xref: /openbmc/qemu/hw/i386/pc.c (revision 08b97f7f)
1 /*
2  * QEMU PC System Emulator
3  *
4  * Copyright (c) 2003-2004 Fabrice Bellard
5  *
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:
12  *
13  * The above copyright notice and this permission notice shall be included in
14  * all copies or substantial portions of the Software.
15  *
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.
23  */
24 
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 "hw/i386/fw_cfg.h"
34 #include "sysemu/cpus.h"
35 #include "hw/block/fdc.h"
36 #include "hw/ide.h"
37 #include "hw/pci/pci.h"
38 #include "hw/pci/pci_bus.h"
39 #include "hw/nvram/fw_cfg.h"
40 #include "hw/timer/hpet.h"
41 #include "hw/firmware/smbios.h"
42 #include "hw/loader.h"
43 #include "elf.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/audio/pcspk.h"
50 #include "hw/pci/msi.h"
51 #include "hw/sysbus.h"
52 #include "sysemu/sysemu.h"
53 #include "sysemu/tcg.h"
54 #include "sysemu/numa.h"
55 #include "sysemu/kvm.h"
56 #include "sysemu/qtest.h"
57 #include "kvm_i386.h"
58 #include "hw/xen/xen.h"
59 #include "hw/xen/start_info.h"
60 #include "ui/qemu-spice.h"
61 #include "exec/memory.h"
62 #include "exec/address-spaces.h"
63 #include "sysemu/arch_init.h"
64 #include "qemu/bitmap.h"
65 #include "qemu/config-file.h"
66 #include "qemu/error-report.h"
67 #include "qemu/option.h"
68 #include "hw/acpi/acpi.h"
69 #include "hw/acpi/cpu_hotplug.h"
70 #include "hw/boards.h"
71 #include "acpi-build.h"
72 #include "hw/mem/pc-dimm.h"
73 #include "qapi/error.h"
74 #include "qapi/qapi-visit-common.h"
75 #include "qapi/visitor.h"
76 #include "qom/cpu.h"
77 #include "hw/nmi.h"
78 #include "hw/usb.h"
79 #include "hw/i386/intel_iommu.h"
80 #include "hw/net/ne2000-isa.h"
81 #include "standard-headers/asm-x86/bootparam.h"
82 #include "hw/virtio/virtio-pmem-pci.h"
83 #include "hw/mem/memory-device.h"
84 #include "sysemu/replay.h"
85 #include "qapi/qmp/qerror.h"
86 
87 /* debug PC/ISA interrupts */
88 //#define DEBUG_IRQ
89 
90 #ifdef DEBUG_IRQ
91 #define DPRINTF(fmt, ...)                                       \
92     do { printf("CPUIRQ: " fmt , ## __VA_ARGS__); } while (0)
93 #else
94 #define DPRINTF(fmt, ...)
95 #endif
96 
97 #define E820_NR_ENTRIES		16
98 
99 struct e820_entry {
100     uint64_t address;
101     uint64_t length;
102     uint32_t type;
103 } QEMU_PACKED __attribute((__aligned__(4)));
104 
105 struct e820_table {
106     uint32_t count;
107     struct e820_entry entry[E820_NR_ENTRIES];
108 } QEMU_PACKED __attribute((__aligned__(4)));
109 
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};
114 
115 /* Physical Address of PVH entry point read from kernel ELF NOTE */
116 static size_t pvh_start_addr;
117 
118 GlobalProperty pc_compat_4_0[] = {};
119 const size_t pc_compat_4_0_len = G_N_ELEMENTS(pc_compat_4_0);
120 
121 GlobalProperty pc_compat_3_1[] = {
122     { "intel-iommu", "dma-drain", "off" },
123     { "Opteron_G3" "-" TYPE_X86_CPU, "rdtscp", "off" },
124     { "Opteron_G4" "-" TYPE_X86_CPU, "rdtscp", "off" },
125     { "Opteron_G4" "-" TYPE_X86_CPU, "npt", "off" },
126     { "Opteron_G4" "-" TYPE_X86_CPU, "nrip-save", "off" },
127     { "Opteron_G5" "-" TYPE_X86_CPU, "rdtscp", "off" },
128     { "Opteron_G5" "-" TYPE_X86_CPU, "npt", "off" },
129     { "Opteron_G5" "-" TYPE_X86_CPU, "nrip-save", "off" },
130     { "EPYC" "-" TYPE_X86_CPU, "npt", "off" },
131     { "EPYC" "-" TYPE_X86_CPU, "nrip-save", "off" },
132     { "EPYC-IBPB" "-" TYPE_X86_CPU, "npt", "off" },
133     { "EPYC-IBPB" "-" TYPE_X86_CPU, "nrip-save", "off" },
134     { "Skylake-Client" "-" TYPE_X86_CPU,      "mpx", "on" },
135     { "Skylake-Client-IBRS" "-" TYPE_X86_CPU, "mpx", "on" },
136     { "Skylake-Server" "-" TYPE_X86_CPU,      "mpx", "on" },
137     { "Skylake-Server-IBRS" "-" TYPE_X86_CPU, "mpx", "on" },
138     { "Cascadelake-Server" "-" TYPE_X86_CPU,  "mpx", "on" },
139     { "Icelake-Client" "-" TYPE_X86_CPU,      "mpx", "on" },
140     { "Icelake-Server" "-" TYPE_X86_CPU,      "mpx", "on" },
141     { "Cascadelake-Server" "-" TYPE_X86_CPU, "stepping", "5" },
142     { TYPE_X86_CPU, "x-intel-pt-auto-level", "off" },
143 };
144 const size_t pc_compat_3_1_len = G_N_ELEMENTS(pc_compat_3_1);
145 
146 GlobalProperty pc_compat_3_0[] = {
147     { TYPE_X86_CPU, "x-hv-synic-kvm-only", "on" },
148     { "Skylake-Server" "-" TYPE_X86_CPU, "pku", "off" },
149     { "Skylake-Server-IBRS" "-" TYPE_X86_CPU, "pku", "off" },
150 };
151 const size_t pc_compat_3_0_len = G_N_ELEMENTS(pc_compat_3_0);
152 
153 GlobalProperty pc_compat_2_12[] = {
154     { TYPE_X86_CPU, "legacy-cache", "on" },
155     { TYPE_X86_CPU, "topoext", "off" },
156     { "EPYC-" TYPE_X86_CPU, "xlevel", "0x8000000a" },
157     { "EPYC-IBPB-" TYPE_X86_CPU, "xlevel", "0x8000000a" },
158 };
159 const size_t pc_compat_2_12_len = G_N_ELEMENTS(pc_compat_2_12);
160 
161 GlobalProperty pc_compat_2_11[] = {
162     { TYPE_X86_CPU, "x-migrate-smi-count", "off" },
163     { "Skylake-Server" "-" TYPE_X86_CPU, "clflushopt", "off" },
164 };
165 const size_t pc_compat_2_11_len = G_N_ELEMENTS(pc_compat_2_11);
166 
167 GlobalProperty pc_compat_2_10[] = {
168     { TYPE_X86_CPU, "x-hv-max-vps", "0x40" },
169     { "i440FX-pcihost", "x-pci-hole64-fix", "off" },
170     { "q35-pcihost", "x-pci-hole64-fix", "off" },
171 };
172 const size_t pc_compat_2_10_len = G_N_ELEMENTS(pc_compat_2_10);
173 
174 GlobalProperty pc_compat_2_9[] = {
175     { "mch", "extended-tseg-mbytes", "0" },
176 };
177 const size_t pc_compat_2_9_len = G_N_ELEMENTS(pc_compat_2_9);
178 
179 GlobalProperty pc_compat_2_8[] = {
180     { TYPE_X86_CPU, "tcg-cpuid", "off" },
181     { "kvmclock", "x-mach-use-reliable-get-clock", "off" },
182     { "ICH9-LPC", "x-smi-broadcast", "off" },
183     { TYPE_X86_CPU, "vmware-cpuid-freq", "off" },
184     { "Haswell-" TYPE_X86_CPU, "stepping", "1" },
185 };
186 const size_t pc_compat_2_8_len = G_N_ELEMENTS(pc_compat_2_8);
187 
188 GlobalProperty pc_compat_2_7[] = {
189     { TYPE_X86_CPU, "l3-cache", "off" },
190     { TYPE_X86_CPU, "full-cpuid-auto-level", "off" },
191     { "Opteron_G3" "-" TYPE_X86_CPU, "family", "15" },
192     { "Opteron_G3" "-" TYPE_X86_CPU, "model", "6" },
193     { "Opteron_G3" "-" TYPE_X86_CPU, "stepping", "1" },
194     { "isa-pcspk", "migrate", "off" },
195 };
196 const size_t pc_compat_2_7_len = G_N_ELEMENTS(pc_compat_2_7);
197 
198 GlobalProperty pc_compat_2_6[] = {
199     { TYPE_X86_CPU, "cpuid-0xb", "off" },
200     { "vmxnet3", "romfile", "" },
201     { TYPE_X86_CPU, "fill-mtrr-mask", "off" },
202     { "apic-common", "legacy-instance-id", "on", }
203 };
204 const size_t pc_compat_2_6_len = G_N_ELEMENTS(pc_compat_2_6);
205 
206 GlobalProperty pc_compat_2_5[] = {};
207 const size_t pc_compat_2_5_len = G_N_ELEMENTS(pc_compat_2_5);
208 
209 GlobalProperty pc_compat_2_4[] = {
210     PC_CPU_MODEL_IDS("2.4.0")
211     { "Haswell-" TYPE_X86_CPU, "abm", "off" },
212     { "Haswell-noTSX-" TYPE_X86_CPU, "abm", "off" },
213     { "Broadwell-" TYPE_X86_CPU, "abm", "off" },
214     { "Broadwell-noTSX-" TYPE_X86_CPU, "abm", "off" },
215     { "host" "-" TYPE_X86_CPU, "host-cache-info", "on" },
216     { TYPE_X86_CPU, "check", "off" },
217     { "qemu64" "-" TYPE_X86_CPU, "sse4a", "on" },
218     { "qemu64" "-" TYPE_X86_CPU, "abm", "on" },
219     { "qemu64" "-" TYPE_X86_CPU, "popcnt", "on" },
220     { "qemu32" "-" TYPE_X86_CPU, "popcnt", "on" },
221     { "Opteron_G2" "-" TYPE_X86_CPU, "rdtscp", "on" },
222     { "Opteron_G3" "-" TYPE_X86_CPU, "rdtscp", "on" },
223     { "Opteron_G4" "-" TYPE_X86_CPU, "rdtscp", "on" },
224     { "Opteron_G5" "-" TYPE_X86_CPU, "rdtscp", "on", }
225 };
226 const size_t pc_compat_2_4_len = G_N_ELEMENTS(pc_compat_2_4);
227 
228 GlobalProperty pc_compat_2_3[] = {
229     PC_CPU_MODEL_IDS("2.3.0")
230     { TYPE_X86_CPU, "arat", "off" },
231     { "qemu64" "-" TYPE_X86_CPU, "min-level", "4" },
232     { "kvm64" "-" TYPE_X86_CPU, "min-level", "5" },
233     { "pentium3" "-" TYPE_X86_CPU, "min-level", "2" },
234     { "n270" "-" TYPE_X86_CPU, "min-level", "5" },
235     { "Conroe" "-" TYPE_X86_CPU, "min-level", "4" },
236     { "Penryn" "-" TYPE_X86_CPU, "min-level", "4" },
237     { "Nehalem" "-" TYPE_X86_CPU, "min-level", "4" },
238     { "n270" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
239     { "Penryn" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
240     { "Conroe" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
241     { "Nehalem" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
242     { "Westmere" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
243     { "SandyBridge" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
244     { "IvyBridge" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
245     { "Haswell" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
246     { "Haswell-noTSX" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
247     { "Broadwell" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
248     { "Broadwell-noTSX" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
249     { TYPE_X86_CPU, "kvm-no-smi-migration", "on" },
250 };
251 const size_t pc_compat_2_3_len = G_N_ELEMENTS(pc_compat_2_3);
252 
253 GlobalProperty pc_compat_2_2[] = {
254     PC_CPU_MODEL_IDS("2.2.0")
255     { "kvm64" "-" TYPE_X86_CPU, "vme", "off" },
256     { "kvm32" "-" TYPE_X86_CPU, "vme", "off" },
257     { "Conroe" "-" TYPE_X86_CPU, "vme", "off" },
258     { "Penryn" "-" TYPE_X86_CPU, "vme", "off" },
259     { "Nehalem" "-" TYPE_X86_CPU, "vme", "off" },
260     { "Westmere" "-" TYPE_X86_CPU, "vme", "off" },
261     { "SandyBridge" "-" TYPE_X86_CPU, "vme", "off" },
262     { "Haswell" "-" TYPE_X86_CPU, "vme", "off" },
263     { "Broadwell" "-" TYPE_X86_CPU, "vme", "off" },
264     { "Opteron_G1" "-" TYPE_X86_CPU, "vme", "off" },
265     { "Opteron_G2" "-" TYPE_X86_CPU, "vme", "off" },
266     { "Opteron_G3" "-" TYPE_X86_CPU, "vme", "off" },
267     { "Opteron_G4" "-" TYPE_X86_CPU, "vme", "off" },
268     { "Opteron_G5" "-" TYPE_X86_CPU, "vme", "off" },
269     { "Haswell" "-" TYPE_X86_CPU, "f16c", "off" },
270     { "Haswell" "-" TYPE_X86_CPU, "rdrand", "off" },
271     { "Broadwell" "-" TYPE_X86_CPU, "f16c", "off" },
272     { "Broadwell" "-" TYPE_X86_CPU, "rdrand", "off" },
273 };
274 const size_t pc_compat_2_2_len = G_N_ELEMENTS(pc_compat_2_2);
275 
276 GlobalProperty pc_compat_2_1[] = {
277     PC_CPU_MODEL_IDS("2.1.0")
278     { "coreduo" "-" TYPE_X86_CPU, "vmx", "on" },
279     { "core2duo" "-" TYPE_X86_CPU, "vmx", "on" },
280 };
281 const size_t pc_compat_2_1_len = G_N_ELEMENTS(pc_compat_2_1);
282 
283 GlobalProperty pc_compat_2_0[] = {
284     PC_CPU_MODEL_IDS("2.0.0")
285     { "virtio-scsi-pci", "any_layout", "off" },
286     { "PIIX4_PM", "memory-hotplug-support", "off" },
287     { "apic", "version", "0x11" },
288     { "nec-usb-xhci", "superspeed-ports-first", "off" },
289     { "nec-usb-xhci", "force-pcie-endcap", "on" },
290     { "pci-serial", "prog_if", "0" },
291     { "pci-serial-2x", "prog_if", "0" },
292     { "pci-serial-4x", "prog_if", "0" },
293     { "virtio-net-pci", "guest_announce", "off" },
294     { "ICH9-LPC", "memory-hotplug-support", "off" },
295     { "xio3130-downstream", COMPAT_PROP_PCP, "off" },
296     { "ioh3420", COMPAT_PROP_PCP, "off" },
297 };
298 const size_t pc_compat_2_0_len = G_N_ELEMENTS(pc_compat_2_0);
299 
300 GlobalProperty pc_compat_1_7[] = {
301     PC_CPU_MODEL_IDS("1.7.0")
302     { TYPE_USB_DEVICE, "msos-desc", "no" },
303     { "PIIX4_PM", "acpi-pci-hotplug-with-bridge-support", "off" },
304     { "hpet", HPET_INTCAP, "4" },
305 };
306 const size_t pc_compat_1_7_len = G_N_ELEMENTS(pc_compat_1_7);
307 
308 GlobalProperty pc_compat_1_6[] = {
309     PC_CPU_MODEL_IDS("1.6.0")
310     { "e1000", "mitigation", "off" },
311     { "qemu64-" TYPE_X86_CPU, "model", "2" },
312     { "qemu32-" TYPE_X86_CPU, "model", "3" },
313     { "i440FX-pcihost", "short_root_bus", "1" },
314     { "q35-pcihost", "short_root_bus", "1" },
315 };
316 const size_t pc_compat_1_6_len = G_N_ELEMENTS(pc_compat_1_6);
317 
318 GlobalProperty pc_compat_1_5[] = {
319     PC_CPU_MODEL_IDS("1.5.0")
320     { "Conroe-" TYPE_X86_CPU, "model", "2" },
321     { "Conroe-" TYPE_X86_CPU, "min-level", "2" },
322     { "Penryn-" TYPE_X86_CPU, "model", "2" },
323     { "Penryn-" TYPE_X86_CPU, "min-level", "2" },
324     { "Nehalem-" TYPE_X86_CPU, "model", "2" },
325     { "Nehalem-" TYPE_X86_CPU, "min-level", "2" },
326     { "virtio-net-pci", "any_layout", "off" },
327     { TYPE_X86_CPU, "pmu", "on" },
328     { "i440FX-pcihost", "short_root_bus", "0" },
329     { "q35-pcihost", "short_root_bus", "0" },
330 };
331 const size_t pc_compat_1_5_len = G_N_ELEMENTS(pc_compat_1_5);
332 
333 GlobalProperty pc_compat_1_4[] = {
334     PC_CPU_MODEL_IDS("1.4.0")
335     { "scsi-hd", "discard_granularity", "0" },
336     { "scsi-cd", "discard_granularity", "0" },
337     { "scsi-disk", "discard_granularity", "0" },
338     { "ide-hd", "discard_granularity", "0" },
339     { "ide-cd", "discard_granularity", "0" },
340     { "ide-drive", "discard_granularity", "0" },
341     { "virtio-blk-pci", "discard_granularity", "0" },
342     /* DEV_NVECTORS_UNSPECIFIED as a uint32_t string: */
343     { "virtio-serial-pci", "vectors", "0xFFFFFFFF" },
344     { "virtio-net-pci", "ctrl_guest_offloads", "off" },
345     { "e1000", "romfile", "pxe-e1000.rom" },
346     { "ne2k_pci", "romfile", "pxe-ne2k_pci.rom" },
347     { "pcnet", "romfile", "pxe-pcnet.rom" },
348     { "rtl8139", "romfile", "pxe-rtl8139.rom" },
349     { "virtio-net-pci", "romfile", "pxe-virtio.rom" },
350     { "486-" TYPE_X86_CPU, "model", "0" },
351     { "n270" "-" TYPE_X86_CPU, "movbe", "off" },
352     { "Westmere" "-" TYPE_X86_CPU, "pclmulqdq", "off" },
353 };
354 const size_t pc_compat_1_4_len = G_N_ELEMENTS(pc_compat_1_4);
355 
356 void gsi_handler(void *opaque, int n, int level)
357 {
358     GSIState *s = opaque;
359 
360     DPRINTF("pc: %s GSI %d\n", level ? "raising" : "lowering", n);
361     if (n < ISA_NUM_IRQS) {
362         qemu_set_irq(s->i8259_irq[n], level);
363     }
364     qemu_set_irq(s->ioapic_irq[n], level);
365 }
366 
367 static void ioport80_write(void *opaque, hwaddr addr, uint64_t data,
368                            unsigned size)
369 {
370 }
371 
372 static uint64_t ioport80_read(void *opaque, hwaddr addr, unsigned size)
373 {
374     return 0xffffffffffffffffULL;
375 }
376 
377 /* MSDOS compatibility mode FPU exception support */
378 static qemu_irq ferr_irq;
379 
380 void pc_register_ferr_irq(qemu_irq irq)
381 {
382     ferr_irq = irq;
383 }
384 
385 /* XXX: add IGNNE support */
386 void cpu_set_ferr(CPUX86State *s)
387 {
388     qemu_irq_raise(ferr_irq);
389 }
390 
391 static void ioportF0_write(void *opaque, hwaddr addr, uint64_t data,
392                            unsigned size)
393 {
394     qemu_irq_lower(ferr_irq);
395 }
396 
397 static uint64_t ioportF0_read(void *opaque, hwaddr addr, unsigned size)
398 {
399     return 0xffffffffffffffffULL;
400 }
401 
402 /* TSC handling */
403 uint64_t cpu_get_tsc(CPUX86State *env)
404 {
405     return cpu_get_ticks();
406 }
407 
408 /* IRQ handling */
409 int cpu_get_pic_interrupt(CPUX86State *env)
410 {
411     X86CPU *cpu = env_archcpu(env);
412     int intno;
413 
414     if (!kvm_irqchip_in_kernel()) {
415         intno = apic_get_interrupt(cpu->apic_state);
416         if (intno >= 0) {
417             return intno;
418         }
419         /* read the irq from the PIC */
420         if (!apic_accept_pic_intr(cpu->apic_state)) {
421             return -1;
422         }
423     }
424 
425     intno = pic_read_irq(isa_pic);
426     return intno;
427 }
428 
429 static void pic_irq_request(void *opaque, int irq, int level)
430 {
431     CPUState *cs = first_cpu;
432     X86CPU *cpu = X86_CPU(cs);
433 
434     DPRINTF("pic_irqs: %s irq %d\n", level? "raise" : "lower", irq);
435     if (cpu->apic_state && !kvm_irqchip_in_kernel()) {
436         CPU_FOREACH(cs) {
437             cpu = X86_CPU(cs);
438             if (apic_accept_pic_intr(cpu->apic_state)) {
439                 apic_deliver_pic_intr(cpu->apic_state, level);
440             }
441         }
442     } else {
443         if (level) {
444             cpu_interrupt(cs, CPU_INTERRUPT_HARD);
445         } else {
446             cpu_reset_interrupt(cs, CPU_INTERRUPT_HARD);
447         }
448     }
449 }
450 
451 /* PC cmos mappings */
452 
453 #define REG_EQUIPMENT_BYTE          0x14
454 
455 int cmos_get_fd_drive_type(FloppyDriveType fd0)
456 {
457     int val;
458 
459     switch (fd0) {
460     case FLOPPY_DRIVE_TYPE_144:
461         /* 1.44 Mb 3"5 drive */
462         val = 4;
463         break;
464     case FLOPPY_DRIVE_TYPE_288:
465         /* 2.88 Mb 3"5 drive */
466         val = 5;
467         break;
468     case FLOPPY_DRIVE_TYPE_120:
469         /* 1.2 Mb 5"5 drive */
470         val = 2;
471         break;
472     case FLOPPY_DRIVE_TYPE_NONE:
473     default:
474         val = 0;
475         break;
476     }
477     return val;
478 }
479 
480 static void cmos_init_hd(ISADevice *s, int type_ofs, int info_ofs,
481                          int16_t cylinders, int8_t heads, int8_t sectors)
482 {
483     rtc_set_memory(s, type_ofs, 47);
484     rtc_set_memory(s, info_ofs, cylinders);
485     rtc_set_memory(s, info_ofs + 1, cylinders >> 8);
486     rtc_set_memory(s, info_ofs + 2, heads);
487     rtc_set_memory(s, info_ofs + 3, 0xff);
488     rtc_set_memory(s, info_ofs + 4, 0xff);
489     rtc_set_memory(s, info_ofs + 5, 0xc0 | ((heads > 8) << 3));
490     rtc_set_memory(s, info_ofs + 6, cylinders);
491     rtc_set_memory(s, info_ofs + 7, cylinders >> 8);
492     rtc_set_memory(s, info_ofs + 8, sectors);
493 }
494 
495 /* convert boot_device letter to something recognizable by the bios */
496 static int boot_device2nibble(char boot_device)
497 {
498     switch(boot_device) {
499     case 'a':
500     case 'b':
501         return 0x01; /* floppy boot */
502     case 'c':
503         return 0x02; /* hard drive boot */
504     case 'd':
505         return 0x03; /* CD-ROM boot */
506     case 'n':
507         return 0x04; /* Network boot */
508     }
509     return 0;
510 }
511 
512 static void set_boot_dev(ISADevice *s, const char *boot_device, Error **errp)
513 {
514 #define PC_MAX_BOOT_DEVICES 3
515     int nbds, bds[3] = { 0, };
516     int i;
517 
518     nbds = strlen(boot_device);
519     if (nbds > PC_MAX_BOOT_DEVICES) {
520         error_setg(errp, "Too many boot devices for PC");
521         return;
522     }
523     for (i = 0; i < nbds; i++) {
524         bds[i] = boot_device2nibble(boot_device[i]);
525         if (bds[i] == 0) {
526             error_setg(errp, "Invalid boot device for PC: '%c'",
527                        boot_device[i]);
528             return;
529         }
530     }
531     rtc_set_memory(s, 0x3d, (bds[1] << 4) | bds[0]);
532     rtc_set_memory(s, 0x38, (bds[2] << 4) | (fd_bootchk ? 0x0 : 0x1));
533 }
534 
535 static void pc_boot_set(void *opaque, const char *boot_device, Error **errp)
536 {
537     set_boot_dev(opaque, boot_device, errp);
538 }
539 
540 static void pc_cmos_init_floppy(ISADevice *rtc_state, ISADevice *floppy)
541 {
542     int val, nb, i;
543     FloppyDriveType fd_type[2] = { FLOPPY_DRIVE_TYPE_NONE,
544                                    FLOPPY_DRIVE_TYPE_NONE };
545 
546     /* floppy type */
547     if (floppy) {
548         for (i = 0; i < 2; i++) {
549             fd_type[i] = isa_fdc_get_drive_type(floppy, i);
550         }
551     }
552     val = (cmos_get_fd_drive_type(fd_type[0]) << 4) |
553         cmos_get_fd_drive_type(fd_type[1]);
554     rtc_set_memory(rtc_state, 0x10, val);
555 
556     val = rtc_get_memory(rtc_state, REG_EQUIPMENT_BYTE);
557     nb = 0;
558     if (fd_type[0] != FLOPPY_DRIVE_TYPE_NONE) {
559         nb++;
560     }
561     if (fd_type[1] != FLOPPY_DRIVE_TYPE_NONE) {
562         nb++;
563     }
564     switch (nb) {
565     case 0:
566         break;
567     case 1:
568         val |= 0x01; /* 1 drive, ready for boot */
569         break;
570     case 2:
571         val |= 0x41; /* 2 drives, ready for boot */
572         break;
573     }
574     rtc_set_memory(rtc_state, REG_EQUIPMENT_BYTE, val);
575 }
576 
577 typedef struct pc_cmos_init_late_arg {
578     ISADevice *rtc_state;
579     BusState *idebus[2];
580 } pc_cmos_init_late_arg;
581 
582 typedef struct check_fdc_state {
583     ISADevice *floppy;
584     bool multiple;
585 } CheckFdcState;
586 
587 static int check_fdc(Object *obj, void *opaque)
588 {
589     CheckFdcState *state = opaque;
590     Object *fdc;
591     uint32_t iobase;
592     Error *local_err = NULL;
593 
594     fdc = object_dynamic_cast(obj, TYPE_ISA_FDC);
595     if (!fdc) {
596         return 0;
597     }
598 
599     iobase = object_property_get_uint(obj, "iobase", &local_err);
600     if (local_err || iobase != 0x3f0) {
601         error_free(local_err);
602         return 0;
603     }
604 
605     if (state->floppy) {
606         state->multiple = true;
607     } else {
608         state->floppy = ISA_DEVICE(obj);
609     }
610     return 0;
611 }
612 
613 static const char * const fdc_container_path[] = {
614     "/unattached", "/peripheral", "/peripheral-anon"
615 };
616 
617 /*
618  * Locate the FDC at IO address 0x3f0, in order to configure the CMOS registers
619  * and ACPI objects.
620  */
621 ISADevice *pc_find_fdc0(void)
622 {
623     int i;
624     Object *container;
625     CheckFdcState state = { 0 };
626 
627     for (i = 0; i < ARRAY_SIZE(fdc_container_path); i++) {
628         container = container_get(qdev_get_machine(), fdc_container_path[i]);
629         object_child_foreach(container, check_fdc, &state);
630     }
631 
632     if (state.multiple) {
633         warn_report("multiple floppy disk controllers with "
634                     "iobase=0x3f0 have been found");
635         error_printf("the one being picked for CMOS setup might not reflect "
636                      "your intent");
637     }
638 
639     return state.floppy;
640 }
641 
642 static void pc_cmos_init_late(void *opaque)
643 {
644     pc_cmos_init_late_arg *arg = opaque;
645     ISADevice *s = arg->rtc_state;
646     int16_t cylinders;
647     int8_t heads, sectors;
648     int val;
649     int i, trans;
650 
651     val = 0;
652     if (arg->idebus[0] && ide_get_geometry(arg->idebus[0], 0,
653                                            &cylinders, &heads, &sectors) >= 0) {
654         cmos_init_hd(s, 0x19, 0x1b, cylinders, heads, sectors);
655         val |= 0xf0;
656     }
657     if (arg->idebus[0] && ide_get_geometry(arg->idebus[0], 1,
658                                            &cylinders, &heads, &sectors) >= 0) {
659         cmos_init_hd(s, 0x1a, 0x24, cylinders, heads, sectors);
660         val |= 0x0f;
661     }
662     rtc_set_memory(s, 0x12, val);
663 
664     val = 0;
665     for (i = 0; i < 4; i++) {
666         /* NOTE: ide_get_geometry() returns the physical
667            geometry.  It is always such that: 1 <= sects <= 63, 1
668            <= heads <= 16, 1 <= cylinders <= 16383. The BIOS
669            geometry can be different if a translation is done. */
670         if (arg->idebus[i / 2] &&
671             ide_get_geometry(arg->idebus[i / 2], i % 2,
672                              &cylinders, &heads, &sectors) >= 0) {
673             trans = ide_get_bios_chs_trans(arg->idebus[i / 2], i % 2) - 1;
674             assert((trans & ~3) == 0);
675             val |= trans << (i * 2);
676         }
677     }
678     rtc_set_memory(s, 0x39, val);
679 
680     pc_cmos_init_floppy(s, pc_find_fdc0());
681 
682     qemu_unregister_reset(pc_cmos_init_late, opaque);
683 }
684 
685 void pc_cmos_init(PCMachineState *pcms,
686                   BusState *idebus0, BusState *idebus1,
687                   ISADevice *s)
688 {
689     int val;
690     static pc_cmos_init_late_arg arg;
691 
692     /* various important CMOS locations needed by PC/Bochs bios */
693 
694     /* memory size */
695     /* base memory (first MiB) */
696     val = MIN(pcms->below_4g_mem_size / KiB, 640);
697     rtc_set_memory(s, 0x15, val);
698     rtc_set_memory(s, 0x16, val >> 8);
699     /* extended memory (next 64MiB) */
700     if (pcms->below_4g_mem_size > 1 * MiB) {
701         val = (pcms->below_4g_mem_size - 1 * MiB) / KiB;
702     } else {
703         val = 0;
704     }
705     if (val > 65535)
706         val = 65535;
707     rtc_set_memory(s, 0x17, val);
708     rtc_set_memory(s, 0x18, val >> 8);
709     rtc_set_memory(s, 0x30, val);
710     rtc_set_memory(s, 0x31, val >> 8);
711     /* memory between 16MiB and 4GiB */
712     if (pcms->below_4g_mem_size > 16 * MiB) {
713         val = (pcms->below_4g_mem_size - 16 * MiB) / (64 * KiB);
714     } else {
715         val = 0;
716     }
717     if (val > 65535)
718         val = 65535;
719     rtc_set_memory(s, 0x34, val);
720     rtc_set_memory(s, 0x35, val >> 8);
721     /* memory above 4GiB */
722     val = pcms->above_4g_mem_size / 65536;
723     rtc_set_memory(s, 0x5b, val);
724     rtc_set_memory(s, 0x5c, val >> 8);
725     rtc_set_memory(s, 0x5d, val >> 16);
726 
727     object_property_add_link(OBJECT(pcms), "rtc_state",
728                              TYPE_ISA_DEVICE,
729                              (Object **)&pcms->rtc,
730                              object_property_allow_set_link,
731                              OBJ_PROP_LINK_STRONG, &error_abort);
732     object_property_set_link(OBJECT(pcms), OBJECT(s),
733                              "rtc_state", &error_abort);
734 
735     set_boot_dev(s, MACHINE(pcms)->boot_order, &error_fatal);
736 
737     val = 0;
738     val |= 0x02; /* FPU is there */
739     val |= 0x04; /* PS/2 mouse installed */
740     rtc_set_memory(s, REG_EQUIPMENT_BYTE, val);
741 
742     /* hard drives and FDC */
743     arg.rtc_state = s;
744     arg.idebus[0] = idebus0;
745     arg.idebus[1] = idebus1;
746     qemu_register_reset(pc_cmos_init_late, &arg);
747 }
748 
749 #define TYPE_PORT92 "port92"
750 #define PORT92(obj) OBJECT_CHECK(Port92State, (obj), TYPE_PORT92)
751 
752 /* port 92 stuff: could be split off */
753 typedef struct Port92State {
754     ISADevice parent_obj;
755 
756     MemoryRegion io;
757     uint8_t outport;
758     qemu_irq a20_out;
759 } Port92State;
760 
761 static void port92_write(void *opaque, hwaddr addr, uint64_t val,
762                          unsigned size)
763 {
764     Port92State *s = opaque;
765     int oldval = s->outport;
766 
767     DPRINTF("port92: write 0x%02" PRIx64 "\n", val);
768     s->outport = val;
769     qemu_set_irq(s->a20_out, (val >> 1) & 1);
770     if ((val & 1) && !(oldval & 1)) {
771         qemu_system_reset_request(SHUTDOWN_CAUSE_GUEST_RESET);
772     }
773 }
774 
775 static uint64_t port92_read(void *opaque, hwaddr addr,
776                             unsigned size)
777 {
778     Port92State *s = opaque;
779     uint32_t ret;
780 
781     ret = s->outport;
782     DPRINTF("port92: read 0x%02x\n", ret);
783     return ret;
784 }
785 
786 static void port92_init(ISADevice *dev, qemu_irq a20_out)
787 {
788     qdev_connect_gpio_out_named(DEVICE(dev), PORT92_A20_LINE, 0, a20_out);
789 }
790 
791 static const VMStateDescription vmstate_port92_isa = {
792     .name = "port92",
793     .version_id = 1,
794     .minimum_version_id = 1,
795     .fields = (VMStateField[]) {
796         VMSTATE_UINT8(outport, Port92State),
797         VMSTATE_END_OF_LIST()
798     }
799 };
800 
801 static void port92_reset(DeviceState *d)
802 {
803     Port92State *s = PORT92(d);
804 
805     s->outport &= ~1;
806 }
807 
808 static const MemoryRegionOps port92_ops = {
809     .read = port92_read,
810     .write = port92_write,
811     .impl = {
812         .min_access_size = 1,
813         .max_access_size = 1,
814     },
815     .endianness = DEVICE_LITTLE_ENDIAN,
816 };
817 
818 static void port92_initfn(Object *obj)
819 {
820     Port92State *s = PORT92(obj);
821 
822     memory_region_init_io(&s->io, OBJECT(s), &port92_ops, s, "port92", 1);
823 
824     s->outport = 0;
825 
826     qdev_init_gpio_out_named(DEVICE(obj), &s->a20_out, PORT92_A20_LINE, 1);
827 }
828 
829 static void port92_realizefn(DeviceState *dev, Error **errp)
830 {
831     ISADevice *isadev = ISA_DEVICE(dev);
832     Port92State *s = PORT92(dev);
833 
834     isa_register_ioport(isadev, &s->io, 0x92);
835 }
836 
837 static void port92_class_initfn(ObjectClass *klass, void *data)
838 {
839     DeviceClass *dc = DEVICE_CLASS(klass);
840 
841     dc->realize = port92_realizefn;
842     dc->reset = port92_reset;
843     dc->vmsd = &vmstate_port92_isa;
844     /*
845      * Reason: unlike ordinary ISA devices, this one needs additional
846      * wiring: its A20 output line needs to be wired up by
847      * port92_init().
848      */
849     dc->user_creatable = false;
850 }
851 
852 static const TypeInfo port92_info = {
853     .name          = TYPE_PORT92,
854     .parent        = TYPE_ISA_DEVICE,
855     .instance_size = sizeof(Port92State),
856     .instance_init = port92_initfn,
857     .class_init    = port92_class_initfn,
858 };
859 
860 static void port92_register_types(void)
861 {
862     type_register_static(&port92_info);
863 }
864 
865 type_init(port92_register_types)
866 
867 static void handle_a20_line_change(void *opaque, int irq, int level)
868 {
869     X86CPU *cpu = opaque;
870 
871     /* XXX: send to all CPUs ? */
872     /* XXX: add logic to handle multiple A20 line sources */
873     x86_cpu_set_a20(cpu, level);
874 }
875 
876 int e820_add_entry(uint64_t address, uint64_t length, uint32_t type)
877 {
878     int index = le32_to_cpu(e820_reserve.count);
879     struct e820_entry *entry;
880 
881     if (type != E820_RAM) {
882         /* old FW_CFG_E820_TABLE entry -- reservations only */
883         if (index >= E820_NR_ENTRIES) {
884             return -EBUSY;
885         }
886         entry = &e820_reserve.entry[index++];
887 
888         entry->address = cpu_to_le64(address);
889         entry->length = cpu_to_le64(length);
890         entry->type = cpu_to_le32(type);
891 
892         e820_reserve.count = cpu_to_le32(index);
893     }
894 
895     /* new "etc/e820" file -- include ram too */
896     e820_table = g_renew(struct e820_entry, e820_table, e820_entries + 1);
897     e820_table[e820_entries].address = cpu_to_le64(address);
898     e820_table[e820_entries].length = cpu_to_le64(length);
899     e820_table[e820_entries].type = cpu_to_le32(type);
900     e820_entries++;
901 
902     return e820_entries;
903 }
904 
905 int e820_get_num_entries(void)
906 {
907     return e820_entries;
908 }
909 
910 bool e820_get_entry(int idx, uint32_t type, uint64_t *address, uint64_t *length)
911 {
912     if (idx < e820_entries && e820_table[idx].type == cpu_to_le32(type)) {
913         *address = le64_to_cpu(e820_table[idx].address);
914         *length = le64_to_cpu(e820_table[idx].length);
915         return true;
916     }
917     return false;
918 }
919 
920 /* Calculates initial APIC ID for a specific CPU index
921  *
922  * Currently we need to be able to calculate the APIC ID from the CPU index
923  * alone (without requiring a CPU object), as the QEMU<->Seabios interfaces have
924  * no concept of "CPU index", and the NUMA tables on fw_cfg need the APIC ID of
925  * all CPUs up to max_cpus.
926  */
927 static uint32_t x86_cpu_apic_id_from_index(PCMachineState *pcms,
928                                            unsigned int cpu_index)
929 {
930     MachineState *ms = MACHINE(pcms);
931     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
932     uint32_t correct_id;
933     static bool warned;
934 
935     correct_id = x86_apicid_from_cpu_idx(pcms->smp_dies, ms->smp.cores,
936                                          ms->smp.threads, cpu_index);
937     if (pcmc->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;
942         }
943         return cpu_index;
944     } else {
945         return correct_id;
946     }
947 }
948 
949 static void pc_build_smbios(PCMachineState *pcms)
950 {
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);
957 
958     /* tell smbios about cpuid version and features */
959     smbios_set_cpuid(cpu->env.cpuid_version, cpu->env.features[FEAT_1_EDX]);
960 
961     smbios_tables = smbios_get_table_legacy(ms, &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);
965     }
966 
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;
971 
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++;
976         }
977     }
978     smbios_get_tables(ms, mem_array, array_count,
979                       &smbios_tables, &smbios_tables_len,
980                       &smbios_anchor, &smbios_anchor_len);
981     g_free(mem_array);
982 
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);
988     }
989 }
990 
991 static FWCfgState *bochs_bios_init(AddressSpace *as, PCMachineState *pcms)
992 {
993     FWCfgState *fw_cfg;
994     uint64_t *numa_fw_cfg;
995     int i;
996     const CPUArchIdList *cpus;
997     MachineClass *mc = MACHINE_GET_CLASS(pcms);
998 
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);
1001 
1002     /* FW_CFG_MAX_CPUS is a bit confusing/problematic on x86:
1003      *
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".
1010      *
1011      * So for compatibility reasons with old BIOSes we are stuck with
1012      * "etc/max-cpus" actually being apic_id_limit
1013      */
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());
1019 
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);
1024 
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.
1029      */
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);
1037     }
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);
1041     }
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));
1045 
1046     return fw_cfg;
1047 }
1048 
1049 static long get_file_size(FILE *f)
1050 {
1051     long where, size;
1052 
1053     /* XXX: on Unix systems, using fstat() probably makes more sense */
1054 
1055     where = ftell(f);
1056     fseek(f, 0, SEEK_END);
1057     size = ftell(f);
1058     fseek(f, where, SEEK_SET);
1059 
1060     return size;
1061 }
1062 
1063 struct setup_data {
1064     uint64_t next;
1065     uint32_t type;
1066     uint32_t len;
1067     uint8_t data[0];
1068 } __attribute__((packed));
1069 
1070 
1071 /*
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.
1075  *
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.
1080  *
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.
1084  *
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).
1088  */
1089 static uint64_t read_pvh_start_addr(void *arg1, void *arg2, bool is64)
1090 {
1091     size_t *elf_note_data_addr;
1092 
1093     /* Check if ELF Note header passed in is valid */
1094     if (arg1 == NULL) {
1095         return 0;
1096     }
1097 
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;
1103 
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;
1112 
1113         elf_note_data_addr =
1114             ((void *)nhdr32) + nhdr_size32 +
1115             QEMU_ALIGN_UP(nhdr_namesz, phdr_align);
1116     }
1117 
1118     pvh_start_addr = *elf_note_data_addr;
1119 
1120     return pvh_start_addr;
1121 }
1122 
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)
1128 {
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;
1135 
1136     if (ldl_p(header) != 0x464c457f) {
1137         return false; /* no elfboot */
1138     }
1139 
1140     bool elf_is64 = header[EI_CLASS] == ELFCLASS64;
1141     flags = elf_is64 ?
1142         ((Elf64_Ehdr *)header)->e_flags : ((Elf32_Ehdr *)header)->e_flags;
1143 
1144     if (flags & 0x00010004) { /* LOAD_ELF_HEADER_HAS_ADDR */
1145         error_report("elfboot unsupported flags = %x", flags);
1146         exit(1);
1147     }
1148 
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);
1154 
1155     if (kernel_size < 0) {
1156         error_report("Error while loading elf kernel");
1157         exit(1);
1158     }
1159     mh_load_addr = elf_low;
1160     elf_kernel_size = elf_high - elf_low;
1161 
1162     if (pvh_start_addr == 0) {
1163         error_report("Error loading uncompressed kernel without PVH ELF Note");
1164         exit(1);
1165     }
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);
1169 
1170     return true;
1171 }
1172 
1173 static void load_linux(PCMachineState *pcms,
1174                        FWCfgState *fw_cfg)
1175 {
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;
1191 
1192     /* Align to 16 bytes as a paranoia measure */
1193     cmdline_size = (strlen(kernel_cmdline)+16) & ~15;
1194 
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);
1203     }
1204 
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 {
1212         /*
1213          * This could be a multiboot kernel. If it is, let's stop treating it
1214          * like a Linux kernel.
1215          * Note: some multiboot images could be in the ELF format (the same of
1216          * PVH), so we try multiboot first since we check the multiboot magic
1217          * header before to load it.
1218          */
1219         if (load_multiboot(fw_cfg, f, kernel_filename, initrd_filename,
1220                            kernel_cmdline, kernel_size, header)) {
1221             return;
1222         }
1223         /*
1224          * Check if the file is an uncompressed kernel file (ELF) and load it,
1225          * saving the PVH entry point used by the x86/HVM direct boot ABI.
1226          * If load_elfboot() is successful, populate the fw_cfg info.
1227          */
1228         if (pcmc->pvh_enabled &&
1229             load_elfboot(kernel_filename, kernel_size,
1230                          header, pvh_start_addr, fw_cfg)) {
1231             fclose(f);
1232 
1233             fw_cfg_add_i32(fw_cfg, FW_CFG_CMDLINE_SIZE,
1234                 strlen(kernel_cmdline) + 1);
1235             fw_cfg_add_string(fw_cfg, FW_CFG_CMDLINE_DATA, kernel_cmdline);
1236 
1237             fw_cfg_add_i32(fw_cfg, FW_CFG_SETUP_SIZE, sizeof(header));
1238             fw_cfg_add_bytes(fw_cfg, FW_CFG_SETUP_DATA,
1239                              header, sizeof(header));
1240 
1241             /* load initrd */
1242             if (initrd_filename) {
1243                 gsize initrd_size;
1244                 gchar *initrd_data;
1245                 GError *gerr = NULL;
1246 
1247                 if (!g_file_get_contents(initrd_filename, &initrd_data,
1248                             &initrd_size, &gerr)) {
1249                     fprintf(stderr, "qemu: error reading initrd %s: %s\n",
1250                             initrd_filename, gerr->message);
1251                     exit(1);
1252                 }
1253 
1254                 initrd_max = pcms->below_4g_mem_size - pcmc->acpi_data_size - 1;
1255                 if (initrd_size >= initrd_max) {
1256                     fprintf(stderr, "qemu: initrd is too large, cannot support."
1257                             "(max: %"PRIu32", need %"PRId64")\n",
1258                             initrd_max, (uint64_t)initrd_size);
1259                     exit(1);
1260                 }
1261 
1262                 initrd_addr = (initrd_max - initrd_size) & ~4095;
1263 
1264                 fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_ADDR, initrd_addr);
1265                 fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_SIZE, initrd_size);
1266                 fw_cfg_add_bytes(fw_cfg, FW_CFG_INITRD_DATA, initrd_data,
1267                                  initrd_size);
1268             }
1269 
1270             option_rom[nb_option_roms].bootindex = 0;
1271             option_rom[nb_option_roms].name = "pvh.bin";
1272             nb_option_roms++;
1273 
1274             return;
1275         }
1276         protocol = 0;
1277     }
1278 
1279     if (protocol < 0x200 || !(header[0x211] & 0x01)) {
1280         /* Low kernel */
1281         real_addr    = 0x90000;
1282         cmdline_addr = 0x9a000 - cmdline_size;
1283         prot_addr    = 0x10000;
1284     } else if (protocol < 0x202) {
1285         /* High but ancient kernel */
1286         real_addr    = 0x90000;
1287         cmdline_addr = 0x9a000 - cmdline_size;
1288         prot_addr    = 0x100000;
1289     } else {
1290         /* High and recent kernel */
1291         real_addr    = 0x10000;
1292         cmdline_addr = 0x20000;
1293         prot_addr    = 0x100000;
1294     }
1295 
1296 #if 0
1297     fprintf(stderr,
1298             "qemu: real_addr     = 0x" TARGET_FMT_plx "\n"
1299             "qemu: cmdline_addr  = 0x" TARGET_FMT_plx "\n"
1300             "qemu: prot_addr     = 0x" TARGET_FMT_plx "\n",
1301             real_addr,
1302             cmdline_addr,
1303             prot_addr);
1304 #endif
1305 
1306     /* highest address for loading the initrd */
1307     if (protocol >= 0x20c &&
1308         lduw_p(header+0x236) & XLF_CAN_BE_LOADED_ABOVE_4G) {
1309         /*
1310          * Linux has supported initrd up to 4 GB for a very long time (2007,
1311          * long before XLF_CAN_BE_LOADED_ABOVE_4G which was added in 2013),
1312          * though it only sets initrd_max to 2 GB to "work around bootloader
1313          * bugs". Luckily, QEMU firmware(which does something like bootloader)
1314          * has supported this.
1315          *
1316          * It's believed that if XLF_CAN_BE_LOADED_ABOVE_4G is set, initrd can
1317          * be loaded into any address.
1318          *
1319          * In addition, initrd_max is uint32_t simply because QEMU doesn't
1320          * support the 64-bit boot protocol (specifically the ext_ramdisk_image
1321          * field).
1322          *
1323          * Therefore here just limit initrd_max to UINT32_MAX simply as well.
1324          */
1325         initrd_max = UINT32_MAX;
1326     } else if (protocol >= 0x203) {
1327         initrd_max = ldl_p(header+0x22c);
1328     } else {
1329         initrd_max = 0x37ffffff;
1330     }
1331 
1332     if (initrd_max >= pcms->below_4g_mem_size - pcmc->acpi_data_size) {
1333         initrd_max = pcms->below_4g_mem_size - pcmc->acpi_data_size - 1;
1334     }
1335 
1336     fw_cfg_add_i32(fw_cfg, FW_CFG_CMDLINE_ADDR, cmdline_addr);
1337     fw_cfg_add_i32(fw_cfg, FW_CFG_CMDLINE_SIZE, strlen(kernel_cmdline)+1);
1338     fw_cfg_add_string(fw_cfg, FW_CFG_CMDLINE_DATA, kernel_cmdline);
1339 
1340     if (protocol >= 0x202) {
1341         stl_p(header+0x228, cmdline_addr);
1342     } else {
1343         stw_p(header+0x20, 0xA33F);
1344         stw_p(header+0x22, cmdline_addr-real_addr);
1345     }
1346 
1347     /* handle vga= parameter */
1348     vmode = strstr(kernel_cmdline, "vga=");
1349     if (vmode) {
1350         unsigned int video_mode;
1351         /* skip "vga=" */
1352         vmode += 4;
1353         if (!strncmp(vmode, "normal", 6)) {
1354             video_mode = 0xffff;
1355         } else if (!strncmp(vmode, "ext", 3)) {
1356             video_mode = 0xfffe;
1357         } else if (!strncmp(vmode, "ask", 3)) {
1358             video_mode = 0xfffd;
1359         } else {
1360             video_mode = strtol(vmode, NULL, 0);
1361         }
1362         stw_p(header+0x1fa, video_mode);
1363     }
1364 
1365     /* loader type */
1366     /* High nybble = B reserved for QEMU; low nybble is revision number.
1367        If this code is substantially changed, you may want to consider
1368        incrementing the revision. */
1369     if (protocol >= 0x200) {
1370         header[0x210] = 0xB0;
1371     }
1372     /* heap */
1373     if (protocol >= 0x201) {
1374         header[0x211] |= 0x80;	/* CAN_USE_HEAP */
1375         stw_p(header+0x224, cmdline_addr-real_addr-0x200);
1376     }
1377 
1378     /* load initrd */
1379     if (initrd_filename) {
1380         gsize initrd_size;
1381         gchar *initrd_data;
1382         GError *gerr = NULL;
1383 
1384         if (protocol < 0x200) {
1385             fprintf(stderr, "qemu: linux kernel too old to load a ram disk\n");
1386             exit(1);
1387         }
1388 
1389         if (!g_file_get_contents(initrd_filename, &initrd_data,
1390                                  &initrd_size, &gerr)) {
1391             fprintf(stderr, "qemu: error reading initrd %s: %s\n",
1392                     initrd_filename, gerr->message);
1393             exit(1);
1394         }
1395         if (initrd_size >= initrd_max) {
1396             fprintf(stderr, "qemu: initrd is too large, cannot support."
1397                     "(max: %"PRIu32", need %"PRId64")\n",
1398                     initrd_max, (uint64_t)initrd_size);
1399             exit(1);
1400         }
1401 
1402         initrd_addr = (initrd_max-initrd_size) & ~4095;
1403 
1404         fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_ADDR, initrd_addr);
1405         fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_SIZE, initrd_size);
1406         fw_cfg_add_bytes(fw_cfg, FW_CFG_INITRD_DATA, initrd_data, initrd_size);
1407 
1408         stl_p(header+0x218, initrd_addr);
1409         stl_p(header+0x21c, initrd_size);
1410     }
1411 
1412     /* load kernel and setup */
1413     setup_size = header[0x1f1];
1414     if (setup_size == 0) {
1415         setup_size = 4;
1416     }
1417     setup_size = (setup_size+1)*512;
1418     if (setup_size > kernel_size) {
1419         fprintf(stderr, "qemu: invalid kernel header\n");
1420         exit(1);
1421     }
1422     kernel_size -= setup_size;
1423 
1424     setup  = g_malloc(setup_size);
1425     kernel = g_malloc(kernel_size);
1426     fseek(f, 0, SEEK_SET);
1427     if (fread(setup, 1, setup_size, f) != setup_size) {
1428         fprintf(stderr, "fread() failed\n");
1429         exit(1);
1430     }
1431     if (fread(kernel, 1, kernel_size, f) != kernel_size) {
1432         fprintf(stderr, "fread() failed\n");
1433         exit(1);
1434     }
1435     fclose(f);
1436 
1437     /* append dtb to kernel */
1438     if (dtb_filename) {
1439         if (protocol < 0x209) {
1440             fprintf(stderr, "qemu: Linux kernel too old to load a dtb\n");
1441             exit(1);
1442         }
1443 
1444         dtb_size = get_image_size(dtb_filename);
1445         if (dtb_size <= 0) {
1446             fprintf(stderr, "qemu: error reading dtb %s: %s\n",
1447                     dtb_filename, strerror(errno));
1448             exit(1);
1449         }
1450 
1451         setup_data_offset = QEMU_ALIGN_UP(kernel_size, 16);
1452         kernel_size = setup_data_offset + sizeof(struct setup_data) + dtb_size;
1453         kernel = g_realloc(kernel, kernel_size);
1454 
1455         stq_p(header+0x250, prot_addr + setup_data_offset);
1456 
1457         setup_data = (struct setup_data *)(kernel + setup_data_offset);
1458         setup_data->next = 0;
1459         setup_data->type = cpu_to_le32(SETUP_DTB);
1460         setup_data->len = cpu_to_le32(dtb_size);
1461 
1462         load_image_size(dtb_filename, setup_data->data, dtb_size);
1463     }
1464 
1465     memcpy(setup, header, MIN(sizeof(header), setup_size));
1466 
1467     fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_ADDR, prot_addr);
1468     fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_SIZE, kernel_size);
1469     fw_cfg_add_bytes(fw_cfg, FW_CFG_KERNEL_DATA, kernel, kernel_size);
1470 
1471     fw_cfg_add_i32(fw_cfg, FW_CFG_SETUP_ADDR, real_addr);
1472     fw_cfg_add_i32(fw_cfg, FW_CFG_SETUP_SIZE, setup_size);
1473     fw_cfg_add_bytes(fw_cfg, FW_CFG_SETUP_DATA, setup, setup_size);
1474 
1475     option_rom[nb_option_roms].bootindex = 0;
1476     option_rom[nb_option_roms].name = "linuxboot.bin";
1477     if (pcmc->linuxboot_dma_enabled && fw_cfg_dma_enabled(fw_cfg)) {
1478         option_rom[nb_option_roms].name = "linuxboot_dma.bin";
1479     }
1480     nb_option_roms++;
1481 }
1482 
1483 #define NE2000_NB_MAX 6
1484 
1485 static const int ne2000_io[NE2000_NB_MAX] = { 0x300, 0x320, 0x340, 0x360,
1486                                               0x280, 0x380 };
1487 static const int ne2000_irq[NE2000_NB_MAX] = { 9, 10, 11, 3, 4, 5 };
1488 
1489 void pc_init_ne2k_isa(ISABus *bus, NICInfo *nd)
1490 {
1491     static int nb_ne2k = 0;
1492 
1493     if (nb_ne2k == NE2000_NB_MAX)
1494         return;
1495     isa_ne2000_init(bus, ne2000_io[nb_ne2k],
1496                     ne2000_irq[nb_ne2k], nd);
1497     nb_ne2k++;
1498 }
1499 
1500 DeviceState *cpu_get_current_apic(void)
1501 {
1502     if (current_cpu) {
1503         X86CPU *cpu = X86_CPU(current_cpu);
1504         return cpu->apic_state;
1505     } else {
1506         return NULL;
1507     }
1508 }
1509 
1510 void pc_acpi_smi_interrupt(void *opaque, int irq, int level)
1511 {
1512     X86CPU *cpu = opaque;
1513 
1514     if (level) {
1515         cpu_interrupt(CPU(cpu), CPU_INTERRUPT_SMI);
1516     }
1517 }
1518 
1519 static void pc_new_cpu(PCMachineState *pcms, int64_t apic_id, Error **errp)
1520 {
1521     Object *cpu = NULL;
1522     Error *local_err = NULL;
1523     CPUX86State *env = NULL;
1524 
1525     cpu = object_new(MACHINE(pcms)->cpu_type);
1526 
1527     env = &X86_CPU(cpu)->env;
1528     env->nr_dies = pcms->smp_dies;
1529 
1530     object_property_set_uint(cpu, apic_id, "apic-id", &local_err);
1531     object_property_set_bool(cpu, true, "realized", &local_err);
1532 
1533     object_unref(cpu);
1534     error_propagate(errp, local_err);
1535 }
1536 
1537 /*
1538  * This function is very similar to smp_parse()
1539  * in hw/core/machine.c but includes CPU die support.
1540  */
1541 void pc_smp_parse(MachineState *ms, QemuOpts *opts)
1542 {
1543     PCMachineState *pcms = PC_MACHINE(ms);
1544 
1545     if (opts) {
1546         unsigned cpus    = qemu_opt_get_number(opts, "cpus", 0);
1547         unsigned sockets = qemu_opt_get_number(opts, "sockets", 0);
1548         unsigned dies = qemu_opt_get_number(opts, "dies", 1);
1549         unsigned cores   = qemu_opt_get_number(opts, "cores", 0);
1550         unsigned threads = qemu_opt_get_number(opts, "threads", 0);
1551 
1552         /* compute missing values, prefer sockets over cores over threads */
1553         if (cpus == 0 || sockets == 0) {
1554             cores = cores > 0 ? cores : 1;
1555             threads = threads > 0 ? threads : 1;
1556             if (cpus == 0) {
1557                 sockets = sockets > 0 ? sockets : 1;
1558                 cpus = cores * threads * dies * sockets;
1559             } else {
1560                 ms->smp.max_cpus =
1561                         qemu_opt_get_number(opts, "maxcpus", cpus);
1562                 sockets = ms->smp.max_cpus / (cores * threads * dies);
1563             }
1564         } else if (cores == 0) {
1565             threads = threads > 0 ? threads : 1;
1566             cores = cpus / (sockets * dies * threads);
1567             cores = cores > 0 ? cores : 1;
1568         } else if (threads == 0) {
1569             threads = cpus / (cores * dies * sockets);
1570             threads = threads > 0 ? threads : 1;
1571         } else if (sockets * dies * cores * threads < cpus) {
1572             error_report("cpu topology: "
1573                          "sockets (%u) * dies (%u) * cores (%u) * threads (%u) < "
1574                          "smp_cpus (%u)",
1575                          sockets, dies, cores, threads, cpus);
1576             exit(1);
1577         }
1578 
1579         ms->smp.max_cpus =
1580                 qemu_opt_get_number(opts, "maxcpus", cpus);
1581 
1582         if (ms->smp.max_cpus < cpus) {
1583             error_report("maxcpus must be equal to or greater than smp");
1584             exit(1);
1585         }
1586 
1587         if (sockets * dies * cores * threads > ms->smp.max_cpus) {
1588             error_report("cpu topology: "
1589                          "sockets (%u) * dies (%u) * cores (%u) * threads (%u) > "
1590                          "maxcpus (%u)",
1591                          sockets, dies, cores, threads,
1592                          ms->smp.max_cpus);
1593             exit(1);
1594         }
1595 
1596         if (sockets * dies * cores * threads != ms->smp.max_cpus) {
1597             warn_report("Invalid CPU topology deprecated: "
1598                         "sockets (%u) * dies (%u) * cores (%u) * threads (%u) "
1599                         "!= maxcpus (%u)",
1600                         sockets, dies, cores, threads,
1601                         ms->smp.max_cpus);
1602         }
1603 
1604         ms->smp.cpus = cpus;
1605         ms->smp.cores = cores;
1606         ms->smp.threads = threads;
1607         pcms->smp_dies = dies;
1608     }
1609 
1610     if (ms->smp.cpus > 1) {
1611         Error *blocker = NULL;
1612         error_setg(&blocker, QERR_REPLAY_NOT_SUPPORTED, "smp");
1613         replay_add_blocker(blocker);
1614     }
1615 }
1616 
1617 void pc_hot_add_cpu(MachineState *ms, const int64_t id, Error **errp)
1618 {
1619     PCMachineState *pcms = PC_MACHINE(ms);
1620     int64_t apic_id = x86_cpu_apic_id_from_index(pcms, id);
1621     Error *local_err = NULL;
1622 
1623     if (id < 0) {
1624         error_setg(errp, "Invalid CPU id: %" PRIi64, id);
1625         return;
1626     }
1627 
1628     if (apic_id >= ACPI_CPU_HOTPLUG_ID_LIMIT) {
1629         error_setg(errp, "Unable to add CPU: %" PRIi64
1630                    ", resulting APIC ID (%" PRIi64 ") is too large",
1631                    id, apic_id);
1632         return;
1633     }
1634 
1635     pc_new_cpu(PC_MACHINE(ms), apic_id, &local_err);
1636     if (local_err) {
1637         error_propagate(errp, local_err);
1638         return;
1639     }
1640 }
1641 
1642 void pc_cpus_init(PCMachineState *pcms)
1643 {
1644     int i;
1645     const CPUArchIdList *possible_cpus;
1646     MachineState *ms = MACHINE(pcms);
1647     MachineClass *mc = MACHINE_GET_CLASS(pcms);
1648     PCMachineClass *pcmc = PC_MACHINE_CLASS(mc);
1649 
1650     x86_cpu_set_default_version(pcmc->default_cpu_version);
1651 
1652     /* Calculates the limit to CPU APIC ID values
1653      *
1654      * Limit for the APIC ID value, so that all
1655      * CPU APIC IDs are < pcms->apic_id_limit.
1656      *
1657      * This is used for FW_CFG_MAX_CPUS. See comments on bochs_bios_init().
1658      */
1659     pcms->apic_id_limit = x86_cpu_apic_id_from_index(pcms,
1660                                                      ms->smp.max_cpus - 1) + 1;
1661     possible_cpus = mc->possible_cpu_arch_ids(ms);
1662     for (i = 0; i < ms->smp.cpus; i++) {
1663         pc_new_cpu(pcms, possible_cpus->cpus[i].arch_id, &error_fatal);
1664     }
1665 }
1666 
1667 static void pc_build_feature_control_file(PCMachineState *pcms)
1668 {
1669     MachineState *ms = MACHINE(pcms);
1670     X86CPU *cpu = X86_CPU(ms->possible_cpus->cpus[0].cpu);
1671     CPUX86State *env = &cpu->env;
1672     uint32_t unused, ecx, edx;
1673     uint64_t feature_control_bits = 0;
1674     uint64_t *val;
1675 
1676     cpu_x86_cpuid(env, 1, 0, &unused, &unused, &ecx, &edx);
1677     if (ecx & CPUID_EXT_VMX) {
1678         feature_control_bits |= FEATURE_CONTROL_VMXON_ENABLED_OUTSIDE_SMX;
1679     }
1680 
1681     if ((edx & (CPUID_EXT2_MCE | CPUID_EXT2_MCA)) ==
1682         (CPUID_EXT2_MCE | CPUID_EXT2_MCA) &&
1683         (env->mcg_cap & MCG_LMCE_P)) {
1684         feature_control_bits |= FEATURE_CONTROL_LMCE;
1685     }
1686 
1687     if (!feature_control_bits) {
1688         return;
1689     }
1690 
1691     val = g_malloc(sizeof(*val));
1692     *val = cpu_to_le64(feature_control_bits | FEATURE_CONTROL_LOCKED);
1693     fw_cfg_add_file(pcms->fw_cfg, "etc/msr_feature_control", val, sizeof(*val));
1694 }
1695 
1696 static void rtc_set_cpus_count(ISADevice *rtc, uint16_t cpus_count)
1697 {
1698     if (cpus_count > 0xff) {
1699         /* If the number of CPUs can't be represented in 8 bits, the
1700          * BIOS must use "FW_CFG_NB_CPUS". Set RTC field to 0 just
1701          * to make old BIOSes fail more predictably.
1702          */
1703         rtc_set_memory(rtc, 0x5f, 0);
1704     } else {
1705         rtc_set_memory(rtc, 0x5f, cpus_count - 1);
1706     }
1707 }
1708 
1709 static
1710 void pc_machine_done(Notifier *notifier, void *data)
1711 {
1712     PCMachineState *pcms = container_of(notifier,
1713                                         PCMachineState, machine_done);
1714     PCIBus *bus = pcms->bus;
1715 
1716     /* set the number of CPUs */
1717     rtc_set_cpus_count(pcms->rtc, pcms->boot_cpus);
1718 
1719     if (bus) {
1720         int extra_hosts = 0;
1721 
1722         QLIST_FOREACH(bus, &bus->child, sibling) {
1723             /* look for expander root buses */
1724             if (pci_bus_is_root(bus)) {
1725                 extra_hosts++;
1726             }
1727         }
1728         if (extra_hosts && pcms->fw_cfg) {
1729             uint64_t *val = g_malloc(sizeof(*val));
1730             *val = cpu_to_le64(extra_hosts);
1731             fw_cfg_add_file(pcms->fw_cfg,
1732                     "etc/extra-pci-roots", val, sizeof(*val));
1733         }
1734     }
1735 
1736     acpi_setup();
1737     if (pcms->fw_cfg) {
1738         pc_build_smbios(pcms);
1739         pc_build_feature_control_file(pcms);
1740         /* update FW_CFG_NB_CPUS to account for -device added CPUs */
1741         fw_cfg_modify_i16(pcms->fw_cfg, FW_CFG_NB_CPUS, pcms->boot_cpus);
1742     }
1743 
1744     if (pcms->apic_id_limit > 255 && !xen_enabled()) {
1745         IntelIOMMUState *iommu = INTEL_IOMMU_DEVICE(x86_iommu_get_default());
1746 
1747         if (!iommu || !x86_iommu_ir_supported(X86_IOMMU_DEVICE(iommu)) ||
1748             iommu->intr_eim != ON_OFF_AUTO_ON) {
1749             error_report("current -smp configuration requires "
1750                          "Extended Interrupt Mode enabled. "
1751                          "You can add an IOMMU using: "
1752                          "-device intel-iommu,intremap=on,eim=on");
1753             exit(EXIT_FAILURE);
1754         }
1755     }
1756 }
1757 
1758 void pc_guest_info_init(PCMachineState *pcms)
1759 {
1760     int i;
1761 
1762     pcms->apic_xrupt_override = kvm_allows_irq0_override();
1763     pcms->numa_nodes = nb_numa_nodes;
1764     pcms->node_mem = g_malloc0(pcms->numa_nodes *
1765                                     sizeof *pcms->node_mem);
1766     for (i = 0; i < nb_numa_nodes; i++) {
1767         pcms->node_mem[i] = numa_info[i].node_mem;
1768     }
1769 
1770     pcms->machine_done.notify = pc_machine_done;
1771     qemu_add_machine_init_done_notifier(&pcms->machine_done);
1772 }
1773 
1774 /* setup pci memory address space mapping into system address space */
1775 void pc_pci_as_mapping_init(Object *owner, MemoryRegion *system_memory,
1776                             MemoryRegion *pci_address_space)
1777 {
1778     /* Set to lower priority than RAM */
1779     memory_region_add_subregion_overlap(system_memory, 0x0,
1780                                         pci_address_space, -1);
1781 }
1782 
1783 void xen_load_linux(PCMachineState *pcms)
1784 {
1785     int i;
1786     FWCfgState *fw_cfg;
1787 
1788     assert(MACHINE(pcms)->kernel_filename != NULL);
1789 
1790     fw_cfg = fw_cfg_init_io(FW_CFG_IO_BASE);
1791     fw_cfg_add_i16(fw_cfg, FW_CFG_NB_CPUS, pcms->boot_cpus);
1792     rom_set_fw(fw_cfg);
1793 
1794     load_linux(pcms, fw_cfg);
1795     for (i = 0; i < nb_option_roms; i++) {
1796         assert(!strcmp(option_rom[i].name, "linuxboot.bin") ||
1797                !strcmp(option_rom[i].name, "linuxboot_dma.bin") ||
1798                !strcmp(option_rom[i].name, "pvh.bin") ||
1799                !strcmp(option_rom[i].name, "multiboot.bin"));
1800         rom_add_option(option_rom[i].name, option_rom[i].bootindex);
1801     }
1802     pcms->fw_cfg = fw_cfg;
1803 }
1804 
1805 void pc_memory_init(PCMachineState *pcms,
1806                     MemoryRegion *system_memory,
1807                     MemoryRegion *rom_memory,
1808                     MemoryRegion **ram_memory)
1809 {
1810     int linux_boot, i;
1811     MemoryRegion *ram, *option_rom_mr;
1812     MemoryRegion *ram_below_4g, *ram_above_4g;
1813     FWCfgState *fw_cfg;
1814     MachineState *machine = MACHINE(pcms);
1815     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
1816 
1817     assert(machine->ram_size == pcms->below_4g_mem_size +
1818                                 pcms->above_4g_mem_size);
1819 
1820     linux_boot = (machine->kernel_filename != NULL);
1821 
1822     /* Allocate RAM.  We allocate it as a single memory region and use
1823      * aliases to address portions of it, mostly for backwards compatibility
1824      * with older qemus that used qemu_ram_alloc().
1825      */
1826     ram = g_malloc(sizeof(*ram));
1827     memory_region_allocate_system_memory(ram, NULL, "pc.ram",
1828                                          machine->ram_size);
1829     *ram_memory = ram;
1830     ram_below_4g = g_malloc(sizeof(*ram_below_4g));
1831     memory_region_init_alias(ram_below_4g, NULL, "ram-below-4g", ram,
1832                              0, pcms->below_4g_mem_size);
1833     memory_region_add_subregion(system_memory, 0, ram_below_4g);
1834     e820_add_entry(0, pcms->below_4g_mem_size, E820_RAM);
1835     if (pcms->above_4g_mem_size > 0) {
1836         ram_above_4g = g_malloc(sizeof(*ram_above_4g));
1837         memory_region_init_alias(ram_above_4g, NULL, "ram-above-4g", ram,
1838                                  pcms->below_4g_mem_size,
1839                                  pcms->above_4g_mem_size);
1840         memory_region_add_subregion(system_memory, 0x100000000ULL,
1841                                     ram_above_4g);
1842         e820_add_entry(0x100000000ULL, pcms->above_4g_mem_size, E820_RAM);
1843     }
1844 
1845     if (!pcmc->has_reserved_memory &&
1846         (machine->ram_slots ||
1847          (machine->maxram_size > machine->ram_size))) {
1848         MachineClass *mc = MACHINE_GET_CLASS(machine);
1849 
1850         error_report("\"-memory 'slots|maxmem'\" is not supported by: %s",
1851                      mc->name);
1852         exit(EXIT_FAILURE);
1853     }
1854 
1855     /* always allocate the device memory information */
1856     machine->device_memory = g_malloc0(sizeof(*machine->device_memory));
1857 
1858     /* initialize device memory address space */
1859     if (pcmc->has_reserved_memory &&
1860         (machine->ram_size < machine->maxram_size)) {
1861         ram_addr_t device_mem_size = machine->maxram_size - machine->ram_size;
1862 
1863         if (machine->ram_slots > ACPI_MAX_RAM_SLOTS) {
1864             error_report("unsupported amount of memory slots: %"PRIu64,
1865                          machine->ram_slots);
1866             exit(EXIT_FAILURE);
1867         }
1868 
1869         if (QEMU_ALIGN_UP(machine->maxram_size,
1870                           TARGET_PAGE_SIZE) != machine->maxram_size) {
1871             error_report("maximum memory size must by aligned to multiple of "
1872                          "%d bytes", TARGET_PAGE_SIZE);
1873             exit(EXIT_FAILURE);
1874         }
1875 
1876         machine->device_memory->base =
1877             ROUND_UP(0x100000000ULL + pcms->above_4g_mem_size, 1 * GiB);
1878 
1879         if (pcmc->enforce_aligned_dimm) {
1880             /* size device region assuming 1G page max alignment per slot */
1881             device_mem_size += (1 * GiB) * machine->ram_slots;
1882         }
1883 
1884         if ((machine->device_memory->base + device_mem_size) <
1885             device_mem_size) {
1886             error_report("unsupported amount of maximum memory: " RAM_ADDR_FMT,
1887                          machine->maxram_size);
1888             exit(EXIT_FAILURE);
1889         }
1890 
1891         memory_region_init(&machine->device_memory->mr, OBJECT(pcms),
1892                            "device-memory", device_mem_size);
1893         memory_region_add_subregion(system_memory, machine->device_memory->base,
1894                                     &machine->device_memory->mr);
1895     }
1896 
1897     /* Initialize PC system firmware */
1898     pc_system_firmware_init(pcms, rom_memory);
1899 
1900     option_rom_mr = g_malloc(sizeof(*option_rom_mr));
1901     memory_region_init_ram(option_rom_mr, NULL, "pc.rom", PC_ROM_SIZE,
1902                            &error_fatal);
1903     if (pcmc->pci_enabled) {
1904         memory_region_set_readonly(option_rom_mr, true);
1905     }
1906     memory_region_add_subregion_overlap(rom_memory,
1907                                         PC_ROM_MIN_VGA,
1908                                         option_rom_mr,
1909                                         1);
1910 
1911     fw_cfg = bochs_bios_init(&address_space_memory, pcms);
1912 
1913     rom_set_fw(fw_cfg);
1914 
1915     if (pcmc->has_reserved_memory && machine->device_memory->base) {
1916         uint64_t *val = g_malloc(sizeof(*val));
1917         PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
1918         uint64_t res_mem_end = machine->device_memory->base;
1919 
1920         if (!pcmc->broken_reserved_end) {
1921             res_mem_end += memory_region_size(&machine->device_memory->mr);
1922         }
1923         *val = cpu_to_le64(ROUND_UP(res_mem_end, 1 * GiB));
1924         fw_cfg_add_file(fw_cfg, "etc/reserved-memory-end", val, sizeof(*val));
1925     }
1926 
1927     if (linux_boot) {
1928         load_linux(pcms, fw_cfg);
1929     }
1930 
1931     for (i = 0; i < nb_option_roms; i++) {
1932         rom_add_option(option_rom[i].name, option_rom[i].bootindex);
1933     }
1934     pcms->fw_cfg = fw_cfg;
1935 
1936     /* Init default IOAPIC address space */
1937     pcms->ioapic_as = &address_space_memory;
1938 }
1939 
1940 /*
1941  * The 64bit pci hole starts after "above 4G RAM" and
1942  * potentially the space reserved for memory hotplug.
1943  */
1944 uint64_t pc_pci_hole64_start(void)
1945 {
1946     PCMachineState *pcms = PC_MACHINE(qdev_get_machine());
1947     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
1948     MachineState *ms = MACHINE(pcms);
1949     uint64_t hole64_start = 0;
1950 
1951     if (pcmc->has_reserved_memory && ms->device_memory->base) {
1952         hole64_start = ms->device_memory->base;
1953         if (!pcmc->broken_reserved_end) {
1954             hole64_start += memory_region_size(&ms->device_memory->mr);
1955         }
1956     } else {
1957         hole64_start = 0x100000000ULL + pcms->above_4g_mem_size;
1958     }
1959 
1960     return ROUND_UP(hole64_start, 1 * GiB);
1961 }
1962 
1963 qemu_irq pc_allocate_cpu_irq(void)
1964 {
1965     return qemu_allocate_irq(pic_irq_request, NULL, 0);
1966 }
1967 
1968 DeviceState *pc_vga_init(ISABus *isa_bus, PCIBus *pci_bus)
1969 {
1970     DeviceState *dev = NULL;
1971 
1972     rom_set_order_override(FW_CFG_ORDER_OVERRIDE_VGA);
1973     if (pci_bus) {
1974         PCIDevice *pcidev = pci_vga_init(pci_bus);
1975         dev = pcidev ? &pcidev->qdev : NULL;
1976     } else if (isa_bus) {
1977         ISADevice *isadev = isa_vga_init(isa_bus);
1978         dev = isadev ? DEVICE(isadev) : NULL;
1979     }
1980     rom_reset_order_override();
1981     return dev;
1982 }
1983 
1984 static const MemoryRegionOps ioport80_io_ops = {
1985     .write = ioport80_write,
1986     .read = ioport80_read,
1987     .endianness = DEVICE_NATIVE_ENDIAN,
1988     .impl = {
1989         .min_access_size = 1,
1990         .max_access_size = 1,
1991     },
1992 };
1993 
1994 static const MemoryRegionOps ioportF0_io_ops = {
1995     .write = ioportF0_write,
1996     .read = ioportF0_read,
1997     .endianness = DEVICE_NATIVE_ENDIAN,
1998     .impl = {
1999         .min_access_size = 1,
2000         .max_access_size = 1,
2001     },
2002 };
2003 
2004 static void pc_superio_init(ISABus *isa_bus, bool create_fdctrl, bool no_vmport)
2005 {
2006     int i;
2007     DriveInfo *fd[MAX_FD];
2008     qemu_irq *a20_line;
2009     ISADevice *i8042, *port92, *vmmouse;
2010 
2011     serial_hds_isa_init(isa_bus, 0, MAX_ISA_SERIAL_PORTS);
2012     parallel_hds_isa_init(isa_bus, MAX_PARALLEL_PORTS);
2013 
2014     for (i = 0; i < MAX_FD; i++) {
2015         fd[i] = drive_get(IF_FLOPPY, 0, i);
2016         create_fdctrl |= !!fd[i];
2017     }
2018     if (create_fdctrl) {
2019         fdctrl_init_isa(isa_bus, fd);
2020     }
2021 
2022     i8042 = isa_create_simple(isa_bus, "i8042");
2023     if (!no_vmport) {
2024         vmport_init(isa_bus);
2025         vmmouse = isa_try_create(isa_bus, "vmmouse");
2026     } else {
2027         vmmouse = NULL;
2028     }
2029     if (vmmouse) {
2030         DeviceState *dev = DEVICE(vmmouse);
2031         qdev_prop_set_ptr(dev, "ps2_mouse", i8042);
2032         qdev_init_nofail(dev);
2033     }
2034     port92 = isa_create_simple(isa_bus, "port92");
2035 
2036     a20_line = qemu_allocate_irqs(handle_a20_line_change, first_cpu, 2);
2037     i8042_setup_a20_line(i8042, a20_line[0]);
2038     port92_init(port92, a20_line[1]);
2039     g_free(a20_line);
2040 }
2041 
2042 void pc_basic_device_init(ISABus *isa_bus, qemu_irq *gsi,
2043                           ISADevice **rtc_state,
2044                           bool create_fdctrl,
2045                           bool no_vmport,
2046                           bool has_pit,
2047                           uint32_t hpet_irqs)
2048 {
2049     int i;
2050     DeviceState *hpet = NULL;
2051     int pit_isa_irq = 0;
2052     qemu_irq pit_alt_irq = NULL;
2053     qemu_irq rtc_irq = NULL;
2054     ISADevice *pit = NULL;
2055     MemoryRegion *ioport80_io = g_new(MemoryRegion, 1);
2056     MemoryRegion *ioportF0_io = g_new(MemoryRegion, 1);
2057 
2058     memory_region_init_io(ioport80_io, NULL, &ioport80_io_ops, NULL, "ioport80", 1);
2059     memory_region_add_subregion(isa_bus->address_space_io, 0x80, ioport80_io);
2060 
2061     memory_region_init_io(ioportF0_io, NULL, &ioportF0_io_ops, NULL, "ioportF0", 1);
2062     memory_region_add_subregion(isa_bus->address_space_io, 0xf0, ioportF0_io);
2063 
2064     /*
2065      * Check if an HPET shall be created.
2066      *
2067      * Without KVM_CAP_PIT_STATE2, we cannot switch off the in-kernel PIT
2068      * when the HPET wants to take over. Thus we have to disable the latter.
2069      */
2070     if (!no_hpet && (!kvm_irqchip_in_kernel() || kvm_has_pit_state2())) {
2071         /* In order to set property, here not using sysbus_try_create_simple */
2072         hpet = qdev_try_create(NULL, TYPE_HPET);
2073         if (hpet) {
2074             /* For pc-piix-*, hpet's intcap is always IRQ2. For pc-q35-1.7
2075              * and earlier, use IRQ2 for compat. Otherwise, use IRQ16~23,
2076              * IRQ8 and IRQ2.
2077              */
2078             uint8_t compat = object_property_get_uint(OBJECT(hpet),
2079                     HPET_INTCAP, NULL);
2080             if (!compat) {
2081                 qdev_prop_set_uint32(hpet, HPET_INTCAP, hpet_irqs);
2082             }
2083             qdev_init_nofail(hpet);
2084             sysbus_mmio_map(SYS_BUS_DEVICE(hpet), 0, HPET_BASE);
2085 
2086             for (i = 0; i < GSI_NUM_PINS; i++) {
2087                 sysbus_connect_irq(SYS_BUS_DEVICE(hpet), i, gsi[i]);
2088             }
2089             pit_isa_irq = -1;
2090             pit_alt_irq = qdev_get_gpio_in(hpet, HPET_LEGACY_PIT_INT);
2091             rtc_irq = qdev_get_gpio_in(hpet, HPET_LEGACY_RTC_INT);
2092         }
2093     }
2094     *rtc_state = mc146818_rtc_init(isa_bus, 2000, rtc_irq);
2095 
2096     qemu_register_boot_set(pc_boot_set, *rtc_state);
2097 
2098     if (!xen_enabled() && has_pit) {
2099         if (kvm_pit_in_kernel()) {
2100             pit = kvm_pit_init(isa_bus, 0x40);
2101         } else {
2102             pit = i8254_pit_init(isa_bus, 0x40, pit_isa_irq, pit_alt_irq);
2103         }
2104         if (hpet) {
2105             /* connect PIT to output control line of the HPET */
2106             qdev_connect_gpio_out(hpet, 0, qdev_get_gpio_in(DEVICE(pit), 0));
2107         }
2108         pcspk_init(isa_bus, pit);
2109     }
2110 
2111     i8257_dma_init(isa_bus, 0);
2112 
2113     /* Super I/O */
2114     pc_superio_init(isa_bus, create_fdctrl, no_vmport);
2115 }
2116 
2117 void pc_nic_init(PCMachineClass *pcmc, ISABus *isa_bus, PCIBus *pci_bus)
2118 {
2119     int i;
2120 
2121     rom_set_order_override(FW_CFG_ORDER_OVERRIDE_NIC);
2122     for (i = 0; i < nb_nics; i++) {
2123         NICInfo *nd = &nd_table[i];
2124         const char *model = nd->model ? nd->model : pcmc->default_nic_model;
2125 
2126         if (g_str_equal(model, "ne2k_isa")) {
2127             pc_init_ne2k_isa(isa_bus, nd);
2128         } else {
2129             pci_nic_init_nofail(nd, pci_bus, model, NULL);
2130         }
2131     }
2132     rom_reset_order_override();
2133 }
2134 
2135 void ioapic_init_gsi(GSIState *gsi_state, const char *parent_name)
2136 {
2137     DeviceState *dev;
2138     SysBusDevice *d;
2139     unsigned int i;
2140 
2141     if (kvm_ioapic_in_kernel()) {
2142         dev = qdev_create(NULL, TYPE_KVM_IOAPIC);
2143     } else {
2144         dev = qdev_create(NULL, TYPE_IOAPIC);
2145     }
2146     if (parent_name) {
2147         object_property_add_child(object_resolve_path(parent_name, NULL),
2148                                   "ioapic", OBJECT(dev), NULL);
2149     }
2150     qdev_init_nofail(dev);
2151     d = SYS_BUS_DEVICE(dev);
2152     sysbus_mmio_map(d, 0, IO_APIC_DEFAULT_ADDRESS);
2153 
2154     for (i = 0; i < IOAPIC_NUM_PINS; i++) {
2155         gsi_state->ioapic_irq[i] = qdev_get_gpio_in(dev, i);
2156     }
2157 }
2158 
2159 static void pc_memory_pre_plug(HotplugHandler *hotplug_dev, DeviceState *dev,
2160                                Error **errp)
2161 {
2162     const PCMachineState *pcms = PC_MACHINE(hotplug_dev);
2163     const PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
2164     const MachineState *ms = MACHINE(hotplug_dev);
2165     const bool is_nvdimm = object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM);
2166     const uint64_t legacy_align = TARGET_PAGE_SIZE;
2167     Error *local_err = NULL;
2168 
2169     /*
2170      * When -no-acpi is used with Q35 machine type, no ACPI is built,
2171      * but pcms->acpi_dev is still created. Check !acpi_enabled in
2172      * addition to cover this case.
2173      */
2174     if (!pcms->acpi_dev || !acpi_enabled) {
2175         error_setg(errp,
2176                    "memory hotplug is not enabled: missing acpi device or acpi disabled");
2177         return;
2178     }
2179 
2180     if (is_nvdimm && !ms->nvdimms_state->is_enabled) {
2181         error_setg(errp, "nvdimm is not enabled: missing 'nvdimm' in '-M'");
2182         return;
2183     }
2184 
2185     hotplug_handler_pre_plug(pcms->acpi_dev, dev, &local_err);
2186     if (local_err) {
2187         error_propagate(errp, local_err);
2188         return;
2189     }
2190 
2191     pc_dimm_pre_plug(PC_DIMM(dev), MACHINE(hotplug_dev),
2192                      pcmc->enforce_aligned_dimm ? NULL : &legacy_align, errp);
2193 }
2194 
2195 static void pc_memory_plug(HotplugHandler *hotplug_dev,
2196                            DeviceState *dev, Error **errp)
2197 {
2198     Error *local_err = NULL;
2199     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
2200     MachineState *ms = MACHINE(hotplug_dev);
2201     bool is_nvdimm = object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM);
2202 
2203     pc_dimm_plug(PC_DIMM(dev), MACHINE(pcms), &local_err);
2204     if (local_err) {
2205         goto out;
2206     }
2207 
2208     if (is_nvdimm) {
2209         nvdimm_plug(ms->nvdimms_state);
2210     }
2211 
2212     hotplug_handler_plug(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &error_abort);
2213 out:
2214     error_propagate(errp, local_err);
2215 }
2216 
2217 static void pc_memory_unplug_request(HotplugHandler *hotplug_dev,
2218                                      DeviceState *dev, Error **errp)
2219 {
2220     Error *local_err = NULL;
2221     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
2222 
2223     /*
2224      * When -no-acpi is used with Q35 machine type, no ACPI is built,
2225      * but pcms->acpi_dev is still created. Check !acpi_enabled in
2226      * addition to cover this case.
2227      */
2228     if (!pcms->acpi_dev || !acpi_enabled) {
2229         error_setg(&local_err,
2230                    "memory hotplug is not enabled: missing acpi device or acpi disabled");
2231         goto out;
2232     }
2233 
2234     if (object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM)) {
2235         error_setg(&local_err,
2236                    "nvdimm device hot unplug is not supported yet.");
2237         goto out;
2238     }
2239 
2240     hotplug_handler_unplug_request(HOTPLUG_HANDLER(pcms->acpi_dev), dev,
2241                                    &local_err);
2242 out:
2243     error_propagate(errp, local_err);
2244 }
2245 
2246 static void pc_memory_unplug(HotplugHandler *hotplug_dev,
2247                              DeviceState *dev, Error **errp)
2248 {
2249     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
2250     Error *local_err = NULL;
2251 
2252     hotplug_handler_unplug(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &local_err);
2253     if (local_err) {
2254         goto out;
2255     }
2256 
2257     pc_dimm_unplug(PC_DIMM(dev), MACHINE(pcms));
2258     object_property_set_bool(OBJECT(dev), false, "realized", NULL);
2259  out:
2260     error_propagate(errp, local_err);
2261 }
2262 
2263 static int pc_apic_cmp(const void *a, const void *b)
2264 {
2265    CPUArchId *apic_a = (CPUArchId *)a;
2266    CPUArchId *apic_b = (CPUArchId *)b;
2267 
2268    return apic_a->arch_id - apic_b->arch_id;
2269 }
2270 
2271 /* returns pointer to CPUArchId descriptor that matches CPU's apic_id
2272  * in ms->possible_cpus->cpus, if ms->possible_cpus->cpus has no
2273  * entry corresponding to CPU's apic_id returns NULL.
2274  */
2275 static CPUArchId *pc_find_cpu_slot(MachineState *ms, uint32_t id, int *idx)
2276 {
2277     CPUArchId apic_id, *found_cpu;
2278 
2279     apic_id.arch_id = id;
2280     found_cpu = bsearch(&apic_id, ms->possible_cpus->cpus,
2281         ms->possible_cpus->len, sizeof(*ms->possible_cpus->cpus),
2282         pc_apic_cmp);
2283     if (found_cpu && idx) {
2284         *idx = found_cpu - ms->possible_cpus->cpus;
2285     }
2286     return found_cpu;
2287 }
2288 
2289 static void pc_cpu_plug(HotplugHandler *hotplug_dev,
2290                         DeviceState *dev, Error **errp)
2291 {
2292     CPUArchId *found_cpu;
2293     Error *local_err = NULL;
2294     X86CPU *cpu = X86_CPU(dev);
2295     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
2296 
2297     if (pcms->acpi_dev) {
2298         hotplug_handler_plug(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &local_err);
2299         if (local_err) {
2300             goto out;
2301         }
2302     }
2303 
2304     /* increment the number of CPUs */
2305     pcms->boot_cpus++;
2306     if (pcms->rtc) {
2307         rtc_set_cpus_count(pcms->rtc, pcms->boot_cpus);
2308     }
2309     if (pcms->fw_cfg) {
2310         fw_cfg_modify_i16(pcms->fw_cfg, FW_CFG_NB_CPUS, pcms->boot_cpus);
2311     }
2312 
2313     found_cpu = pc_find_cpu_slot(MACHINE(pcms), cpu->apic_id, NULL);
2314     found_cpu->cpu = OBJECT(dev);
2315 out:
2316     error_propagate(errp, local_err);
2317 }
2318 static void pc_cpu_unplug_request_cb(HotplugHandler *hotplug_dev,
2319                                      DeviceState *dev, Error **errp)
2320 {
2321     int idx = -1;
2322     Error *local_err = NULL;
2323     X86CPU *cpu = X86_CPU(dev);
2324     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
2325 
2326     if (!pcms->acpi_dev) {
2327         error_setg(&local_err, "CPU hot unplug not supported without ACPI");
2328         goto out;
2329     }
2330 
2331     pc_find_cpu_slot(MACHINE(pcms), cpu->apic_id, &idx);
2332     assert(idx != -1);
2333     if (idx == 0) {
2334         error_setg(&local_err, "Boot CPU is unpluggable");
2335         goto out;
2336     }
2337 
2338     hotplug_handler_unplug_request(HOTPLUG_HANDLER(pcms->acpi_dev), dev,
2339                                    &local_err);
2340     if (local_err) {
2341         goto out;
2342     }
2343 
2344  out:
2345     error_propagate(errp, local_err);
2346 
2347 }
2348 
2349 static void pc_cpu_unplug_cb(HotplugHandler *hotplug_dev,
2350                              DeviceState *dev, Error **errp)
2351 {
2352     CPUArchId *found_cpu;
2353     Error *local_err = NULL;
2354     X86CPU *cpu = X86_CPU(dev);
2355     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
2356 
2357     hotplug_handler_unplug(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &local_err);
2358     if (local_err) {
2359         goto out;
2360     }
2361 
2362     found_cpu = pc_find_cpu_slot(MACHINE(pcms), cpu->apic_id, NULL);
2363     found_cpu->cpu = NULL;
2364     object_property_set_bool(OBJECT(dev), false, "realized", NULL);
2365 
2366     /* decrement the number of CPUs */
2367     pcms->boot_cpus--;
2368     /* Update the number of CPUs in CMOS */
2369     rtc_set_cpus_count(pcms->rtc, pcms->boot_cpus);
2370     fw_cfg_modify_i16(pcms->fw_cfg, FW_CFG_NB_CPUS, pcms->boot_cpus);
2371  out:
2372     error_propagate(errp, local_err);
2373 }
2374 
2375 static void pc_cpu_pre_plug(HotplugHandler *hotplug_dev,
2376                             DeviceState *dev, Error **errp)
2377 {
2378     int idx;
2379     CPUState *cs;
2380     CPUArchId *cpu_slot;
2381     X86CPUTopoInfo topo;
2382     X86CPU *cpu = X86_CPU(dev);
2383     CPUX86State *env = &cpu->env;
2384     MachineState *ms = MACHINE(hotplug_dev);
2385     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
2386     unsigned int smp_cores = ms->smp.cores;
2387     unsigned int smp_threads = ms->smp.threads;
2388 
2389     if(!object_dynamic_cast(OBJECT(cpu), ms->cpu_type)) {
2390         error_setg(errp, "Invalid CPU type, expected cpu type: '%s'",
2391                    ms->cpu_type);
2392         return;
2393     }
2394 
2395     env->nr_dies = pcms->smp_dies;
2396 
2397     /*
2398      * If APIC ID is not set,
2399      * set it based on socket/die/core/thread properties.
2400      */
2401     if (cpu->apic_id == UNASSIGNED_APIC_ID) {
2402         int max_socket = (ms->smp.max_cpus - 1) /
2403                                 smp_threads / smp_cores / pcms->smp_dies;
2404 
2405         if (cpu->socket_id < 0) {
2406             error_setg(errp, "CPU socket-id is not set");
2407             return;
2408         } else if (cpu->socket_id > max_socket) {
2409             error_setg(errp, "Invalid CPU socket-id: %u must be in range 0:%u",
2410                        cpu->socket_id, max_socket);
2411             return;
2412         } else if (cpu->die_id > pcms->smp_dies - 1) {
2413             error_setg(errp, "Invalid CPU die-id: %u must be in range 0:%u",
2414                        cpu->die_id, max_socket);
2415             return;
2416         }
2417         if (cpu->core_id < 0) {
2418             error_setg(errp, "CPU core-id is not set");
2419             return;
2420         } else if (cpu->core_id > (smp_cores - 1)) {
2421             error_setg(errp, "Invalid CPU core-id: %u must be in range 0:%u",
2422                        cpu->core_id, smp_cores - 1);
2423             return;
2424         }
2425         if (cpu->thread_id < 0) {
2426             error_setg(errp, "CPU thread-id is not set");
2427             return;
2428         } else if (cpu->thread_id > (smp_threads - 1)) {
2429             error_setg(errp, "Invalid CPU thread-id: %u must be in range 0:%u",
2430                        cpu->thread_id, smp_threads - 1);
2431             return;
2432         }
2433 
2434         topo.pkg_id = cpu->socket_id;
2435         topo.die_id = cpu->die_id;
2436         topo.core_id = cpu->core_id;
2437         topo.smt_id = cpu->thread_id;
2438         cpu->apic_id = apicid_from_topo_ids(pcms->smp_dies, smp_cores,
2439                                             smp_threads, &topo);
2440     }
2441 
2442     cpu_slot = pc_find_cpu_slot(MACHINE(pcms), cpu->apic_id, &idx);
2443     if (!cpu_slot) {
2444         MachineState *ms = MACHINE(pcms);
2445 
2446         x86_topo_ids_from_apicid(cpu->apic_id, pcms->smp_dies,
2447                                  smp_cores, smp_threads, &topo);
2448         error_setg(errp,
2449             "Invalid CPU [socket: %u, die: %u, core: %u, thread: %u] with"
2450             " APIC ID %" PRIu32 ", valid index range 0:%d",
2451             topo.pkg_id, topo.die_id, topo.core_id, topo.smt_id,
2452             cpu->apic_id, ms->possible_cpus->len - 1);
2453         return;
2454     }
2455 
2456     if (cpu_slot->cpu) {
2457         error_setg(errp, "CPU[%d] with APIC ID %" PRIu32 " exists",
2458                    idx, cpu->apic_id);
2459         return;
2460     }
2461 
2462     /* if 'address' properties socket-id/core-id/thread-id are not set, set them
2463      * so that machine_query_hotpluggable_cpus would show correct values
2464      */
2465     /* TODO: move socket_id/core_id/thread_id checks into x86_cpu_realizefn()
2466      * once -smp refactoring is complete and there will be CPU private
2467      * CPUState::nr_cores and CPUState::nr_threads fields instead of globals */
2468     x86_topo_ids_from_apicid(cpu->apic_id, pcms->smp_dies,
2469                              smp_cores, smp_threads, &topo);
2470     if (cpu->socket_id != -1 && cpu->socket_id != topo.pkg_id) {
2471         error_setg(errp, "property socket-id: %u doesn't match set apic-id:"
2472             " 0x%x (socket-id: %u)", cpu->socket_id, cpu->apic_id, topo.pkg_id);
2473         return;
2474     }
2475     cpu->socket_id = topo.pkg_id;
2476 
2477     if (cpu->die_id != -1 && cpu->die_id != topo.die_id) {
2478         error_setg(errp, "property die-id: %u doesn't match set apic-id:"
2479             " 0x%x (die-id: %u)", cpu->die_id, cpu->apic_id, topo.die_id);
2480         return;
2481     }
2482     cpu->die_id = topo.die_id;
2483 
2484     if (cpu->core_id != -1 && cpu->core_id != topo.core_id) {
2485         error_setg(errp, "property core-id: %u doesn't match set apic-id:"
2486             " 0x%x (core-id: %u)", cpu->core_id, cpu->apic_id, topo.core_id);
2487         return;
2488     }
2489     cpu->core_id = topo.core_id;
2490 
2491     if (cpu->thread_id != -1 && cpu->thread_id != topo.smt_id) {
2492         error_setg(errp, "property thread-id: %u doesn't match set apic-id:"
2493             " 0x%x (thread-id: %u)", cpu->thread_id, cpu->apic_id, topo.smt_id);
2494         return;
2495     }
2496     cpu->thread_id = topo.smt_id;
2497 
2498     if (hyperv_feat_enabled(cpu, HYPERV_FEAT_VPINDEX) &&
2499         !kvm_hv_vpindex_settable()) {
2500         error_setg(errp, "kernel doesn't allow setting HyperV VP_INDEX");
2501         return;
2502     }
2503 
2504     cs = CPU(cpu);
2505     cs->cpu_index = idx;
2506 
2507     numa_cpu_pre_plug(cpu_slot, dev, errp);
2508 }
2509 
2510 static void pc_virtio_pmem_pci_pre_plug(HotplugHandler *hotplug_dev,
2511                                         DeviceState *dev, Error **errp)
2512 {
2513     HotplugHandler *hotplug_dev2 = qdev_get_bus_hotplug_handler(dev);
2514     Error *local_err = NULL;
2515 
2516     if (!hotplug_dev2) {
2517         /*
2518          * Without a bus hotplug handler, we cannot control the plug/unplug
2519          * order. This should never be the case on x86, however better add
2520          * a safety net.
2521          */
2522         error_setg(errp, "virtio-pmem-pci not supported on this bus.");
2523         return;
2524     }
2525     /*
2526      * First, see if we can plug this memory device at all. If that
2527      * succeeds, branch of to the actual hotplug handler.
2528      */
2529     memory_device_pre_plug(MEMORY_DEVICE(dev), MACHINE(hotplug_dev), NULL,
2530                            &local_err);
2531     if (!local_err) {
2532         hotplug_handler_pre_plug(hotplug_dev2, dev, &local_err);
2533     }
2534     error_propagate(errp, local_err);
2535 }
2536 
2537 static void pc_virtio_pmem_pci_plug(HotplugHandler *hotplug_dev,
2538                                     DeviceState *dev, Error **errp)
2539 {
2540     HotplugHandler *hotplug_dev2 = qdev_get_bus_hotplug_handler(dev);
2541     Error *local_err = NULL;
2542 
2543     /*
2544      * Plug the memory device first and then branch off to the actual
2545      * hotplug handler. If that one fails, we can easily undo the memory
2546      * device bits.
2547      */
2548     memory_device_plug(MEMORY_DEVICE(dev), MACHINE(hotplug_dev));
2549     hotplug_handler_plug(hotplug_dev2, dev, &local_err);
2550     if (local_err) {
2551         memory_device_unplug(MEMORY_DEVICE(dev), MACHINE(hotplug_dev));
2552     }
2553     error_propagate(errp, local_err);
2554 }
2555 
2556 static void pc_virtio_pmem_pci_unplug_request(HotplugHandler *hotplug_dev,
2557                                               DeviceState *dev, Error **errp)
2558 {
2559     /* We don't support virtio pmem hot unplug */
2560     error_setg(errp, "virtio pmem device unplug not supported.");
2561 }
2562 
2563 static void pc_virtio_pmem_pci_unplug(HotplugHandler *hotplug_dev,
2564                                       DeviceState *dev, Error **errp)
2565 {
2566     /* We don't support virtio pmem hot unplug */
2567 }
2568 
2569 static void pc_machine_device_pre_plug_cb(HotplugHandler *hotplug_dev,
2570                                           DeviceState *dev, Error **errp)
2571 {
2572     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2573         pc_memory_pre_plug(hotplug_dev, dev, errp);
2574     } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
2575         pc_cpu_pre_plug(hotplug_dev, dev, errp);
2576     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_PMEM_PCI)) {
2577         pc_virtio_pmem_pci_pre_plug(hotplug_dev, dev, errp);
2578     }
2579 }
2580 
2581 static void pc_machine_device_plug_cb(HotplugHandler *hotplug_dev,
2582                                       DeviceState *dev, Error **errp)
2583 {
2584     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2585         pc_memory_plug(hotplug_dev, dev, errp);
2586     } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
2587         pc_cpu_plug(hotplug_dev, dev, errp);
2588     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_PMEM_PCI)) {
2589         pc_virtio_pmem_pci_plug(hotplug_dev, dev, errp);
2590     }
2591 }
2592 
2593 static void pc_machine_device_unplug_request_cb(HotplugHandler *hotplug_dev,
2594                                                 DeviceState *dev, Error **errp)
2595 {
2596     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2597         pc_memory_unplug_request(hotplug_dev, dev, errp);
2598     } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
2599         pc_cpu_unplug_request_cb(hotplug_dev, dev, errp);
2600     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_PMEM_PCI)) {
2601         pc_virtio_pmem_pci_unplug_request(hotplug_dev, dev, errp);
2602     } else {
2603         error_setg(errp, "acpi: device unplug request for not supported device"
2604                    " type: %s", object_get_typename(OBJECT(dev)));
2605     }
2606 }
2607 
2608 static void pc_machine_device_unplug_cb(HotplugHandler *hotplug_dev,
2609                                         DeviceState *dev, Error **errp)
2610 {
2611     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2612         pc_memory_unplug(hotplug_dev, dev, errp);
2613     } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
2614         pc_cpu_unplug_cb(hotplug_dev, dev, errp);
2615     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_PMEM_PCI)) {
2616         pc_virtio_pmem_pci_unplug(hotplug_dev, dev, errp);
2617     } else {
2618         error_setg(errp, "acpi: device unplug for not supported device"
2619                    " type: %s", object_get_typename(OBJECT(dev)));
2620     }
2621 }
2622 
2623 static HotplugHandler *pc_get_hotplug_handler(MachineState *machine,
2624                                              DeviceState *dev)
2625 {
2626     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM) ||
2627         object_dynamic_cast(OBJECT(dev), TYPE_CPU) ||
2628         object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_PMEM_PCI)) {
2629         return HOTPLUG_HANDLER(machine);
2630     }
2631 
2632     return NULL;
2633 }
2634 
2635 static void
2636 pc_machine_get_device_memory_region_size(Object *obj, Visitor *v,
2637                                          const char *name, void *opaque,
2638                                          Error **errp)
2639 {
2640     MachineState *ms = MACHINE(obj);
2641     int64_t value = 0;
2642 
2643     if (ms->device_memory) {
2644         value = memory_region_size(&ms->device_memory->mr);
2645     }
2646 
2647     visit_type_int(v, name, &value, errp);
2648 }
2649 
2650 static void pc_machine_get_max_ram_below_4g(Object *obj, Visitor *v,
2651                                             const char *name, void *opaque,
2652                                             Error **errp)
2653 {
2654     PCMachineState *pcms = PC_MACHINE(obj);
2655     uint64_t value = pcms->max_ram_below_4g;
2656 
2657     visit_type_size(v, name, &value, errp);
2658 }
2659 
2660 static void pc_machine_set_max_ram_below_4g(Object *obj, Visitor *v,
2661                                             const char *name, void *opaque,
2662                                             Error **errp)
2663 {
2664     PCMachineState *pcms = PC_MACHINE(obj);
2665     Error *error = NULL;
2666     uint64_t value;
2667 
2668     visit_type_size(v, name, &value, &error);
2669     if (error) {
2670         error_propagate(errp, error);
2671         return;
2672     }
2673     if (value > 4 * GiB) {
2674         error_setg(&error,
2675                    "Machine option 'max-ram-below-4g=%"PRIu64
2676                    "' expects size less than or equal to 4G", value);
2677         error_propagate(errp, error);
2678         return;
2679     }
2680 
2681     if (value < 1 * MiB) {
2682         warn_report("Only %" PRIu64 " bytes of RAM below the 4GiB boundary,"
2683                     "BIOS may not work with less than 1MiB", value);
2684     }
2685 
2686     pcms->max_ram_below_4g = value;
2687 }
2688 
2689 static void pc_machine_get_vmport(Object *obj, Visitor *v, const char *name,
2690                                   void *opaque, Error **errp)
2691 {
2692     PCMachineState *pcms = PC_MACHINE(obj);
2693     OnOffAuto vmport = pcms->vmport;
2694 
2695     visit_type_OnOffAuto(v, name, &vmport, errp);
2696 }
2697 
2698 static void pc_machine_set_vmport(Object *obj, Visitor *v, const char *name,
2699                                   void *opaque, Error **errp)
2700 {
2701     PCMachineState *pcms = PC_MACHINE(obj);
2702 
2703     visit_type_OnOffAuto(v, name, &pcms->vmport, errp);
2704 }
2705 
2706 bool pc_machine_is_smm_enabled(PCMachineState *pcms)
2707 {
2708     bool smm_available = false;
2709 
2710     if (pcms->smm == ON_OFF_AUTO_OFF) {
2711         return false;
2712     }
2713 
2714     if (tcg_enabled() || qtest_enabled()) {
2715         smm_available = true;
2716     } else if (kvm_enabled()) {
2717         smm_available = kvm_has_smm();
2718     }
2719 
2720     if (smm_available) {
2721         return true;
2722     }
2723 
2724     if (pcms->smm == ON_OFF_AUTO_ON) {
2725         error_report("System Management Mode not supported by this hypervisor.");
2726         exit(1);
2727     }
2728     return false;
2729 }
2730 
2731 static void pc_machine_get_smm(Object *obj, Visitor *v, const char *name,
2732                                void *opaque, Error **errp)
2733 {
2734     PCMachineState *pcms = PC_MACHINE(obj);
2735     OnOffAuto smm = pcms->smm;
2736 
2737     visit_type_OnOffAuto(v, name, &smm, errp);
2738 }
2739 
2740 static void pc_machine_set_smm(Object *obj, Visitor *v, const char *name,
2741                                void *opaque, Error **errp)
2742 {
2743     PCMachineState *pcms = PC_MACHINE(obj);
2744 
2745     visit_type_OnOffAuto(v, name, &pcms->smm, errp);
2746 }
2747 
2748 static bool pc_machine_get_smbus(Object *obj, Error **errp)
2749 {
2750     PCMachineState *pcms = PC_MACHINE(obj);
2751 
2752     return pcms->smbus_enabled;
2753 }
2754 
2755 static void pc_machine_set_smbus(Object *obj, bool value, Error **errp)
2756 {
2757     PCMachineState *pcms = PC_MACHINE(obj);
2758 
2759     pcms->smbus_enabled = value;
2760 }
2761 
2762 static bool pc_machine_get_sata(Object *obj, Error **errp)
2763 {
2764     PCMachineState *pcms = PC_MACHINE(obj);
2765 
2766     return pcms->sata_enabled;
2767 }
2768 
2769 static void pc_machine_set_sata(Object *obj, bool value, Error **errp)
2770 {
2771     PCMachineState *pcms = PC_MACHINE(obj);
2772 
2773     pcms->sata_enabled = value;
2774 }
2775 
2776 static bool pc_machine_get_pit(Object *obj, Error **errp)
2777 {
2778     PCMachineState *pcms = PC_MACHINE(obj);
2779 
2780     return pcms->pit_enabled;
2781 }
2782 
2783 static void pc_machine_set_pit(Object *obj, bool value, Error **errp)
2784 {
2785     PCMachineState *pcms = PC_MACHINE(obj);
2786 
2787     pcms->pit_enabled = value;
2788 }
2789 
2790 static void pc_machine_initfn(Object *obj)
2791 {
2792     PCMachineState *pcms = PC_MACHINE(obj);
2793 
2794     pcms->max_ram_below_4g = 0; /* use default */
2795     pcms->smm = ON_OFF_AUTO_AUTO;
2796     pcms->vmport = ON_OFF_AUTO_AUTO;
2797     /* acpi build is enabled by default if machine supports it */
2798     pcms->acpi_build_enabled = PC_MACHINE_GET_CLASS(pcms)->has_acpi_build;
2799     pcms->smbus_enabled = true;
2800     pcms->sata_enabled = true;
2801     pcms->pit_enabled = true;
2802     pcms->smp_dies = 1;
2803 
2804     pc_system_flash_create(pcms);
2805 }
2806 
2807 static void pc_machine_reset(MachineState *machine)
2808 {
2809     CPUState *cs;
2810     X86CPU *cpu;
2811 
2812     qemu_devices_reset();
2813 
2814     /* Reset APIC after devices have been reset to cancel
2815      * any changes that qemu_devices_reset() might have done.
2816      */
2817     CPU_FOREACH(cs) {
2818         cpu = X86_CPU(cs);
2819 
2820         if (cpu->apic_state) {
2821             device_reset(cpu->apic_state);
2822         }
2823     }
2824 }
2825 
2826 static CpuInstanceProperties
2827 pc_cpu_index_to_props(MachineState *ms, unsigned cpu_index)
2828 {
2829     MachineClass *mc = MACHINE_GET_CLASS(ms);
2830     const CPUArchIdList *possible_cpus = mc->possible_cpu_arch_ids(ms);
2831 
2832     assert(cpu_index < possible_cpus->len);
2833     return possible_cpus->cpus[cpu_index].props;
2834 }
2835 
2836 static int64_t pc_get_default_cpu_node_id(const MachineState *ms, int idx)
2837 {
2838    X86CPUTopoInfo topo;
2839    PCMachineState *pcms = PC_MACHINE(ms);
2840 
2841    assert(idx < ms->possible_cpus->len);
2842    x86_topo_ids_from_apicid(ms->possible_cpus->cpus[idx].arch_id,
2843                             pcms->smp_dies, ms->smp.cores,
2844                             ms->smp.threads, &topo);
2845    return topo.pkg_id % nb_numa_nodes;
2846 }
2847 
2848 static const CPUArchIdList *pc_possible_cpu_arch_ids(MachineState *ms)
2849 {
2850     PCMachineState *pcms = PC_MACHINE(ms);
2851     int i;
2852     unsigned int max_cpus = ms->smp.max_cpus;
2853 
2854     if (ms->possible_cpus) {
2855         /*
2856          * make sure that max_cpus hasn't changed since the first use, i.e.
2857          * -smp hasn't been parsed after it
2858         */
2859         assert(ms->possible_cpus->len == max_cpus);
2860         return ms->possible_cpus;
2861     }
2862 
2863     ms->possible_cpus = g_malloc0(sizeof(CPUArchIdList) +
2864                                   sizeof(CPUArchId) * max_cpus);
2865     ms->possible_cpus->len = max_cpus;
2866     for (i = 0; i < ms->possible_cpus->len; i++) {
2867         X86CPUTopoInfo topo;
2868 
2869         ms->possible_cpus->cpus[i].type = ms->cpu_type;
2870         ms->possible_cpus->cpus[i].vcpus_count = 1;
2871         ms->possible_cpus->cpus[i].arch_id = x86_cpu_apic_id_from_index(pcms, i);
2872         x86_topo_ids_from_apicid(ms->possible_cpus->cpus[i].arch_id,
2873                                  pcms->smp_dies, ms->smp.cores,
2874                                  ms->smp.threads, &topo);
2875         ms->possible_cpus->cpus[i].props.has_socket_id = true;
2876         ms->possible_cpus->cpus[i].props.socket_id = topo.pkg_id;
2877         ms->possible_cpus->cpus[i].props.has_die_id = true;
2878         ms->possible_cpus->cpus[i].props.die_id = topo.die_id;
2879         ms->possible_cpus->cpus[i].props.has_core_id = true;
2880         ms->possible_cpus->cpus[i].props.core_id = topo.core_id;
2881         ms->possible_cpus->cpus[i].props.has_thread_id = true;
2882         ms->possible_cpus->cpus[i].props.thread_id = topo.smt_id;
2883     }
2884     return ms->possible_cpus;
2885 }
2886 
2887 static void x86_nmi(NMIState *n, int cpu_index, Error **errp)
2888 {
2889     /* cpu index isn't used */
2890     CPUState *cs;
2891 
2892     CPU_FOREACH(cs) {
2893         X86CPU *cpu = X86_CPU(cs);
2894 
2895         if (!cpu->apic_state) {
2896             cpu_interrupt(cs, CPU_INTERRUPT_NMI);
2897         } else {
2898             apic_deliver_nmi(cpu->apic_state);
2899         }
2900     }
2901 }
2902 
2903 static void pc_machine_class_init(ObjectClass *oc, void *data)
2904 {
2905     MachineClass *mc = MACHINE_CLASS(oc);
2906     PCMachineClass *pcmc = PC_MACHINE_CLASS(oc);
2907     HotplugHandlerClass *hc = HOTPLUG_HANDLER_CLASS(oc);
2908     NMIClass *nc = NMI_CLASS(oc);
2909 
2910     pcmc->pci_enabled = true;
2911     pcmc->has_acpi_build = true;
2912     pcmc->rsdp_in_ram = true;
2913     pcmc->smbios_defaults = true;
2914     pcmc->smbios_uuid_encoded = true;
2915     pcmc->gigabyte_align = true;
2916     pcmc->has_reserved_memory = true;
2917     pcmc->kvmclock_enabled = true;
2918     pcmc->enforce_aligned_dimm = true;
2919     /* BIOS ACPI tables: 128K. Other BIOS datastructures: less than 4K reported
2920      * to be used at the moment, 32K should be enough for a while.  */
2921     pcmc->acpi_data_size = 0x20000 + 0x8000;
2922     pcmc->save_tsc_khz = true;
2923     pcmc->linuxboot_dma_enabled = true;
2924     pcmc->pvh_enabled = true;
2925     assert(!mc->get_hotplug_handler);
2926     mc->get_hotplug_handler = pc_get_hotplug_handler;
2927     mc->cpu_index_to_instance_props = pc_cpu_index_to_props;
2928     mc->get_default_cpu_node_id = pc_get_default_cpu_node_id;
2929     mc->possible_cpu_arch_ids = pc_possible_cpu_arch_ids;
2930     mc->auto_enable_numa_with_memhp = true;
2931     mc->has_hotpluggable_cpus = true;
2932     mc->default_boot_order = "cad";
2933     mc->hot_add_cpu = pc_hot_add_cpu;
2934     mc->smp_parse = pc_smp_parse;
2935     mc->block_default_type = IF_IDE;
2936     mc->max_cpus = 255;
2937     mc->reset = pc_machine_reset;
2938     hc->pre_plug = pc_machine_device_pre_plug_cb;
2939     hc->plug = pc_machine_device_plug_cb;
2940     hc->unplug_request = pc_machine_device_unplug_request_cb;
2941     hc->unplug = pc_machine_device_unplug_cb;
2942     nc->nmi_monitor_handler = x86_nmi;
2943     mc->default_cpu_type = TARGET_DEFAULT_CPU_TYPE;
2944     mc->nvdimm_supported = true;
2945     mc->numa_mem_supported = true;
2946 
2947     object_class_property_add(oc, PC_MACHINE_DEVMEM_REGION_SIZE, "int",
2948         pc_machine_get_device_memory_region_size, NULL,
2949         NULL, NULL, &error_abort);
2950 
2951     object_class_property_add(oc, PC_MACHINE_MAX_RAM_BELOW_4G, "size",
2952         pc_machine_get_max_ram_below_4g, pc_machine_set_max_ram_below_4g,
2953         NULL, NULL, &error_abort);
2954 
2955     object_class_property_set_description(oc, PC_MACHINE_MAX_RAM_BELOW_4G,
2956         "Maximum ram below the 4G boundary (32bit boundary)", &error_abort);
2957 
2958     object_class_property_add(oc, PC_MACHINE_SMM, "OnOffAuto",
2959         pc_machine_get_smm, pc_machine_set_smm,
2960         NULL, NULL, &error_abort);
2961     object_class_property_set_description(oc, PC_MACHINE_SMM,
2962         "Enable SMM (pc & q35)", &error_abort);
2963 
2964     object_class_property_add(oc, PC_MACHINE_VMPORT, "OnOffAuto",
2965         pc_machine_get_vmport, pc_machine_set_vmport,
2966         NULL, NULL, &error_abort);
2967     object_class_property_set_description(oc, PC_MACHINE_VMPORT,
2968         "Enable vmport (pc & q35)", &error_abort);
2969 
2970     object_class_property_add_bool(oc, PC_MACHINE_SMBUS,
2971         pc_machine_get_smbus, pc_machine_set_smbus, &error_abort);
2972 
2973     object_class_property_add_bool(oc, PC_MACHINE_SATA,
2974         pc_machine_get_sata, pc_machine_set_sata, &error_abort);
2975 
2976     object_class_property_add_bool(oc, PC_MACHINE_PIT,
2977         pc_machine_get_pit, pc_machine_set_pit, &error_abort);
2978 }
2979 
2980 static const TypeInfo pc_machine_info = {
2981     .name = TYPE_PC_MACHINE,
2982     .parent = TYPE_MACHINE,
2983     .abstract = true,
2984     .instance_size = sizeof(PCMachineState),
2985     .instance_init = pc_machine_initfn,
2986     .class_size = sizeof(PCMachineClass),
2987     .class_init = pc_machine_class_init,
2988     .interfaces = (InterfaceInfo[]) {
2989          { TYPE_HOTPLUG_HANDLER },
2990          { TYPE_NMI },
2991          { }
2992     },
2993 };
2994 
2995 static void pc_machine_register_types(void)
2996 {
2997     type_register_static(&pc_machine_info);
2998 }
2999 
3000 type_init(pc_machine_register_types)
3001