(unsigned int cpu_index)
933 uint32_t correct_id;
934 static bool warned;
936 correct_id = x86_apicid_from_cpu_idx(smp_cores, smp_threads, cpu_index);
937 if (compat_apic_id_mode) {
938 if (cpu_index != correct_id && !warned && !qtest_enabled()) {
939 error_report("APIC IDs set in compatibility mode, "
940 "CPU topology won't match the configuration");
941 warned = true;
943 return cpu_index;
944 } else {
945 return correct_id;
949 static void pc_build_smbios(PCMachineState *pcms)
951 uint8_t *smbios_tables, *smbios_anchor;
952 size_t smbios_tables_len, smbios_anchor_len;
953 struct smbios_phys_mem_area *mem_array;
954 unsigned i, array_count;
955 MachineState *ms = MACHINE(pcms);
956 X86CPU *cpu = X86_CPU(ms->possible_cpus->cpus[0].cpu);
958 /* tell smbios about cpuid version and features */
959 smbios_set_cpuid(cpu->env.cpuid_version, cpu->env.features[FEAT_1_EDX]);
961 smbios_tables = smbios_get_table_legacy(&smbios_tables_len);
962 if (smbios_tables) {
963 fw_cfg_add_bytes(pcms->fw_cfg, FW_CFG_SMBIOS_ENTRIES,
964 smbios_tables, smbios_tables_len);
967 /* build the array of physical mem area from e820 table */
968 mem_array = g_malloc0(sizeof(*mem_array) * e820_get_num_entries());
969 for (i = 0, array_count = 0; i < e820_get_num_entries(); i++) {
970 uint64_t addr, len;
972 if (e820_get_entry(i, E820_RAM, &addr, &len)) {
973 mem_array[array_count].address = addr;
974 mem_array[array_count].length = len;
975 array_count++;
978 smbios_get_tables(mem_array, array_count,
979 &smbios_tables, &smbios_tables_len,
980 &smbios_anchor, &smbios_anchor_len);
981 g_free(mem_array);
983 if (smbios_anchor) {
984 fw_cfg_add_file(pcms->fw_cfg, "etc/smbios/smbios-tables",
985 smbios_tables, smbios_tables_len);
986 fw_cfg_add_file(pcms->fw_cfg, "etc/smbios/smbios-anchor",
987 smbios_anchor, smbios_anchor_len);
991 static FWCfgState *bochs_bios_init(AddressSpace *as, PCMachineState *pcms)
993 FWCfgState *fw_cfg;
994 uint64_t *numa_fw_cfg;
995 int i;
996 const CPUArchIdList *cpus;
997 MachineClass *mc = MACHINE_GET_CLASS(pcms);
999 fw_cfg = fw_cfg_init_io_dma(FW_CFG_IO_BASE, FW_CFG_IO_BASE + 4, as);
1000 fw_cfg_add_i16(fw_cfg, FW_CFG_NB_CPUS, pcms->boot_cpus);
1002 /* FW_CFG_MAX_CPUS is a bit confusing/problematic on x86:
1004 * For machine types prior to 1.8, SeaBIOS needs FW_CFG_MAX_CPUS for
1005 * building MPTable, ACPI MADT, ACPI CPU hotplug and ACPI SRAT table,
1006 * that tables are based on xAPIC ID and QEMU<->SeaBIOS interface
1007 * for CPU hotplug also uses APIC ID and not "CPU index".
1008 * This means that FW_CFG_MAX_CPUS is not the "maximum number of CPUs",
1009 * but the "limit to the APIC ID values SeaBIOS may see".
1011 * So for compatibility reasons with old BIOSes we are stuck with
1012 * "etc/max-cpus" actually being apic_id_limit
1014 fw_cfg_add_i16(fw_cfg, FW_CFG_MAX_CPUS, (uint16_t)pcms->apic_id_limit);
1015 fw_cfg_add_i64(fw_cfg, FW_CFG_RAM_SIZE, (uint64_t)ram_size);
1016 fw_cfg_add_bytes(fw_cfg, FW_CFG_ACPI_TABLES,
1017 acpi_tables, acpi_tables_len);
1018 fw_cfg_add_i32(fw_cfg, FW_CFG_IRQ0_OVERRIDE, kvm_allows_irq0_override());
1020 fw_cfg_add_bytes(fw_cfg, FW_CFG_E820_TABLE,
1021 &e820_reserve, sizeof(e820_reserve));
1022 fw_cfg_add_file(fw_cfg, "etc/e820", e820_table,
1023 sizeof(struct e820_entry) * e820_entries);
1025 fw_cfg_add_bytes(fw_cfg, FW_CFG_HPET, &hpet_cfg, sizeof(hpet_cfg));
1026 /* allocate memory for the NUMA channel: one (64bit) word for the number
1027 * of nodes, one word for each VCPU->node and one word for each node to
1028 * hold the amount of memory.
1030 numa_fw_cfg = g_new0(uint64_t, 1 + pcms->apic_id_limit + nb_numa_nodes);
1031 numa_fw_cfg[0] = cpu_to_le64(nb_numa_nodes);
1032 cpus = mc->possible_cpu_arch_ids(MACHINE(pcms));
1033 for (i = 0; i < cpus->len; i++) {
1034 unsigned int apic_id = cpus->cpus[i].arch_id;
1035 assert(apic_id < pcms->apic_id_limit);
1036 numa_fw_cfg[apic_id + 1] = cpu_to_le64(cpus->cpus[i].props.node_id);
1038 for (i = 0; i < nb_numa_nodes; i++) {
1039 numa_fw_cfg[pcms->apic_id_limit + 1 + i] =
1040 cpu_to_le64(numa_info[i].node_mem);
1042 fw_cfg_add_bytes(fw_cfg, FW_CFG_NUMA, numa_fw_cfg,
1043 (1 + pcms->apic_id_limit + nb_numa_nodes) *
1044 sizeof(*numa_fw_cfg));
1046 return fw_cfg;
1049 static long get_file_size(FILE *f)
1051 long where, size;
1053 /* XXX: on Unix systems, using fstat() probably makes more sense */
1055 where = ftell(f);
1056 fseek(f, 0, SEEK_END);
1057 size = ftell(f);
1058 fseek(f, where, SEEK_SET);
1060 return size;
1063 struct setup_data {
1064 uint64_t next;
1065 uint32_t type;
1066 uint32_t len;
1067 uint8_t data[0];
1068 } __attribute__((packed));
1072 * The entry point into the kernel for PVH boot is different from
1073 * the native entry point. The PVH entry is defined by the x86/HVM
1074 * direct boot ABI and is available in an ELFNOTE in the kernel binary.
1076 * This function is passed to load_elf() when it is called from
1077 * load_elfboot() which then additionally checks for an ELF Note of
1078 * type XEN_ELFNOTE_PHYS32_ENTRY and passes it to this function to
1079 * parse the PVH entry address from the ELF Note.
1081 * Due to trickery in elf_opts.h, load_elf() is actually available as
1082 * load_elf32() or load_elf64() and this routine needs to be able
1083 * to deal with being called as 32 or 64 bit.
1085 * The address of the PVH entry point is saved to the 'pvh_start_addr'
1086 * global variable. (although the entry point is 32-bit, the kernel
1087 * binary can be either 32-bit or 64-bit).
1089 static uint64_t read_pvh_start_addr(void *arg1, void *arg2, bool is64)
1091 size_t *elf_note_data_addr;
1093 /* Check if ELF Note header passed in is valid */
1094 if (arg1 == NULL) {
1095 return 0;
1098 if (is64) {
1099 struct elf64_note *nhdr64 = (struct elf64_note *)arg1;
1100 uint64_t nhdr_size64 = sizeof(struct elf64_note);
1101 uint64_t phdr_align = *(uint64_t *)arg2;
1102 uint64_t nhdr_namesz = nhdr64->n_namesz;
1104 elf_note_data_addr =
1105 ((void *)nhdr64) + nhdr_size64 +
1106 QEMU_ALIGN_UP(nhdr_namesz, phdr_align);
1107 } else {
1108 struct elf32_note *nhdr32 = (struct elf32_note *)arg1;
1109 uint32_t nhdr_size32 = sizeof(struct elf32_note);
1110 uint32_t phdr_align = *(uint32_t *)arg2;
1111 uint32_t nhdr_namesz = nhdr32->n_namesz;
1113 elf_note_data_addr =
1114 ((void *)nhdr32) + nhdr_size32 +
1115 QEMU_ALIGN_UP(nhdr_namesz, phdr_align);
1118 pvh_start_addr = *elf_note_data_addr;
1120 return pvh_start_addr;
1123 static bool load_elfboot(const char *kernel_filename,
1124 int kernel_file_size,
1125 uint8_t *header,
1126 size_t pvh_xen_start_addr,
1127 FWCfgState *fw_cfg)
1129 uint32_t flags = 0;
1130 uint32_t mh_load_addr = 0;
1131 uint32_t elf_kernel_size = 0;
1132 uint64_t elf_entry;
1133 uint64_t elf_low, elf_high;
1134 int kernel_size;
1136 if (ldl_p(header) != 0x464c457f) {
1137 return false; /* no elfboot */
1140 bool elf_is64 = header[EI_CLASS] == ELFCLASS64;
1141 flags = elf_is64 ?
1142 ((Elf64_Ehdr *)header)->e_flags : ((Elf32_Ehdr *)header)->e_flags;
1144 if (flags & 0x00010004) { /* LOAD_ELF_HEADER_HAS_ADDR */
1145 error_report("elfboot unsupported flags = %x", flags);
1146 exit(1);
1149 uint64_t elf_note_type = XEN_ELFNOTE_PHYS32_ENTRY;
1150 kernel_size = load_elf(kernel_filename, read_pvh_start_addr,
1151 NULL, &elf_note_type, &elf_entry,
1152 &elf_low, &elf_high, 0, I386_ELF_MACHINE,
1153 0, 0);
1155 if (kernel_size < 0) {
1156 error_report("Error while loading elf kernel");
1157 exit(1);
1159 mh_load_addr = elf_low;
1160 elf_kernel_size = elf_high - elf_low;
1162 if (pvh_start_addr == 0) {
1163 error_report("Error loading uncompressed kernel without PVH ELF Note");
1164 exit(1);
1166 fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_ENTRY, pvh_start_addr);
1167 fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_ADDR, mh_load_addr);
1168 fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_SIZE, elf_kernel_size);
1170 return true;
1173 static void load_linux(PCMachineState *pcms,
1174 FWCfgState *fw_cfg)
1176 uint16_t protocol;
1177 int setup_size, kernel_size, cmdline_size;
1178 int dtb_size, setup_data_offset;
1179 uint32_t initrd_max;
1180 uint8_t header[8192], *setup, *kernel;
1181 hwaddr real_addr, prot_addr, cmdline_addr, initrd_addr = 0;
1182 FILE *f;
1183 char *vmode;
1184 MachineState *machine = MACHINE(pcms);
1185 PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
1186 struct setup_data *setup_data;
1187 const char *kernel_filename = machine->kernel_filename;
1188 const char *initrd_filename = machine->initrd_filename;
1189 const char *dtb_filename = machine->dtb;
1190 const char *kernel_cmdline = machine->kernel_cmdline;
1192 /* Align to 16 bytes as a paranoia measure */
1193 cmdline_size = (strlen(kernel_cmdline)+16) & ~15;
1195 /* load the kernel header */
1196 f = fopen(kernel_filename, "rb");
1197 if (!f || !(kernel_size = get_file_size(f)) ||
1198 fread(header, 1, MIN(ARRAY_SIZE(header), kernel_size), f) !=
1199 MIN(ARRAY_SIZE(header), kernel_size)) {
1200 fprintf(stderr, "qemu: could not load kernel '%s': %s\n",
1201 kernel_filename, strerror(errno));
1202 exit(1);
1205 /* kernel protocol version */
1206 #if 0
1207 fprintf(stderr, "header magic: %#x\n", ldl_p(header+0x202));
1208 #endif
1209 if (ldl_p(header+0x202) == 0x53726448) {
1210 protocol = lduw_p(header+0x206);
1211 } else {
1213 * Check if the file is an uncompressed kernel file (ELF) and load it,
1214 * saving the PVH entry point used by the x86/HVM direct boot ABI.
1215 * If load_elfboot() is successful, populate the fw_cfg info.
1217 if (pcmc->pvh_enabled &&
1218 load_elfboot(kernel_filename, kernel_size,
1219 header, pvh_start_addr, fw_cfg)) {
1220 fclose(f);
1222 fw_cfg_add_i32(fw_cfg, FW_CFG_CMDLINE_SIZE,
1223 strlen(kernel_cmdline) + 1);
1224 fw_cfg_add_string(fw_cfg, FW_CFG_CMDLINE_DATA, kernel_cmdline);
1226 fw_cfg_add_i32(fw_cfg, FW_CFG_SETUP_SIZE, sizeof(header));
1227 fw_cfg_add_bytes(fw_cfg, FW_CFG_SETUP_DATA,
1228 header, sizeof(header));
1230 /* load initrd */
1231 if (initrd_filename) {
1232 gsize initrd_size;
1233 gchar *initrd_data;
1234 GError *gerr = NULL;
1236 if (!g_file_get_contents(initrd_filename, &initrd_data,
1237 &initrd_size, &gerr)) {
1238 fprintf(stderr, "qemu: error reading initrd %s: %s\n",
1239 initrd_filename, gerr->message);
1240 exit(1);
1243 initrd_max = pcms->below_4g_mem_size - pcmc->acpi_data_size - 1;
1244 if (initrd_size >= initrd_max) {
1245 fprintf(stderr, "qemu: initrd is too large, cannot support."
1246 "(max: %"PRIu32", need %"PRId64")\n",
1247 initrd_max, (uint64_t)initrd_size);
1248 exit(1);
1251 initrd_addr = (initrd_max - initrd_size) & ~4095;
1253 fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_ADDR, initrd_addr);
1254 fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_SIZE, initrd_size);
1255 fw_cfg_add_bytes(fw_cfg, FW_CFG_INITRD_DATA, initrd_data,
1256 initrd_size);
1259 option_rom[nb_option_roms].bootindex = 0;
1260 option_rom[nb_option_roms].name = "pvh.bin";
1261 nb_option_roms++;
1263 return;
1265 /* This looks like a multiboot kernel. If it is, let's stop
1266 treating it like a Linux kernel. */
1267 if (load_multiboot(fw_cfg, f, kernel_filename, initrd_filename,
1268 kernel_cmdline, kernel_size, header)) {
1269 return;
1271 protocol = 0;
1274 if (protocol < 0x200 || !(header[0x211] & 0x01)) {
1275 /* Low kernel */
1276 real_addr = 0x90000;
1277 cmdline_addr = 0x9a000 - cmdline_size;
1278 prot_addr = 0x10000;
1279 } else if (protocol < 0x202) {
1280 /* High but ancient kernel */
1281 real_addr = 0x90000;
1282 cmdline_addr = 0x9a000 - cmdline_size;
1283 prot_addr = 0x100000;
1284 } else {
1285 /* High and recent kernel */
1286 real_addr = 0x10000;
1287 cmdline_addr = 0x20000;
1288 prot_addr = 0x100000;
1291 #if 0
1292 fprintf(stderr,
1293 "qemu: real_addr = 0x" TARGET_FMT_plx "\n"
1294 "qemu: cmdline_addr = 0x" TARGET_FMT_plx "\n"
1295 "qemu: prot_addr = 0x" TARGET_FMT_plx "\n",
1296 real_addr,
1297 cmdline_addr,
1298 prot_addr);
1299 #endif
1301 /* highest address for loading the initrd */
1302 if (protocol >= 0x20c &&
1303 lduw_p(header+0x236) & XLF_CAN_BE_LOADED_ABOVE_4G) {
1305 * Linux has supported initrd up to 4 GB for a very long time (2007,
1306 * long before XLF_CAN_BE_LOADED_ABOVE_4G which was added in 2013),
1307 * though it only sets initrd_max to 2 GB to "work around bootloader
1308 * bugs". Luckily, QEMU firmware(which does something like bootloader)
1309 * has supported this.
1311 * It's believed that if XLF_CAN_BE_LOADED_ABOVE_4G is set, initrd can
1312 * be loaded into any address.
1314 * In addition, initrd_max is uint32_t simply because QEMU doesn't
1315 * support the 64-bit boot protocol (specifically the ext_ramdisk_image
1316 * field).
1318 * Therefore here just limit initrd_max to UINT32_MAX simply as well.
1320 initrd_max = UINT32_MAX;
1321 } else if (protocol >= 0x203) {
1322 initrd_max = ldl_p(header+0x22c);
1323 } else {
1324 initrd_max = 0x37ffffff;
1327 if (initrd_max >= pcms->below_4g_mem_size - pcmc->acpi_data_size) {
1328 initrd_max = pcms->below_4g_mem_size - pcmc->acpi_data_size - 1;
1331 fw_cfg_add_i32(fw_cfg, FW_CFG_CMDLINE_ADDR, cmdline_addr);
1332 fw_cfg_add_i32(fw_cfg, FW_CFG_CMDLINE_SIZE, strlen(kernel_cmdline)+1);
1333 fw_cfg_add_string(fw_cfg, FW_CFG_CMDLINE_DATA, kernel_cmdline);
1335 if (protocol >= 0x202) {
1336 stl_p(header+0x228, cmdline_addr);
1337 } else {
1338 stw_p(header+0x20, 0xA33F);
1339 stw_p(header+0x22, cmdline_addr-real_addr);
1342 /* handle vga= parameter */
1343 vmode = strstr(kernel_cmdline, "vga=");
1344 if (vmode) {
1345 unsigned int video_mode;
1346 /* skip "vga=" */
1347 vmode += 4;
1348 if (!strncmp(vmode, "normal", 6)) {
1349 video_mode = 0xffff;
1350 } else if (!strncmp(vmode, "ext", 3)) {
1351 video_mode = 0xfffe;
1352 } else if (!strncmp(vmode, "ask", 3)) {
1353 video_mode = 0xfffd;
1354 } else {
1355 video_mode = strtol(vmode, NULL, 0);
1357 stw_p(header+0x1fa, video_mode);
1360 /* loader type */
1361 /* High nybble = B reserved for QEMU; low nybble is revision number.
1362 If this code is substantially changed, you may want to consider
1363 incrementing the revision. */
1364 if (protocol >= 0x200) {
1365 header[0x210] = 0xB0;
1367 /* heap */
1368 if (protocol >= 0x201) {
1369 header[0x211] |= 0x80; /* CAN_USE_HEAP */
1370 stw_p(header+0x224, cmdline_addr-real_addr-0x200);
1373 /* load initrd */
1374 if (initrd_filename) {
1375 gsize initrd_size;
1376 gchar *initrd_data;
1377 GError *gerr = NULL;
1379 if (protocol < 0x200) {
1380 fprintf(stderr, "qemu: linux kernel too old to load a ram disk\n");
1381 exit(1);
1384 if (!g_file_get_contents(initrd_filename, &initrd_data,
1385 &initrd_size, &gerr)) {
1386 fprintf(stderr, "qemu: error reading initrd %s: %s\n",
1387 initrd_filename, gerr->message);
1388 exit(1);
1390 if (initrd_size >= initrd_max) {
1391 fprintf(stderr, "qemu: initrd is too large, cannot support."
1392 "(max: %"PRIu32", need %"PRId64")\n",
1393 initrd_max, (uint64_t)initrd_size);
1394 exit(1);
1397 initrd_addr = (initrd_max-initrd_size) & ~4095;
1399 fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_ADDR, initrd_addr);
1400 fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_SIZE, initrd_size);
1401 fw_cfg_add_bytes(fw_cfg, FW_CFG_INITRD_DATA, initrd_data, initrd_size);
1403 stl_p(header+0x218, initrd_addr);
1404 stl_p(header+0x21c, initrd_size);
1407 /* load kernel and setup */
1408 setup_size = header[0x1f1];
1409 if (setup_size == 0) {
1410 setup_size = 4;
1412 setup_size = (setup_size+1)*512;
1413 if (setup_size > kernel_size) {
1414 fprintf(stderr, "qemu: invalid kernel header\n");
1415 exit(1);
1417 kernel_size -= setup_size;
1419 setup = g_malloc(setup_size);
1420 kernel = g_malloc(kernel_size);
1421 fseek(f, 0, SEEK_SET);
1422 if (fread(setup, 1, setup_size, f) != setup_size) {
1423 fprintf(stderr, "fread() failed\n");
1424 exit(1);
1426 if (fread(kernel, 1, kernel_size, f) != kernel_size) {
1427 fprintf(stderr, "fread() failed\n");
1428 exit(1);
1430 fclose(f);
1432 /* append dtb to kernel */
1433 if (dtb_filename) {
1434 if (protocol < 0x209) {
1435 fprintf(stderr, "qemu: Linux kernel too old to load a dtb\n");
1436 exit(1);
1439 dtb_size = get_image_size(dtb_filename);
1440 if (dtb_size <= 0) {
1441 fprintf(stderr, "qemu: error reading dtb %s: %s\n",
1442 dtb_filename, strerror(errno));
1443 exit(1);
1446 setup_data_offset = QEMU_ALIGN_UP(kernel_size, 16);
1447 kernel_size = setup_data_offset + sizeof(struct setup_data) + dtb_size;
1448 kernel = g_realloc(kernel, kernel_size);
1450 stq_p(header+0x250, prot_addr + setup_data_offset);
1452 setup_data = (struct setup_data *)(kernel + setup_data_offset);
1453 setup_data->next = 0;
1454 setup_data->type = cpu_to_le32(SETUP_DTB);
1455 setup_data->len = cpu_to_le32(dtb_size);
1457 load_image_size(dtb_filename, setup_data->data, dtb_size);
1460 memcpy(setup, header, MIN(sizeof(header), setup_size));
1462 fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_ADDR, prot_addr);
1463 fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_SIZE, kernel_size);
1464 fw_cfg_add_bytes(fw_cfg, FW_CFG_KERNEL_DATA, kernel, kernel_size);
1466 fw_cfg_add_i32(fw_cfg, FW_CFG_SETUP_ADDR, real_addr);
1467 fw_cfg_add_i32(fw_cfg, FW_CFG_SETUP_SIZE, setup_size);
1468 fw_cfg_add_bytes(fw_cfg, FW_CFG_SETUP_DATA, setup, setup_size);
1470 option_rom[nb_option_roms].bootindex = 0;
1471 option_rom[nb_option_roms].name = "linuxboot.bin";
1472 if (pcmc->linuxboot_dma_enabled && fw_cfg_dma_enabled(fw_cfg)) {
1473 option_rom[nb_option_roms].name = "linuxboot_dma.bin";
1475 nb_option_roms++;
1478 #define NE2000_NB_MAX 6
1480 static const int ne2000_io[NE2000_NB_MAX] = { 0x300, 0x320, 0x340, 0x360,
1481 0x280, 0x380 };
1482 static const int ne2000_irq[NE2000_NB_MAX] = { 9, 10, 11, 3, 4, 5 };
1484 void pc_init_ne2k_isa(ISABus *bus, NICInfo *nd)
1486 static int nb_ne2k = 0;
1488 if (nb_ne2k == NE2000_NB_MAX)
1489 return;
1490 isa_ne2000_init(bus, ne2000_io[nb_ne2k],
1491 ne2000_irq[nb_ne2k], nd);
1492 nb_ne2k++;
1495 DeviceState *cpu_get_current_apic(void)
1497 if (current_cpu) {
1498 X86CPU *cpu = X86_CPU(current_cpu);
1499 return cpu->apic_state;
1500 } else {
1501 return NULL;
1505 void pc_acpi_smi_interrupt(void *opaque, int irq, int level)
1507 X86CPU *cpu = opaque;
1509 if (level) {
1510 cpu_interrupt(CPU(cpu), CPU_INTERRUPT_SMI);
1514 static void pc_new_cpu(const char *typename, int64_t apic_id, Error **errp)
1516 Object *cpu = NULL;
1517 Error *local_err = NULL;
1519 cpu = object_new(typename);
1521 object_property_set_uint(cpu, apic_id, "apic-id", &local_err);
1522 object_property_set_bool(cpu, true, "realized", &local_err);
1524 object_unref(cpu);
1525 error_propagate(errp, local_err);
1528 void pc_hot_add_cpu(const int64_t id, Error **errp)
1530 MachineState *ms = MACHINE(qdev_get_machine());
1531 int64_t apic_id = x86_cpu_apic_id_from_index(id);
1532 Error *local_err = NULL;
1534 if (id < 0) {
1535 error_setg(errp, "Invalid CPU id: %" PRIi64, id);
1536 return;
1539 if (apic_id >= ACPI_CPU_HOTPLUG_ID_LIMIT) {
1540 error_setg(errp, "Unable to add CPU: %" PRIi64
1541 ", resulting APIC ID (%" PRIi64 ") is too large",
1542 id, apic_id);
1543 return;
1546 pc_new_cpu(ms->cpu_type, apic_id, &local_err);
1547 if (local_err) {
1548 error_propagate(errp, local_err);
1549 return;
1553 void pc_cpus_init(PCMachineState *pcms)
1555 int i;
1556 const CPUArchIdList *possible_cpus;
1557 MachineState *ms = MACHINE(pcms);
1558 MachineClass *mc = MACHINE_GET_CLASS(pcms);
1560 /* Calculates the limit to CPU APIC ID values
1562 * Limit for the APIC ID value, so that all
1563 * CPU APIC IDs are < pcms->apic_id_limit.
1565 * This is used for FW_CFG_MAX_CPUS. See comments on bochs_bios_init().
1567 pcms->apic_id_limit = x86_cpu_apic_id_from_index(max_cpus - 1) + 1;
1568 possible_cpus = mc->possible_cpu_arch_ids(ms);
1569 for (i = 0; i < smp_cpus; i++) {
1570 pc_new_cpu(possible_cpus->cpus[i].type, possible_cpus->cpus[i].arch_id,
1571 &error_fatal);
1575 static void pc_build_feature_control_file(PCMachineState *pcms)
1577 MachineState *ms = MACHINE(pcms);
1578 X86CPU *cpu = X86_CPU(ms->possible_cpus->cpus[0].cpu);
1579 CPUX86State *env = &cpu->env;
1580 uint32_t unused, ecx, edx;
1581 uint64_t feature_control_bits = 0;
1582 uint64_t *val;
1584 cpu_x86_cpuid(env, 1, 0, &unused, &unused, &ecx, &edx);
1585 if (ecx & CPUID_EXT_VMX) {
1586 feature_control_bits |= FEATURE_CONTROL_VMXON_ENABLED_OUTSIDE_SMX;
1589 if ((edx & (CPUID_EXT2_MCE | CPUID_EXT2_MCA)) ==
1590 (CPUID_EXT2_MCE | CPUID_EXT2_MCA) &&
1591 (env->mcg_cap & MCG_LMCE_P)) {
1592 feature_control_bits |= FEATURE_CONTROL_LMCE;
1595 if (!feature_control_bits) {
1596 return;
1599 val = g_malloc(sizeof(*val));
1600 *val = cpu_to_le64(feature_control_bits | FEATURE_CONTROL_LOCKED);
1601 fw_cfg_add_file(pcms->fw_cfg, "etc/msr_feature_control", val, sizeof(*val));
1604 static void rtc_set_cpus_count(ISADevice *rtc, uint16_t cpus_count)
1606 if (cpus_count > 0xff) {
1607 /* If the number of CPUs can't be represented in 8 bits, the
1608 * BIOS must use "FW_CFG_NB_CPUS". Set RTC field to 0 just
1609 * to make old BIOSes fail more predictably.
1611 rtc_set_memory(rtc, 0x5f, 0);
1612 } else {
1613 rtc_set_memory(rtc, 0x5f, cpus_count - 1);
1617 static
1618 void pc_machine_done(Notifier *notifier, void *data)
1620 PCMachineState *pcms = container_of(notifier,
1621 PCMachineState, machine_done);
1622 PCIBus *bus = pcms->bus;
1624 /* set the number of CPUs */
1625 rtc_set_cpus_count(pcms->rtc, pcms->boot_cpus);
1627 if (bus) {
1628 int extra_hosts = 0;
1630 QLIST_FOREACH(bus, &bus->child, sibling) {
1631 /* look for expander root buses */
1632 if (pci_bus_is_root(bus)) {
1633 extra_hosts++;
1636 if (extra_hosts && pcms->fw_cfg) {
1637 uint64_t *val = g_malloc(sizeof(*val));
1638 *val = cpu_to_le64(extra_hosts);
1639 fw_cfg_add_file(pcms->fw_cfg,
1640 "etc/extra-pci-roots", val, sizeof(*val));
1644 acpi_setup();
1645 if (pcms->fw_cfg) {
1646 pc_build_smbios(pcms);
1647 pc_build_feature_control_file(pcms);
1648 /* update FW_CFG_NB_CPUS to account for -device added CPUs */
1649 fw_cfg_modify_i16(pcms->fw_cfg, FW_CFG_NB_CPUS, pcms->boot_cpus);
1652 if (pcms->apic_id_limit > 255 && !xen_enabled()) {
1653 IntelIOMMUState *iommu = INTEL_IOMMU_DEVICE(x86_iommu_get_default());
1655 if (!iommu || !x86_iommu_ir_supported(X86_IOMMU_DEVICE(iommu)) ||
1656 iommu->intr_eim != ON_OFF_AUTO_ON) {
1657 error_report("current -smp configuration requires "
1658 "Extended Interrupt Mode enabled. "
1659 "You can add an IOMMU using: "
1660 "-device intel-iommu,intremap=on,eim=on");
1661 exit(EXIT_FAILURE);
1666 void pc_guest_info_init(PCMachineState *pcms)
1668 int i;
1670 pcms->apic_xrupt_override = kvm_allows_irq0_override();
1671 pcms->numa_nodes = nb_numa_nodes;
1672 pcms->node_mem = g_malloc0(pcms->numa_nodes *
1673 sizeof *pcms->node_mem);
1674 for (i = 0; i < nb_numa_nodes; i++) {
1675 pcms->node_mem[i] = numa_info[i].node_mem;
1678 pcms->machine_done.notify = pc_machine_done;
1679 qemu_add_machine_init_done_notifier(&pcms->machine_done);
1682 /* setup pci memory address space mapping into system address space */
1683 void pc_pci_as_mapping_init(Object *owner, MemoryRegion *system_memory,
1684 MemoryRegion *pci_address_space)
1686 /* Set to lower priority than RAM */
1687 memory_region_add_subregion_overlap(system_memory, 0x0,
1688 pci_address_space, -1);
1691 void pc_acpi_init(const char *default_dsdt)
1693 char *filename;
1695 if (acpi_tables != NULL) {
1696 /* manually set via -acpitable, leave it alone */
1697 return;
1700 filename = qemu_find_file(QEMU_FILE_TYPE_BIOS, default_dsdt);
1701 if (filename == NULL) {
1702 warn_report("failed to find %s", default_dsdt);
1703 } else {
1704 QemuOpts *opts = qemu_opts_create(qemu_find_opts("acpi"), NULL, 0,
1705 &error_abort);
1706 Error *err = NULL;
1708 qemu_opt_set(opts, "file", filename, &error_abort);
1710 acpi_table_add_builtin(opts, &err);
1711 if (err) {
1712 warn_reportf_err(err, "failed to load %s: ", filename);
1714 g_free(filename);
1718 void xen_load_linux(PCMachineState *pcms)
1720 int i;
1721 FWCfgState *fw_cfg;
1723 assert(MACHINE(pcms)->kernel_filename != NULL);
1725 fw_cfg = fw_cfg_init_io(FW_CFG_IO_BASE);
1726 fw_cfg_add_i16(fw_cfg, FW_CFG_NB_CPUS, pcms->boot_cpus);
1727 rom_set_fw(fw_cfg);
1729 load_linux(pcms, fw_cfg);
1730 for (i = 0; i < nb_option_roms; i++) {
1731 assert(!strcmp(option_rom[i].name, "linuxboot.bin") ||
1732 !strcmp(option_rom[i].name, "linuxboot_dma.bin") ||
1733 !strcmp(option_rom[i].name, "pvh.bin") ||
1734 !strcmp(option_rom[i].name, "multiboot.bin"));
1735 rom_add_option(option_rom[i].name, option_rom[i].bootindex);
1737 pcms->fw_cfg = fw_cfg;
1740 void pc_memory_init(PCMachineState *pcms,
1741 MemoryRegion *system_memory,
1742 MemoryRegion *rom_memory,
1743 MemoryRegion **ram_memory)
1745 int linux_boot, i;
1746 MemoryRegion *ram, *option_rom_mr;
1747 MemoryRegion *ram_below_4g, *ram_above_4g;
1748 FWCfgState *fw_cfg;
1749 MachineState *machine = MACHINE(pcms);
1750 PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
1752 assert(machine->ram_size == pcms->below_4g_mem_size +
1753 pcms->above_4g_mem_size);
1755 linux_boot = (machine->kernel_filename != NULL);
1757 /* Allocate RAM. We allocate it as a single memory region and use
1758 * aliases to address portions of it, mostly for backwards compatibility
1759 * with older qemus that used qemu_ram_alloc().
1761 ram = g_malloc(sizeof(*ram));
1762 memory_region_allocate_system_memory(ram, NULL, "pc.ram",
1763 machine->ram_size);
1764 *ram_memory = ram;
1765 ram_below_4g = g_malloc(sizeof(*ram_below_4g));
1766 memory_region_init_alias(ram_below_4g, NULL, "ram-below-4g", ram,
1767 0, pcms->below_4g_mem_size);
1768 memory_region_add_subregion(system_memory, 0, ram_below_4g);
1769 e820_add_entry(0, pcms->below_4g_mem_size, E820_RAM);
1770 if (pcms->above_4g_mem_size > 0) {
1771 ram_above_4g = g_malloc(sizeof(*ram_above_4g));
1772 memory_region_init_alias(ram_above_4g, NULL, "ram-above-4g", ram,
1773 pcms->below_4g_mem_size,
1774 pcms->above_4g_mem_size);
1775 memory_region_add_subregion(system_memory, 0x100000000ULL,
1776 ram_above_4g);
1777 e820_add_entry(0x100000000ULL, pcms->above_4g_mem_size, E820_RAM);
1780 if (!pcmc->has_reserved_memory &&
1781 (machine->ram_slots ||
1782 (machine->maxram_size > machine->ram_size))) {
1783 MachineClass *mc = MACHINE_GET_CLASS(machine);
1785 error_report("\"-memory 'slots|maxmem'\" is not supported by: %s",
1786 mc->name);
1787 exit(EXIT_FAILURE);
1790 /* always allocate the device memory information */
1791 machine->device_memory = g_malloc0(sizeof(*machine->device_memory));
1793 /* initialize device memory address space */
1794 if (pcmc->has_reserved_memory &&
1795 (machine->ram_size < machine->maxram_size)) {
1796 ram_addr_t device_mem_size = machine->maxram_size - machine->ram_size;
1798 if (machine->ram_slots > ACPI_MAX_RAM_SLOTS) {
1799 error_report("unsupported amount of memory slots: %"PRIu64,
1800 machine->ram_slots);
1801 exit(EXIT_FAILURE);
1804 if (QEMU_ALIGN_UP(machine->maxram_size,
1805 TARGET_PAGE_SIZE) != machine->maxram_size) {
1806 error_report("maximum memory size must by aligned to multiple of "
1807 "%d bytes", TARGET_PAGE_SIZE);
1808 exit(EXIT_FAILURE);
1811 machine->device_memory->base =
1812 ROUND_UP(0x100000000ULL + pcms->above_4g_mem_size, 1 * GiB);
1814 if (pcmc->enforce_aligned_dimm) {
1815 /* size device region assuming 1G page max alignment per slot */
1816 device_mem_size += (1 * GiB) * machine->ram_slots;
1819 if ((machine->device_memory->base + device_mem_size) <
1820 device_mem_size) {
1821 error_report("unsupported amount of maximum memory: " RAM_ADDR_FMT,
1822 machine->maxram_size);
1823 exit(EXIT_FAILURE);
1826 memory_region_init(&machine->device_memory->mr, OBJECT(pcms),
1827 "device-memory", device_mem_size);
1828 memory_region_add_subregion(system_memory, machine->device_memory->base,
1829 &machine->device_memory->mr);
1832 /* Initialize PC system firmware */
1833 pc_system_firmware_init(rom_memory, !pcmc->pci_enabled);
1835 option_rom_mr = g_malloc(sizeof(*option_rom_mr));
1836 memory_region_init_ram(option_rom_mr, NULL, "pc.rom", PC_ROM_SIZE,
1837 &error_fatal);
1838 if (pcmc->pci_enabled) {
1839 memory_region_set_readonly(option_rom_mr, true);
1841 memory_region_add_subregion_overlap(rom_memory,
1842 PC_ROM_MIN_VGA,
1843 option_rom_mr,
1846 fw_cfg = bochs_bios_init(&address_space_memory, pcms);
1848 rom_set_fw(fw_cfg);
1850 if (pcmc->has_reserved_memory && machine->device_memory->base) {
1851 uint64_t *val = g_malloc(sizeof(*val));
1852 PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
1853 uint64_t res_mem_end = machine->device_memory->base;
1855 if (!pcmc->broken_reserved_end) {
1856 res_mem_end += memory_region_size(&machine->device_memory->mr);
1858 *val = cpu_to_le64(ROUND_UP(res_mem_end, 1 * GiB));
1859 fw_cfg_add_file(fw_cfg, "etc/reserved-memory-end", val, sizeof(*val));
1862 if (linux_boot) {
1863 load_linux(pcms, fw_cfg);
1866 for (i = 0; i < nb_option_roms; i++) {
1867 rom_add_option(option_rom[i].name, option_rom[i].bootindex);
1869 pcms->fw_cfg = fw_cfg;
1871 /* Init default IOAPIC address space */
1872 pcms->ioapic_as = &address_space_memory;
1876 * The 64bit pci hole starts after "above 4G RAM" and
1877 * potentially the space reserved for memory hotplug.
1879 uint64_t pc_pci_hole64_start(void)
1881 PCMachineState *pcms = PC_MACHINE(qdev_get_machine());
1882 PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
1883 MachineState *ms = MACHINE(pcms);
1884 uint64_t hole64_start = 0;
1886 if (pcmc->has_reserved_memory && ms->device_memory->base) {
1887 hole64_start = ms->device_memory->base;
1888 if (!pcmc->broken_reserved_end) {
1889 hole64_start += memory_region_size(&ms->device_memory->mr);
1891 } else {
1892 hole64_start = 0x100000000ULL + pcms->above_4g_mem_size;
1895 return ROUND_UP(hole64_start, 1 * GiB);
1898 qemu_irq pc_allocate_cpu_irq(void)
1900 return qemu_allocate_irq(pic_irq_request, NULL, 0);
1903 DeviceState *pc_vga_init(ISABus *isa_bus, PCIBus *pci_bus)
1905 DeviceState *dev = NULL;
1907 rom_set_order_override(FW_CFG_ORDER_OVERRIDE_VGA);
1908 if (pci_bus) {
1909 PCIDevice *pcidev = pci_vga_init(pci_bus);
1910 dev = pcidev ? &pcidev->qdev : NULL;
1911 } else if (isa_bus) {
1912 ISADevice *isadev = isa_vga_init(isa_bus);
1913 dev = isadev ? DEVICE(isadev) : NULL;
1915 rom_reset_order_override();
1916 return dev;
1919 static const MemoryRegionOps ioport80_io_ops = {
1920 .write = ioport80_write,
1921 .read = ioport80_read,
1922 .endianness = DEVICE_NATIVE_ENDIAN,
1923 .impl = {
1924 .min_access_size = 1,
1925 .max_access_size = 1,
1929 static const MemoryRegionOps ioportF0_io_ops = {
1930 .write = ioportF0_write,
1931 .read = ioportF0_read,
1932 .endianness = DEVICE_NATIVE_ENDIAN,
1933 .impl = {
1934 .min_access_size = 1,
1935 .max_access_size = 1,
1939 static void pc_superio_init(ISABus *isa_bus, bool create_fdctrl, bool no_vmport)
1941 int i;
1942 DriveInfo *fd[MAX_FD];
1943 qemu_irq *a20_line;
1944 ISADevice *i8042, *port92, *vmmouse;
1946 serial_hds_isa_init(isa_bus, 0, MAX_ISA_SERIAL_PORTS);
1947 parallel_hds_isa_init(isa_bus, MAX_PARALLEL_PORTS);
1949 for (i = 0; i < MAX_FD; i++) {
1950 fd[i] = drive_get(IF_FLOPPY, 0, i);
1951 create_fdctrl |= !!fd[i];
1953 if (create_fdctrl) {
1954 fdctrl_init_isa(isa_bus, fd);
1957 i8042 = isa_create_simple(isa_bus, "i8042");
1958 if (!no_vmport) {
1959 vmport_init(isa_bus);
1960 vmmouse = isa_try_create(isa_bus, "vmmouse");
1961 } else {
1962 vmmouse = NULL;
1964 if (vmmouse) {
1965 DeviceState *dev = DEVICE(vmmouse);
1966 qdev_prop_set_ptr(dev, "ps2_mouse", i8042);
1967 qdev_init_nofail(dev);
1969 port92 = isa_create_simple(isa_bus, "port92");
1971 a20_line = qemu_allocate_irqs(handle_a20_line_change, first_cpu, 2);
1972 i8042_setup_a20_line(i8042, a20_line[0]);
1973 port92_init(port92, a20_line[1]);
1974 g_free(a20_line);
1977 void pc_basic_device_init(ISABus *isa_bus, qemu_irq *gsi,
1978 ISADevice **rtc_state,
1979 bool create_fdctrl,
1980 bool no_vmport,
1981 bool has_pit,
1982 uint32_t hpet_irqs)
1984 int i;
1985 DeviceState *hpet = NULL;
1986 int pit_isa_irq = 0;
1987 qemu_irq pit_alt_irq = NULL;
1988 qemu_irq rtc_irq = NULL;
1989 ISADevice *pit = NULL;
1990 MemoryRegion *ioport80_io = g_new(MemoryRegion, 1);
1991 MemoryRegion *ioportF0_io = g_new(MemoryRegion, 1);
1993 memory_region_init_io(ioport80_io, NULL, &ioport80_io_ops, NULL, "ioport80", 1);
1994 memory_region_add_subregion(isa_bus->address_space_io, 0x80, ioport80_io);
1996 memory_region_init_io(ioportF0_io, NULL, &ioportF0_io_ops, NULL, "ioportF0", 1);
1997 memory_region_add_subregion(isa_bus->address_space_io, 0xf0, ioportF0_io);
2000 * Check if an HPET shall be created.
2002 * Without KVM_CAP_PIT_STATE2, we cannot switch off the in-kernel PIT
2003 * when the HPET wants to take over. Thus we have to disable the latter.
2005 if (!no_hpet && (!kvm_irqchip_in_kernel() || kvm_has_pit_state2())) {
2006 /* In order to set property, here not using sysbus_try_create_simple */
2007 hpet = qdev_try_create(NULL, TYPE_HPET);
2008 if (hpet) {
2009 /* For pc-piix-*, hpet's intcap is always IRQ2. For pc-q35-1.7
2010 * and earlier, use IRQ2 for compat. Otherwise, use IRQ16~23,
2011 * IRQ8 and IRQ2.
2013 uint8_t compat = object_property_get_uint(OBJECT(hpet),
2014 HPET_INTCAP, NULL);
2015 if (!compat) {
2016 qdev_prop_set_uint32(hpet, HPET_INTCAP, hpet_irqs);
2018 qdev_init_nofail(hpet);
2019 sysbus_mmio_map(SYS_BUS_DEVICE(hpet), 0, HPET_BASE);
2021 for (i = 0; i < GSI_NUM_PINS; i++) {
2022 sysbus_connect_irq(SYS_BUS_DEVICE(hpet), i, gsi[i]);
2024 pit_isa_irq = -1;
2025 pit_alt_irq = qdev_get_gpio_in(hpet, HPET_LEGACY_PIT_INT);
2026 rtc_irq = qdev_get_gpio_in(hpet, HPET_LEGACY_RTC_INT);
2029 *rtc_state = mc146818_rtc_init(isa_bus, 2000, rtc_irq);
2031 qemu_register_boot_set(pc_boot_set, *rtc_state);
2033 if (!xen_enabled() && has_pit) {
2034 if (kvm_pit_in_kernel()) {
2035 pit = kvm_pit_init(isa_bus, 0x40);
2036 } else {
2037 pit = i8254_pit_init(isa_bus, 0x40, pit_isa_irq, pit_alt_irq);
2039 if (hpet) {
2040 /* connect PIT to output control line of the HPET */
2041 qdev_connect_gpio_out(hpet, 0, qdev_get_gpio_in(DEVICE(pit), 0));
2043 pcspk_init(isa_bus, pit);
2046 i8257_dma_init(isa_bus, 0);
2048 /* Super I/O */
2049 pc_superio_init(isa_bus, create_fdctrl, no_vmport);
2052 void pc_nic_init(PCMachineClass *pcmc, ISABus *isa_bus, PCIBus *pci_bus)
2054 int i;
2056 rom_set_order_override(FW_CFG_ORDER_OVERRIDE_NIC);
2057 for (i = 0; i < nb_nics; i++) {
2058 NICInfo *nd = &nd_table[i];
2059 const char *model = nd->model ? nd->model : pcmc->default_nic_model;
2061 if (g_str_equal(model, "ne2k_isa")) {
2062 pc_init_ne2k_isa(isa_bus, nd);
2063 } else {
2064 pci_nic_init_nofail(nd, pci_bus, model, NULL);
2067 rom_reset_order_override();
2070 void ioapic_init_gsi(GSIState *gsi_state, const char *parent_name)
2072 DeviceState *dev;
2073 SysBusDevice *d;
2074 unsigned int i;
2076 if (kvm_ioapic_in_kernel()) {
2077 dev = qdev_create(NULL, TYPE_KVM_IOAPIC);
2078 } else {
2079 dev = qdev_create(NULL, TYPE_IOAPIC);
2081 if (parent_name) {
2082 object_property_add_child(object_resolve_path(parent_name, NULL),
2083 "ioapic", OBJECT(dev), NULL);
2085 qdev_init_nofail(dev);
2086 d = SYS_BUS_DEVICE(dev);
2087 sysbus_mmio_map(d, 0, IO_APIC_DEFAULT_ADDRESS);
2089 for (i = 0; i < IOAPIC_NUM_PINS; i++) {
2090 gsi_state->ioapic_irq[i] = qdev_get_gpio_in(dev, i);
2094 static void pc_memory_pre_plug(HotplugHandler *hotplug_dev, DeviceState *dev,
2095 Error **errp)
2097 const PCMachineState *pcms = PC_MACHINE(hotplug_dev);
2098 const PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
2099 const bool is_nvdimm = object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM);
2100 const uint64_t legacy_align = TARGET_PAGE_SIZE;
2103 * When -no-acpi is used with Q35 machine type, no ACPI is built,
2104 * but pcms->acpi_dev is still created. Check !acpi_enabled in
2105 * addition to cover this case.
2107 if (!pcms->acpi_dev || !acpi_enabled) {
2108 error_setg(errp,
2109 "memory hotplug is not enabled: missing acpi device or acpi disabled");
2110 return;
2113 if (is_nvdimm && !pcms->acpi_nvdimm_state.is_enabled) {
2114 error_setg(errp, "nvdimm is not enabled: missing 'nvdimm' in '-M'");
2115 return;
2118 pc_dimm_pre_plug(PC_DIMM(dev), MACHINE(hotplug_dev),
2119 pcmc->enforce_aligned_dimm ? NULL : &legacy_align, errp);
2122 static void pc_memory_plug(HotplugHandler *hotplug_dev,
2123 DeviceState *dev, Error **errp)
2125 Error *local_err = NULL;
2126 PCMachineState *pcms = PC_MACHINE(hotplug_dev);
2127 bool is_nvdimm = object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM);
2129 pc_dimm_plug(PC_DIMM(dev), MACHINE(pcms), &local_err);
2130 if (local_err) {
2131 goto out;
2134 if (is_nvdimm) {
2135 nvdimm_plug(&pcms->acpi_nvdimm_state);
2138 hotplug_handler_plug(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &error_abort);
2139 out:
2140 error_propagate(errp, local_err);
2143 static void pc_memory_unplug_request(HotplugHandler *hotplug_dev,
2144 DeviceState *dev, Error **errp)
2146 Error *local_err = NULL;
2147 PCMachineState *pcms = PC_MACHINE(hotplug_dev);
2150 * When -no-acpi is used with Q35 machine type, no ACPI is built,
2151 * but pcms->acpi_dev is still created. Check !acpi_enabled in
2152 * addition to cover this case.
2154 if (!pcms->acpi_dev || !acpi_enabled) {
2155 error_setg(&local_err,
2156 "memory hotplug is not enabled: missing acpi device or acpi disabled");
2157 goto out;
2160 if (object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM)) {
2161 error_setg(&local_err,
2162 "nvdimm device hot unplug is not supported yet.");
2163 goto out;
2166 hotplug_handler_unplug_request(HOTPLUG_HANDLER(pcms->acpi_dev), dev,
2167 &local_err);
2168 out:
2169 error_propagate(errp, local_err);
2172 static void pc_memory_unplug(HotplugHandler *hotplug_dev,
2173 DeviceState *dev, Error **errp)
2175 PCMachineState *pcms = PC_MACHINE(hotplug_dev);
2176 Error *local_err = NULL;
2178 hotplug_handler_unplug(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &local_err);
2179 if (local_err) {
2180 goto out;
2183 pc_dimm_unplug(PC_DIMM(dev), MACHINE(pcms));
2184 object_unparent(OBJECT(dev));
2186 out:
2187 error_propagate(errp, local_err);
2190 static int pc_apic_cmp(const void *a, const void *b)
2192 CPUArchId *apic_a = (CPUArchId *)a;
2193 CPUArchId *apic_b = (CPUArchId *)b;
2195 return apic_a->arch_id - apic_b->arch_id;
2198 /* returns pointer to CPUArchId descriptor that matches CPU's apic_id
2199 * in ms->possible_cpus->cpus, if ms->possible_cpus->cpus has no
2200 * entry corresponding to CPU's apic_id returns NULL.
2202 static CPUArchId *pc_find_cpu_slot(MachineState *ms, uint32_t id, int *idx)
2204 CPUArchId apic_id, *found_cpu;
2206 apic_id.arch_id = id;
2207 found_cpu = bsearch(&apic_id, ms->possible_cpus->cpus,
2208 ms->possible_cpus->len, sizeof(*ms->possible_cpus->cpus),
2209 pc_apic_cmp);
2210 if (found_cpu && idx) {
2211 *idx = found_cpu - ms->possible_cpus->cpus;
2213 return found_cpu;
2216 static void pc_cpu_plug(HotplugHandler *hotplug_dev,
2217 DeviceState *dev, Error **errp)
2219 CPUArchId *found_cpu;
2220 Error *local_err = NULL;
2221 X86CPU *cpu = X86_CPU(dev);
2222 PCMachineState *pcms = PC_MACHINE(hotplug_dev);
2224 if (pcms->acpi_dev) {
2225 hotplug_handler_plug(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &local_err);
2226 if (local_err) {
2227 goto out;
2231 /* increment the number of CPUs */
2232 pcms->boot_cpus++;
2233 if (pcms->rtc) {
2234 rtc_set_cpus_count(pcms->rtc, pcms->boot_cpus);
2236 if (pcms->fw_cfg) {
2237 fw_cfg_modify_i16(pcms->fw_cfg, FW_CFG_NB_CPUS, pcms->boot_cpus);
2240 found_cpu = pc_find_cpu_slot(MACHINE(pcms), cpu->apic_id, NULL);
2241 found_cpu->cpu = OBJECT(dev);
2242 out:
2243 error_propagate(errp, local_err);
2245 static void pc_cpu_unplug_request_cb(HotplugHandler *hotplug_dev,
2246 DeviceState *dev, Error **errp)
2248 int idx = -1;
2249 Error *local_err = NULL;
2250 X86CPU *cpu = X86_CPU(dev);
2251 PCMachineState *pcms = PC_MACHINE(hotplug_dev);
2253 if (!pcms->acpi_dev) {
2254 error_setg(&local_err, "CPU hot unplug not supported without ACPI");
2255 goto out;
2258 pc_find_cpu_slot(MACHINE(pcms), cpu->apic_id, &idx);
2259 assert(idx != -1);
2260 if (idx == 0) {
2261 error_setg(&local_err, "Boot CPU is unpluggable");
2262 goto out;
2265 hotplug_handler_unplug_request(HOTPLUG_HANDLER(pcms->acpi_dev), dev,
2266 &local_err);
2267 if (local_err) {
2268 goto out;
2271 out:
2272 error_propagate(errp, local_err);
2276 static void pc_cpu_unplug_cb(HotplugHandler *hotplug_dev,
2277 DeviceState *dev, Error **errp)
2279 CPUArchId *found_cpu;
2280 Error *local_err = NULL;
2281 X86CPU *cpu = X86_CPU(dev);
2282 PCMachineState *pcms = PC_MACHINE(hotplug_dev);
2284 hotplug_handler_unplug(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &local_err);
2285 if (local_err) {
2286 goto out;
2289 found_cpu = pc_find_cpu_slot(MACHINE(pcms), cpu->apic_id, NULL);
2290 found_cpu->cpu = NULL;
2291 object_unparent(OBJECT(dev));
2293 /* decrement the number of CPUs */
2294 pcms->boot_cpus--;
2295 /* Update the number of CPUs in CMOS */
2296 rtc_set_cpus_count(pcms->rtc, pcms->boot_cpus);
2297 fw_cfg_modify_i16(pcms->fw_cfg, FW_CFG_NB_CPUS, pcms->boot_cpus);
2298 out:
2299 error_propagate(errp, local_err);
2302 static void pc_cpu_pre_plug(HotplugHandler *hotplug_dev,
2303 DeviceState *dev, Error **errp)
2305 int idx;
2306 CPUState *cs;
2307 CPUArchId *cpu_slot;
2308 X86CPUTopoInfo topo;
2309 X86CPU *cpu = X86_CPU(dev);
2310 MachineState *ms = MACHINE(hotplug_dev);
2311 PCMachineState *pcms = PC_MACHINE(hotplug_dev);
2313 if(!object_dynamic_cast(OBJECT(cpu), ms->cpu_type)) {
2314 error_setg(errp, "Invalid CPU type, expected cpu type: '%s'",
2315 ms->cpu_type);
2316 return;
2319 /* if APIC ID is not set, set it based on socket/core/thread properties */
2320 if (cpu->apic_id == UNASSIGNED_APIC_ID) {
2321 int max_socket = (max_cpus - 1) / smp_threads / smp_cores;
2323 if (cpu->socket_id < 0) {
2324 error_setg(errp, "CPU socket-id is not set");
2325 return;
2326 } else if (cpu->socket_id > max_socket) {
2327 error_setg(errp, "Invalid CPU socket-id: %u must be in range 0:%u",
2328 cpu->socket_id, max_socket);
2329 return;
2331 if (cpu->core_id < 0) {
2332 error_setg(errp, "CPU core-id is not set");
2333 return;
2334 } else if (cpu->core_id > (smp_cores - 1)) {
2335 error_setg(errp, "Invalid CPU core-id: %u must be in range 0:%u",
2336 cpu->core_id, smp_cores - 1);
2337 return;
2339 if (cpu->thread_id < 0) {
2340 error_setg(errp, "CPU thread-id is not set");
2341 return;
2342 } else if (cpu->thread_id > (smp_threads - 1)) {
2343 error_setg(errp, "Invalid CPU thread-id: %u must be in range 0:%u",
2344 cpu->thread_id, smp_threads - 1);
2345 return;
2348 topo.pkg_id = cpu->socket_id;
2349 topo.core_id = cpu->core_id;
2350 topo.smt_id = cpu->thread_id;
2351 cpu->apic_id = apicid_from_topo_ids(smp_cores, smp_threads, &topo);
2354 cpu_slot = pc_find_cpu_slot(MACHINE(pcms), cpu->apic_id, &idx);
2355 if (!cpu_slot) {
2356 MachineState *ms = MACHINE(pcms);
2358 x86_topo_ids_from_apicid(cpu->apic_id, smp_cores, smp_threads, &topo);
2359 error_setg(errp, "Invalid CPU [socket: %u, core: %u, thread: %u] with"
2360 " APIC ID %" PRIu32 ", valid index range 0:%d",
2361 topo.pkg_id, topo.core_id, topo.smt_id, cpu->apic_id,
2362 ms->possible_cpus->len - 1);
2363 return;
2366 if (cpu_slot->cpu) {
2367 error_setg(errp, "CPU[%d] with APIC ID %" PRIu32 " exists",
2368 idx, cpu->apic_id);
2369 return;
2372 /* if 'address' properties socket-id/core-id/thread-id are not set, set them
2373 * so that machine_query_hotpluggable_cpus would show correct values
2375 /* TODO: move socket_id/core_id/thread_id checks into x86_cpu_realizefn()
2376 * once -smp refactoring is complete and there will be CPU private
2377 * CPUState::nr_cores and CPUState::nr_threads fields instead of globals */
2378 x86_topo_ids_from_apicid(cpu->apic_id, smp_cores, smp_threads, &topo);
2379 if (cpu->socket_id != -1 && cpu->socket_id != topo.pkg_id) {
2380 error_setg(errp, "property socket-id: %u doesn't match set apic-id:"
2381 " 0x%x (socket-id: %u)", cpu->socket_id, cpu->apic_id, topo.pkg_id);
2382 return;
2384 cpu->socket_id = topo.pkg_id;
2386 if (cpu->core_id != -1 && cpu->core_id != topo.core_id) {
2387 error_setg(errp, "property core-id: %u doesn't match set apic-id:"
2388 " 0x%x (core-id: %u)", cpu->core_id, cpu->apic_id, topo.core_id);
2389 return;
2391 cpu->core_id = topo.core_id;
2393 if (cpu->thread_id != -1 && cpu->thread_id != topo.smt_id) {
2394 error_setg(errp, "property thread-id: %u doesn't match set apic-id:"
2395 " 0x%x (thread-id: %u)", cpu->thread_id, cpu->apic_id, topo.smt_id);
2396 return;
2398 cpu->thread_id = topo.smt_id;
2400 if (cpu->hyperv_vpindex && !kvm_hv_vpindex_settable()) {
2401 error_setg(errp, "kernel doesn't allow setting HyperV VP_INDEX");
2402 return;
2405 cs = CPU(cpu);
2406 cs->cpu_index = idx;
2408 numa_cpu_pre_plug(cpu_slot, dev, errp);
2411 static void pc_machine_device_pre_plug_cb(HotplugHandler *hotplug_dev,
2412 DeviceState *dev, Error **errp)
2414 if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2415 pc_memory_pre_plug(hotplug_dev, dev, errp);
2416 } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
2417 pc_cpu_pre_plug(hotplug_dev, dev, errp);
2421 static void pc_machine_device_plug_cb(HotplugHandler *hotplug_dev,
2422 DeviceState *dev, Error **errp)
2424 if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2425 pc_memory_plug(hotplug_dev, dev, errp);
2426 } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
2427 pc_cpu_plug(hotplug_dev, dev, errp);
2431 static void pc_machine_device_unplug_request_cb(HotplugHandler *hotplug_dev,
2432 DeviceState *dev, Error **errp)
2434 if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2435 pc_memory_unplug_request(hotplug_dev, dev, errp);
2436 } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
2437 pc_cpu_unplug_request_cb(hotplug_dev, dev, errp);
2438 } else {
2439 error_setg(errp, "acpi: device unplug request for not supported device"
2440 " type: %s", object_get_typename(OBJECT(dev)));
2444 static void pc_machine_device_unplug_cb(HotplugHandler *hotplug_dev,
2445 DeviceState *dev, Error **errp)
2447 if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2448 pc_memory_unplug(hotplug_dev, dev, errp);
2449 } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
2450 pc_cpu_unplug_cb(hotplug_dev, dev, errp);
2451 } else {
2452 error_setg(errp, "acpi: device unplug for not supported device"
2453 " type: %s", object_get_typename(OBJECT(dev)));
2457 static HotplugHandler *pc_get_hotplug_handler(MachineState *machine,
2458 DeviceState *dev)
2460 if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM) ||
2461 object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
2462 return HOTPLUG_HANDLER(machine);
2465 return NULL;
2468 static void
2469 pc_machine_get_device_memory_region_size(Object *obj, Visitor *v,
2470 const char *name, void *opaque,
2471 Error **errp)
2473 MachineState *ms = MACHINE(obj);
2474 int64_t value = memory_region_size(&ms->device_memory->mr);
2476 visit_type_int(v, name, &value, errp);
2479 static void pc_machine_get_max_ram_below_4g(Object *obj, Visitor *v,
2480 const char *name, void *opaque,
2481 Error **errp)
2483 PCMachineState *pcms = PC_MACHINE(obj);
2484 uint64_t value = pcms->max_ram_below_4g;
2486 visit_type_size(v, name, &value, errp);
2489 static void pc_machine_set_max_ram_below_4g(Object *obj, Visitor *v,
2490 const char *name, void *opaque,
2491 Error **errp)
2493 PCMachineState *pcms = PC_MACHINE(obj);
2494 Error *error = NULL;
2495 uint64_t value;
2497 visit_type_size(v, name, &value, &error);
2498 if (error) {
2499 error_propagate(errp, error);
2500 return;
2502 if (value > 4 * GiB) {
2503 error_setg(&error,
2504 "Machine option 'max-ram-below-4g=%"PRIu64
2505 "' expects size less than or equal to 4G", value);
2506 error_propagate(errp, error);
2507 return;
2510 if (value < 1 * MiB) {
2511 warn_report("Only %" PRIu64 " bytes of RAM below the 4GiB boundary,"
2512 "BIOS may not work with less than 1MiB", value);
2515 pcms->max_ram_below_4g = value;
2518 static void pc_machine_get_vmport(Object *obj, Visitor *v, const char *name,
2519 void *opaque, Error **errp)
2521 PCMachineState *pcms = PC_MACHINE(obj);
2522 OnOffAuto vmport = pcms->vmport;
2524 visit_type_OnOffAuto(v, name, &vmport, errp);
2527 static void pc_machine_set_vmport(Object *obj, Visitor *v, const char *name,
2528 void *opaque, Error **errp)
2530 PCMachineState *pcms = PC_MACHINE(obj);
2532 visit_type_OnOffAuto(v, name, &pcms->vmport, errp);
2535 bool pc_machine_is_smm_enabled(PCMachineState *pcms)
2537 bool smm_available = false;
2539 if (pcms->smm == ON_OFF_AUTO_OFF) {
2540 return false;
2543 if (tcg_enabled() || qtest_enabled()) {
2544 smm_available = true;
2545 } else if (kvm_enabled()) {
2546 smm_available = kvm_has_smm();
2549 if (smm_available) {
2550 return true;
2553 if (pcms->smm == ON_OFF_AUTO_ON) {
2554 error_report("System Management Mode not supported by this hypervisor.");
2555 exit(1);
2557 return false;
2560 static void pc_machine_get_smm(Object *obj, Visitor *v, const char *name,
2561 void *opaque, Error **errp)
2563 PCMachineState *pcms = PC_MACHINE(obj);
2564 OnOffAuto smm = pcms->smm;
2566 visit_type_OnOffAuto(v, name, &smm, errp);
2569 static void pc_machine_set_smm(Object *obj, Visitor *v, const char *name,
2570 void *opaque, Error **errp)
2572 PCMachineState *pcms = PC_MACHINE(obj);
2574 visit_type_OnOffAuto(v, name, &pcms->smm, errp);
2577 static bool pc_machine_get_nvdimm(Object *obj, Error **errp)
2579 PCMachineState *pcms = PC_MACHINE(obj);
2581 return pcms->acpi_nvdimm_state.is_enabled;
2584 static void pc_machine_set_nvdimm(Object *obj, bool value, Error **errp)
2586 PCMachineState *pcms = PC_MACHINE(obj);
2588 pcms->acpi_nvdimm_state.is_enabled = value;
2591 static char *pc_machine_get_nvdimm_persistence(Object *obj, Error **errp)
2593 PCMachineState *pcms = PC_MACHINE(obj);
2595 return g_strdup(pcms->acpi_nvdimm_state.persistence_string);
2598 static void pc_machine_set_nvdimm_persistence(Object *obj, const char *value,
2599 Error **errp)
2601 PCMachineState *pcms = PC_MACHINE(obj);
2602 AcpiNVDIMMState *nvdimm_state = &pcms->acpi_nvdimm_state;
2604 if (strcmp(value, "cpu") == 0)
2605 nvdimm_state->persistence = 3;
2606 else if (strcmp(value, "mem-ctrl") == 0)
2607 nvdimm_state->persistence = 2;
2608 else {
2609 error_setg(errp, "-machine nvdimm-persistence=%s: unsupported option",
2610 value);
2611 return;
2614 g_free(nvdimm_state->persistence_string);
2615 nvdimm_state->persistence_string = g_strdup(value);
2618 static bool pc_machine_get_smbus(Object *obj, Error **errp)
2620 PCMachineState *pcms = PC_MACHINE(obj);
2622 return pcms->smbus_enabled;
2625 static void pc_machine_set_smbus(Object *obj, bool value, Error **errp)
2627 PCMachineState *pcms = PC_MACHINE(obj);
2629 pcms->smbus_enabled = value;
2632 static bool pc_machine_get_sata(Object *obj, Error **errp)
2634 PCMachineState *pcms = PC_MACHINE(obj);
2636 return pcms->sata_enabled;
2639 static void pc_machine_set_sata(Object *obj, bool value, Error **errp)
2641 PCMachineState *pcms = PC_MACHINE(obj);
2643 pcms->sata_enabled = value;
2646 static bool pc_machine_get_pit(Object *obj, Error **errp)
2648 PCMachineState *pcms = PC_MACHINE(obj);
2650 return pcms->pit_enabled;
2653 static void pc_machine_set_pit(Object *obj, bool value, Error **errp)
2655 PCMachineState *pcms = PC_MACHINE(obj);
2657 pcms->pit_enabled = value;
2660 static void pc_machine_initfn(Object *obj)
2662 PCMachineState *pcms = PC_MACHINE(obj);
2664 pcms->max_ram_below_4g = 0; /* use default */
2665 pcms->smm = ON_OFF_AUTO_AUTO;
2666 pcms->vmport = ON_OFF_AUTO_AUTO;
2667 /* nvdimm is disabled on default. */
2668 pcms->acpi_nvdimm_state.is_enabled = false;
2669 /* acpi build is enabled by default if machine supports it */
2670 pcms->acpi_build_enabled = PC_MACHINE_GET_CLASS(pcms)->has_acpi_build;
2671 pcms->smbus_enabled = true;
2672 pcms->sata_enabled = true;
2673 pcms->pit_enabled = true;
2676 static void pc_machine_reset(void)
2678 CPUState *cs;
2679 X86CPU *cpu;
2681 qemu_devices_reset();
2683 /* Reset APIC after devices have been reset to cancel
2684 * any changes that qemu_devices_reset() might have done.
2686 CPU_FOREACH(cs) {
2687 cpu = X86_CPU(cs);
2689 if (cpu->apic_state) {
2690 device_reset(cpu->apic_state);
2695 static CpuInstanceProperties
2696 pc_cpu_index_to_props(MachineState *ms, unsigned cpu_index)
2698 MachineClass *mc = MACHINE_GET_CLASS(ms);
2699 const CPUArchIdList *possible_cpus = mc->possible_cpu_arch_ids(ms);
2701 assert(cpu_index < possible_cpus->len);
2702 return possible_cpus->cpus[cpu_index].props;
2705 static int64_t pc_get_default_cpu_node_id(const MachineState *ms, int idx)
2707 X86CPUTopoInfo topo;
2709 assert(idx < ms->possible_cpus->len);
2710 x86_topo_ids_from_apicid(ms->possible_cpus->cpus[idx].arch_id,
2711 smp_cores, smp_threads, &topo);
2712 return topo.pkg_id % nb_numa_nodes;
2715 static const CPUArchIdList *pc_possible_cpu_arch_ids(MachineState *ms)
2717 int i;
2719 if (ms->possible_cpus) {
2721 * make sure that max_cpus hasn't changed since the first use, i.e.
2722 * -smp hasn't been parsed after it
2724 assert(ms->possible_cpus->len == max_cpus);
2725 return ms->possible_cpus;
2728 ms->possible_cpus = g_malloc0(sizeof(CPUArchIdList) +
2729 sizeof(CPUArchId) * max_cpus);
2730 ms->possible_cpus->len = max_cpus;
2731 for (i = 0; i < ms->possible_cpus->len; i++) {
2732 X86CPUTopoInfo topo;
2734 ms->possible_cpus->cpus[i].type = ms->cpu_type;
2735 ms->possible_cpus->cpus[i].vcpus_count = 1;
2736 ms->possible_cpus->cpus[i].arch_id = x86_cpu_apic_id_from_index(i);
2737 x86_topo_ids_from_apicid(ms->possible_cpus->cpus[i].arch_id,
2738 smp_cores, smp_threads, &topo);
2739 ms->possible_cpus->cpus[i].props.has_socket_id = true;
2740 ms->possible_cpus->cpus[i].props.socket_id = topo.pkg_id;
2741 ms->possible_cpus->cpus[i].props.has_core_id = true;
2742 ms->possible_cpus->cpus[i].props.core_id = topo.core_id;
2743 ms->possible_cpus->cpus[i].props.has_thread_id = true;
2744 ms->possible_cpus->cpus[i].props.thread_id = topo.smt_id;
2746 return ms->possible_cpus;
2749 static void x86_nmi(NMIState *n, int cpu_index, Error **errp)
2751 /* cpu index isn't used */
2752 CPUState *cs;
2754 CPU_FOREACH(cs) {
2755 X86CPU *cpu = X86_CPU(cs);
2757 if (!cpu->apic_state) {
2758 cpu_interrupt(cs, CPU_INTERRUPT_NMI);
2759 } else {
2760 apic_deliver_nmi(cpu->apic_state);
2765 static void pc_machine_class_init(ObjectClass *oc, void *data)
2767 MachineClass *mc = MACHINE_CLASS(oc);
2768 PCMachineClass *pcmc = PC_MACHINE_CLASS(oc);
2769 HotplugHandlerClass *hc = HOTPLUG_HANDLER_CLASS(oc);
2770 NMIClass *nc = NMI_CLASS(oc);
2772 pcmc->pci_enabled = true;
2773 pcmc->has_acpi_build = true;
2774 pcmc->rsdp_in_ram = true;
2775 pcmc->smbios_defaults = true;
2776 pcmc->smbios_uuid_encoded = true;
2777 pcmc->gigabyte_align = true;
2778 pcmc->has_reserved_memory = true;
2779 pcmc->kvmclock_enabled = true;
2780 pcmc->enforce_aligned_dimm = true;
2781 /* BIOS ACPI tables: 128K. Other BIOS datastructures: less than 4K reported
2782 * to be used at the moment, 32K should be enough for a while. */
2783 pcmc->acpi_data_size = 0x20000 + 0x8000;
2784 pcmc->save_tsc_khz = true;
2785 pcmc->linuxboot_dma_enabled = true;
2786 pcmc->pvh_enabled = true;
2787 assert(!mc->get_hotplug_handler);
2788 mc->get_hotplug_handler = pc_get_hotplug_handler;
2789 mc->cpu_index_to_instance_props = pc_cpu_index_to_props;
2790 mc->get_default_cpu_node_id = pc_get_default_cpu_node_id;
2791 mc->possible_cpu_arch_ids = pc_possible_cpu_arch_ids;
2792 mc->auto_enable_numa_with_memhp = true;
2793 mc->has_hotpluggable_cpus = true;
2794 mc->default_boot_order = "cad";
2795 mc->hot_add_cpu = pc_hot_add_cpu;
2796 mc->block_default_type = IF_IDE;
2797 mc->max_cpus = 255;
2798 mc->reset = pc_machine_reset;
2799 hc->pre_plug = pc_machine_device_pre_plug_cb;
2800 hc->plug = pc_machine_device_plug_cb;
2801 hc->unplug_request = pc_machine_device_unplug_request_cb;
2802 hc->unplug = pc_machine_device_unplug_cb;
2803 nc->nmi_monitor_handler = x86_nmi;
2804 mc->default_cpu_type = TARGET_DEFAULT_CPU_TYPE;
2806 object_class_property_add(oc, PC_MACHINE_DEVMEM_REGION_SIZE, "int",
2807 pc_machine_get_device_memory_region_size, NULL,
2808 NULL, NULL, &error_abort);
2810 object_class_property_add(oc, PC_MACHINE_MAX_RAM_BELOW_4G, "size",
2811 pc_machine_get_max_ram_below_4g, pc_machine_set_max_ram_below_4g,
2812 NULL, NULL, &error_abort);
2814 object_class_property_set_description(oc, PC_MACHINE_MAX_RAM_BELOW_4G,
2815 "Maximum ram below the 4G boundary (32bit boundary)", &error_abort);
2817 object_class_property_add(oc, PC_MACHINE_SMM, "OnOffAuto",
2818 pc_machine_get_smm, pc_machine_set_smm,
2819 NULL, NULL, &error_abort);
2820 object_class_property_set_description(oc, PC_MACHINE_SMM,
2821 "Enable SMM (pc & q35)", &error_abort);
2823 object_class_property_add(oc, PC_MACHINE_VMPORT, "OnOffAuto",
2824 pc_machine_get_vmport, pc_machine_set_vmport,
2825 NULL, NULL, &error_abort);
2826 object_class_property_set_description(oc, PC_MACHINE_VMPORT,
2827 "Enable vmport (pc & q35)", &error_abort);
2829 object_class_property_add_bool(oc, PC_MACHINE_NVDIMM,
2830 pc_machine_get_nvdimm, pc_machine_set_nvdimm, &error_abort);
2832 object_class_property_add_str(oc, PC_MACHINE_NVDIMM_PERSIST,
2833 pc_machine_get_nvdimm_persistence,
2834 pc_machine_set_nvdimm_persistence, &error_abort);
2836 object_class_property_add_bool(oc, PC_MACHINE_SMBUS,
2837 pc_machine_get_smbus, pc_machine_set_smbus, &error_abort);
2839 object_class_property_add_bool(oc, PC_MACHINE_SATA,
2840 pc_machine_get_sata, pc_machine_set_sata, &error_abort);
2842 object_class_property_add_bool(oc, PC_MACHINE_PIT,
2843 pc_machine_get_pit, pc_machine_set_pit, &error_abort);
2846 static const TypeInfo pc_machine_info = {
2847 .name = TYPE_PC_MACHINE,
2848 .parent = TYPE_MACHINE,
2849 .abstract = true,
2850 .instance_size = sizeof(PCMachineState),
2851 .instance_init = pc_machine_initfn,
2852 .class_size = sizeof(PCMachineClass),
2853 .class_init = pc_machine_class_init,
2854 .interfaces = (InterfaceInfo[]) {
2855 { TYPE_HOTPLUG_HANDLER },
2856 { TYPE_NMI },
2861 static void pc_machine_register_types(void)
2863 type_register_static(&pc_machine_info);
2866 type_init(pc_machine_register_types)