xref: /openbmc/qemu/hw/i386/pc.c (revision 1c30044e1aba56066eb6a51a788ce6a15a8faa8f)
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 "sysemu/cpus.h"
34 #include "hw/block/fdc.h"
35 #include "hw/ide.h"
36 #include "hw/pci/pci.h"
37 #include "hw/pci/pci_bus.h"
38 #include "hw/nvram/fw_cfg.h"
39 #include "hw/timer/hpet.h"
40 #include "hw/firmware/smbios.h"
41 #include "hw/loader.h"
42 #include "elf.h"
43 #include "multiboot.h"
44 #include "hw/timer/mc146818rtc.h"
45 #include "hw/dma/i8257.h"
46 #include "hw/timer/i8254.h"
47 #include "hw/input/i8042.h"
48 #include "hw/audio/pcspk.h"
49 #include "hw/pci/msi.h"
50 #include "hw/sysbus.h"
51 #include "sysemu/sysemu.h"
52 #include "sysemu/numa.h"
53 #include "sysemu/kvm.h"
54 #include "sysemu/qtest.h"
55 #include "kvm_i386.h"
56 #include "hw/xen/xen.h"
57 #include "ui/qemu-spice.h"
58 #include "exec/memory.h"
59 #include "exec/address-spaces.h"
60 #include "sysemu/arch_init.h"
61 #include "qemu/bitmap.h"
62 #include "qemu/config-file.h"
63 #include "qemu/error-report.h"
64 #include "qemu/option.h"
65 #include "hw/acpi/acpi.h"
66 #include "hw/acpi/cpu_hotplug.h"
67 #include "hw/boards.h"
68 #include "acpi-build.h"
69 #include "hw/mem/pc-dimm.h"
70 #include "qapi/error.h"
71 #include "qapi/qapi-visit-common.h"
72 #include "qapi/visitor.h"
73 #include "qom/cpu.h"
74 #include "hw/nmi.h"
75 #include "hw/i386/intel_iommu.h"
76 #include "hw/net/ne2000-isa.h"
77 
78 /* debug PC/ISA interrupts */
79 //#define DEBUG_IRQ
80 
81 #ifdef DEBUG_IRQ
82 #define DPRINTF(fmt, ...)                                       \
83     do { printf("CPUIRQ: " fmt , ## __VA_ARGS__); } while (0)
84 #else
85 #define DPRINTF(fmt, ...)
86 #endif
87 
88 #define FW_CFG_ACPI_TABLES (FW_CFG_ARCH_LOCAL + 0)
89 #define FW_CFG_SMBIOS_ENTRIES (FW_CFG_ARCH_LOCAL + 1)
90 #define FW_CFG_IRQ0_OVERRIDE (FW_CFG_ARCH_LOCAL + 2)
91 #define FW_CFG_E820_TABLE (FW_CFG_ARCH_LOCAL + 3)
92 #define FW_CFG_HPET (FW_CFG_ARCH_LOCAL + 4)
93 
94 #define E820_NR_ENTRIES		16
95 
96 struct e820_entry {
97     uint64_t address;
98     uint64_t length;
99     uint32_t type;
100 } QEMU_PACKED __attribute((__aligned__(4)));
101 
102 struct e820_table {
103     uint32_t count;
104     struct e820_entry entry[E820_NR_ENTRIES];
105 } QEMU_PACKED __attribute((__aligned__(4)));
106 
107 static struct e820_table e820_reserve;
108 static struct e820_entry *e820_table;
109 static unsigned e820_entries;
110 struct hpet_fw_config hpet_cfg = {.count = UINT8_MAX};
111 
112 GlobalProperty pc_compat_3_1[] = {
113     {
114         .driver   = "intel-iommu",
115         .property = "dma-drain",
116         .value    = "off",
117     },
118 };
119 const size_t pc_compat_3_1_len = G_N_ELEMENTS(pc_compat_3_1);
120 
121 GlobalProperty pc_compat_3_0[] = {
122     {
123         .driver   = TYPE_X86_CPU,
124         .property = "x-hv-synic-kvm-only",
125         .value    = "on",
126     },{
127         .driver   = "Skylake-Server" "-" TYPE_X86_CPU,
128         .property = "pku",
129         .value    = "off",
130     },{
131         .driver   = "Skylake-Server-IBRS" "-" TYPE_X86_CPU,
132         .property = "pku",
133         .value    = "off",
134     },
135 };
136 const size_t pc_compat_3_0_len = G_N_ELEMENTS(pc_compat_3_0);
137 
138 GlobalProperty pc_compat_2_12[] = {
139     {
140         .driver   = TYPE_X86_CPU,
141         .property = "legacy-cache",
142         .value    = "on",
143     },{
144         .driver   = TYPE_X86_CPU,
145         .property = "topoext",
146         .value    = "off",
147     },{
148         .driver   = "EPYC-" TYPE_X86_CPU,
149         .property = "xlevel",
150         .value    = stringify(0x8000000a),
151     },{
152         .driver   = "EPYC-IBPB-" TYPE_X86_CPU,
153         .property = "xlevel",
154         .value    = stringify(0x8000000a),
155     },
156 };
157 const size_t pc_compat_2_12_len = G_N_ELEMENTS(pc_compat_2_12);
158 
159 GlobalProperty pc_compat_2_11[] = {
160     {
161         .driver   = TYPE_X86_CPU,
162         .property = "x-migrate-smi-count",
163         .value    = "off",
164     },{
165         .driver   = "Skylake-Server" "-" TYPE_X86_CPU,
166         .property = "clflushopt",
167         .value    = "off",
168     },
169 };
170 const size_t pc_compat_2_11_len = G_N_ELEMENTS(pc_compat_2_11);
171 
172 GlobalProperty pc_compat_2_10[] = {
173     {
174         .driver   = TYPE_X86_CPU,
175         .property = "x-hv-max-vps",
176         .value    = "0x40",
177     },{
178         .driver   = "i440FX-pcihost",
179         .property = "x-pci-hole64-fix",
180         .value    = "off",
181     },{
182         .driver   = "q35-pcihost",
183         .property = "x-pci-hole64-fix",
184         .value    = "off",
185     },
186 };
187 const size_t pc_compat_2_10_len = G_N_ELEMENTS(pc_compat_2_10);
188 
189 GlobalProperty pc_compat_2_9[] = {
190     {
191         .driver   = "mch",
192         .property = "extended-tseg-mbytes",
193         .value    = stringify(0),
194     },
195 };
196 const size_t pc_compat_2_9_len = G_N_ELEMENTS(pc_compat_2_9);
197 
198 GlobalProperty pc_compat_2_8[] = {
199     {
200         .driver   = TYPE_X86_CPU,
201         .property = "tcg-cpuid",
202         .value    = "off",
203     },
204     {
205         .driver   = "kvmclock",
206         .property = "x-mach-use-reliable-get-clock",
207         .value    = "off",
208     },
209     {
210         .driver   = "ICH9-LPC",
211         .property = "x-smi-broadcast",
212         .value    = "off",
213     },
214     {
215         .driver   = TYPE_X86_CPU,
216         .property = "vmware-cpuid-freq",
217         .value    = "off",
218     },
219     {
220         .driver   = "Haswell-" TYPE_X86_CPU,
221         .property = "stepping",
222         .value    = "1",
223     },
224 };
225 const size_t pc_compat_2_8_len = G_N_ELEMENTS(pc_compat_2_8);
226 
227 GlobalProperty pc_compat_2_7[] = {
228     {
229         .driver   = TYPE_X86_CPU,
230         .property = "l3-cache",
231         .value    = "off",
232     },
233     {
234         .driver   = TYPE_X86_CPU,
235         .property = "full-cpuid-auto-level",
236         .value    = "off",
237     },
238     {
239         .driver   = "Opteron_G3" "-" TYPE_X86_CPU,
240         .property = "family",
241         .value    = "15",
242     },
243     {
244         .driver   = "Opteron_G3" "-" TYPE_X86_CPU,
245         .property = "model",
246         .value    = "6",
247     },
248     {
249         .driver   = "Opteron_G3" "-" TYPE_X86_CPU,
250         .property = "stepping",
251         .value    = "1",
252     },
253     {
254         .driver   = "isa-pcspk",
255         .property = "migrate",
256         .value    = "off",
257     },
258 };
259 const size_t pc_compat_2_7_len = G_N_ELEMENTS(pc_compat_2_7);
260 
261 GlobalProperty pc_compat_2_6[] = {
262     {
263         .driver   = TYPE_X86_CPU,
264         .property = "cpuid-0xb",
265         .value    = "off",
266     },{
267         .driver   = "vmxnet3",
268         .property = "romfile",
269         .value    = "",
270     },
271     {
272         .driver = TYPE_X86_CPU,
273         .property = "fill-mtrr-mask",
274         .value = "off",
275     },
276     {
277         .driver   = "apic-common",
278         .property = "legacy-instance-id",
279         .value    = "on",
280     }
281 };
282 const size_t pc_compat_2_6_len = G_N_ELEMENTS(pc_compat_2_6);
283 
284 GlobalProperty pc_compat_2_5[] = {};
285 const size_t pc_compat_2_5_len = G_N_ELEMENTS(pc_compat_2_5);
286 
287 GlobalProperty pc_compat_2_4[] = {
288     PC_CPU_MODEL_IDS("2.4.0")
289     {
290         .driver   = "Haswell-" TYPE_X86_CPU,
291         .property = "abm",
292         .value    = "off",
293     },
294     {
295         .driver   = "Haswell-noTSX-" TYPE_X86_CPU,
296         .property = "abm",
297         .value    = "off",
298     },
299     {
300         .driver   = "Broadwell-" TYPE_X86_CPU,
301         .property = "abm",
302         .value    = "off",
303     },
304     {
305         .driver   = "Broadwell-noTSX-" TYPE_X86_CPU,
306         .property = "abm",
307         .value    = "off",
308     },
309     {
310         .driver   = "host" "-" TYPE_X86_CPU,
311         .property = "host-cache-info",
312         .value    = "on",
313     },
314     {
315         .driver   = TYPE_X86_CPU,
316         .property = "check",
317         .value    = "off",
318     },
319     {
320         .driver   = "qemu64" "-" TYPE_X86_CPU,
321         .property = "sse4a",
322         .value    = "on",
323     },
324     {
325         .driver   = "qemu64" "-" TYPE_X86_CPU,
326         .property = "abm",
327         .value    = "on",
328     },
329     {
330         .driver   = "qemu64" "-" TYPE_X86_CPU,
331         .property = "popcnt",
332         .value    = "on",
333     },
334     {
335         .driver   = "qemu32" "-" TYPE_X86_CPU,
336         .property = "popcnt",
337         .value    = "on",
338     },{
339         .driver   = "Opteron_G2" "-" TYPE_X86_CPU,
340         .property = "rdtscp",
341         .value    = "on",
342     },{
343         .driver   = "Opteron_G3" "-" TYPE_X86_CPU,
344         .property = "rdtscp",
345         .value    = "on",
346     },{
347         .driver   = "Opteron_G4" "-" TYPE_X86_CPU,
348         .property = "rdtscp",
349         .value    = "on",
350     },{
351         .driver   = "Opteron_G5" "-" TYPE_X86_CPU,
352         .property = "rdtscp",
353         .value    = "on",
354     }
355 };
356 const size_t pc_compat_2_4_len = G_N_ELEMENTS(pc_compat_2_4);
357 
358 GlobalProperty pc_compat_2_3[] = {
359     PC_CPU_MODEL_IDS("2.3.0")
360     {
361         .driver   = TYPE_X86_CPU,
362         .property = "arat",
363         .value    = "off",
364     },{
365         .driver   = "qemu64" "-" TYPE_X86_CPU,
366         .property = "min-level",
367         .value    = stringify(4),
368     },{
369         .driver   = "kvm64" "-" TYPE_X86_CPU,
370         .property = "min-level",
371         .value    = stringify(5),
372     },{
373         .driver   = "pentium3" "-" TYPE_X86_CPU,
374         .property = "min-level",
375         .value    = stringify(2),
376     },{
377         .driver   = "n270" "-" TYPE_X86_CPU,
378         .property = "min-level",
379         .value    = stringify(5),
380     },{
381         .driver   = "Conroe" "-" TYPE_X86_CPU,
382         .property = "min-level",
383         .value    = stringify(4),
384     },{
385         .driver   = "Penryn" "-" TYPE_X86_CPU,
386         .property = "min-level",
387         .value    = stringify(4),
388     },{
389         .driver   = "Nehalem" "-" TYPE_X86_CPU,
390         .property = "min-level",
391         .value    = stringify(4),
392     },{
393         .driver   = "n270" "-" TYPE_X86_CPU,
394         .property = "min-xlevel",
395         .value    = stringify(0x8000000a),
396     },{
397         .driver   = "Penryn" "-" TYPE_X86_CPU,
398         .property = "min-xlevel",
399         .value    = stringify(0x8000000a),
400     },{
401         .driver   = "Conroe" "-" TYPE_X86_CPU,
402         .property = "min-xlevel",
403         .value    = stringify(0x8000000a),
404     },{
405         .driver   = "Nehalem" "-" TYPE_X86_CPU,
406         .property = "min-xlevel",
407         .value    = stringify(0x8000000a),
408     },{
409         .driver   = "Westmere" "-" TYPE_X86_CPU,
410         .property = "min-xlevel",
411         .value    = stringify(0x8000000a),
412     },{
413         .driver   = "SandyBridge" "-" TYPE_X86_CPU,
414         .property = "min-xlevel",
415         .value    = stringify(0x8000000a),
416     },{
417         .driver   = "IvyBridge" "-" TYPE_X86_CPU,
418         .property = "min-xlevel",
419         .value    = stringify(0x8000000a),
420     },{
421         .driver   = "Haswell" "-" TYPE_X86_CPU,
422         .property = "min-xlevel",
423         .value    = stringify(0x8000000a),
424     },{
425         .driver   = "Haswell-noTSX" "-" TYPE_X86_CPU,
426         .property = "min-xlevel",
427         .value    = stringify(0x8000000a),
428     },{
429         .driver   = "Broadwell" "-" TYPE_X86_CPU,
430         .property = "min-xlevel",
431         .value    = stringify(0x8000000a),
432     },{
433         .driver   = "Broadwell-noTSX" "-" TYPE_X86_CPU,
434         .property = "min-xlevel",
435         .value    = stringify(0x8000000a),
436     },{
437         .driver = TYPE_X86_CPU,
438         .property = "kvm-no-smi-migration",
439         .value    = "on",
440     },
441 };
442 const size_t pc_compat_2_3_len = G_N_ELEMENTS(pc_compat_2_3);
443 
444 GlobalProperty pc_compat_2_2[] = {
445     PC_CPU_MODEL_IDS("2.2.0")
446     {
447         .driver = "kvm64" "-" TYPE_X86_CPU,
448         .property = "vme",
449         .value = "off",
450     },
451     {
452         .driver = "kvm32" "-" TYPE_X86_CPU,
453         .property = "vme",
454         .value = "off",
455     },
456     {
457         .driver = "Conroe" "-" TYPE_X86_CPU,
458         .property = "vme",
459         .value = "off",
460     },
461     {
462         .driver = "Penryn" "-" TYPE_X86_CPU,
463         .property = "vme",
464         .value = "off",
465     },
466     {
467         .driver = "Nehalem" "-" TYPE_X86_CPU,
468         .property = "vme",
469         .value = "off",
470     },
471     {
472         .driver = "Westmere" "-" TYPE_X86_CPU,
473         .property = "vme",
474         .value = "off",
475     },
476     {
477         .driver = "SandyBridge" "-" TYPE_X86_CPU,
478         .property = "vme",
479         .value = "off",
480     },
481     {
482         .driver = "Haswell" "-" TYPE_X86_CPU,
483         .property = "vme",
484         .value = "off",
485     },
486     {
487         .driver = "Broadwell" "-" TYPE_X86_CPU,
488         .property = "vme",
489         .value = "off",
490     },
491     {
492         .driver = "Opteron_G1" "-" TYPE_X86_CPU,
493         .property = "vme",
494         .value = "off",
495     },
496     {
497         .driver = "Opteron_G2" "-" TYPE_X86_CPU,
498         .property = "vme",
499         .value = "off",
500     },
501     {
502         .driver = "Opteron_G3" "-" TYPE_X86_CPU,
503         .property = "vme",
504         .value = "off",
505     },
506     {
507         .driver = "Opteron_G4" "-" TYPE_X86_CPU,
508         .property = "vme",
509         .value = "off",
510     },
511     {
512         .driver = "Opteron_G5" "-" TYPE_X86_CPU,
513         .property = "vme",
514         .value = "off",
515     },
516     {
517         .driver = "Haswell" "-" TYPE_X86_CPU,
518         .property = "f16c",
519         .value = "off",
520     },
521     {
522         .driver = "Haswell" "-" TYPE_X86_CPU,
523         .property = "rdrand",
524         .value = "off",
525     },
526     {
527         .driver = "Broadwell" "-" TYPE_X86_CPU,
528         .property = "f16c",
529         .value = "off",
530     },
531     {
532         .driver = "Broadwell" "-" TYPE_X86_CPU,
533         .property = "rdrand",
534         .value = "off",
535     },
536 };
537 const size_t pc_compat_2_2_len = G_N_ELEMENTS(pc_compat_2_2);
538 
539 void gsi_handler(void *opaque, int n, int level)
540 {
541     GSIState *s = opaque;
542 
543     DPRINTF("pc: %s GSI %d\n", level ? "raising" : "lowering", n);
544     if (n < ISA_NUM_IRQS) {
545         qemu_set_irq(s->i8259_irq[n], level);
546     }
547     qemu_set_irq(s->ioapic_irq[n], level);
548 }
549 
550 static void ioport80_write(void *opaque, hwaddr addr, uint64_t data,
551                            unsigned size)
552 {
553 }
554 
555 static uint64_t ioport80_read(void *opaque, hwaddr addr, unsigned size)
556 {
557     return 0xffffffffffffffffULL;
558 }
559 
560 /* MSDOS compatibility mode FPU exception support */
561 static qemu_irq ferr_irq;
562 
563 void pc_register_ferr_irq(qemu_irq irq)
564 {
565     ferr_irq = irq;
566 }
567 
568 /* XXX: add IGNNE support */
569 void cpu_set_ferr(CPUX86State *s)
570 {
571     qemu_irq_raise(ferr_irq);
572 }
573 
574 static void ioportF0_write(void *opaque, hwaddr addr, uint64_t data,
575                            unsigned size)
576 {
577     qemu_irq_lower(ferr_irq);
578 }
579 
580 static uint64_t ioportF0_read(void *opaque, hwaddr addr, unsigned size)
581 {
582     return 0xffffffffffffffffULL;
583 }
584 
585 /* TSC handling */
586 uint64_t cpu_get_tsc(CPUX86State *env)
587 {
588     return cpu_get_ticks();
589 }
590 
591 /* IRQ handling */
592 int cpu_get_pic_interrupt(CPUX86State *env)
593 {
594     X86CPU *cpu = x86_env_get_cpu(env);
595     int intno;
596 
597     if (!kvm_irqchip_in_kernel()) {
598         intno = apic_get_interrupt(cpu->apic_state);
599         if (intno >= 0) {
600             return intno;
601         }
602         /* read the irq from the PIC */
603         if (!apic_accept_pic_intr(cpu->apic_state)) {
604             return -1;
605         }
606     }
607 
608     intno = pic_read_irq(isa_pic);
609     return intno;
610 }
611 
612 static void pic_irq_request(void *opaque, int irq, int level)
613 {
614     CPUState *cs = first_cpu;
615     X86CPU *cpu = X86_CPU(cs);
616 
617     DPRINTF("pic_irqs: %s irq %d\n", level? "raise" : "lower", irq);
618     if (cpu->apic_state && !kvm_irqchip_in_kernel()) {
619         CPU_FOREACH(cs) {
620             cpu = X86_CPU(cs);
621             if (apic_accept_pic_intr(cpu->apic_state)) {
622                 apic_deliver_pic_intr(cpu->apic_state, level);
623             }
624         }
625     } else {
626         if (level) {
627             cpu_interrupt(cs, CPU_INTERRUPT_HARD);
628         } else {
629             cpu_reset_interrupt(cs, CPU_INTERRUPT_HARD);
630         }
631     }
632 }
633 
634 /* PC cmos mappings */
635 
636 #define REG_EQUIPMENT_BYTE          0x14
637 
638 int cmos_get_fd_drive_type(FloppyDriveType fd0)
639 {
640     int val;
641 
642     switch (fd0) {
643     case FLOPPY_DRIVE_TYPE_144:
644         /* 1.44 Mb 3"5 drive */
645         val = 4;
646         break;
647     case FLOPPY_DRIVE_TYPE_288:
648         /* 2.88 Mb 3"5 drive */
649         val = 5;
650         break;
651     case FLOPPY_DRIVE_TYPE_120:
652         /* 1.2 Mb 5"5 drive */
653         val = 2;
654         break;
655     case FLOPPY_DRIVE_TYPE_NONE:
656     default:
657         val = 0;
658         break;
659     }
660     return val;
661 }
662 
663 static void cmos_init_hd(ISADevice *s, int type_ofs, int info_ofs,
664                          int16_t cylinders, int8_t heads, int8_t sectors)
665 {
666     rtc_set_memory(s, type_ofs, 47);
667     rtc_set_memory(s, info_ofs, cylinders);
668     rtc_set_memory(s, info_ofs + 1, cylinders >> 8);
669     rtc_set_memory(s, info_ofs + 2, heads);
670     rtc_set_memory(s, info_ofs + 3, 0xff);
671     rtc_set_memory(s, info_ofs + 4, 0xff);
672     rtc_set_memory(s, info_ofs + 5, 0xc0 | ((heads > 8) << 3));
673     rtc_set_memory(s, info_ofs + 6, cylinders);
674     rtc_set_memory(s, info_ofs + 7, cylinders >> 8);
675     rtc_set_memory(s, info_ofs + 8, sectors);
676 }
677 
678 /* convert boot_device letter to something recognizable by the bios */
679 static int boot_device2nibble(char boot_device)
680 {
681     switch(boot_device) {
682     case 'a':
683     case 'b':
684         return 0x01; /* floppy boot */
685     case 'c':
686         return 0x02; /* hard drive boot */
687     case 'd':
688         return 0x03; /* CD-ROM boot */
689     case 'n':
690         return 0x04; /* Network boot */
691     }
692     return 0;
693 }
694 
695 static void set_boot_dev(ISADevice *s, const char *boot_device, Error **errp)
696 {
697 #define PC_MAX_BOOT_DEVICES 3
698     int nbds, bds[3] = { 0, };
699     int i;
700 
701     nbds = strlen(boot_device);
702     if (nbds > PC_MAX_BOOT_DEVICES) {
703         error_setg(errp, "Too many boot devices for PC");
704         return;
705     }
706     for (i = 0; i < nbds; i++) {
707         bds[i] = boot_device2nibble(boot_device[i]);
708         if (bds[i] == 0) {
709             error_setg(errp, "Invalid boot device for PC: '%c'",
710                        boot_device[i]);
711             return;
712         }
713     }
714     rtc_set_memory(s, 0x3d, (bds[1] << 4) | bds[0]);
715     rtc_set_memory(s, 0x38, (bds[2] << 4) | (fd_bootchk ? 0x0 : 0x1));
716 }
717 
718 static void pc_boot_set(void *opaque, const char *boot_device, Error **errp)
719 {
720     set_boot_dev(opaque, boot_device, errp);
721 }
722 
723 static void pc_cmos_init_floppy(ISADevice *rtc_state, ISADevice *floppy)
724 {
725     int val, nb, i;
726     FloppyDriveType fd_type[2] = { FLOPPY_DRIVE_TYPE_NONE,
727                                    FLOPPY_DRIVE_TYPE_NONE };
728 
729     /* floppy type */
730     if (floppy) {
731         for (i = 0; i < 2; i++) {
732             fd_type[i] = isa_fdc_get_drive_type(floppy, i);
733         }
734     }
735     val = (cmos_get_fd_drive_type(fd_type[0]) << 4) |
736         cmos_get_fd_drive_type(fd_type[1]);
737     rtc_set_memory(rtc_state, 0x10, val);
738 
739     val = rtc_get_memory(rtc_state, REG_EQUIPMENT_BYTE);
740     nb = 0;
741     if (fd_type[0] != FLOPPY_DRIVE_TYPE_NONE) {
742         nb++;
743     }
744     if (fd_type[1] != FLOPPY_DRIVE_TYPE_NONE) {
745         nb++;
746     }
747     switch (nb) {
748     case 0:
749         break;
750     case 1:
751         val |= 0x01; /* 1 drive, ready for boot */
752         break;
753     case 2:
754         val |= 0x41; /* 2 drives, ready for boot */
755         break;
756     }
757     rtc_set_memory(rtc_state, REG_EQUIPMENT_BYTE, val);
758 }
759 
760 typedef struct pc_cmos_init_late_arg {
761     ISADevice *rtc_state;
762     BusState *idebus[2];
763 } pc_cmos_init_late_arg;
764 
765 typedef struct check_fdc_state {
766     ISADevice *floppy;
767     bool multiple;
768 } CheckFdcState;
769 
770 static int check_fdc(Object *obj, void *opaque)
771 {
772     CheckFdcState *state = opaque;
773     Object *fdc;
774     uint32_t iobase;
775     Error *local_err = NULL;
776 
777     fdc = object_dynamic_cast(obj, TYPE_ISA_FDC);
778     if (!fdc) {
779         return 0;
780     }
781 
782     iobase = object_property_get_uint(obj, "iobase", &local_err);
783     if (local_err || iobase != 0x3f0) {
784         error_free(local_err);
785         return 0;
786     }
787 
788     if (state->floppy) {
789         state->multiple = true;
790     } else {
791         state->floppy = ISA_DEVICE(obj);
792     }
793     return 0;
794 }
795 
796 static const char * const fdc_container_path[] = {
797     "/unattached", "/peripheral", "/peripheral-anon"
798 };
799 
800 /*
801  * Locate the FDC at IO address 0x3f0, in order to configure the CMOS registers
802  * and ACPI objects.
803  */
804 ISADevice *pc_find_fdc0(void)
805 {
806     int i;
807     Object *container;
808     CheckFdcState state = { 0 };
809 
810     for (i = 0; i < ARRAY_SIZE(fdc_container_path); i++) {
811         container = container_get(qdev_get_machine(), fdc_container_path[i]);
812         object_child_foreach(container, check_fdc, &state);
813     }
814 
815     if (state.multiple) {
816         warn_report("multiple floppy disk controllers with "
817                     "iobase=0x3f0 have been found");
818         error_printf("the one being picked for CMOS setup might not reflect "
819                      "your intent");
820     }
821 
822     return state.floppy;
823 }
824 
825 static void pc_cmos_init_late(void *opaque)
826 {
827     pc_cmos_init_late_arg *arg = opaque;
828     ISADevice *s = arg->rtc_state;
829     int16_t cylinders;
830     int8_t heads, sectors;
831     int val;
832     int i, trans;
833 
834     val = 0;
835     if (arg->idebus[0] && ide_get_geometry(arg->idebus[0], 0,
836                                            &cylinders, &heads, &sectors) >= 0) {
837         cmos_init_hd(s, 0x19, 0x1b, cylinders, heads, sectors);
838         val |= 0xf0;
839     }
840     if (arg->idebus[0] && ide_get_geometry(arg->idebus[0], 1,
841                                            &cylinders, &heads, &sectors) >= 0) {
842         cmos_init_hd(s, 0x1a, 0x24, cylinders, heads, sectors);
843         val |= 0x0f;
844     }
845     rtc_set_memory(s, 0x12, val);
846 
847     val = 0;
848     for (i = 0; i < 4; i++) {
849         /* NOTE: ide_get_geometry() returns the physical
850            geometry.  It is always such that: 1 <= sects <= 63, 1
851            <= heads <= 16, 1 <= cylinders <= 16383. The BIOS
852            geometry can be different if a translation is done. */
853         if (arg->idebus[i / 2] &&
854             ide_get_geometry(arg->idebus[i / 2], i % 2,
855                              &cylinders, &heads, &sectors) >= 0) {
856             trans = ide_get_bios_chs_trans(arg->idebus[i / 2], i % 2) - 1;
857             assert((trans & ~3) == 0);
858             val |= trans << (i * 2);
859         }
860     }
861     rtc_set_memory(s, 0x39, val);
862 
863     pc_cmos_init_floppy(s, pc_find_fdc0());
864 
865     qemu_unregister_reset(pc_cmos_init_late, opaque);
866 }
867 
868 void pc_cmos_init(PCMachineState *pcms,
869                   BusState *idebus0, BusState *idebus1,
870                   ISADevice *s)
871 {
872     int val;
873     static pc_cmos_init_late_arg arg;
874 
875     /* various important CMOS locations needed by PC/Bochs bios */
876 
877     /* memory size */
878     /* base memory (first MiB) */
879     val = MIN(pcms->below_4g_mem_size / KiB, 640);
880     rtc_set_memory(s, 0x15, val);
881     rtc_set_memory(s, 0x16, val >> 8);
882     /* extended memory (next 64MiB) */
883     if (pcms->below_4g_mem_size > 1 * MiB) {
884         val = (pcms->below_4g_mem_size - 1 * MiB) / KiB;
885     } else {
886         val = 0;
887     }
888     if (val > 65535)
889         val = 65535;
890     rtc_set_memory(s, 0x17, val);
891     rtc_set_memory(s, 0x18, val >> 8);
892     rtc_set_memory(s, 0x30, val);
893     rtc_set_memory(s, 0x31, val >> 8);
894     /* memory between 16MiB and 4GiB */
895     if (pcms->below_4g_mem_size > 16 * MiB) {
896         val = (pcms->below_4g_mem_size - 16 * MiB) / (64 * KiB);
897     } else {
898         val = 0;
899     }
900     if (val > 65535)
901         val = 65535;
902     rtc_set_memory(s, 0x34, val);
903     rtc_set_memory(s, 0x35, val >> 8);
904     /* memory above 4GiB */
905     val = pcms->above_4g_mem_size / 65536;
906     rtc_set_memory(s, 0x5b, val);
907     rtc_set_memory(s, 0x5c, val >> 8);
908     rtc_set_memory(s, 0x5d, val >> 16);
909 
910     object_property_add_link(OBJECT(pcms), "rtc_state",
911                              TYPE_ISA_DEVICE,
912                              (Object **)&pcms->rtc,
913                              object_property_allow_set_link,
914                              OBJ_PROP_LINK_STRONG, &error_abort);
915     object_property_set_link(OBJECT(pcms), OBJECT(s),
916                              "rtc_state", &error_abort);
917 
918     set_boot_dev(s, MACHINE(pcms)->boot_order, &error_fatal);
919 
920     val = 0;
921     val |= 0x02; /* FPU is there */
922     val |= 0x04; /* PS/2 mouse installed */
923     rtc_set_memory(s, REG_EQUIPMENT_BYTE, val);
924 
925     /* hard drives and FDC */
926     arg.rtc_state = s;
927     arg.idebus[0] = idebus0;
928     arg.idebus[1] = idebus1;
929     qemu_register_reset(pc_cmos_init_late, &arg);
930 }
931 
932 #define TYPE_PORT92 "port92"
933 #define PORT92(obj) OBJECT_CHECK(Port92State, (obj), TYPE_PORT92)
934 
935 /* port 92 stuff: could be split off */
936 typedef struct Port92State {
937     ISADevice parent_obj;
938 
939     MemoryRegion io;
940     uint8_t outport;
941     qemu_irq a20_out;
942 } Port92State;
943 
944 static void port92_write(void *opaque, hwaddr addr, uint64_t val,
945                          unsigned size)
946 {
947     Port92State *s = opaque;
948     int oldval = s->outport;
949 
950     DPRINTF("port92: write 0x%02" PRIx64 "\n", val);
951     s->outport = val;
952     qemu_set_irq(s->a20_out, (val >> 1) & 1);
953     if ((val & 1) && !(oldval & 1)) {
954         qemu_system_reset_request(SHUTDOWN_CAUSE_GUEST_RESET);
955     }
956 }
957 
958 static uint64_t port92_read(void *opaque, hwaddr addr,
959                             unsigned size)
960 {
961     Port92State *s = opaque;
962     uint32_t ret;
963 
964     ret = s->outport;
965     DPRINTF("port92: read 0x%02x\n", ret);
966     return ret;
967 }
968 
969 static void port92_init(ISADevice *dev, qemu_irq a20_out)
970 {
971     qdev_connect_gpio_out_named(DEVICE(dev), PORT92_A20_LINE, 0, a20_out);
972 }
973 
974 static const VMStateDescription vmstate_port92_isa = {
975     .name = "port92",
976     .version_id = 1,
977     .minimum_version_id = 1,
978     .fields = (VMStateField[]) {
979         VMSTATE_UINT8(outport, Port92State),
980         VMSTATE_END_OF_LIST()
981     }
982 };
983 
984 static void port92_reset(DeviceState *d)
985 {
986     Port92State *s = PORT92(d);
987 
988     s->outport &= ~1;
989 }
990 
991 static const MemoryRegionOps port92_ops = {
992     .read = port92_read,
993     .write = port92_write,
994     .impl = {
995         .min_access_size = 1,
996         .max_access_size = 1,
997     },
998     .endianness = DEVICE_LITTLE_ENDIAN,
999 };
1000 
1001 static void port92_initfn(Object *obj)
1002 {
1003     Port92State *s = PORT92(obj);
1004 
1005     memory_region_init_io(&s->io, OBJECT(s), &port92_ops, s, "port92", 1);
1006 
1007     s->outport = 0;
1008 
1009     qdev_init_gpio_out_named(DEVICE(obj), &s->a20_out, PORT92_A20_LINE, 1);
1010 }
1011 
1012 static void port92_realizefn(DeviceState *dev, Error **errp)
1013 {
1014     ISADevice *isadev = ISA_DEVICE(dev);
1015     Port92State *s = PORT92(dev);
1016 
1017     isa_register_ioport(isadev, &s->io, 0x92);
1018 }
1019 
1020 static void port92_class_initfn(ObjectClass *klass, void *data)
1021 {
1022     DeviceClass *dc = DEVICE_CLASS(klass);
1023 
1024     dc->realize = port92_realizefn;
1025     dc->reset = port92_reset;
1026     dc->vmsd = &vmstate_port92_isa;
1027     /*
1028      * Reason: unlike ordinary ISA devices, this one needs additional
1029      * wiring: its A20 output line needs to be wired up by
1030      * port92_init().
1031      */
1032     dc->user_creatable = false;
1033 }
1034 
1035 static const TypeInfo port92_info = {
1036     .name          = TYPE_PORT92,
1037     .parent        = TYPE_ISA_DEVICE,
1038     .instance_size = sizeof(Port92State),
1039     .instance_init = port92_initfn,
1040     .class_init    = port92_class_initfn,
1041 };
1042 
1043 static void port92_register_types(void)
1044 {
1045     type_register_static(&port92_info);
1046 }
1047 
1048 type_init(port92_register_types)
1049 
1050 static void handle_a20_line_change(void *opaque, int irq, int level)
1051 {
1052     X86CPU *cpu = opaque;
1053 
1054     /* XXX: send to all CPUs ? */
1055     /* XXX: add logic to handle multiple A20 line sources */
1056     x86_cpu_set_a20(cpu, level);
1057 }
1058 
1059 int e820_add_entry(uint64_t address, uint64_t length, uint32_t type)
1060 {
1061     int index = le32_to_cpu(e820_reserve.count);
1062     struct e820_entry *entry;
1063 
1064     if (type != E820_RAM) {
1065         /* old FW_CFG_E820_TABLE entry -- reservations only */
1066         if (index >= E820_NR_ENTRIES) {
1067             return -EBUSY;
1068         }
1069         entry = &e820_reserve.entry[index++];
1070 
1071         entry->address = cpu_to_le64(address);
1072         entry->length = cpu_to_le64(length);
1073         entry->type = cpu_to_le32(type);
1074 
1075         e820_reserve.count = cpu_to_le32(index);
1076     }
1077 
1078     /* new "etc/e820" file -- include ram too */
1079     e820_table = g_renew(struct e820_entry, e820_table, e820_entries + 1);
1080     e820_table[e820_entries].address = cpu_to_le64(address);
1081     e820_table[e820_entries].length = cpu_to_le64(length);
1082     e820_table[e820_entries].type = cpu_to_le32(type);
1083     e820_entries++;
1084 
1085     return e820_entries;
1086 }
1087 
1088 int e820_get_num_entries(void)
1089 {
1090     return e820_entries;
1091 }
1092 
1093 bool e820_get_entry(int idx, uint32_t type, uint64_t *address, uint64_t *length)
1094 {
1095     if (idx < e820_entries && e820_table[idx].type == cpu_to_le32(type)) {
1096         *address = le64_to_cpu(e820_table[idx].address);
1097         *length = le64_to_cpu(e820_table[idx].length);
1098         return true;
1099     }
1100     return false;
1101 }
1102 
1103 /* Enables contiguous-apic-ID mode, for compatibility */
1104 static bool compat_apic_id_mode;
1105 
1106 void enable_compat_apic_id_mode(void)
1107 {
1108     compat_apic_id_mode = true;
1109 }
1110 
1111 /* Calculates initial APIC ID for a specific CPU index
1112  *
1113  * Currently we need to be able to calculate the APIC ID from the CPU index
1114  * alone (without requiring a CPU object), as the QEMU<->Seabios interfaces have
1115  * no concept of "CPU index", and the NUMA tables on fw_cfg need the APIC ID of
1116  * all CPUs up to max_cpus.
1117  */
1118 static uint32_t x86_cpu_apic_id_from_index(unsigned int cpu_index)
1119 {
1120     uint32_t correct_id;
1121     static bool warned;
1122 
1123     correct_id = x86_apicid_from_cpu_idx(smp_cores, smp_threads, cpu_index);
1124     if (compat_apic_id_mode) {
1125         if (cpu_index != correct_id && !warned && !qtest_enabled()) {
1126             error_report("APIC IDs set in compatibility mode, "
1127                          "CPU topology won't match the configuration");
1128             warned = true;
1129         }
1130         return cpu_index;
1131     } else {
1132         return correct_id;
1133     }
1134 }
1135 
1136 static void pc_build_smbios(PCMachineState *pcms)
1137 {
1138     uint8_t *smbios_tables, *smbios_anchor;
1139     size_t smbios_tables_len, smbios_anchor_len;
1140     struct smbios_phys_mem_area *mem_array;
1141     unsigned i, array_count;
1142     MachineState *ms = MACHINE(pcms);
1143     X86CPU *cpu = X86_CPU(ms->possible_cpus->cpus[0].cpu);
1144 
1145     /* tell smbios about cpuid version and features */
1146     smbios_set_cpuid(cpu->env.cpuid_version, cpu->env.features[FEAT_1_EDX]);
1147 
1148     smbios_tables = smbios_get_table_legacy(&smbios_tables_len);
1149     if (smbios_tables) {
1150         fw_cfg_add_bytes(pcms->fw_cfg, FW_CFG_SMBIOS_ENTRIES,
1151                          smbios_tables, smbios_tables_len);
1152     }
1153 
1154     /* build the array of physical mem area from e820 table */
1155     mem_array = g_malloc0(sizeof(*mem_array) * e820_get_num_entries());
1156     for (i = 0, array_count = 0; i < e820_get_num_entries(); i++) {
1157         uint64_t addr, len;
1158 
1159         if (e820_get_entry(i, E820_RAM, &addr, &len)) {
1160             mem_array[array_count].address = addr;
1161             mem_array[array_count].length = len;
1162             array_count++;
1163         }
1164     }
1165     smbios_get_tables(mem_array, array_count,
1166                       &smbios_tables, &smbios_tables_len,
1167                       &smbios_anchor, &smbios_anchor_len);
1168     g_free(mem_array);
1169 
1170     if (smbios_anchor) {
1171         fw_cfg_add_file(pcms->fw_cfg, "etc/smbios/smbios-tables",
1172                         smbios_tables, smbios_tables_len);
1173         fw_cfg_add_file(pcms->fw_cfg, "etc/smbios/smbios-anchor",
1174                         smbios_anchor, smbios_anchor_len);
1175     }
1176 }
1177 
1178 static FWCfgState *bochs_bios_init(AddressSpace *as, PCMachineState *pcms)
1179 {
1180     FWCfgState *fw_cfg;
1181     uint64_t *numa_fw_cfg;
1182     int i;
1183     const CPUArchIdList *cpus;
1184     MachineClass *mc = MACHINE_GET_CLASS(pcms);
1185 
1186     fw_cfg = fw_cfg_init_io_dma(FW_CFG_IO_BASE, FW_CFG_IO_BASE + 4, as);
1187     fw_cfg_add_i16(fw_cfg, FW_CFG_NB_CPUS, pcms->boot_cpus);
1188 
1189     /* FW_CFG_MAX_CPUS is a bit confusing/problematic on x86:
1190      *
1191      * For machine types prior to 1.8, SeaBIOS needs FW_CFG_MAX_CPUS for
1192      * building MPTable, ACPI MADT, ACPI CPU hotplug and ACPI SRAT table,
1193      * that tables are based on xAPIC ID and QEMU<->SeaBIOS interface
1194      * for CPU hotplug also uses APIC ID and not "CPU index".
1195      * This means that FW_CFG_MAX_CPUS is not the "maximum number of CPUs",
1196      * but the "limit to the APIC ID values SeaBIOS may see".
1197      *
1198      * So for compatibility reasons with old BIOSes we are stuck with
1199      * "etc/max-cpus" actually being apic_id_limit
1200      */
1201     fw_cfg_add_i16(fw_cfg, FW_CFG_MAX_CPUS, (uint16_t)pcms->apic_id_limit);
1202     fw_cfg_add_i64(fw_cfg, FW_CFG_RAM_SIZE, (uint64_t)ram_size);
1203     fw_cfg_add_bytes(fw_cfg, FW_CFG_ACPI_TABLES,
1204                      acpi_tables, acpi_tables_len);
1205     fw_cfg_add_i32(fw_cfg, FW_CFG_IRQ0_OVERRIDE, kvm_allows_irq0_override());
1206 
1207     fw_cfg_add_bytes(fw_cfg, FW_CFG_E820_TABLE,
1208                      &e820_reserve, sizeof(e820_reserve));
1209     fw_cfg_add_file(fw_cfg, "etc/e820", e820_table,
1210                     sizeof(struct e820_entry) * e820_entries);
1211 
1212     fw_cfg_add_bytes(fw_cfg, FW_CFG_HPET, &hpet_cfg, sizeof(hpet_cfg));
1213     /* allocate memory for the NUMA channel: one (64bit) word for the number
1214      * of nodes, one word for each VCPU->node and one word for each node to
1215      * hold the amount of memory.
1216      */
1217     numa_fw_cfg = g_new0(uint64_t, 1 + pcms->apic_id_limit + nb_numa_nodes);
1218     numa_fw_cfg[0] = cpu_to_le64(nb_numa_nodes);
1219     cpus = mc->possible_cpu_arch_ids(MACHINE(pcms));
1220     for (i = 0; i < cpus->len; i++) {
1221         unsigned int apic_id = cpus->cpus[i].arch_id;
1222         assert(apic_id < pcms->apic_id_limit);
1223         numa_fw_cfg[apic_id + 1] = cpu_to_le64(cpus->cpus[i].props.node_id);
1224     }
1225     for (i = 0; i < nb_numa_nodes; i++) {
1226         numa_fw_cfg[pcms->apic_id_limit + 1 + i] =
1227             cpu_to_le64(numa_info[i].node_mem);
1228     }
1229     fw_cfg_add_bytes(fw_cfg, FW_CFG_NUMA, numa_fw_cfg,
1230                      (1 + pcms->apic_id_limit + nb_numa_nodes) *
1231                      sizeof(*numa_fw_cfg));
1232 
1233     return fw_cfg;
1234 }
1235 
1236 static long get_file_size(FILE *f)
1237 {
1238     long where, size;
1239 
1240     /* XXX: on Unix systems, using fstat() probably makes more sense */
1241 
1242     where = ftell(f);
1243     fseek(f, 0, SEEK_END);
1244     size = ftell(f);
1245     fseek(f, where, SEEK_SET);
1246 
1247     return size;
1248 }
1249 
1250 /* setup_data types */
1251 #define SETUP_NONE     0
1252 #define SETUP_E820_EXT 1
1253 #define SETUP_DTB      2
1254 #define SETUP_PCI      3
1255 #define SETUP_EFI      4
1256 
1257 struct setup_data {
1258     uint64_t next;
1259     uint32_t type;
1260     uint32_t len;
1261     uint8_t data[0];
1262 } __attribute__((packed));
1263 
1264 static void load_linux(PCMachineState *pcms,
1265                        FWCfgState *fw_cfg)
1266 {
1267     uint16_t protocol;
1268     int setup_size, kernel_size, cmdline_size;
1269     int dtb_size, setup_data_offset;
1270     uint32_t initrd_max;
1271     uint8_t header[8192], *setup, *kernel;
1272     hwaddr real_addr, prot_addr, cmdline_addr, initrd_addr = 0;
1273     FILE *f;
1274     char *vmode;
1275     MachineState *machine = MACHINE(pcms);
1276     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
1277     struct setup_data *setup_data;
1278     const char *kernel_filename = machine->kernel_filename;
1279     const char *initrd_filename = machine->initrd_filename;
1280     const char *dtb_filename = machine->dtb;
1281     const char *kernel_cmdline = machine->kernel_cmdline;
1282 
1283     /* Align to 16 bytes as a paranoia measure */
1284     cmdline_size = (strlen(kernel_cmdline)+16) & ~15;
1285 
1286     /* load the kernel header */
1287     f = fopen(kernel_filename, "rb");
1288     if (!f || !(kernel_size = get_file_size(f)) ||
1289         fread(header, 1, MIN(ARRAY_SIZE(header), kernel_size), f) !=
1290         MIN(ARRAY_SIZE(header), kernel_size)) {
1291         fprintf(stderr, "qemu: could not load kernel '%s': %s\n",
1292                 kernel_filename, strerror(errno));
1293         exit(1);
1294     }
1295 
1296     /* kernel protocol version */
1297 #if 0
1298     fprintf(stderr, "header magic: %#x\n", ldl_p(header+0x202));
1299 #endif
1300     if (ldl_p(header+0x202) == 0x53726448) {
1301         protocol = lduw_p(header+0x206);
1302     } else {
1303         /* This looks like a multiboot kernel. If it is, let's stop
1304            treating it like a Linux kernel. */
1305         if (load_multiboot(fw_cfg, f, kernel_filename, initrd_filename,
1306                            kernel_cmdline, kernel_size, header)) {
1307             return;
1308         }
1309         protocol = 0;
1310     }
1311 
1312     if (protocol < 0x200 || !(header[0x211] & 0x01)) {
1313         /* Low kernel */
1314         real_addr    = 0x90000;
1315         cmdline_addr = 0x9a000 - cmdline_size;
1316         prot_addr    = 0x10000;
1317     } else if (protocol < 0x202) {
1318         /* High but ancient kernel */
1319         real_addr    = 0x90000;
1320         cmdline_addr = 0x9a000 - cmdline_size;
1321         prot_addr    = 0x100000;
1322     } else {
1323         /* High and recent kernel */
1324         real_addr    = 0x10000;
1325         cmdline_addr = 0x20000;
1326         prot_addr    = 0x100000;
1327     }
1328 
1329 #if 0
1330     fprintf(stderr,
1331             "qemu: real_addr     = 0x" TARGET_FMT_plx "\n"
1332             "qemu: cmdline_addr  = 0x" TARGET_FMT_plx "\n"
1333             "qemu: prot_addr     = 0x" TARGET_FMT_plx "\n",
1334             real_addr,
1335             cmdline_addr,
1336             prot_addr);
1337 #endif
1338 
1339     /* highest address for loading the initrd */
1340     if (protocol >= 0x203) {
1341         initrd_max = ldl_p(header+0x22c);
1342     } else {
1343         initrd_max = 0x37ffffff;
1344     }
1345 
1346     if (initrd_max >= pcms->below_4g_mem_size - pcmc->acpi_data_size) {
1347         initrd_max = pcms->below_4g_mem_size - pcmc->acpi_data_size - 1;
1348     }
1349 
1350     fw_cfg_add_i32(fw_cfg, FW_CFG_CMDLINE_ADDR, cmdline_addr);
1351     fw_cfg_add_i32(fw_cfg, FW_CFG_CMDLINE_SIZE, strlen(kernel_cmdline)+1);
1352     fw_cfg_add_string(fw_cfg, FW_CFG_CMDLINE_DATA, kernel_cmdline);
1353 
1354     if (protocol >= 0x202) {
1355         stl_p(header+0x228, cmdline_addr);
1356     } else {
1357         stw_p(header+0x20, 0xA33F);
1358         stw_p(header+0x22, cmdline_addr-real_addr);
1359     }
1360 
1361     /* handle vga= parameter */
1362     vmode = strstr(kernel_cmdline, "vga=");
1363     if (vmode) {
1364         unsigned int video_mode;
1365         /* skip "vga=" */
1366         vmode += 4;
1367         if (!strncmp(vmode, "normal", 6)) {
1368             video_mode = 0xffff;
1369         } else if (!strncmp(vmode, "ext", 3)) {
1370             video_mode = 0xfffe;
1371         } else if (!strncmp(vmode, "ask", 3)) {
1372             video_mode = 0xfffd;
1373         } else {
1374             video_mode = strtol(vmode, NULL, 0);
1375         }
1376         stw_p(header+0x1fa, video_mode);
1377     }
1378 
1379     /* loader type */
1380     /* High nybble = B reserved for QEMU; low nybble is revision number.
1381        If this code is substantially changed, you may want to consider
1382        incrementing the revision. */
1383     if (protocol >= 0x200) {
1384         header[0x210] = 0xB0;
1385     }
1386     /* heap */
1387     if (protocol >= 0x201) {
1388         header[0x211] |= 0x80;	/* CAN_USE_HEAP */
1389         stw_p(header+0x224, cmdline_addr-real_addr-0x200);
1390     }
1391 
1392     /* load initrd */
1393     if (initrd_filename) {
1394         gsize initrd_size;
1395         gchar *initrd_data;
1396         GError *gerr = NULL;
1397 
1398         if (protocol < 0x200) {
1399             fprintf(stderr, "qemu: linux kernel too old to load a ram disk\n");
1400             exit(1);
1401         }
1402 
1403         if (!g_file_get_contents(initrd_filename, &initrd_data,
1404                                  &initrd_size, &gerr)) {
1405             fprintf(stderr, "qemu: error reading initrd %s: %s\n",
1406                     initrd_filename, gerr->message);
1407             exit(1);
1408         }
1409         if (initrd_size >= initrd_max) {
1410             fprintf(stderr, "qemu: initrd is too large, cannot support."
1411                     "(max: %"PRIu32", need %"PRId64")\n",
1412                     initrd_max, (uint64_t)initrd_size);
1413             exit(1);
1414         }
1415 
1416         initrd_addr = (initrd_max-initrd_size) & ~4095;
1417 
1418         fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_ADDR, initrd_addr);
1419         fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_SIZE, initrd_size);
1420         fw_cfg_add_bytes(fw_cfg, FW_CFG_INITRD_DATA, initrd_data, initrd_size);
1421 
1422         stl_p(header+0x218, initrd_addr);
1423         stl_p(header+0x21c, initrd_size);
1424     }
1425 
1426     /* load kernel and setup */
1427     setup_size = header[0x1f1];
1428     if (setup_size == 0) {
1429         setup_size = 4;
1430     }
1431     setup_size = (setup_size+1)*512;
1432     if (setup_size > kernel_size) {
1433         fprintf(stderr, "qemu: invalid kernel header\n");
1434         exit(1);
1435     }
1436     kernel_size -= setup_size;
1437 
1438     setup  = g_malloc(setup_size);
1439     kernel = g_malloc(kernel_size);
1440     fseek(f, 0, SEEK_SET);
1441     if (fread(setup, 1, setup_size, f) != setup_size) {
1442         fprintf(stderr, "fread() failed\n");
1443         exit(1);
1444     }
1445     if (fread(kernel, 1, kernel_size, f) != kernel_size) {
1446         fprintf(stderr, "fread() failed\n");
1447         exit(1);
1448     }
1449     fclose(f);
1450 
1451     /* append dtb to kernel */
1452     if (dtb_filename) {
1453         if (protocol < 0x209) {
1454             fprintf(stderr, "qemu: Linux kernel too old to load a dtb\n");
1455             exit(1);
1456         }
1457 
1458         dtb_size = get_image_size(dtb_filename);
1459         if (dtb_size <= 0) {
1460             fprintf(stderr, "qemu: error reading dtb %s: %s\n",
1461                     dtb_filename, strerror(errno));
1462             exit(1);
1463         }
1464 
1465         setup_data_offset = QEMU_ALIGN_UP(kernel_size, 16);
1466         kernel_size = setup_data_offset + sizeof(struct setup_data) + dtb_size;
1467         kernel = g_realloc(kernel, kernel_size);
1468 
1469         stq_p(header+0x250, prot_addr + setup_data_offset);
1470 
1471         setup_data = (struct setup_data *)(kernel + setup_data_offset);
1472         setup_data->next = 0;
1473         setup_data->type = cpu_to_le32(SETUP_DTB);
1474         setup_data->len = cpu_to_le32(dtb_size);
1475 
1476         load_image_size(dtb_filename, setup_data->data, dtb_size);
1477     }
1478 
1479     memcpy(setup, header, MIN(sizeof(header), setup_size));
1480 
1481     fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_ADDR, prot_addr);
1482     fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_SIZE, kernel_size);
1483     fw_cfg_add_bytes(fw_cfg, FW_CFG_KERNEL_DATA, kernel, kernel_size);
1484 
1485     fw_cfg_add_i32(fw_cfg, FW_CFG_SETUP_ADDR, real_addr);
1486     fw_cfg_add_i32(fw_cfg, FW_CFG_SETUP_SIZE, setup_size);
1487     fw_cfg_add_bytes(fw_cfg, FW_CFG_SETUP_DATA, setup, setup_size);
1488 
1489     option_rom[nb_option_roms].bootindex = 0;
1490     option_rom[nb_option_roms].name = "linuxboot.bin";
1491     if (pcmc->linuxboot_dma_enabled && fw_cfg_dma_enabled(fw_cfg)) {
1492         option_rom[nb_option_roms].name = "linuxboot_dma.bin";
1493     }
1494     nb_option_roms++;
1495 }
1496 
1497 #define NE2000_NB_MAX 6
1498 
1499 static const int ne2000_io[NE2000_NB_MAX] = { 0x300, 0x320, 0x340, 0x360,
1500                                               0x280, 0x380 };
1501 static const int ne2000_irq[NE2000_NB_MAX] = { 9, 10, 11, 3, 4, 5 };
1502 
1503 void pc_init_ne2k_isa(ISABus *bus, NICInfo *nd)
1504 {
1505     static int nb_ne2k = 0;
1506 
1507     if (nb_ne2k == NE2000_NB_MAX)
1508         return;
1509     isa_ne2000_init(bus, ne2000_io[nb_ne2k],
1510                     ne2000_irq[nb_ne2k], nd);
1511     nb_ne2k++;
1512 }
1513 
1514 DeviceState *cpu_get_current_apic(void)
1515 {
1516     if (current_cpu) {
1517         X86CPU *cpu = X86_CPU(current_cpu);
1518         return cpu->apic_state;
1519     } else {
1520         return NULL;
1521     }
1522 }
1523 
1524 void pc_acpi_smi_interrupt(void *opaque, int irq, int level)
1525 {
1526     X86CPU *cpu = opaque;
1527 
1528     if (level) {
1529         cpu_interrupt(CPU(cpu), CPU_INTERRUPT_SMI);
1530     }
1531 }
1532 
1533 static void pc_new_cpu(const char *typename, int64_t apic_id, Error **errp)
1534 {
1535     Object *cpu = NULL;
1536     Error *local_err = NULL;
1537 
1538     cpu = object_new(typename);
1539 
1540     object_property_set_uint(cpu, apic_id, "apic-id", &local_err);
1541     object_property_set_bool(cpu, true, "realized", &local_err);
1542 
1543     object_unref(cpu);
1544     error_propagate(errp, local_err);
1545 }
1546 
1547 void pc_hot_add_cpu(const int64_t id, Error **errp)
1548 {
1549     MachineState *ms = MACHINE(qdev_get_machine());
1550     int64_t apic_id = x86_cpu_apic_id_from_index(id);
1551     Error *local_err = NULL;
1552 
1553     if (id < 0) {
1554         error_setg(errp, "Invalid CPU id: %" PRIi64, id);
1555         return;
1556     }
1557 
1558     if (apic_id >= ACPI_CPU_HOTPLUG_ID_LIMIT) {
1559         error_setg(errp, "Unable to add CPU: %" PRIi64
1560                    ", resulting APIC ID (%" PRIi64 ") is too large",
1561                    id, apic_id);
1562         return;
1563     }
1564 
1565     pc_new_cpu(ms->cpu_type, apic_id, &local_err);
1566     if (local_err) {
1567         error_propagate(errp, local_err);
1568         return;
1569     }
1570 }
1571 
1572 void pc_cpus_init(PCMachineState *pcms)
1573 {
1574     int i;
1575     const CPUArchIdList *possible_cpus;
1576     MachineState *ms = MACHINE(pcms);
1577     MachineClass *mc = MACHINE_GET_CLASS(pcms);
1578 
1579     /* Calculates the limit to CPU APIC ID values
1580      *
1581      * Limit for the APIC ID value, so that all
1582      * CPU APIC IDs are < pcms->apic_id_limit.
1583      *
1584      * This is used for FW_CFG_MAX_CPUS. See comments on bochs_bios_init().
1585      */
1586     pcms->apic_id_limit = x86_cpu_apic_id_from_index(max_cpus - 1) + 1;
1587     possible_cpus = mc->possible_cpu_arch_ids(ms);
1588     for (i = 0; i < smp_cpus; i++) {
1589         pc_new_cpu(possible_cpus->cpus[i].type, possible_cpus->cpus[i].arch_id,
1590                    &error_fatal);
1591     }
1592 }
1593 
1594 static void pc_build_feature_control_file(PCMachineState *pcms)
1595 {
1596     MachineState *ms = MACHINE(pcms);
1597     X86CPU *cpu = X86_CPU(ms->possible_cpus->cpus[0].cpu);
1598     CPUX86State *env = &cpu->env;
1599     uint32_t unused, ecx, edx;
1600     uint64_t feature_control_bits = 0;
1601     uint64_t *val;
1602 
1603     cpu_x86_cpuid(env, 1, 0, &unused, &unused, &ecx, &edx);
1604     if (ecx & CPUID_EXT_VMX) {
1605         feature_control_bits |= FEATURE_CONTROL_VMXON_ENABLED_OUTSIDE_SMX;
1606     }
1607 
1608     if ((edx & (CPUID_EXT2_MCE | CPUID_EXT2_MCA)) ==
1609         (CPUID_EXT2_MCE | CPUID_EXT2_MCA) &&
1610         (env->mcg_cap & MCG_LMCE_P)) {
1611         feature_control_bits |= FEATURE_CONTROL_LMCE;
1612     }
1613 
1614     if (!feature_control_bits) {
1615         return;
1616     }
1617 
1618     val = g_malloc(sizeof(*val));
1619     *val = cpu_to_le64(feature_control_bits | FEATURE_CONTROL_LOCKED);
1620     fw_cfg_add_file(pcms->fw_cfg, "etc/msr_feature_control", val, sizeof(*val));
1621 }
1622 
1623 static void rtc_set_cpus_count(ISADevice *rtc, uint16_t cpus_count)
1624 {
1625     if (cpus_count > 0xff) {
1626         /* If the number of CPUs can't be represented in 8 bits, the
1627          * BIOS must use "FW_CFG_NB_CPUS". Set RTC field to 0 just
1628          * to make old BIOSes fail more predictably.
1629          */
1630         rtc_set_memory(rtc, 0x5f, 0);
1631     } else {
1632         rtc_set_memory(rtc, 0x5f, cpus_count - 1);
1633     }
1634 }
1635 
1636 static
1637 void pc_machine_done(Notifier *notifier, void *data)
1638 {
1639     PCMachineState *pcms = container_of(notifier,
1640                                         PCMachineState, machine_done);
1641     PCIBus *bus = pcms->bus;
1642 
1643     /* set the number of CPUs */
1644     rtc_set_cpus_count(pcms->rtc, pcms->boot_cpus);
1645 
1646     if (bus) {
1647         int extra_hosts = 0;
1648 
1649         QLIST_FOREACH(bus, &bus->child, sibling) {
1650             /* look for expander root buses */
1651             if (pci_bus_is_root(bus)) {
1652                 extra_hosts++;
1653             }
1654         }
1655         if (extra_hosts && pcms->fw_cfg) {
1656             uint64_t *val = g_malloc(sizeof(*val));
1657             *val = cpu_to_le64(extra_hosts);
1658             fw_cfg_add_file(pcms->fw_cfg,
1659                     "etc/extra-pci-roots", val, sizeof(*val));
1660         }
1661     }
1662 
1663     acpi_setup();
1664     if (pcms->fw_cfg) {
1665         pc_build_smbios(pcms);
1666         pc_build_feature_control_file(pcms);
1667         /* update FW_CFG_NB_CPUS to account for -device added CPUs */
1668         fw_cfg_modify_i16(pcms->fw_cfg, FW_CFG_NB_CPUS, pcms->boot_cpus);
1669     }
1670 
1671     if (pcms->apic_id_limit > 255 && !xen_enabled()) {
1672         IntelIOMMUState *iommu = INTEL_IOMMU_DEVICE(x86_iommu_get_default());
1673 
1674         if (!iommu || !x86_iommu_ir_supported(X86_IOMMU_DEVICE(iommu)) ||
1675             iommu->intr_eim != ON_OFF_AUTO_ON) {
1676             error_report("current -smp configuration requires "
1677                          "Extended Interrupt Mode enabled. "
1678                          "You can add an IOMMU using: "
1679                          "-device intel-iommu,intremap=on,eim=on");
1680             exit(EXIT_FAILURE);
1681         }
1682     }
1683 }
1684 
1685 void pc_guest_info_init(PCMachineState *pcms)
1686 {
1687     int i;
1688 
1689     pcms->apic_xrupt_override = kvm_allows_irq0_override();
1690     pcms->numa_nodes = nb_numa_nodes;
1691     pcms->node_mem = g_malloc0(pcms->numa_nodes *
1692                                     sizeof *pcms->node_mem);
1693     for (i = 0; i < nb_numa_nodes; i++) {
1694         pcms->node_mem[i] = numa_info[i].node_mem;
1695     }
1696 
1697     pcms->machine_done.notify = pc_machine_done;
1698     qemu_add_machine_init_done_notifier(&pcms->machine_done);
1699 }
1700 
1701 /* setup pci memory address space mapping into system address space */
1702 void pc_pci_as_mapping_init(Object *owner, MemoryRegion *system_memory,
1703                             MemoryRegion *pci_address_space)
1704 {
1705     /* Set to lower priority than RAM */
1706     memory_region_add_subregion_overlap(system_memory, 0x0,
1707                                         pci_address_space, -1);
1708 }
1709 
1710 void pc_acpi_init(const char *default_dsdt)
1711 {
1712     char *filename;
1713 
1714     if (acpi_tables != NULL) {
1715         /* manually set via -acpitable, leave it alone */
1716         return;
1717     }
1718 
1719     filename = qemu_find_file(QEMU_FILE_TYPE_BIOS, default_dsdt);
1720     if (filename == NULL) {
1721         warn_report("failed to find %s", default_dsdt);
1722     } else {
1723         QemuOpts *opts = qemu_opts_create(qemu_find_opts("acpi"), NULL, 0,
1724                                           &error_abort);
1725         Error *err = NULL;
1726 
1727         qemu_opt_set(opts, "file", filename, &error_abort);
1728 
1729         acpi_table_add_builtin(opts, &err);
1730         if (err) {
1731             warn_reportf_err(err, "failed to load %s: ", filename);
1732         }
1733         g_free(filename);
1734     }
1735 }
1736 
1737 void xen_load_linux(PCMachineState *pcms)
1738 {
1739     int i;
1740     FWCfgState *fw_cfg;
1741 
1742     assert(MACHINE(pcms)->kernel_filename != NULL);
1743 
1744     fw_cfg = fw_cfg_init_io(FW_CFG_IO_BASE);
1745     fw_cfg_add_i16(fw_cfg, FW_CFG_NB_CPUS, pcms->boot_cpus);
1746     rom_set_fw(fw_cfg);
1747 
1748     load_linux(pcms, fw_cfg);
1749     for (i = 0; i < nb_option_roms; i++) {
1750         assert(!strcmp(option_rom[i].name, "linuxboot.bin") ||
1751                !strcmp(option_rom[i].name, "linuxboot_dma.bin") ||
1752                !strcmp(option_rom[i].name, "multiboot.bin"));
1753         rom_add_option(option_rom[i].name, option_rom[i].bootindex);
1754     }
1755     pcms->fw_cfg = fw_cfg;
1756 }
1757 
1758 void pc_memory_init(PCMachineState *pcms,
1759                     MemoryRegion *system_memory,
1760                     MemoryRegion *rom_memory,
1761                     MemoryRegion **ram_memory)
1762 {
1763     int linux_boot, i;
1764     MemoryRegion *ram, *option_rom_mr;
1765     MemoryRegion *ram_below_4g, *ram_above_4g;
1766     FWCfgState *fw_cfg;
1767     MachineState *machine = MACHINE(pcms);
1768     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
1769 
1770     assert(machine->ram_size == pcms->below_4g_mem_size +
1771                                 pcms->above_4g_mem_size);
1772 
1773     linux_boot = (machine->kernel_filename != NULL);
1774 
1775     /* Allocate RAM.  We allocate it as a single memory region and use
1776      * aliases to address portions of it, mostly for backwards compatibility
1777      * with older qemus that used qemu_ram_alloc().
1778      */
1779     ram = g_malloc(sizeof(*ram));
1780     memory_region_allocate_system_memory(ram, NULL, "pc.ram",
1781                                          machine->ram_size);
1782     *ram_memory = ram;
1783     ram_below_4g = g_malloc(sizeof(*ram_below_4g));
1784     memory_region_init_alias(ram_below_4g, NULL, "ram-below-4g", ram,
1785                              0, pcms->below_4g_mem_size);
1786     memory_region_add_subregion(system_memory, 0, ram_below_4g);
1787     e820_add_entry(0, pcms->below_4g_mem_size, E820_RAM);
1788     if (pcms->above_4g_mem_size > 0) {
1789         ram_above_4g = g_malloc(sizeof(*ram_above_4g));
1790         memory_region_init_alias(ram_above_4g, NULL, "ram-above-4g", ram,
1791                                  pcms->below_4g_mem_size,
1792                                  pcms->above_4g_mem_size);
1793         memory_region_add_subregion(system_memory, 0x100000000ULL,
1794                                     ram_above_4g);
1795         e820_add_entry(0x100000000ULL, pcms->above_4g_mem_size, E820_RAM);
1796     }
1797 
1798     if (!pcmc->has_reserved_memory &&
1799         (machine->ram_slots ||
1800          (machine->maxram_size > machine->ram_size))) {
1801         MachineClass *mc = MACHINE_GET_CLASS(machine);
1802 
1803         error_report("\"-memory 'slots|maxmem'\" is not supported by: %s",
1804                      mc->name);
1805         exit(EXIT_FAILURE);
1806     }
1807 
1808     /* always allocate the device memory information */
1809     machine->device_memory = g_malloc0(sizeof(*machine->device_memory));
1810 
1811     /* initialize device memory address space */
1812     if (pcmc->has_reserved_memory &&
1813         (machine->ram_size < machine->maxram_size)) {
1814         ram_addr_t device_mem_size = machine->maxram_size - machine->ram_size;
1815 
1816         if (machine->ram_slots > ACPI_MAX_RAM_SLOTS) {
1817             error_report("unsupported amount of memory slots: %"PRIu64,
1818                          machine->ram_slots);
1819             exit(EXIT_FAILURE);
1820         }
1821 
1822         if (QEMU_ALIGN_UP(machine->maxram_size,
1823                           TARGET_PAGE_SIZE) != machine->maxram_size) {
1824             error_report("maximum memory size must by aligned to multiple of "
1825                          "%d bytes", TARGET_PAGE_SIZE);
1826             exit(EXIT_FAILURE);
1827         }
1828 
1829         machine->device_memory->base =
1830             ROUND_UP(0x100000000ULL + pcms->above_4g_mem_size, 1 * GiB);
1831 
1832         if (pcmc->enforce_aligned_dimm) {
1833             /* size device region assuming 1G page max alignment per slot */
1834             device_mem_size += (1 * GiB) * machine->ram_slots;
1835         }
1836 
1837         if ((machine->device_memory->base + device_mem_size) <
1838             device_mem_size) {
1839             error_report("unsupported amount of maximum memory: " RAM_ADDR_FMT,
1840                          machine->maxram_size);
1841             exit(EXIT_FAILURE);
1842         }
1843 
1844         memory_region_init(&machine->device_memory->mr, OBJECT(pcms),
1845                            "device-memory", device_mem_size);
1846         memory_region_add_subregion(system_memory, machine->device_memory->base,
1847                                     &machine->device_memory->mr);
1848     }
1849 
1850     /* Initialize PC system firmware */
1851     pc_system_firmware_init(rom_memory, !pcmc->pci_enabled);
1852 
1853     option_rom_mr = g_malloc(sizeof(*option_rom_mr));
1854     memory_region_init_ram(option_rom_mr, NULL, "pc.rom", PC_ROM_SIZE,
1855                            &error_fatal);
1856     if (pcmc->pci_enabled) {
1857         memory_region_set_readonly(option_rom_mr, true);
1858     }
1859     memory_region_add_subregion_overlap(rom_memory,
1860                                         PC_ROM_MIN_VGA,
1861                                         option_rom_mr,
1862                                         1);
1863 
1864     fw_cfg = bochs_bios_init(&address_space_memory, pcms);
1865 
1866     rom_set_fw(fw_cfg);
1867 
1868     if (pcmc->has_reserved_memory && machine->device_memory->base) {
1869         uint64_t *val = g_malloc(sizeof(*val));
1870         PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
1871         uint64_t res_mem_end = machine->device_memory->base;
1872 
1873         if (!pcmc->broken_reserved_end) {
1874             res_mem_end += memory_region_size(&machine->device_memory->mr);
1875         }
1876         *val = cpu_to_le64(ROUND_UP(res_mem_end, 1 * GiB));
1877         fw_cfg_add_file(fw_cfg, "etc/reserved-memory-end", val, sizeof(*val));
1878     }
1879 
1880     if (linux_boot) {
1881         load_linux(pcms, fw_cfg);
1882     }
1883 
1884     for (i = 0; i < nb_option_roms; i++) {
1885         rom_add_option(option_rom[i].name, option_rom[i].bootindex);
1886     }
1887     pcms->fw_cfg = fw_cfg;
1888 
1889     /* Init default IOAPIC address space */
1890     pcms->ioapic_as = &address_space_memory;
1891 }
1892 
1893 /*
1894  * The 64bit pci hole starts after "above 4G RAM" and
1895  * potentially the space reserved for memory hotplug.
1896  */
1897 uint64_t pc_pci_hole64_start(void)
1898 {
1899     PCMachineState *pcms = PC_MACHINE(qdev_get_machine());
1900     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
1901     MachineState *ms = MACHINE(pcms);
1902     uint64_t hole64_start = 0;
1903 
1904     if (pcmc->has_reserved_memory && ms->device_memory->base) {
1905         hole64_start = ms->device_memory->base;
1906         if (!pcmc->broken_reserved_end) {
1907             hole64_start += memory_region_size(&ms->device_memory->mr);
1908         }
1909     } else {
1910         hole64_start = 0x100000000ULL + pcms->above_4g_mem_size;
1911     }
1912 
1913     return ROUND_UP(hole64_start, 1 * GiB);
1914 }
1915 
1916 qemu_irq pc_allocate_cpu_irq(void)
1917 {
1918     return qemu_allocate_irq(pic_irq_request, NULL, 0);
1919 }
1920 
1921 DeviceState *pc_vga_init(ISABus *isa_bus, PCIBus *pci_bus)
1922 {
1923     DeviceState *dev = NULL;
1924 
1925     rom_set_order_override(FW_CFG_ORDER_OVERRIDE_VGA);
1926     if (pci_bus) {
1927         PCIDevice *pcidev = pci_vga_init(pci_bus);
1928         dev = pcidev ? &pcidev->qdev : NULL;
1929     } else if (isa_bus) {
1930         ISADevice *isadev = isa_vga_init(isa_bus);
1931         dev = isadev ? DEVICE(isadev) : NULL;
1932     }
1933     rom_reset_order_override();
1934     return dev;
1935 }
1936 
1937 static const MemoryRegionOps ioport80_io_ops = {
1938     .write = ioport80_write,
1939     .read = ioport80_read,
1940     .endianness = DEVICE_NATIVE_ENDIAN,
1941     .impl = {
1942         .min_access_size = 1,
1943         .max_access_size = 1,
1944     },
1945 };
1946 
1947 static const MemoryRegionOps ioportF0_io_ops = {
1948     .write = ioportF0_write,
1949     .read = ioportF0_read,
1950     .endianness = DEVICE_NATIVE_ENDIAN,
1951     .impl = {
1952         .min_access_size = 1,
1953         .max_access_size = 1,
1954     },
1955 };
1956 
1957 static void pc_superio_init(ISABus *isa_bus, bool create_fdctrl, bool no_vmport)
1958 {
1959     int i;
1960     DriveInfo *fd[MAX_FD];
1961     qemu_irq *a20_line;
1962     ISADevice *i8042, *port92, *vmmouse;
1963 
1964     serial_hds_isa_init(isa_bus, 0, MAX_ISA_SERIAL_PORTS);
1965     parallel_hds_isa_init(isa_bus, MAX_PARALLEL_PORTS);
1966 
1967     for (i = 0; i < MAX_FD; i++) {
1968         fd[i] = drive_get(IF_FLOPPY, 0, i);
1969         create_fdctrl |= !!fd[i];
1970     }
1971     if (create_fdctrl) {
1972         fdctrl_init_isa(isa_bus, fd);
1973     }
1974 
1975     i8042 = isa_create_simple(isa_bus, "i8042");
1976     if (!no_vmport) {
1977         vmport_init(isa_bus);
1978         vmmouse = isa_try_create(isa_bus, "vmmouse");
1979     } else {
1980         vmmouse = NULL;
1981     }
1982     if (vmmouse) {
1983         DeviceState *dev = DEVICE(vmmouse);
1984         qdev_prop_set_ptr(dev, "ps2_mouse", i8042);
1985         qdev_init_nofail(dev);
1986     }
1987     port92 = isa_create_simple(isa_bus, "port92");
1988 
1989     a20_line = qemu_allocate_irqs(handle_a20_line_change, first_cpu, 2);
1990     i8042_setup_a20_line(i8042, a20_line[0]);
1991     port92_init(port92, a20_line[1]);
1992     g_free(a20_line);
1993 }
1994 
1995 void pc_basic_device_init(ISABus *isa_bus, qemu_irq *gsi,
1996                           ISADevice **rtc_state,
1997                           bool create_fdctrl,
1998                           bool no_vmport,
1999                           bool has_pit,
2000                           uint32_t hpet_irqs)
2001 {
2002     int i;
2003     DeviceState *hpet = NULL;
2004     int pit_isa_irq = 0;
2005     qemu_irq pit_alt_irq = NULL;
2006     qemu_irq rtc_irq = NULL;
2007     ISADevice *pit = NULL;
2008     MemoryRegion *ioport80_io = g_new(MemoryRegion, 1);
2009     MemoryRegion *ioportF0_io = g_new(MemoryRegion, 1);
2010 
2011     memory_region_init_io(ioport80_io, NULL, &ioport80_io_ops, NULL, "ioport80", 1);
2012     memory_region_add_subregion(isa_bus->address_space_io, 0x80, ioport80_io);
2013 
2014     memory_region_init_io(ioportF0_io, NULL, &ioportF0_io_ops, NULL, "ioportF0", 1);
2015     memory_region_add_subregion(isa_bus->address_space_io, 0xf0, ioportF0_io);
2016 
2017     /*
2018      * Check if an HPET shall be created.
2019      *
2020      * Without KVM_CAP_PIT_STATE2, we cannot switch off the in-kernel PIT
2021      * when the HPET wants to take over. Thus we have to disable the latter.
2022      */
2023     if (!no_hpet && (!kvm_irqchip_in_kernel() || kvm_has_pit_state2())) {
2024         /* In order to set property, here not using sysbus_try_create_simple */
2025         hpet = qdev_try_create(NULL, TYPE_HPET);
2026         if (hpet) {
2027             /* For pc-piix-*, hpet's intcap is always IRQ2. For pc-q35-1.7
2028              * and earlier, use IRQ2 for compat. Otherwise, use IRQ16~23,
2029              * IRQ8 and IRQ2.
2030              */
2031             uint8_t compat = object_property_get_uint(OBJECT(hpet),
2032                     HPET_INTCAP, NULL);
2033             if (!compat) {
2034                 qdev_prop_set_uint32(hpet, HPET_INTCAP, hpet_irqs);
2035             }
2036             qdev_init_nofail(hpet);
2037             sysbus_mmio_map(SYS_BUS_DEVICE(hpet), 0, HPET_BASE);
2038 
2039             for (i = 0; i < GSI_NUM_PINS; i++) {
2040                 sysbus_connect_irq(SYS_BUS_DEVICE(hpet), i, gsi[i]);
2041             }
2042             pit_isa_irq = -1;
2043             pit_alt_irq = qdev_get_gpio_in(hpet, HPET_LEGACY_PIT_INT);
2044             rtc_irq = qdev_get_gpio_in(hpet, HPET_LEGACY_RTC_INT);
2045         }
2046     }
2047     *rtc_state = mc146818_rtc_init(isa_bus, 2000, rtc_irq);
2048 
2049     qemu_register_boot_set(pc_boot_set, *rtc_state);
2050 
2051     if (!xen_enabled() && has_pit) {
2052         if (kvm_pit_in_kernel()) {
2053             pit = kvm_pit_init(isa_bus, 0x40);
2054         } else {
2055             pit = i8254_pit_init(isa_bus, 0x40, pit_isa_irq, pit_alt_irq);
2056         }
2057         if (hpet) {
2058             /* connect PIT to output control line of the HPET */
2059             qdev_connect_gpio_out(hpet, 0, qdev_get_gpio_in(DEVICE(pit), 0));
2060         }
2061         pcspk_init(isa_bus, pit);
2062     }
2063 
2064     i8257_dma_init(isa_bus, 0);
2065 
2066     /* Super I/O */
2067     pc_superio_init(isa_bus, create_fdctrl, no_vmport);
2068 }
2069 
2070 void pc_nic_init(PCMachineClass *pcmc, ISABus *isa_bus, PCIBus *pci_bus)
2071 {
2072     int i;
2073 
2074     rom_set_order_override(FW_CFG_ORDER_OVERRIDE_NIC);
2075     for (i = 0; i < nb_nics; i++) {
2076         NICInfo *nd = &nd_table[i];
2077         const char *model = nd->model ? nd->model : pcmc->default_nic_model;
2078 
2079         if (g_str_equal(model, "ne2k_isa")) {
2080             pc_init_ne2k_isa(isa_bus, nd);
2081         } else {
2082             pci_nic_init_nofail(nd, pci_bus, model, NULL);
2083         }
2084     }
2085     rom_reset_order_override();
2086 }
2087 
2088 void ioapic_init_gsi(GSIState *gsi_state, const char *parent_name)
2089 {
2090     DeviceState *dev;
2091     SysBusDevice *d;
2092     unsigned int i;
2093 
2094     if (kvm_ioapic_in_kernel()) {
2095         dev = qdev_create(NULL, "kvm-ioapic");
2096     } else {
2097         dev = qdev_create(NULL, "ioapic");
2098     }
2099     if (parent_name) {
2100         object_property_add_child(object_resolve_path(parent_name, NULL),
2101                                   "ioapic", OBJECT(dev), NULL);
2102     }
2103     qdev_init_nofail(dev);
2104     d = SYS_BUS_DEVICE(dev);
2105     sysbus_mmio_map(d, 0, IO_APIC_DEFAULT_ADDRESS);
2106 
2107     for (i = 0; i < IOAPIC_NUM_PINS; i++) {
2108         gsi_state->ioapic_irq[i] = qdev_get_gpio_in(dev, i);
2109     }
2110 }
2111 
2112 static void pc_memory_pre_plug(HotplugHandler *hotplug_dev, DeviceState *dev,
2113                                Error **errp)
2114 {
2115     const PCMachineState *pcms = PC_MACHINE(hotplug_dev);
2116     const PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
2117     const bool is_nvdimm = object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM);
2118     const uint64_t legacy_align = TARGET_PAGE_SIZE;
2119 
2120     /*
2121      * When -no-acpi is used with Q35 machine type, no ACPI is built,
2122      * but pcms->acpi_dev is still created. Check !acpi_enabled in
2123      * addition to cover this case.
2124      */
2125     if (!pcms->acpi_dev || !acpi_enabled) {
2126         error_setg(errp,
2127                    "memory hotplug is not enabled: missing acpi device or acpi disabled");
2128         return;
2129     }
2130 
2131     if (is_nvdimm && !pcms->acpi_nvdimm_state.is_enabled) {
2132         error_setg(errp, "nvdimm is not enabled: missing 'nvdimm' in '-M'");
2133         return;
2134     }
2135 
2136     pc_dimm_pre_plug(PC_DIMM(dev), MACHINE(hotplug_dev),
2137                      pcmc->enforce_aligned_dimm ? NULL : &legacy_align, errp);
2138 }
2139 
2140 static void pc_memory_plug(HotplugHandler *hotplug_dev,
2141                            DeviceState *dev, Error **errp)
2142 {
2143     HotplugHandlerClass *hhc;
2144     Error *local_err = NULL;
2145     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
2146     bool is_nvdimm = object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM);
2147 
2148     pc_dimm_plug(PC_DIMM(dev), MACHINE(pcms), &local_err);
2149     if (local_err) {
2150         goto out;
2151     }
2152 
2153     if (is_nvdimm) {
2154         nvdimm_plug(&pcms->acpi_nvdimm_state);
2155     }
2156 
2157     hhc = HOTPLUG_HANDLER_GET_CLASS(pcms->acpi_dev);
2158     hhc->plug(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &error_abort);
2159 out:
2160     error_propagate(errp, local_err);
2161 }
2162 
2163 static void pc_memory_unplug_request(HotplugHandler *hotplug_dev,
2164                                      DeviceState *dev, Error **errp)
2165 {
2166     HotplugHandlerClass *hhc;
2167     Error *local_err = NULL;
2168     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
2169 
2170     /*
2171      * When -no-acpi is used with Q35 machine type, no ACPI is built,
2172      * but pcms->acpi_dev is still created. Check !acpi_enabled in
2173      * addition to cover this case.
2174      */
2175     if (!pcms->acpi_dev || !acpi_enabled) {
2176         error_setg(&local_err,
2177                    "memory hotplug is not enabled: missing acpi device or acpi disabled");
2178         goto out;
2179     }
2180 
2181     if (object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM)) {
2182         error_setg(&local_err,
2183                    "nvdimm device hot unplug is not supported yet.");
2184         goto out;
2185     }
2186 
2187     hhc = HOTPLUG_HANDLER_GET_CLASS(pcms->acpi_dev);
2188     hhc->unplug_request(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &local_err);
2189 
2190 out:
2191     error_propagate(errp, local_err);
2192 }
2193 
2194 static void pc_memory_unplug(HotplugHandler *hotplug_dev,
2195                              DeviceState *dev, Error **errp)
2196 {
2197     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
2198     HotplugHandlerClass *hhc;
2199     Error *local_err = NULL;
2200 
2201     hhc = HOTPLUG_HANDLER_GET_CLASS(pcms->acpi_dev);
2202     hhc->unplug(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &local_err);
2203 
2204     if (local_err) {
2205         goto out;
2206     }
2207 
2208     pc_dimm_unplug(PC_DIMM(dev), MACHINE(pcms));
2209     object_unparent(OBJECT(dev));
2210 
2211  out:
2212     error_propagate(errp, local_err);
2213 }
2214 
2215 static int pc_apic_cmp(const void *a, const void *b)
2216 {
2217    CPUArchId *apic_a = (CPUArchId *)a;
2218    CPUArchId *apic_b = (CPUArchId *)b;
2219 
2220    return apic_a->arch_id - apic_b->arch_id;
2221 }
2222 
2223 /* returns pointer to CPUArchId descriptor that matches CPU's apic_id
2224  * in ms->possible_cpus->cpus, if ms->possible_cpus->cpus has no
2225  * entry corresponding to CPU's apic_id returns NULL.
2226  */
2227 static CPUArchId *pc_find_cpu_slot(MachineState *ms, uint32_t id, int *idx)
2228 {
2229     CPUArchId apic_id, *found_cpu;
2230 
2231     apic_id.arch_id = id;
2232     found_cpu = bsearch(&apic_id, ms->possible_cpus->cpus,
2233         ms->possible_cpus->len, sizeof(*ms->possible_cpus->cpus),
2234         pc_apic_cmp);
2235     if (found_cpu && idx) {
2236         *idx = found_cpu - ms->possible_cpus->cpus;
2237     }
2238     return found_cpu;
2239 }
2240 
2241 static void pc_cpu_plug(HotplugHandler *hotplug_dev,
2242                         DeviceState *dev, Error **errp)
2243 {
2244     CPUArchId *found_cpu;
2245     HotplugHandlerClass *hhc;
2246     Error *local_err = NULL;
2247     X86CPU *cpu = X86_CPU(dev);
2248     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
2249 
2250     if (pcms->acpi_dev) {
2251         hhc = HOTPLUG_HANDLER_GET_CLASS(pcms->acpi_dev);
2252         hhc->plug(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &local_err);
2253         if (local_err) {
2254             goto out;
2255         }
2256     }
2257 
2258     /* increment the number of CPUs */
2259     pcms->boot_cpus++;
2260     if (pcms->rtc) {
2261         rtc_set_cpus_count(pcms->rtc, pcms->boot_cpus);
2262     }
2263     if (pcms->fw_cfg) {
2264         fw_cfg_modify_i16(pcms->fw_cfg, FW_CFG_NB_CPUS, pcms->boot_cpus);
2265     }
2266 
2267     found_cpu = pc_find_cpu_slot(MACHINE(pcms), cpu->apic_id, NULL);
2268     found_cpu->cpu = OBJECT(dev);
2269 out:
2270     error_propagate(errp, local_err);
2271 }
2272 static void pc_cpu_unplug_request_cb(HotplugHandler *hotplug_dev,
2273                                      DeviceState *dev, Error **errp)
2274 {
2275     int idx = -1;
2276     HotplugHandlerClass *hhc;
2277     Error *local_err = NULL;
2278     X86CPU *cpu = X86_CPU(dev);
2279     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
2280 
2281     if (!pcms->acpi_dev) {
2282         error_setg(&local_err, "CPU hot unplug not supported without ACPI");
2283         goto out;
2284     }
2285 
2286     pc_find_cpu_slot(MACHINE(pcms), cpu->apic_id, &idx);
2287     assert(idx != -1);
2288     if (idx == 0) {
2289         error_setg(&local_err, "Boot CPU is unpluggable");
2290         goto out;
2291     }
2292 
2293     hhc = HOTPLUG_HANDLER_GET_CLASS(pcms->acpi_dev);
2294     hhc->unplug_request(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &local_err);
2295 
2296     if (local_err) {
2297         goto out;
2298     }
2299 
2300  out:
2301     error_propagate(errp, local_err);
2302 
2303 }
2304 
2305 static void pc_cpu_unplug_cb(HotplugHandler *hotplug_dev,
2306                              DeviceState *dev, Error **errp)
2307 {
2308     CPUArchId *found_cpu;
2309     HotplugHandlerClass *hhc;
2310     Error *local_err = NULL;
2311     X86CPU *cpu = X86_CPU(dev);
2312     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
2313 
2314     hhc = HOTPLUG_HANDLER_GET_CLASS(pcms->acpi_dev);
2315     hhc->unplug(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &local_err);
2316 
2317     if (local_err) {
2318         goto out;
2319     }
2320 
2321     found_cpu = pc_find_cpu_slot(MACHINE(pcms), cpu->apic_id, NULL);
2322     found_cpu->cpu = NULL;
2323     object_unparent(OBJECT(dev));
2324 
2325     /* decrement the number of CPUs */
2326     pcms->boot_cpus--;
2327     /* Update the number of CPUs in CMOS */
2328     rtc_set_cpus_count(pcms->rtc, pcms->boot_cpus);
2329     fw_cfg_modify_i16(pcms->fw_cfg, FW_CFG_NB_CPUS, pcms->boot_cpus);
2330  out:
2331     error_propagate(errp, local_err);
2332 }
2333 
2334 static void pc_cpu_pre_plug(HotplugHandler *hotplug_dev,
2335                             DeviceState *dev, Error **errp)
2336 {
2337     int idx;
2338     CPUState *cs;
2339     CPUArchId *cpu_slot;
2340     X86CPUTopoInfo topo;
2341     X86CPU *cpu = X86_CPU(dev);
2342     MachineState *ms = MACHINE(hotplug_dev);
2343     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
2344 
2345     if(!object_dynamic_cast(OBJECT(cpu), ms->cpu_type)) {
2346         error_setg(errp, "Invalid CPU type, expected cpu type: '%s'",
2347                    ms->cpu_type);
2348         return;
2349     }
2350 
2351     /* if APIC ID is not set, set it based on socket/core/thread properties */
2352     if (cpu->apic_id == UNASSIGNED_APIC_ID) {
2353         int max_socket = (max_cpus - 1) / smp_threads / smp_cores;
2354 
2355         if (cpu->socket_id < 0) {
2356             error_setg(errp, "CPU socket-id is not set");
2357             return;
2358         } else if (cpu->socket_id > max_socket) {
2359             error_setg(errp, "Invalid CPU socket-id: %u must be in range 0:%u",
2360                        cpu->socket_id, max_socket);
2361             return;
2362         }
2363         if (cpu->core_id < 0) {
2364             error_setg(errp, "CPU core-id is not set");
2365             return;
2366         } else if (cpu->core_id > (smp_cores - 1)) {
2367             error_setg(errp, "Invalid CPU core-id: %u must be in range 0:%u",
2368                        cpu->core_id, smp_cores - 1);
2369             return;
2370         }
2371         if (cpu->thread_id < 0) {
2372             error_setg(errp, "CPU thread-id is not set");
2373             return;
2374         } else if (cpu->thread_id > (smp_threads - 1)) {
2375             error_setg(errp, "Invalid CPU thread-id: %u must be in range 0:%u",
2376                        cpu->thread_id, smp_threads - 1);
2377             return;
2378         }
2379 
2380         topo.pkg_id = cpu->socket_id;
2381         topo.core_id = cpu->core_id;
2382         topo.smt_id = cpu->thread_id;
2383         cpu->apic_id = apicid_from_topo_ids(smp_cores, smp_threads, &topo);
2384     }
2385 
2386     cpu_slot = pc_find_cpu_slot(MACHINE(pcms), cpu->apic_id, &idx);
2387     if (!cpu_slot) {
2388         MachineState *ms = MACHINE(pcms);
2389 
2390         x86_topo_ids_from_apicid(cpu->apic_id, smp_cores, smp_threads, &topo);
2391         error_setg(errp, "Invalid CPU [socket: %u, core: %u, thread: %u] with"
2392                   " APIC ID %" PRIu32 ", valid index range 0:%d",
2393                    topo.pkg_id, topo.core_id, topo.smt_id, cpu->apic_id,
2394                    ms->possible_cpus->len - 1);
2395         return;
2396     }
2397 
2398     if (cpu_slot->cpu) {
2399         error_setg(errp, "CPU[%d] with APIC ID %" PRIu32 " exists",
2400                    idx, cpu->apic_id);
2401         return;
2402     }
2403 
2404     /* if 'address' properties socket-id/core-id/thread-id are not set, set them
2405      * so that machine_query_hotpluggable_cpus would show correct values
2406      */
2407     /* TODO: move socket_id/core_id/thread_id checks into x86_cpu_realizefn()
2408      * once -smp refactoring is complete and there will be CPU private
2409      * CPUState::nr_cores and CPUState::nr_threads fields instead of globals */
2410     x86_topo_ids_from_apicid(cpu->apic_id, smp_cores, smp_threads, &topo);
2411     if (cpu->socket_id != -1 && cpu->socket_id != topo.pkg_id) {
2412         error_setg(errp, "property socket-id: %u doesn't match set apic-id:"
2413             " 0x%x (socket-id: %u)", cpu->socket_id, cpu->apic_id, topo.pkg_id);
2414         return;
2415     }
2416     cpu->socket_id = topo.pkg_id;
2417 
2418     if (cpu->core_id != -1 && cpu->core_id != topo.core_id) {
2419         error_setg(errp, "property core-id: %u doesn't match set apic-id:"
2420             " 0x%x (core-id: %u)", cpu->core_id, cpu->apic_id, topo.core_id);
2421         return;
2422     }
2423     cpu->core_id = topo.core_id;
2424 
2425     if (cpu->thread_id != -1 && cpu->thread_id != topo.smt_id) {
2426         error_setg(errp, "property thread-id: %u doesn't match set apic-id:"
2427             " 0x%x (thread-id: %u)", cpu->thread_id, cpu->apic_id, topo.smt_id);
2428         return;
2429     }
2430     cpu->thread_id = topo.smt_id;
2431 
2432     if (cpu->hyperv_vpindex && !kvm_hv_vpindex_settable()) {
2433         error_setg(errp, "kernel doesn't allow setting HyperV VP_INDEX");
2434         return;
2435     }
2436 
2437     cs = CPU(cpu);
2438     cs->cpu_index = idx;
2439 
2440     numa_cpu_pre_plug(cpu_slot, dev, errp);
2441 }
2442 
2443 static void pc_machine_device_pre_plug_cb(HotplugHandler *hotplug_dev,
2444                                           DeviceState *dev, Error **errp)
2445 {
2446     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2447         pc_memory_pre_plug(hotplug_dev, dev, errp);
2448     } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
2449         pc_cpu_pre_plug(hotplug_dev, dev, errp);
2450     }
2451 }
2452 
2453 static void pc_machine_device_plug_cb(HotplugHandler *hotplug_dev,
2454                                       DeviceState *dev, Error **errp)
2455 {
2456     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2457         pc_memory_plug(hotplug_dev, dev, errp);
2458     } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
2459         pc_cpu_plug(hotplug_dev, dev, errp);
2460     }
2461 }
2462 
2463 static void pc_machine_device_unplug_request_cb(HotplugHandler *hotplug_dev,
2464                                                 DeviceState *dev, Error **errp)
2465 {
2466     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2467         pc_memory_unplug_request(hotplug_dev, dev, errp);
2468     } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
2469         pc_cpu_unplug_request_cb(hotplug_dev, dev, errp);
2470     } else {
2471         error_setg(errp, "acpi: device unplug request for not supported device"
2472                    " type: %s", object_get_typename(OBJECT(dev)));
2473     }
2474 }
2475 
2476 static void pc_machine_device_unplug_cb(HotplugHandler *hotplug_dev,
2477                                         DeviceState *dev, Error **errp)
2478 {
2479     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2480         pc_memory_unplug(hotplug_dev, dev, errp);
2481     } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
2482         pc_cpu_unplug_cb(hotplug_dev, dev, errp);
2483     } else {
2484         error_setg(errp, "acpi: device unplug for not supported device"
2485                    " type: %s", object_get_typename(OBJECT(dev)));
2486     }
2487 }
2488 
2489 static HotplugHandler *pc_get_hotpug_handler(MachineState *machine,
2490                                              DeviceState *dev)
2491 {
2492     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM) ||
2493         object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
2494         return HOTPLUG_HANDLER(machine);
2495     }
2496 
2497     return NULL;
2498 }
2499 
2500 static void
2501 pc_machine_get_device_memory_region_size(Object *obj, Visitor *v,
2502                                          const char *name, void *opaque,
2503                                          Error **errp)
2504 {
2505     MachineState *ms = MACHINE(obj);
2506     int64_t value = memory_region_size(&ms->device_memory->mr);
2507 
2508     visit_type_int(v, name, &value, errp);
2509 }
2510 
2511 static void pc_machine_get_max_ram_below_4g(Object *obj, Visitor *v,
2512                                             const char *name, void *opaque,
2513                                             Error **errp)
2514 {
2515     PCMachineState *pcms = PC_MACHINE(obj);
2516     uint64_t value = pcms->max_ram_below_4g;
2517 
2518     visit_type_size(v, name, &value, errp);
2519 }
2520 
2521 static void pc_machine_set_max_ram_below_4g(Object *obj, Visitor *v,
2522                                             const char *name, void *opaque,
2523                                             Error **errp)
2524 {
2525     PCMachineState *pcms = PC_MACHINE(obj);
2526     Error *error = NULL;
2527     uint64_t value;
2528 
2529     visit_type_size(v, name, &value, &error);
2530     if (error) {
2531         error_propagate(errp, error);
2532         return;
2533     }
2534     if (value > 4 * GiB) {
2535         error_setg(&error,
2536                    "Machine option 'max-ram-below-4g=%"PRIu64
2537                    "' expects size less than or equal to 4G", value);
2538         error_propagate(errp, error);
2539         return;
2540     }
2541 
2542     if (value < 1 * MiB) {
2543         warn_report("Only %" PRIu64 " bytes of RAM below the 4GiB boundary,"
2544                     "BIOS may not work with less than 1MiB", value);
2545     }
2546 
2547     pcms->max_ram_below_4g = value;
2548 }
2549 
2550 static void pc_machine_get_vmport(Object *obj, Visitor *v, const char *name,
2551                                   void *opaque, Error **errp)
2552 {
2553     PCMachineState *pcms = PC_MACHINE(obj);
2554     OnOffAuto vmport = pcms->vmport;
2555 
2556     visit_type_OnOffAuto(v, name, &vmport, errp);
2557 }
2558 
2559 static void pc_machine_set_vmport(Object *obj, Visitor *v, const char *name,
2560                                   void *opaque, Error **errp)
2561 {
2562     PCMachineState *pcms = PC_MACHINE(obj);
2563 
2564     visit_type_OnOffAuto(v, name, &pcms->vmport, errp);
2565 }
2566 
2567 bool pc_machine_is_smm_enabled(PCMachineState *pcms)
2568 {
2569     bool smm_available = false;
2570 
2571     if (pcms->smm == ON_OFF_AUTO_OFF) {
2572         return false;
2573     }
2574 
2575     if (tcg_enabled() || qtest_enabled()) {
2576         smm_available = true;
2577     } else if (kvm_enabled()) {
2578         smm_available = kvm_has_smm();
2579     }
2580 
2581     if (smm_available) {
2582         return true;
2583     }
2584 
2585     if (pcms->smm == ON_OFF_AUTO_ON) {
2586         error_report("System Management Mode not supported by this hypervisor.");
2587         exit(1);
2588     }
2589     return false;
2590 }
2591 
2592 static void pc_machine_get_smm(Object *obj, Visitor *v, const char *name,
2593                                void *opaque, Error **errp)
2594 {
2595     PCMachineState *pcms = PC_MACHINE(obj);
2596     OnOffAuto smm = pcms->smm;
2597 
2598     visit_type_OnOffAuto(v, name, &smm, errp);
2599 }
2600 
2601 static void pc_machine_set_smm(Object *obj, Visitor *v, const char *name,
2602                                void *opaque, Error **errp)
2603 {
2604     PCMachineState *pcms = PC_MACHINE(obj);
2605 
2606     visit_type_OnOffAuto(v, name, &pcms->smm, errp);
2607 }
2608 
2609 static bool pc_machine_get_nvdimm(Object *obj, Error **errp)
2610 {
2611     PCMachineState *pcms = PC_MACHINE(obj);
2612 
2613     return pcms->acpi_nvdimm_state.is_enabled;
2614 }
2615 
2616 static void pc_machine_set_nvdimm(Object *obj, bool value, Error **errp)
2617 {
2618     PCMachineState *pcms = PC_MACHINE(obj);
2619 
2620     pcms->acpi_nvdimm_state.is_enabled = value;
2621 }
2622 
2623 static char *pc_machine_get_nvdimm_persistence(Object *obj, Error **errp)
2624 {
2625     PCMachineState *pcms = PC_MACHINE(obj);
2626 
2627     return g_strdup(pcms->acpi_nvdimm_state.persistence_string);
2628 }
2629 
2630 static void pc_machine_set_nvdimm_persistence(Object *obj, const char *value,
2631                                                Error **errp)
2632 {
2633     PCMachineState *pcms = PC_MACHINE(obj);
2634     AcpiNVDIMMState *nvdimm_state = &pcms->acpi_nvdimm_state;
2635 
2636     if (strcmp(value, "cpu") == 0)
2637         nvdimm_state->persistence = 3;
2638     else if (strcmp(value, "mem-ctrl") == 0)
2639         nvdimm_state->persistence = 2;
2640     else {
2641         error_setg(errp, "-machine nvdimm-persistence=%s: unsupported option",
2642                    value);
2643         return;
2644     }
2645 
2646     g_free(nvdimm_state->persistence_string);
2647     nvdimm_state->persistence_string = g_strdup(value);
2648 }
2649 
2650 static bool pc_machine_get_smbus(Object *obj, Error **errp)
2651 {
2652     PCMachineState *pcms = PC_MACHINE(obj);
2653 
2654     return pcms->smbus_enabled;
2655 }
2656 
2657 static void pc_machine_set_smbus(Object *obj, bool value, Error **errp)
2658 {
2659     PCMachineState *pcms = PC_MACHINE(obj);
2660 
2661     pcms->smbus_enabled = value;
2662 }
2663 
2664 static bool pc_machine_get_sata(Object *obj, Error **errp)
2665 {
2666     PCMachineState *pcms = PC_MACHINE(obj);
2667 
2668     return pcms->sata_enabled;
2669 }
2670 
2671 static void pc_machine_set_sata(Object *obj, bool value, Error **errp)
2672 {
2673     PCMachineState *pcms = PC_MACHINE(obj);
2674 
2675     pcms->sata_enabled = value;
2676 }
2677 
2678 static bool pc_machine_get_pit(Object *obj, Error **errp)
2679 {
2680     PCMachineState *pcms = PC_MACHINE(obj);
2681 
2682     return pcms->pit_enabled;
2683 }
2684 
2685 static void pc_machine_set_pit(Object *obj, bool value, Error **errp)
2686 {
2687     PCMachineState *pcms = PC_MACHINE(obj);
2688 
2689     pcms->pit_enabled = value;
2690 }
2691 
2692 static void pc_machine_initfn(Object *obj)
2693 {
2694     PCMachineState *pcms = PC_MACHINE(obj);
2695 
2696     pcms->max_ram_below_4g = 0; /* use default */
2697     pcms->smm = ON_OFF_AUTO_AUTO;
2698     pcms->vmport = ON_OFF_AUTO_AUTO;
2699     /* nvdimm is disabled on default. */
2700     pcms->acpi_nvdimm_state.is_enabled = false;
2701     /* acpi build is enabled by default if machine supports it */
2702     pcms->acpi_build_enabled = PC_MACHINE_GET_CLASS(pcms)->has_acpi_build;
2703     pcms->smbus_enabled = true;
2704     pcms->sata_enabled = true;
2705     pcms->pit_enabled = true;
2706 }
2707 
2708 static void pc_machine_reset(void)
2709 {
2710     CPUState *cs;
2711     X86CPU *cpu;
2712 
2713     qemu_devices_reset();
2714 
2715     /* Reset APIC after devices have been reset to cancel
2716      * any changes that qemu_devices_reset() might have done.
2717      */
2718     CPU_FOREACH(cs) {
2719         cpu = X86_CPU(cs);
2720 
2721         if (cpu->apic_state) {
2722             device_reset(cpu->apic_state);
2723         }
2724     }
2725 }
2726 
2727 static CpuInstanceProperties
2728 pc_cpu_index_to_props(MachineState *ms, unsigned cpu_index)
2729 {
2730     MachineClass *mc = MACHINE_GET_CLASS(ms);
2731     const CPUArchIdList *possible_cpus = mc->possible_cpu_arch_ids(ms);
2732 
2733     assert(cpu_index < possible_cpus->len);
2734     return possible_cpus->cpus[cpu_index].props;
2735 }
2736 
2737 static int64_t pc_get_default_cpu_node_id(const MachineState *ms, int idx)
2738 {
2739    X86CPUTopoInfo topo;
2740 
2741    assert(idx < ms->possible_cpus->len);
2742    x86_topo_ids_from_apicid(ms->possible_cpus->cpus[idx].arch_id,
2743                             smp_cores, smp_threads, &topo);
2744    return topo.pkg_id % nb_numa_nodes;
2745 }
2746 
2747 static const CPUArchIdList *pc_possible_cpu_arch_ids(MachineState *ms)
2748 {
2749     int i;
2750 
2751     if (ms->possible_cpus) {
2752         /*
2753          * make sure that max_cpus hasn't changed since the first use, i.e.
2754          * -smp hasn't been parsed after it
2755         */
2756         assert(ms->possible_cpus->len == max_cpus);
2757         return ms->possible_cpus;
2758     }
2759 
2760     ms->possible_cpus = g_malloc0(sizeof(CPUArchIdList) +
2761                                   sizeof(CPUArchId) * max_cpus);
2762     ms->possible_cpus->len = max_cpus;
2763     for (i = 0; i < ms->possible_cpus->len; i++) {
2764         X86CPUTopoInfo topo;
2765 
2766         ms->possible_cpus->cpus[i].type = ms->cpu_type;
2767         ms->possible_cpus->cpus[i].vcpus_count = 1;
2768         ms->possible_cpus->cpus[i].arch_id = x86_cpu_apic_id_from_index(i);
2769         x86_topo_ids_from_apicid(ms->possible_cpus->cpus[i].arch_id,
2770                                  smp_cores, smp_threads, &topo);
2771         ms->possible_cpus->cpus[i].props.has_socket_id = true;
2772         ms->possible_cpus->cpus[i].props.socket_id = topo.pkg_id;
2773         ms->possible_cpus->cpus[i].props.has_core_id = true;
2774         ms->possible_cpus->cpus[i].props.core_id = topo.core_id;
2775         ms->possible_cpus->cpus[i].props.has_thread_id = true;
2776         ms->possible_cpus->cpus[i].props.thread_id = topo.smt_id;
2777     }
2778     return ms->possible_cpus;
2779 }
2780 
2781 static void x86_nmi(NMIState *n, int cpu_index, Error **errp)
2782 {
2783     /* cpu index isn't used */
2784     CPUState *cs;
2785 
2786     CPU_FOREACH(cs) {
2787         X86CPU *cpu = X86_CPU(cs);
2788 
2789         if (!cpu->apic_state) {
2790             cpu_interrupt(cs, CPU_INTERRUPT_NMI);
2791         } else {
2792             apic_deliver_nmi(cpu->apic_state);
2793         }
2794     }
2795 }
2796 
2797 static void pc_machine_class_init(ObjectClass *oc, void *data)
2798 {
2799     MachineClass *mc = MACHINE_CLASS(oc);
2800     PCMachineClass *pcmc = PC_MACHINE_CLASS(oc);
2801     HotplugHandlerClass *hc = HOTPLUG_HANDLER_CLASS(oc);
2802     NMIClass *nc = NMI_CLASS(oc);
2803 
2804     pcmc->pci_enabled = true;
2805     pcmc->has_acpi_build = true;
2806     pcmc->rsdp_in_ram = true;
2807     pcmc->smbios_defaults = true;
2808     pcmc->smbios_uuid_encoded = true;
2809     pcmc->gigabyte_align = true;
2810     pcmc->has_reserved_memory = true;
2811     pcmc->kvmclock_enabled = true;
2812     pcmc->enforce_aligned_dimm = true;
2813     /* BIOS ACPI tables: 128K. Other BIOS datastructures: less than 4K reported
2814      * to be used at the moment, 32K should be enough for a while.  */
2815     pcmc->acpi_data_size = 0x20000 + 0x8000;
2816     pcmc->save_tsc_khz = true;
2817     pcmc->linuxboot_dma_enabled = true;
2818     assert(!mc->get_hotplug_handler);
2819     mc->get_hotplug_handler = pc_get_hotpug_handler;
2820     mc->cpu_index_to_instance_props = pc_cpu_index_to_props;
2821     mc->get_default_cpu_node_id = pc_get_default_cpu_node_id;
2822     mc->possible_cpu_arch_ids = pc_possible_cpu_arch_ids;
2823     mc->auto_enable_numa_with_memhp = true;
2824     mc->has_hotpluggable_cpus = true;
2825     mc->default_boot_order = "cad";
2826     mc->hot_add_cpu = pc_hot_add_cpu;
2827     mc->block_default_type = IF_IDE;
2828     mc->max_cpus = 255;
2829     mc->reset = pc_machine_reset;
2830     hc->pre_plug = pc_machine_device_pre_plug_cb;
2831     hc->plug = pc_machine_device_plug_cb;
2832     hc->unplug_request = pc_machine_device_unplug_request_cb;
2833     hc->unplug = pc_machine_device_unplug_cb;
2834     nc->nmi_monitor_handler = x86_nmi;
2835     mc->default_cpu_type = TARGET_DEFAULT_CPU_TYPE;
2836 
2837     object_class_property_add(oc, PC_MACHINE_DEVMEM_REGION_SIZE, "int",
2838         pc_machine_get_device_memory_region_size, NULL,
2839         NULL, NULL, &error_abort);
2840 
2841     object_class_property_add(oc, PC_MACHINE_MAX_RAM_BELOW_4G, "size",
2842         pc_machine_get_max_ram_below_4g, pc_machine_set_max_ram_below_4g,
2843         NULL, NULL, &error_abort);
2844 
2845     object_class_property_set_description(oc, PC_MACHINE_MAX_RAM_BELOW_4G,
2846         "Maximum ram below the 4G boundary (32bit boundary)", &error_abort);
2847 
2848     object_class_property_add(oc, PC_MACHINE_SMM, "OnOffAuto",
2849         pc_machine_get_smm, pc_machine_set_smm,
2850         NULL, NULL, &error_abort);
2851     object_class_property_set_description(oc, PC_MACHINE_SMM,
2852         "Enable SMM (pc & q35)", &error_abort);
2853 
2854     object_class_property_add(oc, PC_MACHINE_VMPORT, "OnOffAuto",
2855         pc_machine_get_vmport, pc_machine_set_vmport,
2856         NULL, NULL, &error_abort);
2857     object_class_property_set_description(oc, PC_MACHINE_VMPORT,
2858         "Enable vmport (pc & q35)", &error_abort);
2859 
2860     object_class_property_add_bool(oc, PC_MACHINE_NVDIMM,
2861         pc_machine_get_nvdimm, pc_machine_set_nvdimm, &error_abort);
2862 
2863     object_class_property_add_str(oc, PC_MACHINE_NVDIMM_PERSIST,
2864         pc_machine_get_nvdimm_persistence,
2865         pc_machine_set_nvdimm_persistence, &error_abort);
2866 
2867     object_class_property_add_bool(oc, PC_MACHINE_SMBUS,
2868         pc_machine_get_smbus, pc_machine_set_smbus, &error_abort);
2869 
2870     object_class_property_add_bool(oc, PC_MACHINE_SATA,
2871         pc_machine_get_sata, pc_machine_set_sata, &error_abort);
2872 
2873     object_class_property_add_bool(oc, PC_MACHINE_PIT,
2874         pc_machine_get_pit, pc_machine_set_pit, &error_abort);
2875 }
2876 
2877 static const TypeInfo pc_machine_info = {
2878     .name = TYPE_PC_MACHINE,
2879     .parent = TYPE_MACHINE,
2880     .abstract = true,
2881     .instance_size = sizeof(PCMachineState),
2882     .instance_init = pc_machine_initfn,
2883     .class_size = sizeof(PCMachineClass),
2884     .class_init = pc_machine_class_init,
2885     .interfaces = (InterfaceInfo[]) {
2886          { TYPE_HOTPLUG_HANDLER },
2887          { TYPE_NMI },
2888          { }
2889     },
2890 };
2891 
2892 static void pc_machine_register_types(void)
2893 {
2894     type_register_static(&pc_machine_info);
2895 }
2896 
2897 type_init(pc_machine_register_types)
2898