Merge remote-tracking branch 'remotes/armbru/tags/pull-qapi-2019-10-22-v3' into staging
[qemu.git] / hw / i386 / pc.c
blob4b1904237ec635c02851fee462f39dab3967c2f6
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/i386/pc.h"
28 #include "hw/char/serial.h"
29 #include "hw/char/parallel.h"
30 #include "hw/i386/apic.h"
31 #include "hw/i386/topology.h"
32 #include "hw/i386/fw_cfg.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 "migration/vmstate.h"
44 #include "multiboot.h"
45 #include "hw/timer/mc146818rtc.h"
46 #include "hw/dma/i8257.h"
47 #include "hw/timer/i8254.h"
48 #include "hw/input/i8042.h"
49 #include "hw/irq.h"
50 #include "hw/audio/pcspk.h"
51 #include "hw/pci/msi.h"
52 #include "hw/sysbus.h"
53 #include "sysemu/sysemu.h"
54 #include "sysemu/tcg.h"
55 #include "sysemu/numa.h"
56 #include "sysemu/kvm.h"
57 #include "sysemu/qtest.h"
58 #include "sysemu/reset.h"
59 #include "sysemu/runstate.h"
60 #include "kvm_i386.h"
61 #include "hw/xen/xen.h"
62 #include "hw/xen/start_info.h"
63 #include "ui/qemu-spice.h"
64 #include "exec/memory.h"
65 #include "exec/address-spaces.h"
66 #include "sysemu/arch_init.h"
67 #include "qemu/bitmap.h"
68 #include "qemu/config-file.h"
69 #include "qemu/error-report.h"
70 #include "qemu/option.h"
71 #include "hw/acpi/acpi.h"
72 #include "hw/acpi/cpu_hotplug.h"
73 #include "hw/boards.h"
74 #include "acpi-build.h"
75 #include "hw/mem/pc-dimm.h"
76 #include "qapi/error.h"
77 #include "qapi/qapi-visit-common.h"
78 #include "qapi/visitor.h"
79 #include "hw/core/cpu.h"
80 #include "hw/nmi.h"
81 #include "hw/usb.h"
82 #include "hw/i386/intel_iommu.h"
83 #include "hw/net/ne2000-isa.h"
84 #include "standard-headers/asm-x86/bootparam.h"
85 #include "hw/virtio/virtio-pmem-pci.h"
86 #include "hw/mem/memory-device.h"
87 #include "sysemu/replay.h"
88 #include "qapi/qmp/qerror.h"
89 #include "config-devices.h"
90 #include "e820_memory_layout.h"
91 #include "fw_cfg.h"
93 /* debug PC/ISA interrupts */
94 //#define DEBUG_IRQ
96 #ifdef DEBUG_IRQ
97 #define DPRINTF(fmt, ...) \
98 do { printf("CPUIRQ: " fmt , ## __VA_ARGS__); } while (0)
99 #else
100 #define DPRINTF(fmt, ...)
101 #endif
103 struct hpet_fw_config hpet_cfg = {.count = UINT8_MAX};
105 /* Physical Address of PVH entry point read from kernel ELF NOTE */
106 static size_t pvh_start_addr;
108 GlobalProperty pc_compat_4_1[] = {};
109 const size_t pc_compat_4_1_len = G_N_ELEMENTS(pc_compat_4_1);
111 GlobalProperty pc_compat_4_0[] = {};
112 const size_t pc_compat_4_0_len = G_N_ELEMENTS(pc_compat_4_0);
114 GlobalProperty pc_compat_3_1[] = {
115 { "intel-iommu", "dma-drain", "off" },
116 { "Opteron_G3" "-" TYPE_X86_CPU, "rdtscp", "off" },
117 { "Opteron_G4" "-" TYPE_X86_CPU, "rdtscp", "off" },
118 { "Opteron_G4" "-" TYPE_X86_CPU, "npt", "off" },
119 { "Opteron_G4" "-" TYPE_X86_CPU, "nrip-save", "off" },
120 { "Opteron_G5" "-" TYPE_X86_CPU, "rdtscp", "off" },
121 { "Opteron_G5" "-" TYPE_X86_CPU, "npt", "off" },
122 { "Opteron_G5" "-" TYPE_X86_CPU, "nrip-save", "off" },
123 { "EPYC" "-" TYPE_X86_CPU, "npt", "off" },
124 { "EPYC" "-" TYPE_X86_CPU, "nrip-save", "off" },
125 { "EPYC-IBPB" "-" TYPE_X86_CPU, "npt", "off" },
126 { "EPYC-IBPB" "-" TYPE_X86_CPU, "nrip-save", "off" },
127 { "Skylake-Client" "-" TYPE_X86_CPU, "mpx", "on" },
128 { "Skylake-Client-IBRS" "-" TYPE_X86_CPU, "mpx", "on" },
129 { "Skylake-Server" "-" TYPE_X86_CPU, "mpx", "on" },
130 { "Skylake-Server-IBRS" "-" TYPE_X86_CPU, "mpx", "on" },
131 { "Cascadelake-Server" "-" TYPE_X86_CPU, "mpx", "on" },
132 { "Icelake-Client" "-" TYPE_X86_CPU, "mpx", "on" },
133 { "Icelake-Server" "-" TYPE_X86_CPU, "mpx", "on" },
134 { "Cascadelake-Server" "-" TYPE_X86_CPU, "stepping", "5" },
135 { TYPE_X86_CPU, "x-intel-pt-auto-level", "off" },
137 const size_t pc_compat_3_1_len = G_N_ELEMENTS(pc_compat_3_1);
139 GlobalProperty pc_compat_3_0[] = {
140 { TYPE_X86_CPU, "x-hv-synic-kvm-only", "on" },
141 { "Skylake-Server" "-" TYPE_X86_CPU, "pku", "off" },
142 { "Skylake-Server-IBRS" "-" TYPE_X86_CPU, "pku", "off" },
144 const size_t pc_compat_3_0_len = G_N_ELEMENTS(pc_compat_3_0);
146 GlobalProperty pc_compat_2_12[] = {
147 { TYPE_X86_CPU, "legacy-cache", "on" },
148 { TYPE_X86_CPU, "topoext", "off" },
149 { "EPYC-" TYPE_X86_CPU, "xlevel", "0x8000000a" },
150 { "EPYC-IBPB-" TYPE_X86_CPU, "xlevel", "0x8000000a" },
152 const size_t pc_compat_2_12_len = G_N_ELEMENTS(pc_compat_2_12);
154 GlobalProperty pc_compat_2_11[] = {
155 { TYPE_X86_CPU, "x-migrate-smi-count", "off" },
156 { "Skylake-Server" "-" TYPE_X86_CPU, "clflushopt", "off" },
158 const size_t pc_compat_2_11_len = G_N_ELEMENTS(pc_compat_2_11);
160 GlobalProperty pc_compat_2_10[] = {
161 { TYPE_X86_CPU, "x-hv-max-vps", "0x40" },
162 { "i440FX-pcihost", "x-pci-hole64-fix", "off" },
163 { "q35-pcihost", "x-pci-hole64-fix", "off" },
165 const size_t pc_compat_2_10_len = G_N_ELEMENTS(pc_compat_2_10);
167 GlobalProperty pc_compat_2_9[] = {
168 { "mch", "extended-tseg-mbytes", "0" },
170 const size_t pc_compat_2_9_len = G_N_ELEMENTS(pc_compat_2_9);
172 GlobalProperty pc_compat_2_8[] = {
173 { TYPE_X86_CPU, "tcg-cpuid", "off" },
174 { "kvmclock", "x-mach-use-reliable-get-clock", "off" },
175 { "ICH9-LPC", "x-smi-broadcast", "off" },
176 { TYPE_X86_CPU, "vmware-cpuid-freq", "off" },
177 { "Haswell-" TYPE_X86_CPU, "stepping", "1" },
179 const size_t pc_compat_2_8_len = G_N_ELEMENTS(pc_compat_2_8);
181 GlobalProperty pc_compat_2_7[] = {
182 { TYPE_X86_CPU, "l3-cache", "off" },
183 { TYPE_X86_CPU, "full-cpuid-auto-level", "off" },
184 { "Opteron_G3" "-" TYPE_X86_CPU, "family", "15" },
185 { "Opteron_G3" "-" TYPE_X86_CPU, "model", "6" },
186 { "Opteron_G3" "-" TYPE_X86_CPU, "stepping", "1" },
187 { "isa-pcspk", "migrate", "off" },
189 const size_t pc_compat_2_7_len = G_N_ELEMENTS(pc_compat_2_7);
191 GlobalProperty pc_compat_2_6[] = {
192 { TYPE_X86_CPU, "cpuid-0xb", "off" },
193 { "vmxnet3", "romfile", "" },
194 { TYPE_X86_CPU, "fill-mtrr-mask", "off" },
195 { "apic-common", "legacy-instance-id", "on", }
197 const size_t pc_compat_2_6_len = G_N_ELEMENTS(pc_compat_2_6);
199 GlobalProperty pc_compat_2_5[] = {};
200 const size_t pc_compat_2_5_len = G_N_ELEMENTS(pc_compat_2_5);
202 GlobalProperty pc_compat_2_4[] = {
203 PC_CPU_MODEL_IDS("2.4.0")
204 { "Haswell-" TYPE_X86_CPU, "abm", "off" },
205 { "Haswell-noTSX-" TYPE_X86_CPU, "abm", "off" },
206 { "Broadwell-" TYPE_X86_CPU, "abm", "off" },
207 { "Broadwell-noTSX-" TYPE_X86_CPU, "abm", "off" },
208 { "host" "-" TYPE_X86_CPU, "host-cache-info", "on" },
209 { TYPE_X86_CPU, "check", "off" },
210 { "qemu64" "-" TYPE_X86_CPU, "sse4a", "on" },
211 { "qemu64" "-" TYPE_X86_CPU, "abm", "on" },
212 { "qemu64" "-" TYPE_X86_CPU, "popcnt", "on" },
213 { "qemu32" "-" TYPE_X86_CPU, "popcnt", "on" },
214 { "Opteron_G2" "-" TYPE_X86_CPU, "rdtscp", "on" },
215 { "Opteron_G3" "-" TYPE_X86_CPU, "rdtscp", "on" },
216 { "Opteron_G4" "-" TYPE_X86_CPU, "rdtscp", "on" },
217 { "Opteron_G5" "-" TYPE_X86_CPU, "rdtscp", "on", }
219 const size_t pc_compat_2_4_len = G_N_ELEMENTS(pc_compat_2_4);
221 GlobalProperty pc_compat_2_3[] = {
222 PC_CPU_MODEL_IDS("2.3.0")
223 { TYPE_X86_CPU, "arat", "off" },
224 { "qemu64" "-" TYPE_X86_CPU, "min-level", "4" },
225 { "kvm64" "-" TYPE_X86_CPU, "min-level", "5" },
226 { "pentium3" "-" TYPE_X86_CPU, "min-level", "2" },
227 { "n270" "-" TYPE_X86_CPU, "min-level", "5" },
228 { "Conroe" "-" TYPE_X86_CPU, "min-level", "4" },
229 { "Penryn" "-" TYPE_X86_CPU, "min-level", "4" },
230 { "Nehalem" "-" TYPE_X86_CPU, "min-level", "4" },
231 { "n270" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
232 { "Penryn" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
233 { "Conroe" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
234 { "Nehalem" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
235 { "Westmere" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
236 { "SandyBridge" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
237 { "IvyBridge" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
238 { "Haswell" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
239 { "Haswell-noTSX" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
240 { "Broadwell" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
241 { "Broadwell-noTSX" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
242 { TYPE_X86_CPU, "kvm-no-smi-migration", "on" },
244 const size_t pc_compat_2_3_len = G_N_ELEMENTS(pc_compat_2_3);
246 GlobalProperty pc_compat_2_2[] = {
247 PC_CPU_MODEL_IDS("2.2.0")
248 { "kvm64" "-" TYPE_X86_CPU, "vme", "off" },
249 { "kvm32" "-" TYPE_X86_CPU, "vme", "off" },
250 { "Conroe" "-" TYPE_X86_CPU, "vme", "off" },
251 { "Penryn" "-" TYPE_X86_CPU, "vme", "off" },
252 { "Nehalem" "-" TYPE_X86_CPU, "vme", "off" },
253 { "Westmere" "-" TYPE_X86_CPU, "vme", "off" },
254 { "SandyBridge" "-" TYPE_X86_CPU, "vme", "off" },
255 { "Haswell" "-" TYPE_X86_CPU, "vme", "off" },
256 { "Broadwell" "-" TYPE_X86_CPU, "vme", "off" },
257 { "Opteron_G1" "-" TYPE_X86_CPU, "vme", "off" },
258 { "Opteron_G2" "-" TYPE_X86_CPU, "vme", "off" },
259 { "Opteron_G3" "-" TYPE_X86_CPU, "vme", "off" },
260 { "Opteron_G4" "-" TYPE_X86_CPU, "vme", "off" },
261 { "Opteron_G5" "-" TYPE_X86_CPU, "vme", "off" },
262 { "Haswell" "-" TYPE_X86_CPU, "f16c", "off" },
263 { "Haswell" "-" TYPE_X86_CPU, "rdrand", "off" },
264 { "Broadwell" "-" TYPE_X86_CPU, "f16c", "off" },
265 { "Broadwell" "-" TYPE_X86_CPU, "rdrand", "off" },
267 const size_t pc_compat_2_2_len = G_N_ELEMENTS(pc_compat_2_2);
269 GlobalProperty pc_compat_2_1[] = {
270 PC_CPU_MODEL_IDS("2.1.0")
271 { "coreduo" "-" TYPE_X86_CPU, "vmx", "on" },
272 { "core2duo" "-" TYPE_X86_CPU, "vmx", "on" },
274 const size_t pc_compat_2_1_len = G_N_ELEMENTS(pc_compat_2_1);
276 GlobalProperty pc_compat_2_0[] = {
277 PC_CPU_MODEL_IDS("2.0.0")
278 { "virtio-scsi-pci", "any_layout", "off" },
279 { "PIIX4_PM", "memory-hotplug-support", "off" },
280 { "apic", "version", "0x11" },
281 { "nec-usb-xhci", "superspeed-ports-first", "off" },
282 { "nec-usb-xhci", "force-pcie-endcap", "on" },
283 { "pci-serial", "prog_if", "0" },
284 { "pci-serial-2x", "prog_if", "0" },
285 { "pci-serial-4x", "prog_if", "0" },
286 { "virtio-net-pci", "guest_announce", "off" },
287 { "ICH9-LPC", "memory-hotplug-support", "off" },
288 { "xio3130-downstream", COMPAT_PROP_PCP, "off" },
289 { "ioh3420", COMPAT_PROP_PCP, "off" },
291 const size_t pc_compat_2_0_len = G_N_ELEMENTS(pc_compat_2_0);
293 GlobalProperty pc_compat_1_7[] = {
294 PC_CPU_MODEL_IDS("1.7.0")
295 { TYPE_USB_DEVICE, "msos-desc", "no" },
296 { "PIIX4_PM", "acpi-pci-hotplug-with-bridge-support", "off" },
297 { "hpet", HPET_INTCAP, "4" },
299 const size_t pc_compat_1_7_len = G_N_ELEMENTS(pc_compat_1_7);
301 GlobalProperty pc_compat_1_6[] = {
302 PC_CPU_MODEL_IDS("1.6.0")
303 { "e1000", "mitigation", "off" },
304 { "qemu64-" TYPE_X86_CPU, "model", "2" },
305 { "qemu32-" TYPE_X86_CPU, "model", "3" },
306 { "i440FX-pcihost", "short_root_bus", "1" },
307 { "q35-pcihost", "short_root_bus", "1" },
309 const size_t pc_compat_1_6_len = G_N_ELEMENTS(pc_compat_1_6);
311 GlobalProperty pc_compat_1_5[] = {
312 PC_CPU_MODEL_IDS("1.5.0")
313 { "Conroe-" TYPE_X86_CPU, "model", "2" },
314 { "Conroe-" TYPE_X86_CPU, "min-level", "2" },
315 { "Penryn-" TYPE_X86_CPU, "model", "2" },
316 { "Penryn-" TYPE_X86_CPU, "min-level", "2" },
317 { "Nehalem-" TYPE_X86_CPU, "model", "2" },
318 { "Nehalem-" TYPE_X86_CPU, "min-level", "2" },
319 { "virtio-net-pci", "any_layout", "off" },
320 { TYPE_X86_CPU, "pmu", "on" },
321 { "i440FX-pcihost", "short_root_bus", "0" },
322 { "q35-pcihost", "short_root_bus", "0" },
324 const size_t pc_compat_1_5_len = G_N_ELEMENTS(pc_compat_1_5);
326 GlobalProperty pc_compat_1_4[] = {
327 PC_CPU_MODEL_IDS("1.4.0")
328 { "scsi-hd", "discard_granularity", "0" },
329 { "scsi-cd", "discard_granularity", "0" },
330 { "scsi-disk", "discard_granularity", "0" },
331 { "ide-hd", "discard_granularity", "0" },
332 { "ide-cd", "discard_granularity", "0" },
333 { "ide-drive", "discard_granularity", "0" },
334 { "virtio-blk-pci", "discard_granularity", "0" },
335 /* DEV_NVECTORS_UNSPECIFIED as a uint32_t string: */
336 { "virtio-serial-pci", "vectors", "0xFFFFFFFF" },
337 { "virtio-net-pci", "ctrl_guest_offloads", "off" },
338 { "e1000", "romfile", "pxe-e1000.rom" },
339 { "ne2k_pci", "romfile", "pxe-ne2k_pci.rom" },
340 { "pcnet", "romfile", "pxe-pcnet.rom" },
341 { "rtl8139", "romfile", "pxe-rtl8139.rom" },
342 { "virtio-net-pci", "romfile", "pxe-virtio.rom" },
343 { "486-" TYPE_X86_CPU, "model", "0" },
344 { "n270" "-" TYPE_X86_CPU, "movbe", "off" },
345 { "Westmere" "-" TYPE_X86_CPU, "pclmulqdq", "off" },
347 const size_t pc_compat_1_4_len = G_N_ELEMENTS(pc_compat_1_4);
349 void gsi_handler(void *opaque, int n, int level)
351 GSIState *s = opaque;
353 DPRINTF("pc: %s GSI %d\n", level ? "raising" : "lowering", n);
354 if (n < ISA_NUM_IRQS) {
355 qemu_set_irq(s->i8259_irq[n], level);
357 qemu_set_irq(s->ioapic_irq[n], level);
360 static void ioport80_write(void *opaque, hwaddr addr, uint64_t data,
361 unsigned size)
365 static uint64_t ioport80_read(void *opaque, hwaddr addr, unsigned size)
367 return 0xffffffffffffffffULL;
370 /* MSDOS compatibility mode FPU exception support */
371 static qemu_irq ferr_irq;
373 void pc_register_ferr_irq(qemu_irq irq)
375 ferr_irq = irq;
378 /* XXX: add IGNNE support */
379 void cpu_set_ferr(CPUX86State *s)
381 qemu_irq_raise(ferr_irq);
384 static void ioportF0_write(void *opaque, hwaddr addr, uint64_t data,
385 unsigned size)
387 qemu_irq_lower(ferr_irq);
390 static uint64_t ioportF0_read(void *opaque, hwaddr addr, unsigned size)
392 return 0xffffffffffffffffULL;
395 /* TSC handling */
396 uint64_t cpu_get_tsc(CPUX86State *env)
398 return cpu_get_ticks();
401 /* IRQ handling */
402 int cpu_get_pic_interrupt(CPUX86State *env)
404 X86CPU *cpu = env_archcpu(env);
405 int intno;
407 if (!kvm_irqchip_in_kernel()) {
408 intno = apic_get_interrupt(cpu->apic_state);
409 if (intno >= 0) {
410 return intno;
412 /* read the irq from the PIC */
413 if (!apic_accept_pic_intr(cpu->apic_state)) {
414 return -1;
418 intno = pic_read_irq(isa_pic);
419 return intno;
422 static void pic_irq_request(void *opaque, int irq, int level)
424 CPUState *cs = first_cpu;
425 X86CPU *cpu = X86_CPU(cs);
427 DPRINTF("pic_irqs: %s irq %d\n", level? "raise" : "lower", irq);
428 if (cpu->apic_state && !kvm_irqchip_in_kernel()) {
429 CPU_FOREACH(cs) {
430 cpu = X86_CPU(cs);
431 if (apic_accept_pic_intr(cpu->apic_state)) {
432 apic_deliver_pic_intr(cpu->apic_state, level);
435 } else {
436 if (level) {
437 cpu_interrupt(cs, CPU_INTERRUPT_HARD);
438 } else {
439 cpu_reset_interrupt(cs, CPU_INTERRUPT_HARD);
444 /* PC cmos mappings */
446 #define REG_EQUIPMENT_BYTE 0x14
448 int cmos_get_fd_drive_type(FloppyDriveType fd0)
450 int val;
452 switch (fd0) {
453 case FLOPPY_DRIVE_TYPE_144:
454 /* 1.44 Mb 3"5 drive */
455 val = 4;
456 break;
457 case FLOPPY_DRIVE_TYPE_288:
458 /* 2.88 Mb 3"5 drive */
459 val = 5;
460 break;
461 case FLOPPY_DRIVE_TYPE_120:
462 /* 1.2 Mb 5"5 drive */
463 val = 2;
464 break;
465 case FLOPPY_DRIVE_TYPE_NONE:
466 default:
467 val = 0;
468 break;
470 return val;
473 static void cmos_init_hd(ISADevice *s, int type_ofs, int info_ofs,
474 int16_t cylinders, int8_t heads, int8_t sectors)
476 rtc_set_memory(s, type_ofs, 47);
477 rtc_set_memory(s, info_ofs, cylinders);
478 rtc_set_memory(s, info_ofs + 1, cylinders >> 8);
479 rtc_set_memory(s, info_ofs + 2, heads);
480 rtc_set_memory(s, info_ofs + 3, 0xff);
481 rtc_set_memory(s, info_ofs + 4, 0xff);
482 rtc_set_memory(s, info_ofs + 5, 0xc0 | ((heads > 8) << 3));
483 rtc_set_memory(s, info_ofs + 6, cylinders);
484 rtc_set_memory(s, info_ofs + 7, cylinders >> 8);
485 rtc_set_memory(s, info_ofs + 8, sectors);
488 /* convert boot_device letter to something recognizable by the bios */
489 static int boot_device2nibble(char boot_device)
491 switch(boot_device) {
492 case 'a':
493 case 'b':
494 return 0x01; /* floppy boot */
495 case 'c':
496 return 0x02; /* hard drive boot */
497 case 'd':
498 return 0x03; /* CD-ROM boot */
499 case 'n':
500 return 0x04; /* Network boot */
502 return 0;
505 static void set_boot_dev(ISADevice *s, const char *boot_device, Error **errp)
507 #define PC_MAX_BOOT_DEVICES 3
508 int nbds, bds[3] = { 0, };
509 int i;
511 nbds = strlen(boot_device);
512 if (nbds > PC_MAX_BOOT_DEVICES) {
513 error_setg(errp, "Too many boot devices for PC");
514 return;
516 for (i = 0; i < nbds; i++) {
517 bds[i] = boot_device2nibble(boot_device[i]);
518 if (bds[i] == 0) {
519 error_setg(errp, "Invalid boot device for PC: '%c'",
520 boot_device[i]);
521 return;
524 rtc_set_memory(s, 0x3d, (bds[1] << 4) | bds[0]);
525 rtc_set_memory(s, 0x38, (bds[2] << 4) | (fd_bootchk ? 0x0 : 0x1));
528 static void pc_boot_set(void *opaque, const char *boot_device, Error **errp)
530 set_boot_dev(opaque, boot_device, errp);
533 static void pc_cmos_init_floppy(ISADevice *rtc_state, ISADevice *floppy)
535 int val, nb, i;
536 FloppyDriveType fd_type[2] = { FLOPPY_DRIVE_TYPE_NONE,
537 FLOPPY_DRIVE_TYPE_NONE };
539 /* floppy type */
540 if (floppy) {
541 for (i = 0; i < 2; i++) {
542 fd_type[i] = isa_fdc_get_drive_type(floppy, i);
545 val = (cmos_get_fd_drive_type(fd_type[0]) << 4) |
546 cmos_get_fd_drive_type(fd_type[1]);
547 rtc_set_memory(rtc_state, 0x10, val);
549 val = rtc_get_memory(rtc_state, REG_EQUIPMENT_BYTE);
550 nb = 0;
551 if (fd_type[0] != FLOPPY_DRIVE_TYPE_NONE) {
552 nb++;
554 if (fd_type[1] != FLOPPY_DRIVE_TYPE_NONE) {
555 nb++;
557 switch (nb) {
558 case 0:
559 break;
560 case 1:
561 val |= 0x01; /* 1 drive, ready for boot */
562 break;
563 case 2:
564 val |= 0x41; /* 2 drives, ready for boot */
565 break;
567 rtc_set_memory(rtc_state, REG_EQUIPMENT_BYTE, val);
570 typedef struct pc_cmos_init_late_arg {
571 ISADevice *rtc_state;
572 BusState *idebus[2];
573 } pc_cmos_init_late_arg;
575 typedef struct check_fdc_state {
576 ISADevice *floppy;
577 bool multiple;
578 } CheckFdcState;
580 static int check_fdc(Object *obj, void *opaque)
582 CheckFdcState *state = opaque;
583 Object *fdc;
584 uint32_t iobase;
585 Error *local_err = NULL;
587 fdc = object_dynamic_cast(obj, TYPE_ISA_FDC);
588 if (!fdc) {
589 return 0;
592 iobase = object_property_get_uint(obj, "iobase", &local_err);
593 if (local_err || iobase != 0x3f0) {
594 error_free(local_err);
595 return 0;
598 if (state->floppy) {
599 state->multiple = true;
600 } else {
601 state->floppy = ISA_DEVICE(obj);
603 return 0;
606 static const char * const fdc_container_path[] = {
607 "/unattached", "/peripheral", "/peripheral-anon"
611 * Locate the FDC at IO address 0x3f0, in order to configure the CMOS registers
612 * and ACPI objects.
614 ISADevice *pc_find_fdc0(void)
616 int i;
617 Object *container;
618 CheckFdcState state = { 0 };
620 for (i = 0; i < ARRAY_SIZE(fdc_container_path); i++) {
621 container = container_get(qdev_get_machine(), fdc_container_path[i]);
622 object_child_foreach(container, check_fdc, &state);
625 if (state.multiple) {
626 warn_report("multiple floppy disk controllers with "
627 "iobase=0x3f0 have been found");
628 error_printf("the one being picked for CMOS setup might not reflect "
629 "your intent");
632 return state.floppy;
635 static void pc_cmos_init_late(void *opaque)
637 pc_cmos_init_late_arg *arg = opaque;
638 ISADevice *s = arg->rtc_state;
639 int16_t cylinders;
640 int8_t heads, sectors;
641 int val;
642 int i, trans;
644 val = 0;
645 if (arg->idebus[0] && ide_get_geometry(arg->idebus[0], 0,
646 &cylinders, &heads, &sectors) >= 0) {
647 cmos_init_hd(s, 0x19, 0x1b, cylinders, heads, sectors);
648 val |= 0xf0;
650 if (arg->idebus[0] && ide_get_geometry(arg->idebus[0], 1,
651 &cylinders, &heads, &sectors) >= 0) {
652 cmos_init_hd(s, 0x1a, 0x24, cylinders, heads, sectors);
653 val |= 0x0f;
655 rtc_set_memory(s, 0x12, val);
657 val = 0;
658 for (i = 0; i < 4; i++) {
659 /* NOTE: ide_get_geometry() returns the physical
660 geometry. It is always such that: 1 <= sects <= 63, 1
661 <= heads <= 16, 1 <= cylinders <= 16383. The BIOS
662 geometry can be different if a translation is done. */
663 if (arg->idebus[i / 2] &&
664 ide_get_geometry(arg->idebus[i / 2], i % 2,
665 &cylinders, &heads, &sectors) >= 0) {
666 trans = ide_get_bios_chs_trans(arg->idebus[i / 2], i % 2) - 1;
667 assert((trans & ~3) == 0);
668 val |= trans << (i * 2);
671 rtc_set_memory(s, 0x39, val);
673 pc_cmos_init_floppy(s, pc_find_fdc0());
675 qemu_unregister_reset(pc_cmos_init_late, opaque);
678 void pc_cmos_init(PCMachineState *pcms,
679 BusState *idebus0, BusState *idebus1,
680 ISADevice *s)
682 int val;
683 static pc_cmos_init_late_arg arg;
685 /* various important CMOS locations needed by PC/Bochs bios */
687 /* memory size */
688 /* base memory (first MiB) */
689 val = MIN(pcms->below_4g_mem_size / KiB, 640);
690 rtc_set_memory(s, 0x15, val);
691 rtc_set_memory(s, 0x16, val >> 8);
692 /* extended memory (next 64MiB) */
693 if (pcms->below_4g_mem_size > 1 * MiB) {
694 val = (pcms->below_4g_mem_size - 1 * MiB) / KiB;
695 } else {
696 val = 0;
698 if (val > 65535)
699 val = 65535;
700 rtc_set_memory(s, 0x17, val);
701 rtc_set_memory(s, 0x18, val >> 8);
702 rtc_set_memory(s, 0x30, val);
703 rtc_set_memory(s, 0x31, val >> 8);
704 /* memory between 16MiB and 4GiB */
705 if (pcms->below_4g_mem_size > 16 * MiB) {
706 val = (pcms->below_4g_mem_size - 16 * MiB) / (64 * KiB);
707 } else {
708 val = 0;
710 if (val > 65535)
711 val = 65535;
712 rtc_set_memory(s, 0x34, val);
713 rtc_set_memory(s, 0x35, val >> 8);
714 /* memory above 4GiB */
715 val = pcms->above_4g_mem_size / 65536;
716 rtc_set_memory(s, 0x5b, val);
717 rtc_set_memory(s, 0x5c, val >> 8);
718 rtc_set_memory(s, 0x5d, val >> 16);
720 object_property_add_link(OBJECT(pcms), "rtc_state",
721 TYPE_ISA_DEVICE,
722 (Object **)&pcms->rtc,
723 object_property_allow_set_link,
724 OBJ_PROP_LINK_STRONG, &error_abort);
725 object_property_set_link(OBJECT(pcms), OBJECT(s),
726 "rtc_state", &error_abort);
728 set_boot_dev(s, MACHINE(pcms)->boot_order, &error_fatal);
730 val = 0;
731 val |= 0x02; /* FPU is there */
732 val |= 0x04; /* PS/2 mouse installed */
733 rtc_set_memory(s, REG_EQUIPMENT_BYTE, val);
735 /* hard drives and FDC */
736 arg.rtc_state = s;
737 arg.idebus[0] = idebus0;
738 arg.idebus[1] = idebus1;
739 qemu_register_reset(pc_cmos_init_late, &arg);
742 #define TYPE_PORT92 "port92"
743 #define PORT92(obj) OBJECT_CHECK(Port92State, (obj), TYPE_PORT92)
745 /* port 92 stuff: could be split off */
746 typedef struct Port92State {
747 ISADevice parent_obj;
749 MemoryRegion io;
750 uint8_t outport;
751 qemu_irq a20_out;
752 } Port92State;
754 static void port92_write(void *opaque, hwaddr addr, uint64_t val,
755 unsigned size)
757 Port92State *s = opaque;
758 int oldval = s->outport;
760 DPRINTF("port92: write 0x%02" PRIx64 "\n", val);
761 s->outport = val;
762 qemu_set_irq(s->a20_out, (val >> 1) & 1);
763 if ((val & 1) && !(oldval & 1)) {
764 qemu_system_reset_request(SHUTDOWN_CAUSE_GUEST_RESET);
768 static uint64_t port92_read(void *opaque, hwaddr addr,
769 unsigned size)
771 Port92State *s = opaque;
772 uint32_t ret;
774 ret = s->outport;
775 DPRINTF("port92: read 0x%02x\n", ret);
776 return ret;
779 static void port92_init(ISADevice *dev, qemu_irq a20_out)
781 qdev_connect_gpio_out_named(DEVICE(dev), PORT92_A20_LINE, 0, a20_out);
784 static const VMStateDescription vmstate_port92_isa = {
785 .name = "port92",
786 .version_id = 1,
787 .minimum_version_id = 1,
788 .fields = (VMStateField[]) {
789 VMSTATE_UINT8(outport, Port92State),
790 VMSTATE_END_OF_LIST()
794 static void port92_reset(DeviceState *d)
796 Port92State *s = PORT92(d);
798 s->outport &= ~1;
801 static const MemoryRegionOps port92_ops = {
802 .read = port92_read,
803 .write = port92_write,
804 .impl = {
805 .min_access_size = 1,
806 .max_access_size = 1,
808 .endianness = DEVICE_LITTLE_ENDIAN,
811 static void port92_initfn(Object *obj)
813 Port92State *s = PORT92(obj);
815 memory_region_init_io(&s->io, OBJECT(s), &port92_ops, s, "port92", 1);
817 s->outport = 0;
819 qdev_init_gpio_out_named(DEVICE(obj), &s->a20_out, PORT92_A20_LINE, 1);
822 static void port92_realizefn(DeviceState *dev, Error **errp)
824 ISADevice *isadev = ISA_DEVICE(dev);
825 Port92State *s = PORT92(dev);
827 isa_register_ioport(isadev, &s->io, 0x92);
830 static void port92_class_initfn(ObjectClass *klass, void *data)
832 DeviceClass *dc = DEVICE_CLASS(klass);
834 dc->realize = port92_realizefn;
835 dc->reset = port92_reset;
836 dc->vmsd = &vmstate_port92_isa;
838 * Reason: unlike ordinary ISA devices, this one needs additional
839 * wiring: its A20 output line needs to be wired up by
840 * port92_init().
842 dc->user_creatable = false;
845 static const TypeInfo port92_info = {
846 .name = TYPE_PORT92,
847 .parent = TYPE_ISA_DEVICE,
848 .instance_size = sizeof(Port92State),
849 .instance_init = port92_initfn,
850 .class_init = port92_class_initfn,
853 static void port92_register_types(void)
855 type_register_static(&port92_info);
858 type_init(port92_register_types)
860 static void handle_a20_line_change(void *opaque, int irq, int level)
862 X86CPU *cpu = opaque;
864 /* XXX: send to all CPUs ? */
865 /* XXX: add logic to handle multiple A20 line sources */
866 x86_cpu_set_a20(cpu, level);
869 /* Calculates initial APIC ID for a specific CPU index
871 * Currently we need to be able to calculate the APIC ID from the CPU index
872 * alone (without requiring a CPU object), as the QEMU<->Seabios interfaces have
873 * no concept of "CPU index", and the NUMA tables on fw_cfg need the APIC ID of
874 * all CPUs up to max_cpus.
876 static uint32_t x86_cpu_apic_id_from_index(PCMachineState *pcms,
877 unsigned int cpu_index)
879 MachineState *ms = MACHINE(pcms);
880 PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
881 uint32_t correct_id;
882 static bool warned;
884 correct_id = x86_apicid_from_cpu_idx(pcms->smp_dies, ms->smp.cores,
885 ms->smp.threads, cpu_index);
886 if (pcmc->compat_apic_id_mode) {
887 if (cpu_index != correct_id && !warned && !qtest_enabled()) {
888 error_report("APIC IDs set in compatibility mode, "
889 "CPU topology won't match the configuration");
890 warned = true;
892 return cpu_index;
893 } else {
894 return correct_id;
898 static long get_file_size(FILE *f)
900 long where, size;
902 /* XXX: on Unix systems, using fstat() probably makes more sense */
904 where = ftell(f);
905 fseek(f, 0, SEEK_END);
906 size = ftell(f);
907 fseek(f, where, SEEK_SET);
909 return size;
912 struct setup_data {
913 uint64_t next;
914 uint32_t type;
915 uint32_t len;
916 uint8_t data[0];
917 } __attribute__((packed));
921 * The entry point into the kernel for PVH boot is different from
922 * the native entry point. The PVH entry is defined by the x86/HVM
923 * direct boot ABI and is available in an ELFNOTE in the kernel binary.
925 * This function is passed to load_elf() when it is called from
926 * load_elfboot() which then additionally checks for an ELF Note of
927 * type XEN_ELFNOTE_PHYS32_ENTRY and passes it to this function to
928 * parse the PVH entry address from the ELF Note.
930 * Due to trickery in elf_opts.h, load_elf() is actually available as
931 * load_elf32() or load_elf64() and this routine needs to be able
932 * to deal with being called as 32 or 64 bit.
934 * The address of the PVH entry point is saved to the 'pvh_start_addr'
935 * global variable. (although the entry point is 32-bit, the kernel
936 * binary can be either 32-bit or 64-bit).
938 static uint64_t read_pvh_start_addr(void *arg1, void *arg2, bool is64)
940 size_t *elf_note_data_addr;
942 /* Check if ELF Note header passed in is valid */
943 if (arg1 == NULL) {
944 return 0;
947 if (is64) {
948 struct elf64_note *nhdr64 = (struct elf64_note *)arg1;
949 uint64_t nhdr_size64 = sizeof(struct elf64_note);
950 uint64_t phdr_align = *(uint64_t *)arg2;
951 uint64_t nhdr_namesz = nhdr64->n_namesz;
953 elf_note_data_addr =
954 ((void *)nhdr64) + nhdr_size64 +
955 QEMU_ALIGN_UP(nhdr_namesz, phdr_align);
956 } else {
957 struct elf32_note *nhdr32 = (struct elf32_note *)arg1;
958 uint32_t nhdr_size32 = sizeof(struct elf32_note);
959 uint32_t phdr_align = *(uint32_t *)arg2;
960 uint32_t nhdr_namesz = nhdr32->n_namesz;
962 elf_note_data_addr =
963 ((void *)nhdr32) + nhdr_size32 +
964 QEMU_ALIGN_UP(nhdr_namesz, phdr_align);
967 pvh_start_addr = *elf_note_data_addr;
969 return pvh_start_addr;
972 static bool load_elfboot(const char *kernel_filename,
973 int kernel_file_size,
974 uint8_t *header,
975 size_t pvh_xen_start_addr,
976 FWCfgState *fw_cfg)
978 uint32_t flags = 0;
979 uint32_t mh_load_addr = 0;
980 uint32_t elf_kernel_size = 0;
981 uint64_t elf_entry;
982 uint64_t elf_low, elf_high;
983 int kernel_size;
985 if (ldl_p(header) != 0x464c457f) {
986 return false; /* no elfboot */
989 bool elf_is64 = header[EI_CLASS] == ELFCLASS64;
990 flags = elf_is64 ?
991 ((Elf64_Ehdr *)header)->e_flags : ((Elf32_Ehdr *)header)->e_flags;
993 if (flags & 0x00010004) { /* LOAD_ELF_HEADER_HAS_ADDR */
994 error_report("elfboot unsupported flags = %x", flags);
995 exit(1);
998 uint64_t elf_note_type = XEN_ELFNOTE_PHYS32_ENTRY;
999 kernel_size = load_elf(kernel_filename, read_pvh_start_addr,
1000 NULL, &elf_note_type, &elf_entry,
1001 &elf_low, &elf_high, 0, I386_ELF_MACHINE,
1002 0, 0);
1004 if (kernel_size < 0) {
1005 error_report("Error while loading elf kernel");
1006 exit(1);
1008 mh_load_addr = elf_low;
1009 elf_kernel_size = elf_high - elf_low;
1011 if (pvh_start_addr == 0) {
1012 error_report("Error loading uncompressed kernel without PVH ELF Note");
1013 exit(1);
1015 fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_ENTRY, pvh_start_addr);
1016 fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_ADDR, mh_load_addr);
1017 fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_SIZE, elf_kernel_size);
1019 return true;
1022 static void load_linux(PCMachineState *pcms,
1023 FWCfgState *fw_cfg)
1025 uint16_t protocol;
1026 int setup_size, kernel_size, cmdline_size;
1027 int dtb_size, setup_data_offset;
1028 uint32_t initrd_max;
1029 uint8_t header[8192], *setup, *kernel;
1030 hwaddr real_addr, prot_addr, cmdline_addr, initrd_addr = 0;
1031 FILE *f;
1032 char *vmode;
1033 MachineState *machine = MACHINE(pcms);
1034 PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
1035 struct setup_data *setup_data;
1036 const char *kernel_filename = machine->kernel_filename;
1037 const char *initrd_filename = machine->initrd_filename;
1038 const char *dtb_filename = machine->dtb;
1039 const char *kernel_cmdline = machine->kernel_cmdline;
1041 /* Align to 16 bytes as a paranoia measure */
1042 cmdline_size = (strlen(kernel_cmdline)+16) & ~15;
1044 /* load the kernel header */
1045 f = fopen(kernel_filename, "rb");
1046 if (!f || !(kernel_size = get_file_size(f)) ||
1047 fread(header, 1, MIN(ARRAY_SIZE(header), kernel_size), f) !=
1048 MIN(ARRAY_SIZE(header), kernel_size)) {
1049 fprintf(stderr, "qemu: could not load kernel '%s': %s\n",
1050 kernel_filename, strerror(errno));
1051 exit(1);
1054 /* kernel protocol version */
1055 #if 0
1056 fprintf(stderr, "header magic: %#x\n", ldl_p(header+0x202));
1057 #endif
1058 if (ldl_p(header+0x202) == 0x53726448) {
1059 protocol = lduw_p(header+0x206);
1060 } else {
1062 * This could be a multiboot kernel. If it is, let's stop treating it
1063 * like a Linux kernel.
1064 * Note: some multiboot images could be in the ELF format (the same of
1065 * PVH), so we try multiboot first since we check the multiboot magic
1066 * header before to load it.
1068 if (load_multiboot(fw_cfg, f, kernel_filename, initrd_filename,
1069 kernel_cmdline, kernel_size, header)) {
1070 return;
1073 * Check if the file is an uncompressed kernel file (ELF) and load it,
1074 * saving the PVH entry point used by the x86/HVM direct boot ABI.
1075 * If load_elfboot() is successful, populate the fw_cfg info.
1077 if (pcmc->pvh_enabled &&
1078 load_elfboot(kernel_filename, kernel_size,
1079 header, pvh_start_addr, fw_cfg)) {
1080 fclose(f);
1082 fw_cfg_add_i32(fw_cfg, FW_CFG_CMDLINE_SIZE,
1083 strlen(kernel_cmdline) + 1);
1084 fw_cfg_add_string(fw_cfg, FW_CFG_CMDLINE_DATA, kernel_cmdline);
1086 fw_cfg_add_i32(fw_cfg, FW_CFG_SETUP_SIZE, sizeof(header));
1087 fw_cfg_add_bytes(fw_cfg, FW_CFG_SETUP_DATA,
1088 header, sizeof(header));
1090 /* load initrd */
1091 if (initrd_filename) {
1092 GMappedFile *mapped_file;
1093 gsize initrd_size;
1094 gchar *initrd_data;
1095 GError *gerr = NULL;
1097 mapped_file = g_mapped_file_new(initrd_filename, false, &gerr);
1098 if (!mapped_file) {
1099 fprintf(stderr, "qemu: error reading initrd %s: %s\n",
1100 initrd_filename, gerr->message);
1101 exit(1);
1103 pcms->initrd_mapped_file = mapped_file;
1105 initrd_data = g_mapped_file_get_contents(mapped_file);
1106 initrd_size = g_mapped_file_get_length(mapped_file);
1107 initrd_max = pcms->below_4g_mem_size - pcmc->acpi_data_size - 1;
1108 if (initrd_size >= initrd_max) {
1109 fprintf(stderr, "qemu: initrd is too large, cannot support."
1110 "(max: %"PRIu32", need %"PRId64")\n",
1111 initrd_max, (uint64_t)initrd_size);
1112 exit(1);
1115 initrd_addr = (initrd_max - initrd_size) & ~4095;
1117 fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_ADDR, initrd_addr);
1118 fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_SIZE, initrd_size);
1119 fw_cfg_add_bytes(fw_cfg, FW_CFG_INITRD_DATA, initrd_data,
1120 initrd_size);
1123 option_rom[nb_option_roms].bootindex = 0;
1124 option_rom[nb_option_roms].name = "pvh.bin";
1125 nb_option_roms++;
1127 return;
1129 protocol = 0;
1132 if (protocol < 0x200 || !(header[0x211] & 0x01)) {
1133 /* Low kernel */
1134 real_addr = 0x90000;
1135 cmdline_addr = 0x9a000 - cmdline_size;
1136 prot_addr = 0x10000;
1137 } else if (protocol < 0x202) {
1138 /* High but ancient kernel */
1139 real_addr = 0x90000;
1140 cmdline_addr = 0x9a000 - cmdline_size;
1141 prot_addr = 0x100000;
1142 } else {
1143 /* High and recent kernel */
1144 real_addr = 0x10000;
1145 cmdline_addr = 0x20000;
1146 prot_addr = 0x100000;
1149 #if 0
1150 fprintf(stderr,
1151 "qemu: real_addr = 0x" TARGET_FMT_plx "\n"
1152 "qemu: cmdline_addr = 0x" TARGET_FMT_plx "\n"
1153 "qemu: prot_addr = 0x" TARGET_FMT_plx "\n",
1154 real_addr,
1155 cmdline_addr,
1156 prot_addr);
1157 #endif
1159 /* highest address for loading the initrd */
1160 if (protocol >= 0x20c &&
1161 lduw_p(header+0x236) & XLF_CAN_BE_LOADED_ABOVE_4G) {
1163 * Linux has supported initrd up to 4 GB for a very long time (2007,
1164 * long before XLF_CAN_BE_LOADED_ABOVE_4G which was added in 2013),
1165 * though it only sets initrd_max to 2 GB to "work around bootloader
1166 * bugs". Luckily, QEMU firmware(which does something like bootloader)
1167 * has supported this.
1169 * It's believed that if XLF_CAN_BE_LOADED_ABOVE_4G is set, initrd can
1170 * be loaded into any address.
1172 * In addition, initrd_max is uint32_t simply because QEMU doesn't
1173 * support the 64-bit boot protocol (specifically the ext_ramdisk_image
1174 * field).
1176 * Therefore here just limit initrd_max to UINT32_MAX simply as well.
1178 initrd_max = UINT32_MAX;
1179 } else if (protocol >= 0x203) {
1180 initrd_max = ldl_p(header+0x22c);
1181 } else {
1182 initrd_max = 0x37ffffff;
1185 if (initrd_max >= pcms->below_4g_mem_size - pcmc->acpi_data_size) {
1186 initrd_max = pcms->below_4g_mem_size - pcmc->acpi_data_size - 1;
1189 fw_cfg_add_i32(fw_cfg, FW_CFG_CMDLINE_ADDR, cmdline_addr);
1190 fw_cfg_add_i32(fw_cfg, FW_CFG_CMDLINE_SIZE, strlen(kernel_cmdline)+1);
1191 fw_cfg_add_string(fw_cfg, FW_CFG_CMDLINE_DATA, kernel_cmdline);
1193 if (protocol >= 0x202) {
1194 stl_p(header+0x228, cmdline_addr);
1195 } else {
1196 stw_p(header+0x20, 0xA33F);
1197 stw_p(header+0x22, cmdline_addr-real_addr);
1200 /* handle vga= parameter */
1201 vmode = strstr(kernel_cmdline, "vga=");
1202 if (vmode) {
1203 unsigned int video_mode;
1204 /* skip "vga=" */
1205 vmode += 4;
1206 if (!strncmp(vmode, "normal", 6)) {
1207 video_mode = 0xffff;
1208 } else if (!strncmp(vmode, "ext", 3)) {
1209 video_mode = 0xfffe;
1210 } else if (!strncmp(vmode, "ask", 3)) {
1211 video_mode = 0xfffd;
1212 } else {
1213 video_mode = strtol(vmode, NULL, 0);
1215 stw_p(header+0x1fa, video_mode);
1218 /* loader type */
1219 /* High nybble = B reserved for QEMU; low nybble is revision number.
1220 If this code is substantially changed, you may want to consider
1221 incrementing the revision. */
1222 if (protocol >= 0x200) {
1223 header[0x210] = 0xB0;
1225 /* heap */
1226 if (protocol >= 0x201) {
1227 header[0x211] |= 0x80; /* CAN_USE_HEAP */
1228 stw_p(header+0x224, cmdline_addr-real_addr-0x200);
1231 /* load initrd */
1232 if (initrd_filename) {
1233 GMappedFile *mapped_file;
1234 gsize initrd_size;
1235 gchar *initrd_data;
1236 GError *gerr = NULL;
1238 if (protocol < 0x200) {
1239 fprintf(stderr, "qemu: linux kernel too old to load a ram disk\n");
1240 exit(1);
1243 mapped_file = g_mapped_file_new(initrd_filename, false, &gerr);
1244 if (!mapped_file) {
1245 fprintf(stderr, "qemu: error reading initrd %s: %s\n",
1246 initrd_filename, gerr->message);
1247 exit(1);
1249 pcms->initrd_mapped_file = mapped_file;
1251 initrd_data = g_mapped_file_get_contents(mapped_file);
1252 initrd_size = g_mapped_file_get_length(mapped_file);
1253 if (initrd_size >= initrd_max) {
1254 fprintf(stderr, "qemu: initrd is too large, cannot support."
1255 "(max: %"PRIu32", need %"PRId64")\n",
1256 initrd_max, (uint64_t)initrd_size);
1257 exit(1);
1260 initrd_addr = (initrd_max-initrd_size) & ~4095;
1262 fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_ADDR, initrd_addr);
1263 fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_SIZE, initrd_size);
1264 fw_cfg_add_bytes(fw_cfg, FW_CFG_INITRD_DATA, initrd_data, initrd_size);
1266 stl_p(header+0x218, initrd_addr);
1267 stl_p(header+0x21c, initrd_size);
1270 /* load kernel and setup */
1271 setup_size = header[0x1f1];
1272 if (setup_size == 0) {
1273 setup_size = 4;
1275 setup_size = (setup_size+1)*512;
1276 if (setup_size > kernel_size) {
1277 fprintf(stderr, "qemu: invalid kernel header\n");
1278 exit(1);
1280 kernel_size -= setup_size;
1282 setup = g_malloc(setup_size);
1283 kernel = g_malloc(kernel_size);
1284 fseek(f, 0, SEEK_SET);
1285 if (fread(setup, 1, setup_size, f) != setup_size) {
1286 fprintf(stderr, "fread() failed\n");
1287 exit(1);
1289 if (fread(kernel, 1, kernel_size, f) != kernel_size) {
1290 fprintf(stderr, "fread() failed\n");
1291 exit(1);
1293 fclose(f);
1295 /* append dtb to kernel */
1296 if (dtb_filename) {
1297 if (protocol < 0x209) {
1298 fprintf(stderr, "qemu: Linux kernel too old to load a dtb\n");
1299 exit(1);
1302 dtb_size = get_image_size(dtb_filename);
1303 if (dtb_size <= 0) {
1304 fprintf(stderr, "qemu: error reading dtb %s: %s\n",
1305 dtb_filename, strerror(errno));
1306 exit(1);
1309 setup_data_offset = QEMU_ALIGN_UP(kernel_size, 16);
1310 kernel_size = setup_data_offset + sizeof(struct setup_data) + dtb_size;
1311 kernel = g_realloc(kernel, kernel_size);
1313 stq_p(header+0x250, prot_addr + setup_data_offset);
1315 setup_data = (struct setup_data *)(kernel + setup_data_offset);
1316 setup_data->next = 0;
1317 setup_data->type = cpu_to_le32(SETUP_DTB);
1318 setup_data->len = cpu_to_le32(dtb_size);
1320 load_image_size(dtb_filename, setup_data->data, dtb_size);
1323 memcpy(setup, header, MIN(sizeof(header), setup_size));
1325 fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_ADDR, prot_addr);
1326 fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_SIZE, kernel_size);
1327 fw_cfg_add_bytes(fw_cfg, FW_CFG_KERNEL_DATA, kernel, kernel_size);
1329 fw_cfg_add_i32(fw_cfg, FW_CFG_SETUP_ADDR, real_addr);
1330 fw_cfg_add_i32(fw_cfg, FW_CFG_SETUP_SIZE, setup_size);
1331 fw_cfg_add_bytes(fw_cfg, FW_CFG_SETUP_DATA, setup, setup_size);
1333 option_rom[nb_option_roms].bootindex = 0;
1334 option_rom[nb_option_roms].name = "linuxboot.bin";
1335 if (pcmc->linuxboot_dma_enabled && fw_cfg_dma_enabled(fw_cfg)) {
1336 option_rom[nb_option_roms].name = "linuxboot_dma.bin";
1338 nb_option_roms++;
1341 #define NE2000_NB_MAX 6
1343 static const int ne2000_io[NE2000_NB_MAX] = { 0x300, 0x320, 0x340, 0x360,
1344 0x280, 0x380 };
1345 static const int ne2000_irq[NE2000_NB_MAX] = { 9, 10, 11, 3, 4, 5 };
1347 void pc_init_ne2k_isa(ISABus *bus, NICInfo *nd)
1349 static int nb_ne2k = 0;
1351 if (nb_ne2k == NE2000_NB_MAX)
1352 return;
1353 isa_ne2000_init(bus, ne2000_io[nb_ne2k],
1354 ne2000_irq[nb_ne2k], nd);
1355 nb_ne2k++;
1358 DeviceState *cpu_get_current_apic(void)
1360 if (current_cpu) {
1361 X86CPU *cpu = X86_CPU(current_cpu);
1362 return cpu->apic_state;
1363 } else {
1364 return NULL;
1368 void pc_acpi_smi_interrupt(void *opaque, int irq, int level)
1370 X86CPU *cpu = opaque;
1372 if (level) {
1373 cpu_interrupt(CPU(cpu), CPU_INTERRUPT_SMI);
1377 static void pc_new_cpu(PCMachineState *pcms, int64_t apic_id, Error **errp)
1379 Object *cpu = NULL;
1380 Error *local_err = NULL;
1381 CPUX86State *env = NULL;
1383 cpu = object_new(MACHINE(pcms)->cpu_type);
1385 env = &X86_CPU(cpu)->env;
1386 env->nr_dies = pcms->smp_dies;
1388 object_property_set_uint(cpu, apic_id, "apic-id", &local_err);
1389 object_property_set_bool(cpu, true, "realized", &local_err);
1391 object_unref(cpu);
1392 error_propagate(errp, local_err);
1396 * This function is very similar to smp_parse()
1397 * in hw/core/machine.c but includes CPU die support.
1399 void pc_smp_parse(MachineState *ms, QemuOpts *opts)
1401 PCMachineState *pcms = PC_MACHINE(ms);
1403 if (opts) {
1404 unsigned cpus = qemu_opt_get_number(opts, "cpus", 0);
1405 unsigned sockets = qemu_opt_get_number(opts, "sockets", 0);
1406 unsigned dies = qemu_opt_get_number(opts, "dies", 1);
1407 unsigned cores = qemu_opt_get_number(opts, "cores", 0);
1408 unsigned threads = qemu_opt_get_number(opts, "threads", 0);
1410 /* compute missing values, prefer sockets over cores over threads */
1411 if (cpus == 0 || sockets == 0) {
1412 cores = cores > 0 ? cores : 1;
1413 threads = threads > 0 ? threads : 1;
1414 if (cpus == 0) {
1415 sockets = sockets > 0 ? sockets : 1;
1416 cpus = cores * threads * dies * sockets;
1417 } else {
1418 ms->smp.max_cpus =
1419 qemu_opt_get_number(opts, "maxcpus", cpus);
1420 sockets = ms->smp.max_cpus / (cores * threads * dies);
1422 } else if (cores == 0) {
1423 threads = threads > 0 ? threads : 1;
1424 cores = cpus / (sockets * dies * threads);
1425 cores = cores > 0 ? cores : 1;
1426 } else if (threads == 0) {
1427 threads = cpus / (cores * dies * sockets);
1428 threads = threads > 0 ? threads : 1;
1429 } else if (sockets * dies * cores * threads < cpus) {
1430 error_report("cpu topology: "
1431 "sockets (%u) * dies (%u) * cores (%u) * threads (%u) < "
1432 "smp_cpus (%u)",
1433 sockets, dies, cores, threads, cpus);
1434 exit(1);
1437 ms->smp.max_cpus =
1438 qemu_opt_get_number(opts, "maxcpus", cpus);
1440 if (ms->smp.max_cpus < cpus) {
1441 error_report("maxcpus must be equal to or greater than smp");
1442 exit(1);
1445 if (sockets * dies * cores * threads > ms->smp.max_cpus) {
1446 error_report("cpu topology: "
1447 "sockets (%u) * dies (%u) * cores (%u) * threads (%u) > "
1448 "maxcpus (%u)",
1449 sockets, dies, cores, threads,
1450 ms->smp.max_cpus);
1451 exit(1);
1454 if (sockets * dies * cores * threads != ms->smp.max_cpus) {
1455 warn_report("Invalid CPU topology deprecated: "
1456 "sockets (%u) * dies (%u) * cores (%u) * threads (%u) "
1457 "!= maxcpus (%u)",
1458 sockets, dies, cores, threads,
1459 ms->smp.max_cpus);
1462 ms->smp.cpus = cpus;
1463 ms->smp.cores = cores;
1464 ms->smp.threads = threads;
1465 pcms->smp_dies = dies;
1468 if (ms->smp.cpus > 1) {
1469 Error *blocker = NULL;
1470 error_setg(&blocker, QERR_REPLAY_NOT_SUPPORTED, "smp");
1471 replay_add_blocker(blocker);
1475 void pc_hot_add_cpu(MachineState *ms, const int64_t id, Error **errp)
1477 PCMachineState *pcms = PC_MACHINE(ms);
1478 int64_t apic_id = x86_cpu_apic_id_from_index(pcms, id);
1479 Error *local_err = NULL;
1481 if (id < 0) {
1482 error_setg(errp, "Invalid CPU id: %" PRIi64, id);
1483 return;
1486 if (apic_id >= ACPI_CPU_HOTPLUG_ID_LIMIT) {
1487 error_setg(errp, "Unable to add CPU: %" PRIi64
1488 ", resulting APIC ID (%" PRIi64 ") is too large",
1489 id, apic_id);
1490 return;
1493 pc_new_cpu(PC_MACHINE(ms), apic_id, &local_err);
1494 if (local_err) {
1495 error_propagate(errp, local_err);
1496 return;
1500 void pc_cpus_init(PCMachineState *pcms)
1502 int i;
1503 const CPUArchIdList *possible_cpus;
1504 MachineState *ms = MACHINE(pcms);
1505 MachineClass *mc = MACHINE_GET_CLASS(pcms);
1506 PCMachineClass *pcmc = PC_MACHINE_CLASS(mc);
1508 x86_cpu_set_default_version(pcmc->default_cpu_version);
1510 /* Calculates the limit to CPU APIC ID values
1512 * Limit for the APIC ID value, so that all
1513 * CPU APIC IDs are < pcms->apic_id_limit.
1515 * This is used for FW_CFG_MAX_CPUS. See comments on fw_cfg_arch_create().
1517 pcms->apic_id_limit = x86_cpu_apic_id_from_index(pcms,
1518 ms->smp.max_cpus - 1) + 1;
1519 possible_cpus = mc->possible_cpu_arch_ids(ms);
1520 for (i = 0; i < ms->smp.cpus; i++) {
1521 pc_new_cpu(pcms, possible_cpus->cpus[i].arch_id, &error_fatal);
1525 static void rtc_set_cpus_count(ISADevice *rtc, uint16_t cpus_count)
1527 if (cpus_count > 0xff) {
1528 /* If the number of CPUs can't be represented in 8 bits, the
1529 * BIOS must use "FW_CFG_NB_CPUS". Set RTC field to 0 just
1530 * to make old BIOSes fail more predictably.
1532 rtc_set_memory(rtc, 0x5f, 0);
1533 } else {
1534 rtc_set_memory(rtc, 0x5f, cpus_count - 1);
1538 static
1539 void pc_machine_done(Notifier *notifier, void *data)
1541 PCMachineState *pcms = container_of(notifier,
1542 PCMachineState, machine_done);
1543 PCIBus *bus = pcms->bus;
1545 /* set the number of CPUs */
1546 rtc_set_cpus_count(pcms->rtc, pcms->boot_cpus);
1548 if (bus) {
1549 int extra_hosts = 0;
1551 QLIST_FOREACH(bus, &bus->child, sibling) {
1552 /* look for expander root buses */
1553 if (pci_bus_is_root(bus)) {
1554 extra_hosts++;
1557 if (extra_hosts && pcms->fw_cfg) {
1558 uint64_t *val = g_malloc(sizeof(*val));
1559 *val = cpu_to_le64(extra_hosts);
1560 fw_cfg_add_file(pcms->fw_cfg,
1561 "etc/extra-pci-roots", val, sizeof(*val));
1565 acpi_setup();
1566 if (pcms->fw_cfg) {
1567 fw_cfg_build_smbios(MACHINE(pcms), pcms->fw_cfg);
1568 fw_cfg_build_feature_control(MACHINE(pcms), pcms->fw_cfg);
1569 /* update FW_CFG_NB_CPUS to account for -device added CPUs */
1570 fw_cfg_modify_i16(pcms->fw_cfg, FW_CFG_NB_CPUS, pcms->boot_cpus);
1573 if (pcms->apic_id_limit > 255 && !xen_enabled()) {
1574 IntelIOMMUState *iommu = INTEL_IOMMU_DEVICE(x86_iommu_get_default());
1576 if (!iommu || !x86_iommu_ir_supported(X86_IOMMU_DEVICE(iommu)) ||
1577 iommu->intr_eim != ON_OFF_AUTO_ON) {
1578 error_report("current -smp configuration requires "
1579 "Extended Interrupt Mode enabled. "
1580 "You can add an IOMMU using: "
1581 "-device intel-iommu,intremap=on,eim=on");
1582 exit(EXIT_FAILURE);
1587 void pc_guest_info_init(PCMachineState *pcms)
1589 int i;
1590 MachineState *ms = MACHINE(pcms);
1592 pcms->apic_xrupt_override = kvm_allows_irq0_override();
1593 pcms->numa_nodes = ms->numa_state->num_nodes;
1594 pcms->node_mem = g_malloc0(pcms->numa_nodes *
1595 sizeof *pcms->node_mem);
1596 for (i = 0; i < ms->numa_state->num_nodes; i++) {
1597 pcms->node_mem[i] = ms->numa_state->nodes[i].node_mem;
1600 pcms->machine_done.notify = pc_machine_done;
1601 qemu_add_machine_init_done_notifier(&pcms->machine_done);
1604 /* setup pci memory address space mapping into system address space */
1605 void pc_pci_as_mapping_init(Object *owner, MemoryRegion *system_memory,
1606 MemoryRegion *pci_address_space)
1608 /* Set to lower priority than RAM */
1609 memory_region_add_subregion_overlap(system_memory, 0x0,
1610 pci_address_space, -1);
1613 void xen_load_linux(PCMachineState *pcms)
1615 int i;
1616 FWCfgState *fw_cfg;
1618 assert(MACHINE(pcms)->kernel_filename != NULL);
1620 fw_cfg = fw_cfg_init_io(FW_CFG_IO_BASE);
1621 fw_cfg_add_i16(fw_cfg, FW_CFG_NB_CPUS, pcms->boot_cpus);
1622 rom_set_fw(fw_cfg);
1624 load_linux(pcms, fw_cfg);
1625 for (i = 0; i < nb_option_roms; i++) {
1626 assert(!strcmp(option_rom[i].name, "linuxboot.bin") ||
1627 !strcmp(option_rom[i].name, "linuxboot_dma.bin") ||
1628 !strcmp(option_rom[i].name, "pvh.bin") ||
1629 !strcmp(option_rom[i].name, "multiboot.bin"));
1630 rom_add_option(option_rom[i].name, option_rom[i].bootindex);
1632 pcms->fw_cfg = fw_cfg;
1635 void pc_memory_init(PCMachineState *pcms,
1636 MemoryRegion *system_memory,
1637 MemoryRegion *rom_memory,
1638 MemoryRegion **ram_memory)
1640 int linux_boot, i;
1641 MemoryRegion *ram, *option_rom_mr;
1642 MemoryRegion *ram_below_4g, *ram_above_4g;
1643 FWCfgState *fw_cfg;
1644 MachineState *machine = MACHINE(pcms);
1645 MachineClass *mc = MACHINE_GET_CLASS(machine);
1646 PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
1648 assert(machine->ram_size == pcms->below_4g_mem_size +
1649 pcms->above_4g_mem_size);
1651 linux_boot = (machine->kernel_filename != NULL);
1653 /* Allocate RAM. We allocate it as a single memory region and use
1654 * aliases to address portions of it, mostly for backwards compatibility
1655 * with older qemus that used qemu_ram_alloc().
1657 ram = g_malloc(sizeof(*ram));
1658 memory_region_allocate_system_memory(ram, NULL, "pc.ram",
1659 machine->ram_size);
1660 *ram_memory = ram;
1661 ram_below_4g = g_malloc(sizeof(*ram_below_4g));
1662 memory_region_init_alias(ram_below_4g, NULL, "ram-below-4g", ram,
1663 0, pcms->below_4g_mem_size);
1664 memory_region_add_subregion(system_memory, 0, ram_below_4g);
1665 e820_add_entry(0, pcms->below_4g_mem_size, E820_RAM);
1666 if (pcms->above_4g_mem_size > 0) {
1667 ram_above_4g = g_malloc(sizeof(*ram_above_4g));
1668 memory_region_init_alias(ram_above_4g, NULL, "ram-above-4g", ram,
1669 pcms->below_4g_mem_size,
1670 pcms->above_4g_mem_size);
1671 memory_region_add_subregion(system_memory, 0x100000000ULL,
1672 ram_above_4g);
1673 e820_add_entry(0x100000000ULL, pcms->above_4g_mem_size, E820_RAM);
1676 if (!pcmc->has_reserved_memory &&
1677 (machine->ram_slots ||
1678 (machine->maxram_size > machine->ram_size))) {
1680 error_report("\"-memory 'slots|maxmem'\" is not supported by: %s",
1681 mc->name);
1682 exit(EXIT_FAILURE);
1685 /* always allocate the device memory information */
1686 machine->device_memory = g_malloc0(sizeof(*machine->device_memory));
1688 /* initialize device memory address space */
1689 if (pcmc->has_reserved_memory &&
1690 (machine->ram_size < machine->maxram_size)) {
1691 ram_addr_t device_mem_size = machine->maxram_size - machine->ram_size;
1693 if (machine->ram_slots > ACPI_MAX_RAM_SLOTS) {
1694 error_report("unsupported amount of memory slots: %"PRIu64,
1695 machine->ram_slots);
1696 exit(EXIT_FAILURE);
1699 if (QEMU_ALIGN_UP(machine->maxram_size,
1700 TARGET_PAGE_SIZE) != machine->maxram_size) {
1701 error_report("maximum memory size must by aligned to multiple of "
1702 "%d bytes", TARGET_PAGE_SIZE);
1703 exit(EXIT_FAILURE);
1706 machine->device_memory->base =
1707 ROUND_UP(0x100000000ULL + pcms->above_4g_mem_size, 1 * GiB);
1709 if (pcmc->enforce_aligned_dimm) {
1710 /* size device region assuming 1G page max alignment per slot */
1711 device_mem_size += (1 * GiB) * machine->ram_slots;
1714 if ((machine->device_memory->base + device_mem_size) <
1715 device_mem_size) {
1716 error_report("unsupported amount of maximum memory: " RAM_ADDR_FMT,
1717 machine->maxram_size);
1718 exit(EXIT_FAILURE);
1721 memory_region_init(&machine->device_memory->mr, OBJECT(pcms),
1722 "device-memory", device_mem_size);
1723 memory_region_add_subregion(system_memory, machine->device_memory->base,
1724 &machine->device_memory->mr);
1727 /* Initialize PC system firmware */
1728 pc_system_firmware_init(pcms, rom_memory);
1730 option_rom_mr = g_malloc(sizeof(*option_rom_mr));
1731 memory_region_init_ram(option_rom_mr, NULL, "pc.rom", PC_ROM_SIZE,
1732 &error_fatal);
1733 if (pcmc->pci_enabled) {
1734 memory_region_set_readonly(option_rom_mr, true);
1736 memory_region_add_subregion_overlap(rom_memory,
1737 PC_ROM_MIN_VGA,
1738 option_rom_mr,
1741 fw_cfg = fw_cfg_arch_create(machine,
1742 pcms->boot_cpus, pcms->apic_id_limit);
1744 rom_set_fw(fw_cfg);
1746 if (pcmc->has_reserved_memory && machine->device_memory->base) {
1747 uint64_t *val = g_malloc(sizeof(*val));
1748 PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
1749 uint64_t res_mem_end = machine->device_memory->base;
1751 if (!pcmc->broken_reserved_end) {
1752 res_mem_end += memory_region_size(&machine->device_memory->mr);
1754 *val = cpu_to_le64(ROUND_UP(res_mem_end, 1 * GiB));
1755 fw_cfg_add_file(fw_cfg, "etc/reserved-memory-end", val, sizeof(*val));
1758 if (linux_boot) {
1759 load_linux(pcms, fw_cfg);
1762 for (i = 0; i < nb_option_roms; i++) {
1763 rom_add_option(option_rom[i].name, option_rom[i].bootindex);
1765 pcms->fw_cfg = fw_cfg;
1767 /* Init default IOAPIC address space */
1768 pcms->ioapic_as = &address_space_memory;
1770 /* Init ACPI memory hotplug IO base address */
1771 pcms->memhp_io_base = ACPI_MEMORY_HOTPLUG_BASE;
1775 * The 64bit pci hole starts after "above 4G RAM" and
1776 * potentially the space reserved for memory hotplug.
1778 uint64_t pc_pci_hole64_start(void)
1780 PCMachineState *pcms = PC_MACHINE(qdev_get_machine());
1781 PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
1782 MachineState *ms = MACHINE(pcms);
1783 uint64_t hole64_start = 0;
1785 if (pcmc->has_reserved_memory && ms->device_memory->base) {
1786 hole64_start = ms->device_memory->base;
1787 if (!pcmc->broken_reserved_end) {
1788 hole64_start += memory_region_size(&ms->device_memory->mr);
1790 } else {
1791 hole64_start = 0x100000000ULL + pcms->above_4g_mem_size;
1794 return ROUND_UP(hole64_start, 1 * GiB);
1797 qemu_irq pc_allocate_cpu_irq(void)
1799 return qemu_allocate_irq(pic_irq_request, NULL, 0);
1802 DeviceState *pc_vga_init(ISABus *isa_bus, PCIBus *pci_bus)
1804 DeviceState *dev = NULL;
1806 rom_set_order_override(FW_CFG_ORDER_OVERRIDE_VGA);
1807 if (pci_bus) {
1808 PCIDevice *pcidev = pci_vga_init(pci_bus);
1809 dev = pcidev ? &pcidev->qdev : NULL;
1810 } else if (isa_bus) {
1811 ISADevice *isadev = isa_vga_init(isa_bus);
1812 dev = isadev ? DEVICE(isadev) : NULL;
1814 rom_reset_order_override();
1815 return dev;
1818 static const MemoryRegionOps ioport80_io_ops = {
1819 .write = ioport80_write,
1820 .read = ioport80_read,
1821 .endianness = DEVICE_NATIVE_ENDIAN,
1822 .impl = {
1823 .min_access_size = 1,
1824 .max_access_size = 1,
1828 static const MemoryRegionOps ioportF0_io_ops = {
1829 .write = ioportF0_write,
1830 .read = ioportF0_read,
1831 .endianness = DEVICE_NATIVE_ENDIAN,
1832 .impl = {
1833 .min_access_size = 1,
1834 .max_access_size = 1,
1838 static void pc_superio_init(ISABus *isa_bus, bool create_fdctrl, bool no_vmport)
1840 int i;
1841 DriveInfo *fd[MAX_FD];
1842 qemu_irq *a20_line;
1843 ISADevice *i8042, *port92, *vmmouse;
1845 serial_hds_isa_init(isa_bus, 0, MAX_ISA_SERIAL_PORTS);
1846 parallel_hds_isa_init(isa_bus, MAX_PARALLEL_PORTS);
1848 for (i = 0; i < MAX_FD; i++) {
1849 fd[i] = drive_get(IF_FLOPPY, 0, i);
1850 create_fdctrl |= !!fd[i];
1852 if (create_fdctrl) {
1853 fdctrl_init_isa(isa_bus, fd);
1856 i8042 = isa_create_simple(isa_bus, "i8042");
1857 if (!no_vmport) {
1858 vmport_init(isa_bus);
1859 vmmouse = isa_try_create(isa_bus, "vmmouse");
1860 } else {
1861 vmmouse = NULL;
1863 if (vmmouse) {
1864 DeviceState *dev = DEVICE(vmmouse);
1865 qdev_prop_set_ptr(dev, "ps2_mouse", i8042);
1866 qdev_init_nofail(dev);
1868 port92 = isa_create_simple(isa_bus, "port92");
1870 a20_line = qemu_allocate_irqs(handle_a20_line_change, first_cpu, 2);
1871 i8042_setup_a20_line(i8042, a20_line[0]);
1872 port92_init(port92, a20_line[1]);
1873 g_free(a20_line);
1876 void pc_basic_device_init(ISABus *isa_bus, qemu_irq *gsi,
1877 ISADevice **rtc_state,
1878 bool create_fdctrl,
1879 bool no_vmport,
1880 bool has_pit,
1881 uint32_t hpet_irqs)
1883 int i;
1884 DeviceState *hpet = NULL;
1885 int pit_isa_irq = 0;
1886 qemu_irq pit_alt_irq = NULL;
1887 qemu_irq rtc_irq = NULL;
1888 ISADevice *pit = NULL;
1889 MemoryRegion *ioport80_io = g_new(MemoryRegion, 1);
1890 MemoryRegion *ioportF0_io = g_new(MemoryRegion, 1);
1892 memory_region_init_io(ioport80_io, NULL, &ioport80_io_ops, NULL, "ioport80", 1);
1893 memory_region_add_subregion(isa_bus->address_space_io, 0x80, ioport80_io);
1895 memory_region_init_io(ioportF0_io, NULL, &ioportF0_io_ops, NULL, "ioportF0", 1);
1896 memory_region_add_subregion(isa_bus->address_space_io, 0xf0, ioportF0_io);
1899 * Check if an HPET shall be created.
1901 * Without KVM_CAP_PIT_STATE2, we cannot switch off the in-kernel PIT
1902 * when the HPET wants to take over. Thus we have to disable the latter.
1904 if (!no_hpet && (!kvm_irqchip_in_kernel() || kvm_has_pit_state2())) {
1905 /* In order to set property, here not using sysbus_try_create_simple */
1906 hpet = qdev_try_create(NULL, TYPE_HPET);
1907 if (hpet) {
1908 /* For pc-piix-*, hpet's intcap is always IRQ2. For pc-q35-1.7
1909 * and earlier, use IRQ2 for compat. Otherwise, use IRQ16~23,
1910 * IRQ8 and IRQ2.
1912 uint8_t compat = object_property_get_uint(OBJECT(hpet),
1913 HPET_INTCAP, NULL);
1914 if (!compat) {
1915 qdev_prop_set_uint32(hpet, HPET_INTCAP, hpet_irqs);
1917 qdev_init_nofail(hpet);
1918 sysbus_mmio_map(SYS_BUS_DEVICE(hpet), 0, HPET_BASE);
1920 for (i = 0; i < GSI_NUM_PINS; i++) {
1921 sysbus_connect_irq(SYS_BUS_DEVICE(hpet), i, gsi[i]);
1923 pit_isa_irq = -1;
1924 pit_alt_irq = qdev_get_gpio_in(hpet, HPET_LEGACY_PIT_INT);
1925 rtc_irq = qdev_get_gpio_in(hpet, HPET_LEGACY_RTC_INT);
1928 *rtc_state = mc146818_rtc_init(isa_bus, 2000, rtc_irq);
1930 qemu_register_boot_set(pc_boot_set, *rtc_state);
1932 if (!xen_enabled() && has_pit) {
1933 if (kvm_pit_in_kernel()) {
1934 pit = kvm_pit_init(isa_bus, 0x40);
1935 } else {
1936 pit = i8254_pit_init(isa_bus, 0x40, pit_isa_irq, pit_alt_irq);
1938 if (hpet) {
1939 /* connect PIT to output control line of the HPET */
1940 qdev_connect_gpio_out(hpet, 0, qdev_get_gpio_in(DEVICE(pit), 0));
1942 pcspk_init(isa_bus, pit);
1945 i8257_dma_init(isa_bus, 0);
1947 /* Super I/O */
1948 pc_superio_init(isa_bus, create_fdctrl, no_vmport);
1951 void pc_nic_init(PCMachineClass *pcmc, ISABus *isa_bus, PCIBus *pci_bus)
1953 int i;
1955 rom_set_order_override(FW_CFG_ORDER_OVERRIDE_NIC);
1956 for (i = 0; i < nb_nics; i++) {
1957 NICInfo *nd = &nd_table[i];
1958 const char *model = nd->model ? nd->model : pcmc->default_nic_model;
1960 if (g_str_equal(model, "ne2k_isa")) {
1961 pc_init_ne2k_isa(isa_bus, nd);
1962 } else {
1963 pci_nic_init_nofail(nd, pci_bus, model, NULL);
1966 rom_reset_order_override();
1969 void ioapic_init_gsi(GSIState *gsi_state, const char *parent_name)
1971 DeviceState *dev;
1972 SysBusDevice *d;
1973 unsigned int i;
1975 if (kvm_ioapic_in_kernel()) {
1976 dev = qdev_create(NULL, TYPE_KVM_IOAPIC);
1977 } else {
1978 dev = qdev_create(NULL, TYPE_IOAPIC);
1980 if (parent_name) {
1981 object_property_add_child(object_resolve_path(parent_name, NULL),
1982 "ioapic", OBJECT(dev), NULL);
1984 qdev_init_nofail(dev);
1985 d = SYS_BUS_DEVICE(dev);
1986 sysbus_mmio_map(d, 0, IO_APIC_DEFAULT_ADDRESS);
1988 for (i = 0; i < IOAPIC_NUM_PINS; i++) {
1989 gsi_state->ioapic_irq[i] = qdev_get_gpio_in(dev, i);
1993 static void pc_memory_pre_plug(HotplugHandler *hotplug_dev, DeviceState *dev,
1994 Error **errp)
1996 const PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1997 const PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
1998 const MachineState *ms = MACHINE(hotplug_dev);
1999 const bool is_nvdimm = object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM);
2000 const uint64_t legacy_align = TARGET_PAGE_SIZE;
2001 Error *local_err = NULL;
2004 * When -no-acpi is used with Q35 machine type, no ACPI is built,
2005 * but pcms->acpi_dev is still created. Check !acpi_enabled in
2006 * addition to cover this case.
2008 if (!pcms->acpi_dev || !acpi_enabled) {
2009 error_setg(errp,
2010 "memory hotplug is not enabled: missing acpi device or acpi disabled");
2011 return;
2014 if (is_nvdimm && !ms->nvdimms_state->is_enabled) {
2015 error_setg(errp, "nvdimm is not enabled: missing 'nvdimm' in '-M'");
2016 return;
2019 hotplug_handler_pre_plug(pcms->acpi_dev, dev, &local_err);
2020 if (local_err) {
2021 error_propagate(errp, local_err);
2022 return;
2025 pc_dimm_pre_plug(PC_DIMM(dev), MACHINE(hotplug_dev),
2026 pcmc->enforce_aligned_dimm ? NULL : &legacy_align, errp);
2029 static void pc_memory_plug(HotplugHandler *hotplug_dev,
2030 DeviceState *dev, Error **errp)
2032 Error *local_err = NULL;
2033 PCMachineState *pcms = PC_MACHINE(hotplug_dev);
2034 MachineState *ms = MACHINE(hotplug_dev);
2035 bool is_nvdimm = object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM);
2037 pc_dimm_plug(PC_DIMM(dev), MACHINE(pcms), &local_err);
2038 if (local_err) {
2039 goto out;
2042 if (is_nvdimm) {
2043 nvdimm_plug(ms->nvdimms_state);
2046 hotplug_handler_plug(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &error_abort);
2047 out:
2048 error_propagate(errp, local_err);
2051 static void pc_memory_unplug_request(HotplugHandler *hotplug_dev,
2052 DeviceState *dev, Error **errp)
2054 Error *local_err = NULL;
2055 PCMachineState *pcms = PC_MACHINE(hotplug_dev);
2058 * When -no-acpi is used with Q35 machine type, no ACPI is built,
2059 * but pcms->acpi_dev is still created. Check !acpi_enabled in
2060 * addition to cover this case.
2062 if (!pcms->acpi_dev || !acpi_enabled) {
2063 error_setg(&local_err,
2064 "memory hotplug is not enabled: missing acpi device or acpi disabled");
2065 goto out;
2068 if (object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM)) {
2069 error_setg(&local_err,
2070 "nvdimm device hot unplug is not supported yet.");
2071 goto out;
2074 hotplug_handler_unplug_request(HOTPLUG_HANDLER(pcms->acpi_dev), dev,
2075 &local_err);
2076 out:
2077 error_propagate(errp, local_err);
2080 static void pc_memory_unplug(HotplugHandler *hotplug_dev,
2081 DeviceState *dev, Error **errp)
2083 PCMachineState *pcms = PC_MACHINE(hotplug_dev);
2084 Error *local_err = NULL;
2086 hotplug_handler_unplug(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &local_err);
2087 if (local_err) {
2088 goto out;
2091 pc_dimm_unplug(PC_DIMM(dev), MACHINE(pcms));
2092 object_property_set_bool(OBJECT(dev), false, "realized", NULL);
2093 out:
2094 error_propagate(errp, local_err);
2097 static int pc_apic_cmp(const void *a, const void *b)
2099 CPUArchId *apic_a = (CPUArchId *)a;
2100 CPUArchId *apic_b = (CPUArchId *)b;
2102 return apic_a->arch_id - apic_b->arch_id;
2105 /* returns pointer to CPUArchId descriptor that matches CPU's apic_id
2106 * in ms->possible_cpus->cpus, if ms->possible_cpus->cpus has no
2107 * entry corresponding to CPU's apic_id returns NULL.
2109 static CPUArchId *pc_find_cpu_slot(MachineState *ms, uint32_t id, int *idx)
2111 CPUArchId apic_id, *found_cpu;
2113 apic_id.arch_id = id;
2114 found_cpu = bsearch(&apic_id, ms->possible_cpus->cpus,
2115 ms->possible_cpus->len, sizeof(*ms->possible_cpus->cpus),
2116 pc_apic_cmp);
2117 if (found_cpu && idx) {
2118 *idx = found_cpu - ms->possible_cpus->cpus;
2120 return found_cpu;
2123 static void pc_cpu_plug(HotplugHandler *hotplug_dev,
2124 DeviceState *dev, Error **errp)
2126 CPUArchId *found_cpu;
2127 Error *local_err = NULL;
2128 X86CPU *cpu = X86_CPU(dev);
2129 PCMachineState *pcms = PC_MACHINE(hotplug_dev);
2131 if (pcms->acpi_dev) {
2132 hotplug_handler_plug(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &local_err);
2133 if (local_err) {
2134 goto out;
2138 /* increment the number of CPUs */
2139 pcms->boot_cpus++;
2140 if (pcms->rtc) {
2141 rtc_set_cpus_count(pcms->rtc, pcms->boot_cpus);
2143 if (pcms->fw_cfg) {
2144 fw_cfg_modify_i16(pcms->fw_cfg, FW_CFG_NB_CPUS, pcms->boot_cpus);
2147 found_cpu = pc_find_cpu_slot(MACHINE(pcms), cpu->apic_id, NULL);
2148 found_cpu->cpu = OBJECT(dev);
2149 out:
2150 error_propagate(errp, local_err);
2152 static void pc_cpu_unplug_request_cb(HotplugHandler *hotplug_dev,
2153 DeviceState *dev, Error **errp)
2155 int idx = -1;
2156 Error *local_err = NULL;
2157 X86CPU *cpu = X86_CPU(dev);
2158 PCMachineState *pcms = PC_MACHINE(hotplug_dev);
2160 if (!pcms->acpi_dev) {
2161 error_setg(&local_err, "CPU hot unplug not supported without ACPI");
2162 goto out;
2165 pc_find_cpu_slot(MACHINE(pcms), cpu->apic_id, &idx);
2166 assert(idx != -1);
2167 if (idx == 0) {
2168 error_setg(&local_err, "Boot CPU is unpluggable");
2169 goto out;
2172 hotplug_handler_unplug_request(HOTPLUG_HANDLER(pcms->acpi_dev), dev,
2173 &local_err);
2174 if (local_err) {
2175 goto out;
2178 out:
2179 error_propagate(errp, local_err);
2183 static void pc_cpu_unplug_cb(HotplugHandler *hotplug_dev,
2184 DeviceState *dev, Error **errp)
2186 CPUArchId *found_cpu;
2187 Error *local_err = NULL;
2188 X86CPU *cpu = X86_CPU(dev);
2189 PCMachineState *pcms = PC_MACHINE(hotplug_dev);
2191 hotplug_handler_unplug(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &local_err);
2192 if (local_err) {
2193 goto out;
2196 found_cpu = pc_find_cpu_slot(MACHINE(pcms), cpu->apic_id, NULL);
2197 found_cpu->cpu = NULL;
2198 object_property_set_bool(OBJECT(dev), false, "realized", NULL);
2200 /* decrement the number of CPUs */
2201 pcms->boot_cpus--;
2202 /* Update the number of CPUs in CMOS */
2203 rtc_set_cpus_count(pcms->rtc, pcms->boot_cpus);
2204 fw_cfg_modify_i16(pcms->fw_cfg, FW_CFG_NB_CPUS, pcms->boot_cpus);
2205 out:
2206 error_propagate(errp, local_err);
2209 static void pc_cpu_pre_plug(HotplugHandler *hotplug_dev,
2210 DeviceState *dev, Error **errp)
2212 int idx;
2213 CPUState *cs;
2214 CPUArchId *cpu_slot;
2215 X86CPUTopoInfo topo;
2216 X86CPU *cpu = X86_CPU(dev);
2217 CPUX86State *env = &cpu->env;
2218 MachineState *ms = MACHINE(hotplug_dev);
2219 PCMachineState *pcms = PC_MACHINE(hotplug_dev);
2220 unsigned int smp_cores = ms->smp.cores;
2221 unsigned int smp_threads = ms->smp.threads;
2223 if(!object_dynamic_cast(OBJECT(cpu), ms->cpu_type)) {
2224 error_setg(errp, "Invalid CPU type, expected cpu type: '%s'",
2225 ms->cpu_type);
2226 return;
2229 env->nr_dies = pcms->smp_dies;
2232 * If APIC ID is not set,
2233 * set it based on socket/die/core/thread properties.
2235 if (cpu->apic_id == UNASSIGNED_APIC_ID) {
2236 int max_socket = (ms->smp.max_cpus - 1) /
2237 smp_threads / smp_cores / pcms->smp_dies;
2240 * die-id was optional in QEMU 4.0 and older, so keep it optional
2241 * if there's only one die per socket.
2243 if (cpu->die_id < 0 && pcms->smp_dies == 1) {
2244 cpu->die_id = 0;
2247 if (cpu->socket_id < 0) {
2248 error_setg(errp, "CPU socket-id is not set");
2249 return;
2250 } else if (cpu->socket_id > max_socket) {
2251 error_setg(errp, "Invalid CPU socket-id: %u must be in range 0:%u",
2252 cpu->socket_id, max_socket);
2253 return;
2255 if (cpu->die_id < 0) {
2256 error_setg(errp, "CPU die-id is not set");
2257 return;
2258 } else if (cpu->die_id > pcms->smp_dies - 1) {
2259 error_setg(errp, "Invalid CPU die-id: %u must be in range 0:%u",
2260 cpu->die_id, pcms->smp_dies - 1);
2261 return;
2263 if (cpu->core_id < 0) {
2264 error_setg(errp, "CPU core-id is not set");
2265 return;
2266 } else if (cpu->core_id > (smp_cores - 1)) {
2267 error_setg(errp, "Invalid CPU core-id: %u must be in range 0:%u",
2268 cpu->core_id, smp_cores - 1);
2269 return;
2271 if (cpu->thread_id < 0) {
2272 error_setg(errp, "CPU thread-id is not set");
2273 return;
2274 } else if (cpu->thread_id > (smp_threads - 1)) {
2275 error_setg(errp, "Invalid CPU thread-id: %u must be in range 0:%u",
2276 cpu->thread_id, smp_threads - 1);
2277 return;
2280 topo.pkg_id = cpu->socket_id;
2281 topo.die_id = cpu->die_id;
2282 topo.core_id = cpu->core_id;
2283 topo.smt_id = cpu->thread_id;
2284 cpu->apic_id = apicid_from_topo_ids(pcms->smp_dies, smp_cores,
2285 smp_threads, &topo);
2288 cpu_slot = pc_find_cpu_slot(MACHINE(pcms), cpu->apic_id, &idx);
2289 if (!cpu_slot) {
2290 MachineState *ms = MACHINE(pcms);
2292 x86_topo_ids_from_apicid(cpu->apic_id, pcms->smp_dies,
2293 smp_cores, smp_threads, &topo);
2294 error_setg(errp,
2295 "Invalid CPU [socket: %u, die: %u, core: %u, thread: %u] with"
2296 " APIC ID %" PRIu32 ", valid index range 0:%d",
2297 topo.pkg_id, topo.die_id, topo.core_id, topo.smt_id,
2298 cpu->apic_id, ms->possible_cpus->len - 1);
2299 return;
2302 if (cpu_slot->cpu) {
2303 error_setg(errp, "CPU[%d] with APIC ID %" PRIu32 " exists",
2304 idx, cpu->apic_id);
2305 return;
2308 /* if 'address' properties socket-id/core-id/thread-id are not set, set them
2309 * so that machine_query_hotpluggable_cpus would show correct values
2311 /* TODO: move socket_id/core_id/thread_id checks into x86_cpu_realizefn()
2312 * once -smp refactoring is complete and there will be CPU private
2313 * CPUState::nr_cores and CPUState::nr_threads fields instead of globals */
2314 x86_topo_ids_from_apicid(cpu->apic_id, pcms->smp_dies,
2315 smp_cores, smp_threads, &topo);
2316 if (cpu->socket_id != -1 && cpu->socket_id != topo.pkg_id) {
2317 error_setg(errp, "property socket-id: %u doesn't match set apic-id:"
2318 " 0x%x (socket-id: %u)", cpu->socket_id, cpu->apic_id, topo.pkg_id);
2319 return;
2321 cpu->socket_id = topo.pkg_id;
2323 if (cpu->die_id != -1 && cpu->die_id != topo.die_id) {
2324 error_setg(errp, "property die-id: %u doesn't match set apic-id:"
2325 " 0x%x (die-id: %u)", cpu->die_id, cpu->apic_id, topo.die_id);
2326 return;
2328 cpu->die_id = topo.die_id;
2330 if (cpu->core_id != -1 && cpu->core_id != topo.core_id) {
2331 error_setg(errp, "property core-id: %u doesn't match set apic-id:"
2332 " 0x%x (core-id: %u)", cpu->core_id, cpu->apic_id, topo.core_id);
2333 return;
2335 cpu->core_id = topo.core_id;
2337 if (cpu->thread_id != -1 && cpu->thread_id != topo.smt_id) {
2338 error_setg(errp, "property thread-id: %u doesn't match set apic-id:"
2339 " 0x%x (thread-id: %u)", cpu->thread_id, cpu->apic_id, topo.smt_id);
2340 return;
2342 cpu->thread_id = topo.smt_id;
2344 if (hyperv_feat_enabled(cpu, HYPERV_FEAT_VPINDEX) &&
2345 !kvm_hv_vpindex_settable()) {
2346 error_setg(errp, "kernel doesn't allow setting HyperV VP_INDEX");
2347 return;
2350 cs = CPU(cpu);
2351 cs->cpu_index = idx;
2353 numa_cpu_pre_plug(cpu_slot, dev, errp);
2356 static void pc_virtio_pmem_pci_pre_plug(HotplugHandler *hotplug_dev,
2357 DeviceState *dev, Error **errp)
2359 HotplugHandler *hotplug_dev2 = qdev_get_bus_hotplug_handler(dev);
2360 Error *local_err = NULL;
2362 if (!hotplug_dev2) {
2364 * Without a bus hotplug handler, we cannot control the plug/unplug
2365 * order. This should never be the case on x86, however better add
2366 * a safety net.
2368 error_setg(errp, "virtio-pmem-pci not supported on this bus.");
2369 return;
2372 * First, see if we can plug this memory device at all. If that
2373 * succeeds, branch of to the actual hotplug handler.
2375 memory_device_pre_plug(MEMORY_DEVICE(dev), MACHINE(hotplug_dev), NULL,
2376 &local_err);
2377 if (!local_err) {
2378 hotplug_handler_pre_plug(hotplug_dev2, dev, &local_err);
2380 error_propagate(errp, local_err);
2383 static void pc_virtio_pmem_pci_plug(HotplugHandler *hotplug_dev,
2384 DeviceState *dev, Error **errp)
2386 HotplugHandler *hotplug_dev2 = qdev_get_bus_hotplug_handler(dev);
2387 Error *local_err = NULL;
2390 * Plug the memory device first and then branch off to the actual
2391 * hotplug handler. If that one fails, we can easily undo the memory
2392 * device bits.
2394 memory_device_plug(MEMORY_DEVICE(dev), MACHINE(hotplug_dev));
2395 hotplug_handler_plug(hotplug_dev2, dev, &local_err);
2396 if (local_err) {
2397 memory_device_unplug(MEMORY_DEVICE(dev), MACHINE(hotplug_dev));
2399 error_propagate(errp, local_err);
2402 static void pc_virtio_pmem_pci_unplug_request(HotplugHandler *hotplug_dev,
2403 DeviceState *dev, Error **errp)
2405 /* We don't support virtio pmem hot unplug */
2406 error_setg(errp, "virtio pmem device unplug not supported.");
2409 static void pc_virtio_pmem_pci_unplug(HotplugHandler *hotplug_dev,
2410 DeviceState *dev, Error **errp)
2412 /* We don't support virtio pmem hot unplug */
2415 static void pc_machine_device_pre_plug_cb(HotplugHandler *hotplug_dev,
2416 DeviceState *dev, Error **errp)
2418 if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2419 pc_memory_pre_plug(hotplug_dev, dev, errp);
2420 } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
2421 pc_cpu_pre_plug(hotplug_dev, dev, errp);
2422 } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_PMEM_PCI)) {
2423 pc_virtio_pmem_pci_pre_plug(hotplug_dev, dev, errp);
2427 static void pc_machine_device_plug_cb(HotplugHandler *hotplug_dev,
2428 DeviceState *dev, Error **errp)
2430 if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2431 pc_memory_plug(hotplug_dev, dev, errp);
2432 } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
2433 pc_cpu_plug(hotplug_dev, dev, errp);
2434 } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_PMEM_PCI)) {
2435 pc_virtio_pmem_pci_plug(hotplug_dev, dev, errp);
2439 static void pc_machine_device_unplug_request_cb(HotplugHandler *hotplug_dev,
2440 DeviceState *dev, Error **errp)
2442 if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2443 pc_memory_unplug_request(hotplug_dev, dev, errp);
2444 } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
2445 pc_cpu_unplug_request_cb(hotplug_dev, dev, errp);
2446 } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_PMEM_PCI)) {
2447 pc_virtio_pmem_pci_unplug_request(hotplug_dev, dev, errp);
2448 } else {
2449 error_setg(errp, "acpi: device unplug request for not supported device"
2450 " type: %s", object_get_typename(OBJECT(dev)));
2454 static void pc_machine_device_unplug_cb(HotplugHandler *hotplug_dev,
2455 DeviceState *dev, Error **errp)
2457 if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2458 pc_memory_unplug(hotplug_dev, dev, errp);
2459 } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
2460 pc_cpu_unplug_cb(hotplug_dev, dev, errp);
2461 } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_PMEM_PCI)) {
2462 pc_virtio_pmem_pci_unplug(hotplug_dev, dev, errp);
2463 } else {
2464 error_setg(errp, "acpi: device unplug for not supported device"
2465 " type: %s", object_get_typename(OBJECT(dev)));
2469 static HotplugHandler *pc_get_hotplug_handler(MachineState *machine,
2470 DeviceState *dev)
2472 if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM) ||
2473 object_dynamic_cast(OBJECT(dev), TYPE_CPU) ||
2474 object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_PMEM_PCI)) {
2475 return HOTPLUG_HANDLER(machine);
2478 return NULL;
2481 static void
2482 pc_machine_get_device_memory_region_size(Object *obj, Visitor *v,
2483 const char *name, void *opaque,
2484 Error **errp)
2486 MachineState *ms = MACHINE(obj);
2487 int64_t value = 0;
2489 if (ms->device_memory) {
2490 value = memory_region_size(&ms->device_memory->mr);
2493 visit_type_int(v, name, &value, errp);
2496 static void pc_machine_get_max_ram_below_4g(Object *obj, Visitor *v,
2497 const char *name, void *opaque,
2498 Error **errp)
2500 PCMachineState *pcms = PC_MACHINE(obj);
2501 uint64_t value = pcms->max_ram_below_4g;
2503 visit_type_size(v, name, &value, errp);
2506 static void pc_machine_set_max_ram_below_4g(Object *obj, Visitor *v,
2507 const char *name, void *opaque,
2508 Error **errp)
2510 PCMachineState *pcms = PC_MACHINE(obj);
2511 Error *error = NULL;
2512 uint64_t value;
2514 visit_type_size(v, name, &value, &error);
2515 if (error) {
2516 error_propagate(errp, error);
2517 return;
2519 if (value > 4 * GiB) {
2520 error_setg(&error,
2521 "Machine option 'max-ram-below-4g=%"PRIu64
2522 "' expects size less than or equal to 4G", value);
2523 error_propagate(errp, error);
2524 return;
2527 if (value < 1 * MiB) {
2528 warn_report("Only %" PRIu64 " bytes of RAM below the 4GiB boundary,"
2529 "BIOS may not work with less than 1MiB", value);
2532 pcms->max_ram_below_4g = value;
2535 static void pc_machine_get_vmport(Object *obj, Visitor *v, const char *name,
2536 void *opaque, Error **errp)
2538 PCMachineState *pcms = PC_MACHINE(obj);
2539 OnOffAuto vmport = pcms->vmport;
2541 visit_type_OnOffAuto(v, name, &vmport, errp);
2544 static void pc_machine_set_vmport(Object *obj, Visitor *v, const char *name,
2545 void *opaque, Error **errp)
2547 PCMachineState *pcms = PC_MACHINE(obj);
2549 visit_type_OnOffAuto(v, name, &pcms->vmport, errp);
2552 bool pc_machine_is_smm_enabled(PCMachineState *pcms)
2554 bool smm_available = false;
2556 if (pcms->smm == ON_OFF_AUTO_OFF) {
2557 return false;
2560 if (tcg_enabled() || qtest_enabled()) {
2561 smm_available = true;
2562 } else if (kvm_enabled()) {
2563 smm_available = kvm_has_smm();
2566 if (smm_available) {
2567 return true;
2570 if (pcms->smm == ON_OFF_AUTO_ON) {
2571 error_report("System Management Mode not supported by this hypervisor.");
2572 exit(1);
2574 return false;
2577 static void pc_machine_get_smm(Object *obj, Visitor *v, const char *name,
2578 void *opaque, Error **errp)
2580 PCMachineState *pcms = PC_MACHINE(obj);
2581 OnOffAuto smm = pcms->smm;
2583 visit_type_OnOffAuto(v, name, &smm, errp);
2586 static void pc_machine_set_smm(Object *obj, Visitor *v, const char *name,
2587 void *opaque, Error **errp)
2589 PCMachineState *pcms = PC_MACHINE(obj);
2591 visit_type_OnOffAuto(v, name, &pcms->smm, errp);
2594 static bool pc_machine_get_smbus(Object *obj, Error **errp)
2596 PCMachineState *pcms = PC_MACHINE(obj);
2598 return pcms->smbus_enabled;
2601 static void pc_machine_set_smbus(Object *obj, bool value, Error **errp)
2603 PCMachineState *pcms = PC_MACHINE(obj);
2605 pcms->smbus_enabled = value;
2608 static bool pc_machine_get_sata(Object *obj, Error **errp)
2610 PCMachineState *pcms = PC_MACHINE(obj);
2612 return pcms->sata_enabled;
2615 static void pc_machine_set_sata(Object *obj, bool value, Error **errp)
2617 PCMachineState *pcms = PC_MACHINE(obj);
2619 pcms->sata_enabled = value;
2622 static bool pc_machine_get_pit(Object *obj, Error **errp)
2624 PCMachineState *pcms = PC_MACHINE(obj);
2626 return pcms->pit_enabled;
2629 static void pc_machine_set_pit(Object *obj, bool value, Error **errp)
2631 PCMachineState *pcms = PC_MACHINE(obj);
2633 pcms->pit_enabled = value;
2636 static void pc_machine_initfn(Object *obj)
2638 PCMachineState *pcms = PC_MACHINE(obj);
2640 pcms->max_ram_below_4g = 0; /* use default */
2641 pcms->smm = ON_OFF_AUTO_AUTO;
2642 #ifdef CONFIG_VMPORT
2643 pcms->vmport = ON_OFF_AUTO_AUTO;
2644 #else
2645 pcms->vmport = ON_OFF_AUTO_OFF;
2646 #endif /* CONFIG_VMPORT */
2647 /* acpi build is enabled by default if machine supports it */
2648 pcms->acpi_build_enabled = PC_MACHINE_GET_CLASS(pcms)->has_acpi_build;
2649 pcms->smbus_enabled = true;
2650 pcms->sata_enabled = true;
2651 pcms->pit_enabled = true;
2652 pcms->smp_dies = 1;
2654 pc_system_flash_create(pcms);
2657 static void pc_machine_reset(MachineState *machine)
2659 CPUState *cs;
2660 X86CPU *cpu;
2662 qemu_devices_reset();
2664 /* Reset APIC after devices have been reset to cancel
2665 * any changes that qemu_devices_reset() might have done.
2667 CPU_FOREACH(cs) {
2668 cpu = X86_CPU(cs);
2670 if (cpu->apic_state) {
2671 device_reset(cpu->apic_state);
2676 static void pc_machine_wakeup(MachineState *machine)
2678 cpu_synchronize_all_states();
2679 pc_machine_reset(machine);
2680 cpu_synchronize_all_post_reset();
2683 static CpuInstanceProperties
2684 pc_cpu_index_to_props(MachineState *ms, unsigned cpu_index)
2686 MachineClass *mc = MACHINE_GET_CLASS(ms);
2687 const CPUArchIdList *possible_cpus = mc->possible_cpu_arch_ids(ms);
2689 assert(cpu_index < possible_cpus->len);
2690 return possible_cpus->cpus[cpu_index].props;
2693 static int64_t pc_get_default_cpu_node_id(const MachineState *ms, int idx)
2695 X86CPUTopoInfo topo;
2696 PCMachineState *pcms = PC_MACHINE(ms);
2698 assert(idx < ms->possible_cpus->len);
2699 x86_topo_ids_from_apicid(ms->possible_cpus->cpus[idx].arch_id,
2700 pcms->smp_dies, ms->smp.cores,
2701 ms->smp.threads, &topo);
2702 return topo.pkg_id % ms->numa_state->num_nodes;
2705 static const CPUArchIdList *pc_possible_cpu_arch_ids(MachineState *ms)
2707 PCMachineState *pcms = PC_MACHINE(ms);
2708 int i;
2709 unsigned int max_cpus = ms->smp.max_cpus;
2711 if (ms->possible_cpus) {
2713 * make sure that max_cpus hasn't changed since the first use, i.e.
2714 * -smp hasn't been parsed after it
2716 assert(ms->possible_cpus->len == max_cpus);
2717 return ms->possible_cpus;
2720 ms->possible_cpus = g_malloc0(sizeof(CPUArchIdList) +
2721 sizeof(CPUArchId) * max_cpus);
2722 ms->possible_cpus->len = max_cpus;
2723 for (i = 0; i < ms->possible_cpus->len; i++) {
2724 X86CPUTopoInfo topo;
2726 ms->possible_cpus->cpus[i].type = ms->cpu_type;
2727 ms->possible_cpus->cpus[i].vcpus_count = 1;
2728 ms->possible_cpus->cpus[i].arch_id = x86_cpu_apic_id_from_index(pcms, i);
2729 x86_topo_ids_from_apicid(ms->possible_cpus->cpus[i].arch_id,
2730 pcms->smp_dies, ms->smp.cores,
2731 ms->smp.threads, &topo);
2732 ms->possible_cpus->cpus[i].props.has_socket_id = true;
2733 ms->possible_cpus->cpus[i].props.socket_id = topo.pkg_id;
2734 if (pcms->smp_dies > 1) {
2735 ms->possible_cpus->cpus[i].props.has_die_id = true;
2736 ms->possible_cpus->cpus[i].props.die_id = topo.die_id;
2738 ms->possible_cpus->cpus[i].props.has_core_id = true;
2739 ms->possible_cpus->cpus[i].props.core_id = topo.core_id;
2740 ms->possible_cpus->cpus[i].props.has_thread_id = true;
2741 ms->possible_cpus->cpus[i].props.thread_id = topo.smt_id;
2743 return ms->possible_cpus;
2746 static void x86_nmi(NMIState *n, int cpu_index, Error **errp)
2748 /* cpu index isn't used */
2749 CPUState *cs;
2751 CPU_FOREACH(cs) {
2752 X86CPU *cpu = X86_CPU(cs);
2754 if (!cpu->apic_state) {
2755 cpu_interrupt(cs, CPU_INTERRUPT_NMI);
2756 } else {
2757 apic_deliver_nmi(cpu->apic_state);
2763 static bool pc_hotplug_allowed(MachineState *ms, DeviceState *dev, Error **errp)
2765 X86IOMMUState *iommu = x86_iommu_get_default();
2766 IntelIOMMUState *intel_iommu;
2768 if (iommu &&
2769 object_dynamic_cast((Object *)iommu, TYPE_INTEL_IOMMU_DEVICE) &&
2770 object_dynamic_cast((Object *)dev, "vfio-pci")) {
2771 intel_iommu = INTEL_IOMMU_DEVICE(iommu);
2772 if (!intel_iommu->caching_mode) {
2773 error_setg(errp, "Device assignment is not allowed without "
2774 "enabling caching-mode=on for Intel IOMMU.");
2775 return false;
2779 return true;
2782 static void pc_machine_class_init(ObjectClass *oc, void *data)
2784 MachineClass *mc = MACHINE_CLASS(oc);
2785 PCMachineClass *pcmc = PC_MACHINE_CLASS(oc);
2786 HotplugHandlerClass *hc = HOTPLUG_HANDLER_CLASS(oc);
2787 NMIClass *nc = NMI_CLASS(oc);
2789 pcmc->pci_enabled = true;
2790 pcmc->has_acpi_build = true;
2791 pcmc->rsdp_in_ram = true;
2792 pcmc->smbios_defaults = true;
2793 pcmc->smbios_uuid_encoded = true;
2794 pcmc->gigabyte_align = true;
2795 pcmc->has_reserved_memory = true;
2796 pcmc->kvmclock_enabled = true;
2797 pcmc->enforce_aligned_dimm = true;
2798 /* BIOS ACPI tables: 128K. Other BIOS datastructures: less than 4K reported
2799 * to be used at the moment, 32K should be enough for a while. */
2800 pcmc->acpi_data_size = 0x20000 + 0x8000;
2801 pcmc->save_tsc_khz = true;
2802 pcmc->linuxboot_dma_enabled = true;
2803 pcmc->pvh_enabled = true;
2804 assert(!mc->get_hotplug_handler);
2805 mc->get_hotplug_handler = pc_get_hotplug_handler;
2806 mc->hotplug_allowed = pc_hotplug_allowed;
2807 mc->cpu_index_to_instance_props = pc_cpu_index_to_props;
2808 mc->get_default_cpu_node_id = pc_get_default_cpu_node_id;
2809 mc->possible_cpu_arch_ids = pc_possible_cpu_arch_ids;
2810 mc->auto_enable_numa_with_memhp = true;
2811 mc->has_hotpluggable_cpus = true;
2812 mc->default_boot_order = "cad";
2813 mc->hot_add_cpu = pc_hot_add_cpu;
2814 mc->smp_parse = pc_smp_parse;
2815 mc->block_default_type = IF_IDE;
2816 mc->max_cpus = 255;
2817 mc->reset = pc_machine_reset;
2818 mc->wakeup = pc_machine_wakeup;
2819 hc->pre_plug = pc_machine_device_pre_plug_cb;
2820 hc->plug = pc_machine_device_plug_cb;
2821 hc->unplug_request = pc_machine_device_unplug_request_cb;
2822 hc->unplug = pc_machine_device_unplug_cb;
2823 nc->nmi_monitor_handler = x86_nmi;
2824 mc->default_cpu_type = TARGET_DEFAULT_CPU_TYPE;
2825 mc->nvdimm_supported = true;
2826 mc->numa_mem_supported = true;
2828 object_class_property_add(oc, PC_MACHINE_DEVMEM_REGION_SIZE, "int",
2829 pc_machine_get_device_memory_region_size, NULL,
2830 NULL, NULL, &error_abort);
2832 object_class_property_add(oc, PC_MACHINE_MAX_RAM_BELOW_4G, "size",
2833 pc_machine_get_max_ram_below_4g, pc_machine_set_max_ram_below_4g,
2834 NULL, NULL, &error_abort);
2836 object_class_property_set_description(oc, PC_MACHINE_MAX_RAM_BELOW_4G,
2837 "Maximum ram below the 4G boundary (32bit boundary)", &error_abort);
2839 object_class_property_add(oc, PC_MACHINE_SMM, "OnOffAuto",
2840 pc_machine_get_smm, pc_machine_set_smm,
2841 NULL, NULL, &error_abort);
2842 object_class_property_set_description(oc, PC_MACHINE_SMM,
2843 "Enable SMM (pc & q35)", &error_abort);
2845 object_class_property_add(oc, PC_MACHINE_VMPORT, "OnOffAuto",
2846 pc_machine_get_vmport, pc_machine_set_vmport,
2847 NULL, NULL, &error_abort);
2848 object_class_property_set_description(oc, PC_MACHINE_VMPORT,
2849 "Enable vmport (pc & q35)", &error_abort);
2851 object_class_property_add_bool(oc, PC_MACHINE_SMBUS,
2852 pc_machine_get_smbus, pc_machine_set_smbus, &error_abort);
2854 object_class_property_add_bool(oc, PC_MACHINE_SATA,
2855 pc_machine_get_sata, pc_machine_set_sata, &error_abort);
2857 object_class_property_add_bool(oc, PC_MACHINE_PIT,
2858 pc_machine_get_pit, pc_machine_set_pit, &error_abort);
2861 static const TypeInfo pc_machine_info = {
2862 .name = TYPE_PC_MACHINE,
2863 .parent = TYPE_MACHINE,
2864 .abstract = true,
2865 .instance_size = sizeof(PCMachineState),
2866 .instance_init = pc_machine_initfn,
2867 .class_size = sizeof(PCMachineClass),
2868 .class_init = pc_machine_class_init,
2869 .interfaces = (InterfaceInfo[]) {
2870 { TYPE_HOTPLUG_HANDLER },
2871 { TYPE_NMI },
2876 static void pc_machine_register_types(void)
2878 type_register_static(&pc_machine_info);
2881 type_init(pc_machine_register_types)