xref: /openbmc/qemu/hw/i386/pc.c (revision 8995dd9009a4f0845393c102098dc94e14806c37)
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 void gsi_handler(void *opaque, int n, int level)
445 {
446     GSIState *s = opaque;
447 
448     DPRINTF("pc: %s GSI %d\n", level ? "raising" : "lowering", n);
449     if (n < ISA_NUM_IRQS) {
450         qemu_set_irq(s->i8259_irq[n], level);
451     }
452     qemu_set_irq(s->ioapic_irq[n], level);
453 }
454 
455 static void ioport80_write(void *opaque, hwaddr addr, uint64_t data,
456                            unsigned size)
457 {
458 }
459 
460 static uint64_t ioport80_read(void *opaque, hwaddr addr, unsigned size)
461 {
462     return 0xffffffffffffffffULL;
463 }
464 
465 /* MSDOS compatibility mode FPU exception support */
466 static qemu_irq ferr_irq;
467 
468 void pc_register_ferr_irq(qemu_irq irq)
469 {
470     ferr_irq = irq;
471 }
472 
473 /* XXX: add IGNNE support */
474 void cpu_set_ferr(CPUX86State *s)
475 {
476     qemu_irq_raise(ferr_irq);
477 }
478 
479 static void ioportF0_write(void *opaque, hwaddr addr, uint64_t data,
480                            unsigned size)
481 {
482     qemu_irq_lower(ferr_irq);
483 }
484 
485 static uint64_t ioportF0_read(void *opaque, hwaddr addr, unsigned size)
486 {
487     return 0xffffffffffffffffULL;
488 }
489 
490 /* TSC handling */
491 uint64_t cpu_get_tsc(CPUX86State *env)
492 {
493     return cpu_get_ticks();
494 }
495 
496 /* IRQ handling */
497 int cpu_get_pic_interrupt(CPUX86State *env)
498 {
499     X86CPU *cpu = x86_env_get_cpu(env);
500     int intno;
501 
502     if (!kvm_irqchip_in_kernel()) {
503         intno = apic_get_interrupt(cpu->apic_state);
504         if (intno >= 0) {
505             return intno;
506         }
507         /* read the irq from the PIC */
508         if (!apic_accept_pic_intr(cpu->apic_state)) {
509             return -1;
510         }
511     }
512 
513     intno = pic_read_irq(isa_pic);
514     return intno;
515 }
516 
517 static void pic_irq_request(void *opaque, int irq, int level)
518 {
519     CPUState *cs = first_cpu;
520     X86CPU *cpu = X86_CPU(cs);
521 
522     DPRINTF("pic_irqs: %s irq %d\n", level? "raise" : "lower", irq);
523     if (cpu->apic_state && !kvm_irqchip_in_kernel()) {
524         CPU_FOREACH(cs) {
525             cpu = X86_CPU(cs);
526             if (apic_accept_pic_intr(cpu->apic_state)) {
527                 apic_deliver_pic_intr(cpu->apic_state, level);
528             }
529         }
530     } else {
531         if (level) {
532             cpu_interrupt(cs, CPU_INTERRUPT_HARD);
533         } else {
534             cpu_reset_interrupt(cs, CPU_INTERRUPT_HARD);
535         }
536     }
537 }
538 
539 /* PC cmos mappings */
540 
541 #define REG_EQUIPMENT_BYTE          0x14
542 
543 int cmos_get_fd_drive_type(FloppyDriveType fd0)
544 {
545     int val;
546 
547     switch (fd0) {
548     case FLOPPY_DRIVE_TYPE_144:
549         /* 1.44 Mb 3"5 drive */
550         val = 4;
551         break;
552     case FLOPPY_DRIVE_TYPE_288:
553         /* 2.88 Mb 3"5 drive */
554         val = 5;
555         break;
556     case FLOPPY_DRIVE_TYPE_120:
557         /* 1.2 Mb 5"5 drive */
558         val = 2;
559         break;
560     case FLOPPY_DRIVE_TYPE_NONE:
561     default:
562         val = 0;
563         break;
564     }
565     return val;
566 }
567 
568 static void cmos_init_hd(ISADevice *s, int type_ofs, int info_ofs,
569                          int16_t cylinders, int8_t heads, int8_t sectors)
570 {
571     rtc_set_memory(s, type_ofs, 47);
572     rtc_set_memory(s, info_ofs, cylinders);
573     rtc_set_memory(s, info_ofs + 1, cylinders >> 8);
574     rtc_set_memory(s, info_ofs + 2, heads);
575     rtc_set_memory(s, info_ofs + 3, 0xff);
576     rtc_set_memory(s, info_ofs + 4, 0xff);
577     rtc_set_memory(s, info_ofs + 5, 0xc0 | ((heads > 8) << 3));
578     rtc_set_memory(s, info_ofs + 6, cylinders);
579     rtc_set_memory(s, info_ofs + 7, cylinders >> 8);
580     rtc_set_memory(s, info_ofs + 8, sectors);
581 }
582 
583 /* convert boot_device letter to something recognizable by the bios */
584 static int boot_device2nibble(char boot_device)
585 {
586     switch(boot_device) {
587     case 'a':
588     case 'b':
589         return 0x01; /* floppy boot */
590     case 'c':
591         return 0x02; /* hard drive boot */
592     case 'd':
593         return 0x03; /* CD-ROM boot */
594     case 'n':
595         return 0x04; /* Network boot */
596     }
597     return 0;
598 }
599 
600 static void set_boot_dev(ISADevice *s, const char *boot_device, Error **errp)
601 {
602 #define PC_MAX_BOOT_DEVICES 3
603     int nbds, bds[3] = { 0, };
604     int i;
605 
606     nbds = strlen(boot_device);
607     if (nbds > PC_MAX_BOOT_DEVICES) {
608         error_setg(errp, "Too many boot devices for PC");
609         return;
610     }
611     for (i = 0; i < nbds; i++) {
612         bds[i] = boot_device2nibble(boot_device[i]);
613         if (bds[i] == 0) {
614             error_setg(errp, "Invalid boot device for PC: '%c'",
615                        boot_device[i]);
616             return;
617         }
618     }
619     rtc_set_memory(s, 0x3d, (bds[1] << 4) | bds[0]);
620     rtc_set_memory(s, 0x38, (bds[2] << 4) | (fd_bootchk ? 0x0 : 0x1));
621 }
622 
623 static void pc_boot_set(void *opaque, const char *boot_device, Error **errp)
624 {
625     set_boot_dev(opaque, boot_device, errp);
626 }
627 
628 static void pc_cmos_init_floppy(ISADevice *rtc_state, ISADevice *floppy)
629 {
630     int val, nb, i;
631     FloppyDriveType fd_type[2] = { FLOPPY_DRIVE_TYPE_NONE,
632                                    FLOPPY_DRIVE_TYPE_NONE };
633 
634     /* floppy type */
635     if (floppy) {
636         for (i = 0; i < 2; i++) {
637             fd_type[i] = isa_fdc_get_drive_type(floppy, i);
638         }
639     }
640     val = (cmos_get_fd_drive_type(fd_type[0]) << 4) |
641         cmos_get_fd_drive_type(fd_type[1]);
642     rtc_set_memory(rtc_state, 0x10, val);
643 
644     val = rtc_get_memory(rtc_state, REG_EQUIPMENT_BYTE);
645     nb = 0;
646     if (fd_type[0] != FLOPPY_DRIVE_TYPE_NONE) {
647         nb++;
648     }
649     if (fd_type[1] != FLOPPY_DRIVE_TYPE_NONE) {
650         nb++;
651     }
652     switch (nb) {
653     case 0:
654         break;
655     case 1:
656         val |= 0x01; /* 1 drive, ready for boot */
657         break;
658     case 2:
659         val |= 0x41; /* 2 drives, ready for boot */
660         break;
661     }
662     rtc_set_memory(rtc_state, REG_EQUIPMENT_BYTE, val);
663 }
664 
665 typedef struct pc_cmos_init_late_arg {
666     ISADevice *rtc_state;
667     BusState *idebus[2];
668 } pc_cmos_init_late_arg;
669 
670 typedef struct check_fdc_state {
671     ISADevice *floppy;
672     bool multiple;
673 } CheckFdcState;
674 
675 static int check_fdc(Object *obj, void *opaque)
676 {
677     CheckFdcState *state = opaque;
678     Object *fdc;
679     uint32_t iobase;
680     Error *local_err = NULL;
681 
682     fdc = object_dynamic_cast(obj, TYPE_ISA_FDC);
683     if (!fdc) {
684         return 0;
685     }
686 
687     iobase = object_property_get_uint(obj, "iobase", &local_err);
688     if (local_err || iobase != 0x3f0) {
689         error_free(local_err);
690         return 0;
691     }
692 
693     if (state->floppy) {
694         state->multiple = true;
695     } else {
696         state->floppy = ISA_DEVICE(obj);
697     }
698     return 0;
699 }
700 
701 static const char * const fdc_container_path[] = {
702     "/unattached", "/peripheral", "/peripheral-anon"
703 };
704 
705 /*
706  * Locate the FDC at IO address 0x3f0, in order to configure the CMOS registers
707  * and ACPI objects.
708  */
709 ISADevice *pc_find_fdc0(void)
710 {
711     int i;
712     Object *container;
713     CheckFdcState state = { 0 };
714 
715     for (i = 0; i < ARRAY_SIZE(fdc_container_path); i++) {
716         container = container_get(qdev_get_machine(), fdc_container_path[i]);
717         object_child_foreach(container, check_fdc, &state);
718     }
719 
720     if (state.multiple) {
721         warn_report("multiple floppy disk controllers with "
722                     "iobase=0x3f0 have been found");
723         error_printf("the one being picked for CMOS setup might not reflect "
724                      "your intent");
725     }
726 
727     return state.floppy;
728 }
729 
730 static void pc_cmos_init_late(void *opaque)
731 {
732     pc_cmos_init_late_arg *arg = opaque;
733     ISADevice *s = arg->rtc_state;
734     int16_t cylinders;
735     int8_t heads, sectors;
736     int val;
737     int i, trans;
738 
739     val = 0;
740     if (arg->idebus[0] && ide_get_geometry(arg->idebus[0], 0,
741                                            &cylinders, &heads, &sectors) >= 0) {
742         cmos_init_hd(s, 0x19, 0x1b, cylinders, heads, sectors);
743         val |= 0xf0;
744     }
745     if (arg->idebus[0] && ide_get_geometry(arg->idebus[0], 1,
746                                            &cylinders, &heads, &sectors) >= 0) {
747         cmos_init_hd(s, 0x1a, 0x24, cylinders, heads, sectors);
748         val |= 0x0f;
749     }
750     rtc_set_memory(s, 0x12, val);
751 
752     val = 0;
753     for (i = 0; i < 4; i++) {
754         /* NOTE: ide_get_geometry() returns the physical
755            geometry.  It is always such that: 1 <= sects <= 63, 1
756            <= heads <= 16, 1 <= cylinders <= 16383. The BIOS
757            geometry can be different if a translation is done. */
758         if (arg->idebus[i / 2] &&
759             ide_get_geometry(arg->idebus[i / 2], i % 2,
760                              &cylinders, &heads, &sectors) >= 0) {
761             trans = ide_get_bios_chs_trans(arg->idebus[i / 2], i % 2) - 1;
762             assert((trans & ~3) == 0);
763             val |= trans << (i * 2);
764         }
765     }
766     rtc_set_memory(s, 0x39, val);
767 
768     pc_cmos_init_floppy(s, pc_find_fdc0());
769 
770     qemu_unregister_reset(pc_cmos_init_late, opaque);
771 }
772 
773 void pc_cmos_init(PCMachineState *pcms,
774                   BusState *idebus0, BusState *idebus1,
775                   ISADevice *s)
776 {
777     int val;
778     static pc_cmos_init_late_arg arg;
779 
780     /* various important CMOS locations needed by PC/Bochs bios */
781 
782     /* memory size */
783     /* base memory (first MiB) */
784     val = MIN(pcms->below_4g_mem_size / KiB, 640);
785     rtc_set_memory(s, 0x15, val);
786     rtc_set_memory(s, 0x16, val >> 8);
787     /* extended memory (next 64MiB) */
788     if (pcms->below_4g_mem_size > 1 * MiB) {
789         val = (pcms->below_4g_mem_size - 1 * MiB) / KiB;
790     } else {
791         val = 0;
792     }
793     if (val > 65535)
794         val = 65535;
795     rtc_set_memory(s, 0x17, val);
796     rtc_set_memory(s, 0x18, val >> 8);
797     rtc_set_memory(s, 0x30, val);
798     rtc_set_memory(s, 0x31, val >> 8);
799     /* memory between 16MiB and 4GiB */
800     if (pcms->below_4g_mem_size > 16 * MiB) {
801         val = (pcms->below_4g_mem_size - 16 * MiB) / (64 * KiB);
802     } else {
803         val = 0;
804     }
805     if (val > 65535)
806         val = 65535;
807     rtc_set_memory(s, 0x34, val);
808     rtc_set_memory(s, 0x35, val >> 8);
809     /* memory above 4GiB */
810     val = pcms->above_4g_mem_size / 65536;
811     rtc_set_memory(s, 0x5b, val);
812     rtc_set_memory(s, 0x5c, val >> 8);
813     rtc_set_memory(s, 0x5d, val >> 16);
814 
815     object_property_add_link(OBJECT(pcms), "rtc_state",
816                              TYPE_ISA_DEVICE,
817                              (Object **)&pcms->rtc,
818                              object_property_allow_set_link,
819                              OBJ_PROP_LINK_STRONG, &error_abort);
820     object_property_set_link(OBJECT(pcms), OBJECT(s),
821                              "rtc_state", &error_abort);
822 
823     set_boot_dev(s, MACHINE(pcms)->boot_order, &error_fatal);
824 
825     val = 0;
826     val |= 0x02; /* FPU is there */
827     val |= 0x04; /* PS/2 mouse installed */
828     rtc_set_memory(s, REG_EQUIPMENT_BYTE, val);
829 
830     /* hard drives and FDC */
831     arg.rtc_state = s;
832     arg.idebus[0] = idebus0;
833     arg.idebus[1] = idebus1;
834     qemu_register_reset(pc_cmos_init_late, &arg);
835 }
836 
837 #define TYPE_PORT92 "port92"
838 #define PORT92(obj) OBJECT_CHECK(Port92State, (obj), TYPE_PORT92)
839 
840 /* port 92 stuff: could be split off */
841 typedef struct Port92State {
842     ISADevice parent_obj;
843 
844     MemoryRegion io;
845     uint8_t outport;
846     qemu_irq a20_out;
847 } Port92State;
848 
849 static void port92_write(void *opaque, hwaddr addr, uint64_t val,
850                          unsigned size)
851 {
852     Port92State *s = opaque;
853     int oldval = s->outport;
854 
855     DPRINTF("port92: write 0x%02" PRIx64 "\n", val);
856     s->outport = val;
857     qemu_set_irq(s->a20_out, (val >> 1) & 1);
858     if ((val & 1) && !(oldval & 1)) {
859         qemu_system_reset_request(SHUTDOWN_CAUSE_GUEST_RESET);
860     }
861 }
862 
863 static uint64_t port92_read(void *opaque, hwaddr addr,
864                             unsigned size)
865 {
866     Port92State *s = opaque;
867     uint32_t ret;
868 
869     ret = s->outport;
870     DPRINTF("port92: read 0x%02x\n", ret);
871     return ret;
872 }
873 
874 static void port92_init(ISADevice *dev, qemu_irq a20_out)
875 {
876     qdev_connect_gpio_out_named(DEVICE(dev), PORT92_A20_LINE, 0, a20_out);
877 }
878 
879 static const VMStateDescription vmstate_port92_isa = {
880     .name = "port92",
881     .version_id = 1,
882     .minimum_version_id = 1,
883     .fields = (VMStateField[]) {
884         VMSTATE_UINT8(outport, Port92State),
885         VMSTATE_END_OF_LIST()
886     }
887 };
888 
889 static void port92_reset(DeviceState *d)
890 {
891     Port92State *s = PORT92(d);
892 
893     s->outport &= ~1;
894 }
895 
896 static const MemoryRegionOps port92_ops = {
897     .read = port92_read,
898     .write = port92_write,
899     .impl = {
900         .min_access_size = 1,
901         .max_access_size = 1,
902     },
903     .endianness = DEVICE_LITTLE_ENDIAN,
904 };
905 
906 static void port92_initfn(Object *obj)
907 {
908     Port92State *s = PORT92(obj);
909 
910     memory_region_init_io(&s->io, OBJECT(s), &port92_ops, s, "port92", 1);
911 
912     s->outport = 0;
913 
914     qdev_init_gpio_out_named(DEVICE(obj), &s->a20_out, PORT92_A20_LINE, 1);
915 }
916 
917 static void port92_realizefn(DeviceState *dev, Error **errp)
918 {
919     ISADevice *isadev = ISA_DEVICE(dev);
920     Port92State *s = PORT92(dev);
921 
922     isa_register_ioport(isadev, &s->io, 0x92);
923 }
924 
925 static void port92_class_initfn(ObjectClass *klass, void *data)
926 {
927     DeviceClass *dc = DEVICE_CLASS(klass);
928 
929     dc->realize = port92_realizefn;
930     dc->reset = port92_reset;
931     dc->vmsd = &vmstate_port92_isa;
932     /*
933      * Reason: unlike ordinary ISA devices, this one needs additional
934      * wiring: its A20 output line needs to be wired up by
935      * port92_init().
936      */
937     dc->user_creatable = false;
938 }
939 
940 static const TypeInfo port92_info = {
941     .name          = TYPE_PORT92,
942     .parent        = TYPE_ISA_DEVICE,
943     .instance_size = sizeof(Port92State),
944     .instance_init = port92_initfn,
945     .class_init    = port92_class_initfn,
946 };
947 
948 static void port92_register_types(void)
949 {
950     type_register_static(&port92_info);
951 }
952 
953 type_init(port92_register_types)
954 
955 static void handle_a20_line_change(void *opaque, int irq, int level)
956 {
957     X86CPU *cpu = opaque;
958 
959     /* XXX: send to all CPUs ? */
960     /* XXX: add logic to handle multiple A20 line sources */
961     x86_cpu_set_a20(cpu, level);
962 }
963 
964 int e820_add_entry(uint64_t address, uint64_t length, uint32_t type)
965 {
966     int index = le32_to_cpu(e820_reserve.count);
967     struct e820_entry *entry;
968 
969     if (type != E820_RAM) {
970         /* old FW_CFG_E820_TABLE entry -- reservations only */
971         if (index >= E820_NR_ENTRIES) {
972             return -EBUSY;
973         }
974         entry = &e820_reserve.entry[index++];
975 
976         entry->address = cpu_to_le64(address);
977         entry->length = cpu_to_le64(length);
978         entry->type = cpu_to_le32(type);
979 
980         e820_reserve.count = cpu_to_le32(index);
981     }
982 
983     /* new "etc/e820" file -- include ram too */
984     e820_table = g_renew(struct e820_entry, e820_table, e820_entries + 1);
985     e820_table[e820_entries].address = cpu_to_le64(address);
986     e820_table[e820_entries].length = cpu_to_le64(length);
987     e820_table[e820_entries].type = cpu_to_le32(type);
988     e820_entries++;
989 
990     return e820_entries;
991 }
992 
993 int e820_get_num_entries(void)
994 {
995     return e820_entries;
996 }
997 
998 bool e820_get_entry(int idx, uint32_t type, uint64_t *address, uint64_t *length)
999 {
1000     if (idx < e820_entries && e820_table[idx].type == cpu_to_le32(type)) {
1001         *address = le64_to_cpu(e820_table[idx].address);
1002         *length = le64_to_cpu(e820_table[idx].length);
1003         return true;
1004     }
1005     return false;
1006 }
1007 
1008 /* Enables contiguous-apic-ID mode, for compatibility */
1009 static bool compat_apic_id_mode;
1010 
1011 void enable_compat_apic_id_mode(void)
1012 {
1013     compat_apic_id_mode = true;
1014 }
1015 
1016 /* Calculates initial APIC ID for a specific CPU index
1017  *
1018  * Currently we need to be able to calculate the APIC ID from the CPU index
1019  * alone (without requiring a CPU object), as the QEMU<->Seabios interfaces have
1020  * no concept of "CPU index", and the NUMA tables on fw_cfg need the APIC ID of
1021  * all CPUs up to max_cpus.
1022  */
1023 static uint32_t x86_cpu_apic_id_from_index(unsigned int cpu_index)
1024 {
1025     uint32_t correct_id;
1026     static bool warned;
1027 
1028     correct_id = x86_apicid_from_cpu_idx(smp_cores, smp_threads, cpu_index);
1029     if (compat_apic_id_mode) {
1030         if (cpu_index != correct_id && !warned && !qtest_enabled()) {
1031             error_report("APIC IDs set in compatibility mode, "
1032                          "CPU topology won't match the configuration");
1033             warned = true;
1034         }
1035         return cpu_index;
1036     } else {
1037         return correct_id;
1038     }
1039 }
1040 
1041 static void pc_build_smbios(PCMachineState *pcms)
1042 {
1043     uint8_t *smbios_tables, *smbios_anchor;
1044     size_t smbios_tables_len, smbios_anchor_len;
1045     struct smbios_phys_mem_area *mem_array;
1046     unsigned i, array_count;
1047     MachineState *ms = MACHINE(pcms);
1048     X86CPU *cpu = X86_CPU(ms->possible_cpus->cpus[0].cpu);
1049 
1050     /* tell smbios about cpuid version and features */
1051     smbios_set_cpuid(cpu->env.cpuid_version, cpu->env.features[FEAT_1_EDX]);
1052 
1053     smbios_tables = smbios_get_table_legacy(&smbios_tables_len);
1054     if (smbios_tables) {
1055         fw_cfg_add_bytes(pcms->fw_cfg, FW_CFG_SMBIOS_ENTRIES,
1056                          smbios_tables, smbios_tables_len);
1057     }
1058 
1059     /* build the array of physical mem area from e820 table */
1060     mem_array = g_malloc0(sizeof(*mem_array) * e820_get_num_entries());
1061     for (i = 0, array_count = 0; i < e820_get_num_entries(); i++) {
1062         uint64_t addr, len;
1063 
1064         if (e820_get_entry(i, E820_RAM, &addr, &len)) {
1065             mem_array[array_count].address = addr;
1066             mem_array[array_count].length = len;
1067             array_count++;
1068         }
1069     }
1070     smbios_get_tables(mem_array, array_count,
1071                       &smbios_tables, &smbios_tables_len,
1072                       &smbios_anchor, &smbios_anchor_len);
1073     g_free(mem_array);
1074 
1075     if (smbios_anchor) {
1076         fw_cfg_add_file(pcms->fw_cfg, "etc/smbios/smbios-tables",
1077                         smbios_tables, smbios_tables_len);
1078         fw_cfg_add_file(pcms->fw_cfg, "etc/smbios/smbios-anchor",
1079                         smbios_anchor, smbios_anchor_len);
1080     }
1081 }
1082 
1083 static FWCfgState *bochs_bios_init(AddressSpace *as, PCMachineState *pcms)
1084 {
1085     FWCfgState *fw_cfg;
1086     uint64_t *numa_fw_cfg;
1087     int i;
1088     const CPUArchIdList *cpus;
1089     MachineClass *mc = MACHINE_GET_CLASS(pcms);
1090 
1091     fw_cfg = fw_cfg_init_io_dma(FW_CFG_IO_BASE, FW_CFG_IO_BASE + 4, as);
1092     fw_cfg_add_i16(fw_cfg, FW_CFG_NB_CPUS, pcms->boot_cpus);
1093 
1094     /* FW_CFG_MAX_CPUS is a bit confusing/problematic on x86:
1095      *
1096      * For machine types prior to 1.8, SeaBIOS needs FW_CFG_MAX_CPUS for
1097      * building MPTable, ACPI MADT, ACPI CPU hotplug and ACPI SRAT table,
1098      * that tables are based on xAPIC ID and QEMU<->SeaBIOS interface
1099      * for CPU hotplug also uses APIC ID and not "CPU index".
1100      * This means that FW_CFG_MAX_CPUS is not the "maximum number of CPUs",
1101      * but the "limit to the APIC ID values SeaBIOS may see".
1102      *
1103      * So for compatibility reasons with old BIOSes we are stuck with
1104      * "etc/max-cpus" actually being apic_id_limit
1105      */
1106     fw_cfg_add_i16(fw_cfg, FW_CFG_MAX_CPUS, (uint16_t)pcms->apic_id_limit);
1107     fw_cfg_add_i64(fw_cfg, FW_CFG_RAM_SIZE, (uint64_t)ram_size);
1108     fw_cfg_add_bytes(fw_cfg, FW_CFG_ACPI_TABLES,
1109                      acpi_tables, acpi_tables_len);
1110     fw_cfg_add_i32(fw_cfg, FW_CFG_IRQ0_OVERRIDE, kvm_allows_irq0_override());
1111 
1112     fw_cfg_add_bytes(fw_cfg, FW_CFG_E820_TABLE,
1113                      &e820_reserve, sizeof(e820_reserve));
1114     fw_cfg_add_file(fw_cfg, "etc/e820", e820_table,
1115                     sizeof(struct e820_entry) * e820_entries);
1116 
1117     fw_cfg_add_bytes(fw_cfg, FW_CFG_HPET, &hpet_cfg, sizeof(hpet_cfg));
1118     /* allocate memory for the NUMA channel: one (64bit) word for the number
1119      * of nodes, one word for each VCPU->node and one word for each node to
1120      * hold the amount of memory.
1121      */
1122     numa_fw_cfg = g_new0(uint64_t, 1 + pcms->apic_id_limit + nb_numa_nodes);
1123     numa_fw_cfg[0] = cpu_to_le64(nb_numa_nodes);
1124     cpus = mc->possible_cpu_arch_ids(MACHINE(pcms));
1125     for (i = 0; i < cpus->len; i++) {
1126         unsigned int apic_id = cpus->cpus[i].arch_id;
1127         assert(apic_id < pcms->apic_id_limit);
1128         numa_fw_cfg[apic_id + 1] = cpu_to_le64(cpus->cpus[i].props.node_id);
1129     }
1130     for (i = 0; i < nb_numa_nodes; i++) {
1131         numa_fw_cfg[pcms->apic_id_limit + 1 + i] =
1132             cpu_to_le64(numa_info[i].node_mem);
1133     }
1134     fw_cfg_add_bytes(fw_cfg, FW_CFG_NUMA, numa_fw_cfg,
1135                      (1 + pcms->apic_id_limit + nb_numa_nodes) *
1136                      sizeof(*numa_fw_cfg));
1137 
1138     return fw_cfg;
1139 }
1140 
1141 static long get_file_size(FILE *f)
1142 {
1143     long where, size;
1144 
1145     /* XXX: on Unix systems, using fstat() probably makes more sense */
1146 
1147     where = ftell(f);
1148     fseek(f, 0, SEEK_END);
1149     size = ftell(f);
1150     fseek(f, where, SEEK_SET);
1151 
1152     return size;
1153 }
1154 
1155 /* setup_data types */
1156 #define SETUP_NONE     0
1157 #define SETUP_E820_EXT 1
1158 #define SETUP_DTB      2
1159 #define SETUP_PCI      3
1160 #define SETUP_EFI      4
1161 
1162 struct setup_data {
1163     uint64_t next;
1164     uint32_t type;
1165     uint32_t len;
1166     uint8_t data[0];
1167 } __attribute__((packed));
1168 
1169 static void load_linux(PCMachineState *pcms,
1170                        FWCfgState *fw_cfg)
1171 {
1172     uint16_t protocol;
1173     int setup_size, kernel_size, cmdline_size;
1174     int dtb_size, setup_data_offset;
1175     uint32_t initrd_max;
1176     uint8_t header[8192], *setup, *kernel;
1177     hwaddr real_addr, prot_addr, cmdline_addr, initrd_addr = 0;
1178     FILE *f;
1179     char *vmode;
1180     MachineState *machine = MACHINE(pcms);
1181     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
1182     struct setup_data *setup_data;
1183     const char *kernel_filename = machine->kernel_filename;
1184     const char *initrd_filename = machine->initrd_filename;
1185     const char *dtb_filename = machine->dtb;
1186     const char *kernel_cmdline = machine->kernel_cmdline;
1187 
1188     /* Align to 16 bytes as a paranoia measure */
1189     cmdline_size = (strlen(kernel_cmdline)+16) & ~15;
1190 
1191     /* load the kernel header */
1192     f = fopen(kernel_filename, "rb");
1193     if (!f || !(kernel_size = get_file_size(f)) ||
1194         fread(header, 1, MIN(ARRAY_SIZE(header), kernel_size), f) !=
1195         MIN(ARRAY_SIZE(header), kernel_size)) {
1196         fprintf(stderr, "qemu: could not load kernel '%s': %s\n",
1197                 kernel_filename, strerror(errno));
1198         exit(1);
1199     }
1200 
1201     /* kernel protocol version */
1202 #if 0
1203     fprintf(stderr, "header magic: %#x\n", ldl_p(header+0x202));
1204 #endif
1205     if (ldl_p(header+0x202) == 0x53726448) {
1206         protocol = lduw_p(header+0x206);
1207     } else {
1208         /* This looks like a multiboot kernel. If it is, let's stop
1209            treating it like a Linux kernel. */
1210         if (load_multiboot(fw_cfg, f, kernel_filename, initrd_filename,
1211                            kernel_cmdline, kernel_size, header)) {
1212             return;
1213         }
1214         protocol = 0;
1215     }
1216 
1217     if (protocol < 0x200 || !(header[0x211] & 0x01)) {
1218         /* Low kernel */
1219         real_addr    = 0x90000;
1220         cmdline_addr = 0x9a000 - cmdline_size;
1221         prot_addr    = 0x10000;
1222     } else if (protocol < 0x202) {
1223         /* High but ancient kernel */
1224         real_addr    = 0x90000;
1225         cmdline_addr = 0x9a000 - cmdline_size;
1226         prot_addr    = 0x100000;
1227     } else {
1228         /* High and recent kernel */
1229         real_addr    = 0x10000;
1230         cmdline_addr = 0x20000;
1231         prot_addr    = 0x100000;
1232     }
1233 
1234 #if 0
1235     fprintf(stderr,
1236             "qemu: real_addr     = 0x" TARGET_FMT_plx "\n"
1237             "qemu: cmdline_addr  = 0x" TARGET_FMT_plx "\n"
1238             "qemu: prot_addr     = 0x" TARGET_FMT_plx "\n",
1239             real_addr,
1240             cmdline_addr,
1241             prot_addr);
1242 #endif
1243 
1244     /* highest address for loading the initrd */
1245     if (protocol >= 0x203) {
1246         initrd_max = ldl_p(header+0x22c);
1247     } else {
1248         initrd_max = 0x37ffffff;
1249     }
1250 
1251     if (initrd_max >= pcms->below_4g_mem_size - pcmc->acpi_data_size) {
1252         initrd_max = pcms->below_4g_mem_size - pcmc->acpi_data_size - 1;
1253     }
1254 
1255     fw_cfg_add_i32(fw_cfg, FW_CFG_CMDLINE_ADDR, cmdline_addr);
1256     fw_cfg_add_i32(fw_cfg, FW_CFG_CMDLINE_SIZE, strlen(kernel_cmdline)+1);
1257     fw_cfg_add_string(fw_cfg, FW_CFG_CMDLINE_DATA, kernel_cmdline);
1258 
1259     if (protocol >= 0x202) {
1260         stl_p(header+0x228, cmdline_addr);
1261     } else {
1262         stw_p(header+0x20, 0xA33F);
1263         stw_p(header+0x22, cmdline_addr-real_addr);
1264     }
1265 
1266     /* handle vga= parameter */
1267     vmode = strstr(kernel_cmdline, "vga=");
1268     if (vmode) {
1269         unsigned int video_mode;
1270         /* skip "vga=" */
1271         vmode += 4;
1272         if (!strncmp(vmode, "normal", 6)) {
1273             video_mode = 0xffff;
1274         } else if (!strncmp(vmode, "ext", 3)) {
1275             video_mode = 0xfffe;
1276         } else if (!strncmp(vmode, "ask", 3)) {
1277             video_mode = 0xfffd;
1278         } else {
1279             video_mode = strtol(vmode, NULL, 0);
1280         }
1281         stw_p(header+0x1fa, video_mode);
1282     }
1283 
1284     /* loader type */
1285     /* High nybble = B reserved for QEMU; low nybble is revision number.
1286        If this code is substantially changed, you may want to consider
1287        incrementing the revision. */
1288     if (protocol >= 0x200) {
1289         header[0x210] = 0xB0;
1290     }
1291     /* heap */
1292     if (protocol >= 0x201) {
1293         header[0x211] |= 0x80;	/* CAN_USE_HEAP */
1294         stw_p(header+0x224, cmdline_addr-real_addr-0x200);
1295     }
1296 
1297     /* load initrd */
1298     if (initrd_filename) {
1299         gsize initrd_size;
1300         gchar *initrd_data;
1301         GError *gerr = NULL;
1302 
1303         if (protocol < 0x200) {
1304             fprintf(stderr, "qemu: linux kernel too old to load a ram disk\n");
1305             exit(1);
1306         }
1307 
1308         if (!g_file_get_contents(initrd_filename, &initrd_data,
1309                                  &initrd_size, &gerr)) {
1310             fprintf(stderr, "qemu: error reading initrd %s: %s\n",
1311                     initrd_filename, gerr->message);
1312             exit(1);
1313         }
1314         if (initrd_size >= initrd_max) {
1315             fprintf(stderr, "qemu: initrd is too large, cannot support."
1316                     "(max: %"PRIu32", need %"PRId64")\n",
1317                     initrd_max, (uint64_t)initrd_size);
1318             exit(1);
1319         }
1320 
1321         initrd_addr = (initrd_max-initrd_size) & ~4095;
1322 
1323         fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_ADDR, initrd_addr);
1324         fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_SIZE, initrd_size);
1325         fw_cfg_add_bytes(fw_cfg, FW_CFG_INITRD_DATA, initrd_data, initrd_size);
1326 
1327         stl_p(header+0x218, initrd_addr);
1328         stl_p(header+0x21c, initrd_size);
1329     }
1330 
1331     /* load kernel and setup */
1332     setup_size = header[0x1f1];
1333     if (setup_size == 0) {
1334         setup_size = 4;
1335     }
1336     setup_size = (setup_size+1)*512;
1337     if (setup_size > kernel_size) {
1338         fprintf(stderr, "qemu: invalid kernel header\n");
1339         exit(1);
1340     }
1341     kernel_size -= setup_size;
1342 
1343     setup  = g_malloc(setup_size);
1344     kernel = g_malloc(kernel_size);
1345     fseek(f, 0, SEEK_SET);
1346     if (fread(setup, 1, setup_size, f) != setup_size) {
1347         fprintf(stderr, "fread() failed\n");
1348         exit(1);
1349     }
1350     if (fread(kernel, 1, kernel_size, f) != kernel_size) {
1351         fprintf(stderr, "fread() failed\n");
1352         exit(1);
1353     }
1354     fclose(f);
1355 
1356     /* append dtb to kernel */
1357     if (dtb_filename) {
1358         if (protocol < 0x209) {
1359             fprintf(stderr, "qemu: Linux kernel too old to load a dtb\n");
1360             exit(1);
1361         }
1362 
1363         dtb_size = get_image_size(dtb_filename);
1364         if (dtb_size <= 0) {
1365             fprintf(stderr, "qemu: error reading dtb %s: %s\n",
1366                     dtb_filename, strerror(errno));
1367             exit(1);
1368         }
1369 
1370         setup_data_offset = QEMU_ALIGN_UP(kernel_size, 16);
1371         kernel_size = setup_data_offset + sizeof(struct setup_data) + dtb_size;
1372         kernel = g_realloc(kernel, kernel_size);
1373 
1374         stq_p(header+0x250, prot_addr + setup_data_offset);
1375 
1376         setup_data = (struct setup_data *)(kernel + setup_data_offset);
1377         setup_data->next = 0;
1378         setup_data->type = cpu_to_le32(SETUP_DTB);
1379         setup_data->len = cpu_to_le32(dtb_size);
1380 
1381         load_image_size(dtb_filename, setup_data->data, dtb_size);
1382     }
1383 
1384     memcpy(setup, header, MIN(sizeof(header), setup_size));
1385 
1386     fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_ADDR, prot_addr);
1387     fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_SIZE, kernel_size);
1388     fw_cfg_add_bytes(fw_cfg, FW_CFG_KERNEL_DATA, kernel, kernel_size);
1389 
1390     fw_cfg_add_i32(fw_cfg, FW_CFG_SETUP_ADDR, real_addr);
1391     fw_cfg_add_i32(fw_cfg, FW_CFG_SETUP_SIZE, setup_size);
1392     fw_cfg_add_bytes(fw_cfg, FW_CFG_SETUP_DATA, setup, setup_size);
1393 
1394     option_rom[nb_option_roms].bootindex = 0;
1395     option_rom[nb_option_roms].name = "linuxboot.bin";
1396     if (pcmc->linuxboot_dma_enabled && fw_cfg_dma_enabled(fw_cfg)) {
1397         option_rom[nb_option_roms].name = "linuxboot_dma.bin";
1398     }
1399     nb_option_roms++;
1400 }
1401 
1402 #define NE2000_NB_MAX 6
1403 
1404 static const int ne2000_io[NE2000_NB_MAX] = { 0x300, 0x320, 0x340, 0x360,
1405                                               0x280, 0x380 };
1406 static const int ne2000_irq[NE2000_NB_MAX] = { 9, 10, 11, 3, 4, 5 };
1407 
1408 void pc_init_ne2k_isa(ISABus *bus, NICInfo *nd)
1409 {
1410     static int nb_ne2k = 0;
1411 
1412     if (nb_ne2k == NE2000_NB_MAX)
1413         return;
1414     isa_ne2000_init(bus, ne2000_io[nb_ne2k],
1415                     ne2000_irq[nb_ne2k], nd);
1416     nb_ne2k++;
1417 }
1418 
1419 DeviceState *cpu_get_current_apic(void)
1420 {
1421     if (current_cpu) {
1422         X86CPU *cpu = X86_CPU(current_cpu);
1423         return cpu->apic_state;
1424     } else {
1425         return NULL;
1426     }
1427 }
1428 
1429 void pc_acpi_smi_interrupt(void *opaque, int irq, int level)
1430 {
1431     X86CPU *cpu = opaque;
1432 
1433     if (level) {
1434         cpu_interrupt(CPU(cpu), CPU_INTERRUPT_SMI);
1435     }
1436 }
1437 
1438 static void pc_new_cpu(const char *typename, int64_t apic_id, Error **errp)
1439 {
1440     Object *cpu = NULL;
1441     Error *local_err = NULL;
1442 
1443     cpu = object_new(typename);
1444 
1445     object_property_set_uint(cpu, apic_id, "apic-id", &local_err);
1446     object_property_set_bool(cpu, true, "realized", &local_err);
1447 
1448     object_unref(cpu);
1449     error_propagate(errp, local_err);
1450 }
1451 
1452 void pc_hot_add_cpu(const int64_t id, Error **errp)
1453 {
1454     MachineState *ms = MACHINE(qdev_get_machine());
1455     int64_t apic_id = x86_cpu_apic_id_from_index(id);
1456     Error *local_err = NULL;
1457 
1458     if (id < 0) {
1459         error_setg(errp, "Invalid CPU id: %" PRIi64, id);
1460         return;
1461     }
1462 
1463     if (apic_id >= ACPI_CPU_HOTPLUG_ID_LIMIT) {
1464         error_setg(errp, "Unable to add CPU: %" PRIi64
1465                    ", resulting APIC ID (%" PRIi64 ") is too large",
1466                    id, apic_id);
1467         return;
1468     }
1469 
1470     pc_new_cpu(ms->cpu_type, apic_id, &local_err);
1471     if (local_err) {
1472         error_propagate(errp, local_err);
1473         return;
1474     }
1475 }
1476 
1477 void pc_cpus_init(PCMachineState *pcms)
1478 {
1479     int i;
1480     const CPUArchIdList *possible_cpus;
1481     MachineState *ms = MACHINE(pcms);
1482     MachineClass *mc = MACHINE_GET_CLASS(pcms);
1483 
1484     /* Calculates the limit to CPU APIC ID values
1485      *
1486      * Limit for the APIC ID value, so that all
1487      * CPU APIC IDs are < pcms->apic_id_limit.
1488      *
1489      * This is used for FW_CFG_MAX_CPUS. See comments on bochs_bios_init().
1490      */
1491     pcms->apic_id_limit = x86_cpu_apic_id_from_index(max_cpus - 1) + 1;
1492     possible_cpus = mc->possible_cpu_arch_ids(ms);
1493     for (i = 0; i < smp_cpus; i++) {
1494         pc_new_cpu(possible_cpus->cpus[i].type, possible_cpus->cpus[i].arch_id,
1495                    &error_fatal);
1496     }
1497 }
1498 
1499 static void pc_build_feature_control_file(PCMachineState *pcms)
1500 {
1501     MachineState *ms = MACHINE(pcms);
1502     X86CPU *cpu = X86_CPU(ms->possible_cpus->cpus[0].cpu);
1503     CPUX86State *env = &cpu->env;
1504     uint32_t unused, ecx, edx;
1505     uint64_t feature_control_bits = 0;
1506     uint64_t *val;
1507 
1508     cpu_x86_cpuid(env, 1, 0, &unused, &unused, &ecx, &edx);
1509     if (ecx & CPUID_EXT_VMX) {
1510         feature_control_bits |= FEATURE_CONTROL_VMXON_ENABLED_OUTSIDE_SMX;
1511     }
1512 
1513     if ((edx & (CPUID_EXT2_MCE | CPUID_EXT2_MCA)) ==
1514         (CPUID_EXT2_MCE | CPUID_EXT2_MCA) &&
1515         (env->mcg_cap & MCG_LMCE_P)) {
1516         feature_control_bits |= FEATURE_CONTROL_LMCE;
1517     }
1518 
1519     if (!feature_control_bits) {
1520         return;
1521     }
1522 
1523     val = g_malloc(sizeof(*val));
1524     *val = cpu_to_le64(feature_control_bits | FEATURE_CONTROL_LOCKED);
1525     fw_cfg_add_file(pcms->fw_cfg, "etc/msr_feature_control", val, sizeof(*val));
1526 }
1527 
1528 static void rtc_set_cpus_count(ISADevice *rtc, uint16_t cpus_count)
1529 {
1530     if (cpus_count > 0xff) {
1531         /* If the number of CPUs can't be represented in 8 bits, the
1532          * BIOS must use "FW_CFG_NB_CPUS". Set RTC field to 0 just
1533          * to make old BIOSes fail more predictably.
1534          */
1535         rtc_set_memory(rtc, 0x5f, 0);
1536     } else {
1537         rtc_set_memory(rtc, 0x5f, cpus_count - 1);
1538     }
1539 }
1540 
1541 static
1542 void pc_machine_done(Notifier *notifier, void *data)
1543 {
1544     PCMachineState *pcms = container_of(notifier,
1545                                         PCMachineState, machine_done);
1546     PCIBus *bus = pcms->bus;
1547 
1548     /* set the number of CPUs */
1549     rtc_set_cpus_count(pcms->rtc, pcms->boot_cpus);
1550 
1551     if (bus) {
1552         int extra_hosts = 0;
1553 
1554         QLIST_FOREACH(bus, &bus->child, sibling) {
1555             /* look for expander root buses */
1556             if (pci_bus_is_root(bus)) {
1557                 extra_hosts++;
1558             }
1559         }
1560         if (extra_hosts && pcms->fw_cfg) {
1561             uint64_t *val = g_malloc(sizeof(*val));
1562             *val = cpu_to_le64(extra_hosts);
1563             fw_cfg_add_file(pcms->fw_cfg,
1564                     "etc/extra-pci-roots", val, sizeof(*val));
1565         }
1566     }
1567 
1568     acpi_setup();
1569     if (pcms->fw_cfg) {
1570         pc_build_smbios(pcms);
1571         pc_build_feature_control_file(pcms);
1572         /* update FW_CFG_NB_CPUS to account for -device added CPUs */
1573         fw_cfg_modify_i16(pcms->fw_cfg, FW_CFG_NB_CPUS, pcms->boot_cpus);
1574     }
1575 
1576     if (pcms->apic_id_limit > 255 && !xen_enabled()) {
1577         IntelIOMMUState *iommu = INTEL_IOMMU_DEVICE(x86_iommu_get_default());
1578 
1579         if (!iommu || !x86_iommu_ir_supported(X86_IOMMU_DEVICE(iommu)) ||
1580             iommu->intr_eim != ON_OFF_AUTO_ON) {
1581             error_report("current -smp configuration requires "
1582                          "Extended Interrupt Mode enabled. "
1583                          "You can add an IOMMU using: "
1584                          "-device intel-iommu,intremap=on,eim=on");
1585             exit(EXIT_FAILURE);
1586         }
1587     }
1588 }
1589 
1590 void pc_guest_info_init(PCMachineState *pcms)
1591 {
1592     int i;
1593 
1594     pcms->apic_xrupt_override = kvm_allows_irq0_override();
1595     pcms->numa_nodes = nb_numa_nodes;
1596     pcms->node_mem = g_malloc0(pcms->numa_nodes *
1597                                     sizeof *pcms->node_mem);
1598     for (i = 0; i < nb_numa_nodes; i++) {
1599         pcms->node_mem[i] = numa_info[i].node_mem;
1600     }
1601 
1602     pcms->machine_done.notify = pc_machine_done;
1603     qemu_add_machine_init_done_notifier(&pcms->machine_done);
1604 }
1605 
1606 /* setup pci memory address space mapping into system address space */
1607 void pc_pci_as_mapping_init(Object *owner, MemoryRegion *system_memory,
1608                             MemoryRegion *pci_address_space)
1609 {
1610     /* Set to lower priority than RAM */
1611     memory_region_add_subregion_overlap(system_memory, 0x0,
1612                                         pci_address_space, -1);
1613 }
1614 
1615 void pc_acpi_init(const char *default_dsdt)
1616 {
1617     char *filename;
1618 
1619     if (acpi_tables != NULL) {
1620         /* manually set via -acpitable, leave it alone */
1621         return;
1622     }
1623 
1624     filename = qemu_find_file(QEMU_FILE_TYPE_BIOS, default_dsdt);
1625     if (filename == NULL) {
1626         warn_report("failed to find %s", default_dsdt);
1627     } else {
1628         QemuOpts *opts = qemu_opts_create(qemu_find_opts("acpi"), NULL, 0,
1629                                           &error_abort);
1630         Error *err = NULL;
1631 
1632         qemu_opt_set(opts, "file", filename, &error_abort);
1633 
1634         acpi_table_add_builtin(opts, &err);
1635         if (err) {
1636             warn_reportf_err(err, "failed to load %s: ", filename);
1637         }
1638         g_free(filename);
1639     }
1640 }
1641 
1642 void xen_load_linux(PCMachineState *pcms)
1643 {
1644     int i;
1645     FWCfgState *fw_cfg;
1646 
1647     assert(MACHINE(pcms)->kernel_filename != NULL);
1648 
1649     fw_cfg = fw_cfg_init_io(FW_CFG_IO_BASE);
1650     fw_cfg_add_i16(fw_cfg, FW_CFG_NB_CPUS, pcms->boot_cpus);
1651     rom_set_fw(fw_cfg);
1652 
1653     load_linux(pcms, fw_cfg);
1654     for (i = 0; i < nb_option_roms; i++) {
1655         assert(!strcmp(option_rom[i].name, "linuxboot.bin") ||
1656                !strcmp(option_rom[i].name, "linuxboot_dma.bin") ||
1657                !strcmp(option_rom[i].name, "multiboot.bin"));
1658         rom_add_option(option_rom[i].name, option_rom[i].bootindex);
1659     }
1660     pcms->fw_cfg = fw_cfg;
1661 }
1662 
1663 void pc_memory_init(PCMachineState *pcms,
1664                     MemoryRegion *system_memory,
1665                     MemoryRegion *rom_memory,
1666                     MemoryRegion **ram_memory)
1667 {
1668     int linux_boot, i;
1669     MemoryRegion *ram, *option_rom_mr;
1670     MemoryRegion *ram_below_4g, *ram_above_4g;
1671     FWCfgState *fw_cfg;
1672     MachineState *machine = MACHINE(pcms);
1673     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
1674 
1675     assert(machine->ram_size == pcms->below_4g_mem_size +
1676                                 pcms->above_4g_mem_size);
1677 
1678     linux_boot = (machine->kernel_filename != NULL);
1679 
1680     /* Allocate RAM.  We allocate it as a single memory region and use
1681      * aliases to address portions of it, mostly for backwards compatibility
1682      * with older qemus that used qemu_ram_alloc().
1683      */
1684     ram = g_malloc(sizeof(*ram));
1685     memory_region_allocate_system_memory(ram, NULL, "pc.ram",
1686                                          machine->ram_size);
1687     *ram_memory = ram;
1688     ram_below_4g = g_malloc(sizeof(*ram_below_4g));
1689     memory_region_init_alias(ram_below_4g, NULL, "ram-below-4g", ram,
1690                              0, pcms->below_4g_mem_size);
1691     memory_region_add_subregion(system_memory, 0, ram_below_4g);
1692     e820_add_entry(0, pcms->below_4g_mem_size, E820_RAM);
1693     if (pcms->above_4g_mem_size > 0) {
1694         ram_above_4g = g_malloc(sizeof(*ram_above_4g));
1695         memory_region_init_alias(ram_above_4g, NULL, "ram-above-4g", ram,
1696                                  pcms->below_4g_mem_size,
1697                                  pcms->above_4g_mem_size);
1698         memory_region_add_subregion(system_memory, 0x100000000ULL,
1699                                     ram_above_4g);
1700         e820_add_entry(0x100000000ULL, pcms->above_4g_mem_size, E820_RAM);
1701     }
1702 
1703     if (!pcmc->has_reserved_memory &&
1704         (machine->ram_slots ||
1705          (machine->maxram_size > machine->ram_size))) {
1706         MachineClass *mc = MACHINE_GET_CLASS(machine);
1707 
1708         error_report("\"-memory 'slots|maxmem'\" is not supported by: %s",
1709                      mc->name);
1710         exit(EXIT_FAILURE);
1711     }
1712 
1713     /* always allocate the device memory information */
1714     machine->device_memory = g_malloc0(sizeof(*machine->device_memory));
1715 
1716     /* initialize device memory address space */
1717     if (pcmc->has_reserved_memory &&
1718         (machine->ram_size < machine->maxram_size)) {
1719         ram_addr_t device_mem_size = machine->maxram_size - machine->ram_size;
1720 
1721         if (machine->ram_slots > ACPI_MAX_RAM_SLOTS) {
1722             error_report("unsupported amount of memory slots: %"PRIu64,
1723                          machine->ram_slots);
1724             exit(EXIT_FAILURE);
1725         }
1726 
1727         if (QEMU_ALIGN_UP(machine->maxram_size,
1728                           TARGET_PAGE_SIZE) != machine->maxram_size) {
1729             error_report("maximum memory size must by aligned to multiple of "
1730                          "%d bytes", TARGET_PAGE_SIZE);
1731             exit(EXIT_FAILURE);
1732         }
1733 
1734         machine->device_memory->base =
1735             ROUND_UP(0x100000000ULL + pcms->above_4g_mem_size, 1 * GiB);
1736 
1737         if (pcmc->enforce_aligned_dimm) {
1738             /* size device region assuming 1G page max alignment per slot */
1739             device_mem_size += (1 * GiB) * machine->ram_slots;
1740         }
1741 
1742         if ((machine->device_memory->base + device_mem_size) <
1743             device_mem_size) {
1744             error_report("unsupported amount of maximum memory: " RAM_ADDR_FMT,
1745                          machine->maxram_size);
1746             exit(EXIT_FAILURE);
1747         }
1748 
1749         memory_region_init(&machine->device_memory->mr, OBJECT(pcms),
1750                            "device-memory", device_mem_size);
1751         memory_region_add_subregion(system_memory, machine->device_memory->base,
1752                                     &machine->device_memory->mr);
1753     }
1754 
1755     /* Initialize PC system firmware */
1756     pc_system_firmware_init(rom_memory, !pcmc->pci_enabled);
1757 
1758     option_rom_mr = g_malloc(sizeof(*option_rom_mr));
1759     memory_region_init_ram(option_rom_mr, NULL, "pc.rom", PC_ROM_SIZE,
1760                            &error_fatal);
1761     if (pcmc->pci_enabled) {
1762         memory_region_set_readonly(option_rom_mr, true);
1763     }
1764     memory_region_add_subregion_overlap(rom_memory,
1765                                         PC_ROM_MIN_VGA,
1766                                         option_rom_mr,
1767                                         1);
1768 
1769     fw_cfg = bochs_bios_init(&address_space_memory, pcms);
1770 
1771     rom_set_fw(fw_cfg);
1772 
1773     if (pcmc->has_reserved_memory && machine->device_memory->base) {
1774         uint64_t *val = g_malloc(sizeof(*val));
1775         PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
1776         uint64_t res_mem_end = machine->device_memory->base;
1777 
1778         if (!pcmc->broken_reserved_end) {
1779             res_mem_end += memory_region_size(&machine->device_memory->mr);
1780         }
1781         *val = cpu_to_le64(ROUND_UP(res_mem_end, 1 * GiB));
1782         fw_cfg_add_file(fw_cfg, "etc/reserved-memory-end", val, sizeof(*val));
1783     }
1784 
1785     if (linux_boot) {
1786         load_linux(pcms, fw_cfg);
1787     }
1788 
1789     for (i = 0; i < nb_option_roms; i++) {
1790         rom_add_option(option_rom[i].name, option_rom[i].bootindex);
1791     }
1792     pcms->fw_cfg = fw_cfg;
1793 
1794     /* Init default IOAPIC address space */
1795     pcms->ioapic_as = &address_space_memory;
1796 }
1797 
1798 /*
1799  * The 64bit pci hole starts after "above 4G RAM" and
1800  * potentially the space reserved for memory hotplug.
1801  */
1802 uint64_t pc_pci_hole64_start(void)
1803 {
1804     PCMachineState *pcms = PC_MACHINE(qdev_get_machine());
1805     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
1806     MachineState *ms = MACHINE(pcms);
1807     uint64_t hole64_start = 0;
1808 
1809     if (pcmc->has_reserved_memory && ms->device_memory->base) {
1810         hole64_start = ms->device_memory->base;
1811         if (!pcmc->broken_reserved_end) {
1812             hole64_start += memory_region_size(&ms->device_memory->mr);
1813         }
1814     } else {
1815         hole64_start = 0x100000000ULL + pcms->above_4g_mem_size;
1816     }
1817 
1818     return ROUND_UP(hole64_start, 1 * GiB);
1819 }
1820 
1821 qemu_irq pc_allocate_cpu_irq(void)
1822 {
1823     return qemu_allocate_irq(pic_irq_request, NULL, 0);
1824 }
1825 
1826 DeviceState *pc_vga_init(ISABus *isa_bus, PCIBus *pci_bus)
1827 {
1828     DeviceState *dev = NULL;
1829 
1830     rom_set_order_override(FW_CFG_ORDER_OVERRIDE_VGA);
1831     if (pci_bus) {
1832         PCIDevice *pcidev = pci_vga_init(pci_bus);
1833         dev = pcidev ? &pcidev->qdev : NULL;
1834     } else if (isa_bus) {
1835         ISADevice *isadev = isa_vga_init(isa_bus);
1836         dev = isadev ? DEVICE(isadev) : NULL;
1837     }
1838     rom_reset_order_override();
1839     return dev;
1840 }
1841 
1842 static const MemoryRegionOps ioport80_io_ops = {
1843     .write = ioport80_write,
1844     .read = ioport80_read,
1845     .endianness = DEVICE_NATIVE_ENDIAN,
1846     .impl = {
1847         .min_access_size = 1,
1848         .max_access_size = 1,
1849     },
1850 };
1851 
1852 static const MemoryRegionOps ioportF0_io_ops = {
1853     .write = ioportF0_write,
1854     .read = ioportF0_read,
1855     .endianness = DEVICE_NATIVE_ENDIAN,
1856     .impl = {
1857         .min_access_size = 1,
1858         .max_access_size = 1,
1859     },
1860 };
1861 
1862 static void pc_superio_init(ISABus *isa_bus, bool create_fdctrl, bool no_vmport)
1863 {
1864     int i;
1865     DriveInfo *fd[MAX_FD];
1866     qemu_irq *a20_line;
1867     ISADevice *i8042, *port92, *vmmouse;
1868 
1869     serial_hds_isa_init(isa_bus, 0, MAX_ISA_SERIAL_PORTS);
1870     parallel_hds_isa_init(isa_bus, MAX_PARALLEL_PORTS);
1871 
1872     for (i = 0; i < MAX_FD; i++) {
1873         fd[i] = drive_get(IF_FLOPPY, 0, i);
1874         create_fdctrl |= !!fd[i];
1875     }
1876     if (create_fdctrl) {
1877         fdctrl_init_isa(isa_bus, fd);
1878     }
1879 
1880     i8042 = isa_create_simple(isa_bus, "i8042");
1881     if (!no_vmport) {
1882         vmport_init(isa_bus);
1883         vmmouse = isa_try_create(isa_bus, "vmmouse");
1884     } else {
1885         vmmouse = NULL;
1886     }
1887     if (vmmouse) {
1888         DeviceState *dev = DEVICE(vmmouse);
1889         qdev_prop_set_ptr(dev, "ps2_mouse", i8042);
1890         qdev_init_nofail(dev);
1891     }
1892     port92 = isa_create_simple(isa_bus, "port92");
1893 
1894     a20_line = qemu_allocate_irqs(handle_a20_line_change, first_cpu, 2);
1895     i8042_setup_a20_line(i8042, a20_line[0]);
1896     port92_init(port92, a20_line[1]);
1897     g_free(a20_line);
1898 }
1899 
1900 void pc_basic_device_init(ISABus *isa_bus, qemu_irq *gsi,
1901                           ISADevice **rtc_state,
1902                           bool create_fdctrl,
1903                           bool no_vmport,
1904                           bool has_pit,
1905                           uint32_t hpet_irqs)
1906 {
1907     int i;
1908     DeviceState *hpet = NULL;
1909     int pit_isa_irq = 0;
1910     qemu_irq pit_alt_irq = NULL;
1911     qemu_irq rtc_irq = NULL;
1912     ISADevice *pit = NULL;
1913     MemoryRegion *ioport80_io = g_new(MemoryRegion, 1);
1914     MemoryRegion *ioportF0_io = g_new(MemoryRegion, 1);
1915 
1916     memory_region_init_io(ioport80_io, NULL, &ioport80_io_ops, NULL, "ioport80", 1);
1917     memory_region_add_subregion(isa_bus->address_space_io, 0x80, ioport80_io);
1918 
1919     memory_region_init_io(ioportF0_io, NULL, &ioportF0_io_ops, NULL, "ioportF0", 1);
1920     memory_region_add_subregion(isa_bus->address_space_io, 0xf0, ioportF0_io);
1921 
1922     /*
1923      * Check if an HPET shall be created.
1924      *
1925      * Without KVM_CAP_PIT_STATE2, we cannot switch off the in-kernel PIT
1926      * when the HPET wants to take over. Thus we have to disable the latter.
1927      */
1928     if (!no_hpet && (!kvm_irqchip_in_kernel() || kvm_has_pit_state2())) {
1929         /* In order to set property, here not using sysbus_try_create_simple */
1930         hpet = qdev_try_create(NULL, TYPE_HPET);
1931         if (hpet) {
1932             /* For pc-piix-*, hpet's intcap is always IRQ2. For pc-q35-1.7
1933              * and earlier, use IRQ2 for compat. Otherwise, use IRQ16~23,
1934              * IRQ8 and IRQ2.
1935              */
1936             uint8_t compat = object_property_get_uint(OBJECT(hpet),
1937                     HPET_INTCAP, NULL);
1938             if (!compat) {
1939                 qdev_prop_set_uint32(hpet, HPET_INTCAP, hpet_irqs);
1940             }
1941             qdev_init_nofail(hpet);
1942             sysbus_mmio_map(SYS_BUS_DEVICE(hpet), 0, HPET_BASE);
1943 
1944             for (i = 0; i < GSI_NUM_PINS; i++) {
1945                 sysbus_connect_irq(SYS_BUS_DEVICE(hpet), i, gsi[i]);
1946             }
1947             pit_isa_irq = -1;
1948             pit_alt_irq = qdev_get_gpio_in(hpet, HPET_LEGACY_PIT_INT);
1949             rtc_irq = qdev_get_gpio_in(hpet, HPET_LEGACY_RTC_INT);
1950         }
1951     }
1952     *rtc_state = mc146818_rtc_init(isa_bus, 2000, rtc_irq);
1953 
1954     qemu_register_boot_set(pc_boot_set, *rtc_state);
1955 
1956     if (!xen_enabled() && has_pit) {
1957         if (kvm_pit_in_kernel()) {
1958             pit = kvm_pit_init(isa_bus, 0x40);
1959         } else {
1960             pit = i8254_pit_init(isa_bus, 0x40, pit_isa_irq, pit_alt_irq);
1961         }
1962         if (hpet) {
1963             /* connect PIT to output control line of the HPET */
1964             qdev_connect_gpio_out(hpet, 0, qdev_get_gpio_in(DEVICE(pit), 0));
1965         }
1966         pcspk_init(isa_bus, pit);
1967     }
1968 
1969     i8257_dma_init(isa_bus, 0);
1970 
1971     /* Super I/O */
1972     pc_superio_init(isa_bus, create_fdctrl, no_vmport);
1973 }
1974 
1975 void pc_nic_init(PCMachineClass *pcmc, ISABus *isa_bus, PCIBus *pci_bus)
1976 {
1977     int i;
1978 
1979     rom_set_order_override(FW_CFG_ORDER_OVERRIDE_NIC);
1980     for (i = 0; i < nb_nics; i++) {
1981         NICInfo *nd = &nd_table[i];
1982         const char *model = nd->model ? nd->model : pcmc->default_nic_model;
1983 
1984         if (g_str_equal(model, "ne2k_isa")) {
1985             pc_init_ne2k_isa(isa_bus, nd);
1986         } else {
1987             pci_nic_init_nofail(nd, pci_bus, model, NULL);
1988         }
1989     }
1990     rom_reset_order_override();
1991 }
1992 
1993 void ioapic_init_gsi(GSIState *gsi_state, const char *parent_name)
1994 {
1995     DeviceState *dev;
1996     SysBusDevice *d;
1997     unsigned int i;
1998 
1999     if (kvm_ioapic_in_kernel()) {
2000         dev = qdev_create(NULL, "kvm-ioapic");
2001     } else {
2002         dev = qdev_create(NULL, "ioapic");
2003     }
2004     if (parent_name) {
2005         object_property_add_child(object_resolve_path(parent_name, NULL),
2006                                   "ioapic", OBJECT(dev), NULL);
2007     }
2008     qdev_init_nofail(dev);
2009     d = SYS_BUS_DEVICE(dev);
2010     sysbus_mmio_map(d, 0, IO_APIC_DEFAULT_ADDRESS);
2011 
2012     for (i = 0; i < IOAPIC_NUM_PINS; i++) {
2013         gsi_state->ioapic_irq[i] = qdev_get_gpio_in(dev, i);
2014     }
2015 }
2016 
2017 static void pc_memory_pre_plug(HotplugHandler *hotplug_dev, DeviceState *dev,
2018                                Error **errp)
2019 {
2020     const PCMachineState *pcms = PC_MACHINE(hotplug_dev);
2021     const PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
2022     const bool is_nvdimm = object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM);
2023     const uint64_t legacy_align = TARGET_PAGE_SIZE;
2024 
2025     /*
2026      * When -no-acpi is used with Q35 machine type, no ACPI is built,
2027      * but pcms->acpi_dev is still created. Check !acpi_enabled in
2028      * addition to cover this case.
2029      */
2030     if (!pcms->acpi_dev || !acpi_enabled) {
2031         error_setg(errp,
2032                    "memory hotplug is not enabled: missing acpi device or acpi disabled");
2033         return;
2034     }
2035 
2036     if (is_nvdimm && !pcms->acpi_nvdimm_state.is_enabled) {
2037         error_setg(errp, "nvdimm is not enabled: missing 'nvdimm' in '-M'");
2038         return;
2039     }
2040 
2041     pc_dimm_pre_plug(PC_DIMM(dev), MACHINE(hotplug_dev),
2042                      pcmc->enforce_aligned_dimm ? NULL : &legacy_align, errp);
2043 }
2044 
2045 static void pc_memory_plug(HotplugHandler *hotplug_dev,
2046                            DeviceState *dev, Error **errp)
2047 {
2048     HotplugHandlerClass *hhc;
2049     Error *local_err = NULL;
2050     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
2051     bool is_nvdimm = object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM);
2052 
2053     pc_dimm_plug(PC_DIMM(dev), MACHINE(pcms), &local_err);
2054     if (local_err) {
2055         goto out;
2056     }
2057 
2058     if (is_nvdimm) {
2059         nvdimm_plug(&pcms->acpi_nvdimm_state);
2060     }
2061 
2062     hhc = HOTPLUG_HANDLER_GET_CLASS(pcms->acpi_dev);
2063     hhc->plug(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &error_abort);
2064 out:
2065     error_propagate(errp, local_err);
2066 }
2067 
2068 static void pc_memory_unplug_request(HotplugHandler *hotplug_dev,
2069                                      DeviceState *dev, Error **errp)
2070 {
2071     HotplugHandlerClass *hhc;
2072     Error *local_err = NULL;
2073     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
2074 
2075     /*
2076      * When -no-acpi is used with Q35 machine type, no ACPI is built,
2077      * but pcms->acpi_dev is still created. Check !acpi_enabled in
2078      * addition to cover this case.
2079      */
2080     if (!pcms->acpi_dev || !acpi_enabled) {
2081         error_setg(&local_err,
2082                    "memory hotplug is not enabled: missing acpi device or acpi disabled");
2083         goto out;
2084     }
2085 
2086     if (object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM)) {
2087         error_setg(&local_err,
2088                    "nvdimm device hot unplug is not supported yet.");
2089         goto out;
2090     }
2091 
2092     hhc = HOTPLUG_HANDLER_GET_CLASS(pcms->acpi_dev);
2093     hhc->unplug_request(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &local_err);
2094 
2095 out:
2096     error_propagate(errp, local_err);
2097 }
2098 
2099 static void pc_memory_unplug(HotplugHandler *hotplug_dev,
2100                              DeviceState *dev, Error **errp)
2101 {
2102     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
2103     HotplugHandlerClass *hhc;
2104     Error *local_err = NULL;
2105 
2106     hhc = HOTPLUG_HANDLER_GET_CLASS(pcms->acpi_dev);
2107     hhc->unplug(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &local_err);
2108 
2109     if (local_err) {
2110         goto out;
2111     }
2112 
2113     pc_dimm_unplug(PC_DIMM(dev), MACHINE(pcms));
2114     object_unparent(OBJECT(dev));
2115 
2116  out:
2117     error_propagate(errp, local_err);
2118 }
2119 
2120 static int pc_apic_cmp(const void *a, const void *b)
2121 {
2122    CPUArchId *apic_a = (CPUArchId *)a;
2123    CPUArchId *apic_b = (CPUArchId *)b;
2124 
2125    return apic_a->arch_id - apic_b->arch_id;
2126 }
2127 
2128 /* returns pointer to CPUArchId descriptor that matches CPU's apic_id
2129  * in ms->possible_cpus->cpus, if ms->possible_cpus->cpus has no
2130  * entry corresponding to CPU's apic_id returns NULL.
2131  */
2132 static CPUArchId *pc_find_cpu_slot(MachineState *ms, uint32_t id, int *idx)
2133 {
2134     CPUArchId apic_id, *found_cpu;
2135 
2136     apic_id.arch_id = id;
2137     found_cpu = bsearch(&apic_id, ms->possible_cpus->cpus,
2138         ms->possible_cpus->len, sizeof(*ms->possible_cpus->cpus),
2139         pc_apic_cmp);
2140     if (found_cpu && idx) {
2141         *idx = found_cpu - ms->possible_cpus->cpus;
2142     }
2143     return found_cpu;
2144 }
2145 
2146 static void pc_cpu_plug(HotplugHandler *hotplug_dev,
2147                         DeviceState *dev, Error **errp)
2148 {
2149     CPUArchId *found_cpu;
2150     HotplugHandlerClass *hhc;
2151     Error *local_err = NULL;
2152     X86CPU *cpu = X86_CPU(dev);
2153     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
2154 
2155     if (pcms->acpi_dev) {
2156         hhc = HOTPLUG_HANDLER_GET_CLASS(pcms->acpi_dev);
2157         hhc->plug(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &local_err);
2158         if (local_err) {
2159             goto out;
2160         }
2161     }
2162 
2163     /* increment the number of CPUs */
2164     pcms->boot_cpus++;
2165     if (pcms->rtc) {
2166         rtc_set_cpus_count(pcms->rtc, pcms->boot_cpus);
2167     }
2168     if (pcms->fw_cfg) {
2169         fw_cfg_modify_i16(pcms->fw_cfg, FW_CFG_NB_CPUS, pcms->boot_cpus);
2170     }
2171 
2172     found_cpu = pc_find_cpu_slot(MACHINE(pcms), cpu->apic_id, NULL);
2173     found_cpu->cpu = OBJECT(dev);
2174 out:
2175     error_propagate(errp, local_err);
2176 }
2177 static void pc_cpu_unplug_request_cb(HotplugHandler *hotplug_dev,
2178                                      DeviceState *dev, Error **errp)
2179 {
2180     int idx = -1;
2181     HotplugHandlerClass *hhc;
2182     Error *local_err = NULL;
2183     X86CPU *cpu = X86_CPU(dev);
2184     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
2185 
2186     if (!pcms->acpi_dev) {
2187         error_setg(&local_err, "CPU hot unplug not supported without ACPI");
2188         goto out;
2189     }
2190 
2191     pc_find_cpu_slot(MACHINE(pcms), cpu->apic_id, &idx);
2192     assert(idx != -1);
2193     if (idx == 0) {
2194         error_setg(&local_err, "Boot CPU is unpluggable");
2195         goto out;
2196     }
2197 
2198     hhc = HOTPLUG_HANDLER_GET_CLASS(pcms->acpi_dev);
2199     hhc->unplug_request(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &local_err);
2200 
2201     if (local_err) {
2202         goto out;
2203     }
2204 
2205  out:
2206     error_propagate(errp, local_err);
2207 
2208 }
2209 
2210 static void pc_cpu_unplug_cb(HotplugHandler *hotplug_dev,
2211                              DeviceState *dev, Error **errp)
2212 {
2213     CPUArchId *found_cpu;
2214     HotplugHandlerClass *hhc;
2215     Error *local_err = NULL;
2216     X86CPU *cpu = X86_CPU(dev);
2217     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
2218 
2219     hhc = HOTPLUG_HANDLER_GET_CLASS(pcms->acpi_dev);
2220     hhc->unplug(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &local_err);
2221 
2222     if (local_err) {
2223         goto out;
2224     }
2225 
2226     found_cpu = pc_find_cpu_slot(MACHINE(pcms), cpu->apic_id, NULL);
2227     found_cpu->cpu = NULL;
2228     object_unparent(OBJECT(dev));
2229 
2230     /* decrement the number of CPUs */
2231     pcms->boot_cpus--;
2232     /* Update the number of CPUs in CMOS */
2233     rtc_set_cpus_count(pcms->rtc, pcms->boot_cpus);
2234     fw_cfg_modify_i16(pcms->fw_cfg, FW_CFG_NB_CPUS, pcms->boot_cpus);
2235  out:
2236     error_propagate(errp, local_err);
2237 }
2238 
2239 static void pc_cpu_pre_plug(HotplugHandler *hotplug_dev,
2240                             DeviceState *dev, Error **errp)
2241 {
2242     int idx;
2243     CPUState *cs;
2244     CPUArchId *cpu_slot;
2245     X86CPUTopoInfo topo;
2246     X86CPU *cpu = X86_CPU(dev);
2247     MachineState *ms = MACHINE(hotplug_dev);
2248     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
2249 
2250     if(!object_dynamic_cast(OBJECT(cpu), ms->cpu_type)) {
2251         error_setg(errp, "Invalid CPU type, expected cpu type: '%s'",
2252                    ms->cpu_type);
2253         return;
2254     }
2255 
2256     /* if APIC ID is not set, set it based on socket/core/thread properties */
2257     if (cpu->apic_id == UNASSIGNED_APIC_ID) {
2258         int max_socket = (max_cpus - 1) / smp_threads / smp_cores;
2259 
2260         if (cpu->socket_id < 0) {
2261             error_setg(errp, "CPU socket-id is not set");
2262             return;
2263         } else if (cpu->socket_id > max_socket) {
2264             error_setg(errp, "Invalid CPU socket-id: %u must be in range 0:%u",
2265                        cpu->socket_id, max_socket);
2266             return;
2267         }
2268         if (cpu->core_id < 0) {
2269             error_setg(errp, "CPU core-id is not set");
2270             return;
2271         } else if (cpu->core_id > (smp_cores - 1)) {
2272             error_setg(errp, "Invalid CPU core-id: %u must be in range 0:%u",
2273                        cpu->core_id, smp_cores - 1);
2274             return;
2275         }
2276         if (cpu->thread_id < 0) {
2277             error_setg(errp, "CPU thread-id is not set");
2278             return;
2279         } else if (cpu->thread_id > (smp_threads - 1)) {
2280             error_setg(errp, "Invalid CPU thread-id: %u must be in range 0:%u",
2281                        cpu->thread_id, smp_threads - 1);
2282             return;
2283         }
2284 
2285         topo.pkg_id = cpu->socket_id;
2286         topo.core_id = cpu->core_id;
2287         topo.smt_id = cpu->thread_id;
2288         cpu->apic_id = apicid_from_topo_ids(smp_cores, smp_threads, &topo);
2289     }
2290 
2291     cpu_slot = pc_find_cpu_slot(MACHINE(pcms), cpu->apic_id, &idx);
2292     if (!cpu_slot) {
2293         MachineState *ms = MACHINE(pcms);
2294 
2295         x86_topo_ids_from_apicid(cpu->apic_id, smp_cores, smp_threads, &topo);
2296         error_setg(errp, "Invalid CPU [socket: %u, core: %u, thread: %u] with"
2297                   " APIC ID %" PRIu32 ", valid index range 0:%d",
2298                    topo.pkg_id, topo.core_id, topo.smt_id, cpu->apic_id,
2299                    ms->possible_cpus->len - 1);
2300         return;
2301     }
2302 
2303     if (cpu_slot->cpu) {
2304         error_setg(errp, "CPU[%d] with APIC ID %" PRIu32 " exists",
2305                    idx, cpu->apic_id);
2306         return;
2307     }
2308 
2309     /* if 'address' properties socket-id/core-id/thread-id are not set, set them
2310      * so that machine_query_hotpluggable_cpus would show correct values
2311      */
2312     /* TODO: move socket_id/core_id/thread_id checks into x86_cpu_realizefn()
2313      * once -smp refactoring is complete and there will be CPU private
2314      * CPUState::nr_cores and CPUState::nr_threads fields instead of globals */
2315     x86_topo_ids_from_apicid(cpu->apic_id, smp_cores, smp_threads, &topo);
2316     if (cpu->socket_id != -1 && cpu->socket_id != topo.pkg_id) {
2317         error_setg(errp, "property socket-id: %u doesn't match set apic-id:"
2318             " 0x%x (socket-id: %u)", cpu->socket_id, cpu->apic_id, topo.pkg_id);
2319         return;
2320     }
2321     cpu->socket_id = topo.pkg_id;
2322 
2323     if (cpu->core_id != -1 && cpu->core_id != topo.core_id) {
2324         error_setg(errp, "property core-id: %u doesn't match set apic-id:"
2325             " 0x%x (core-id: %u)", cpu->core_id, cpu->apic_id, topo.core_id);
2326         return;
2327     }
2328     cpu->core_id = topo.core_id;
2329 
2330     if (cpu->thread_id != -1 && cpu->thread_id != topo.smt_id) {
2331         error_setg(errp, "property thread-id: %u doesn't match set apic-id:"
2332             " 0x%x (thread-id: %u)", cpu->thread_id, cpu->apic_id, topo.smt_id);
2333         return;
2334     }
2335     cpu->thread_id = topo.smt_id;
2336 
2337     if (cpu->hyperv_vpindex && !kvm_hv_vpindex_settable()) {
2338         error_setg(errp, "kernel doesn't allow setting HyperV VP_INDEX");
2339         return;
2340     }
2341 
2342     cs = CPU(cpu);
2343     cs->cpu_index = idx;
2344 
2345     numa_cpu_pre_plug(cpu_slot, dev, errp);
2346 }
2347 
2348 static void pc_machine_device_pre_plug_cb(HotplugHandler *hotplug_dev,
2349                                           DeviceState *dev, Error **errp)
2350 {
2351     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2352         pc_memory_pre_plug(hotplug_dev, dev, errp);
2353     } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
2354         pc_cpu_pre_plug(hotplug_dev, dev, errp);
2355     }
2356 }
2357 
2358 static void pc_machine_device_plug_cb(HotplugHandler *hotplug_dev,
2359                                       DeviceState *dev, Error **errp)
2360 {
2361     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2362         pc_memory_plug(hotplug_dev, dev, errp);
2363     } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
2364         pc_cpu_plug(hotplug_dev, dev, errp);
2365     }
2366 }
2367 
2368 static void pc_machine_device_unplug_request_cb(HotplugHandler *hotplug_dev,
2369                                                 DeviceState *dev, Error **errp)
2370 {
2371     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2372         pc_memory_unplug_request(hotplug_dev, dev, errp);
2373     } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
2374         pc_cpu_unplug_request_cb(hotplug_dev, dev, errp);
2375     } else {
2376         error_setg(errp, "acpi: device unplug request for not supported device"
2377                    " type: %s", object_get_typename(OBJECT(dev)));
2378     }
2379 }
2380 
2381 static void pc_machine_device_unplug_cb(HotplugHandler *hotplug_dev,
2382                                         DeviceState *dev, Error **errp)
2383 {
2384     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2385         pc_memory_unplug(hotplug_dev, dev, errp);
2386     } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
2387         pc_cpu_unplug_cb(hotplug_dev, dev, errp);
2388     } else {
2389         error_setg(errp, "acpi: device unplug for not supported device"
2390                    " type: %s", object_get_typename(OBJECT(dev)));
2391     }
2392 }
2393 
2394 static HotplugHandler *pc_get_hotpug_handler(MachineState *machine,
2395                                              DeviceState *dev)
2396 {
2397     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM) ||
2398         object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
2399         return HOTPLUG_HANDLER(machine);
2400     }
2401 
2402     return NULL;
2403 }
2404 
2405 static void
2406 pc_machine_get_device_memory_region_size(Object *obj, Visitor *v,
2407                                          const char *name, void *opaque,
2408                                          Error **errp)
2409 {
2410     MachineState *ms = MACHINE(obj);
2411     int64_t value = memory_region_size(&ms->device_memory->mr);
2412 
2413     visit_type_int(v, name, &value, errp);
2414 }
2415 
2416 static void pc_machine_get_max_ram_below_4g(Object *obj, Visitor *v,
2417                                             const char *name, void *opaque,
2418                                             Error **errp)
2419 {
2420     PCMachineState *pcms = PC_MACHINE(obj);
2421     uint64_t value = pcms->max_ram_below_4g;
2422 
2423     visit_type_size(v, name, &value, errp);
2424 }
2425 
2426 static void pc_machine_set_max_ram_below_4g(Object *obj, Visitor *v,
2427                                             const char *name, void *opaque,
2428                                             Error **errp)
2429 {
2430     PCMachineState *pcms = PC_MACHINE(obj);
2431     Error *error = NULL;
2432     uint64_t value;
2433 
2434     visit_type_size(v, name, &value, &error);
2435     if (error) {
2436         error_propagate(errp, error);
2437         return;
2438     }
2439     if (value > 4 * GiB) {
2440         error_setg(&error,
2441                    "Machine option 'max-ram-below-4g=%"PRIu64
2442                    "' expects size less than or equal to 4G", value);
2443         error_propagate(errp, error);
2444         return;
2445     }
2446 
2447     if (value < 1 * MiB) {
2448         warn_report("Only %" PRIu64 " bytes of RAM below the 4GiB boundary,"
2449                     "BIOS may not work with less than 1MiB", value);
2450     }
2451 
2452     pcms->max_ram_below_4g = value;
2453 }
2454 
2455 static void pc_machine_get_vmport(Object *obj, Visitor *v, const char *name,
2456                                   void *opaque, Error **errp)
2457 {
2458     PCMachineState *pcms = PC_MACHINE(obj);
2459     OnOffAuto vmport = pcms->vmport;
2460 
2461     visit_type_OnOffAuto(v, name, &vmport, errp);
2462 }
2463 
2464 static void pc_machine_set_vmport(Object *obj, Visitor *v, const char *name,
2465                                   void *opaque, Error **errp)
2466 {
2467     PCMachineState *pcms = PC_MACHINE(obj);
2468 
2469     visit_type_OnOffAuto(v, name, &pcms->vmport, errp);
2470 }
2471 
2472 bool pc_machine_is_smm_enabled(PCMachineState *pcms)
2473 {
2474     bool smm_available = false;
2475 
2476     if (pcms->smm == ON_OFF_AUTO_OFF) {
2477         return false;
2478     }
2479 
2480     if (tcg_enabled() || qtest_enabled()) {
2481         smm_available = true;
2482     } else if (kvm_enabled()) {
2483         smm_available = kvm_has_smm();
2484     }
2485 
2486     if (smm_available) {
2487         return true;
2488     }
2489 
2490     if (pcms->smm == ON_OFF_AUTO_ON) {
2491         error_report("System Management Mode not supported by this hypervisor.");
2492         exit(1);
2493     }
2494     return false;
2495 }
2496 
2497 static void pc_machine_get_smm(Object *obj, Visitor *v, const char *name,
2498                                void *opaque, Error **errp)
2499 {
2500     PCMachineState *pcms = PC_MACHINE(obj);
2501     OnOffAuto smm = pcms->smm;
2502 
2503     visit_type_OnOffAuto(v, name, &smm, errp);
2504 }
2505 
2506 static void pc_machine_set_smm(Object *obj, Visitor *v, const char *name,
2507                                void *opaque, Error **errp)
2508 {
2509     PCMachineState *pcms = PC_MACHINE(obj);
2510 
2511     visit_type_OnOffAuto(v, name, &pcms->smm, errp);
2512 }
2513 
2514 static bool pc_machine_get_nvdimm(Object *obj, Error **errp)
2515 {
2516     PCMachineState *pcms = PC_MACHINE(obj);
2517 
2518     return pcms->acpi_nvdimm_state.is_enabled;
2519 }
2520 
2521 static void pc_machine_set_nvdimm(Object *obj, bool value, Error **errp)
2522 {
2523     PCMachineState *pcms = PC_MACHINE(obj);
2524 
2525     pcms->acpi_nvdimm_state.is_enabled = value;
2526 }
2527 
2528 static char *pc_machine_get_nvdimm_persistence(Object *obj, Error **errp)
2529 {
2530     PCMachineState *pcms = PC_MACHINE(obj);
2531 
2532     return g_strdup(pcms->acpi_nvdimm_state.persistence_string);
2533 }
2534 
2535 static void pc_machine_set_nvdimm_persistence(Object *obj, const char *value,
2536                                                Error **errp)
2537 {
2538     PCMachineState *pcms = PC_MACHINE(obj);
2539     AcpiNVDIMMState *nvdimm_state = &pcms->acpi_nvdimm_state;
2540 
2541     if (strcmp(value, "cpu") == 0)
2542         nvdimm_state->persistence = 3;
2543     else if (strcmp(value, "mem-ctrl") == 0)
2544         nvdimm_state->persistence = 2;
2545     else {
2546         error_setg(errp, "-machine nvdimm-persistence=%s: unsupported option",
2547                    value);
2548         return;
2549     }
2550 
2551     g_free(nvdimm_state->persistence_string);
2552     nvdimm_state->persistence_string = g_strdup(value);
2553 }
2554 
2555 static bool pc_machine_get_smbus(Object *obj, Error **errp)
2556 {
2557     PCMachineState *pcms = PC_MACHINE(obj);
2558 
2559     return pcms->smbus_enabled;
2560 }
2561 
2562 static void pc_machine_set_smbus(Object *obj, bool value, Error **errp)
2563 {
2564     PCMachineState *pcms = PC_MACHINE(obj);
2565 
2566     pcms->smbus_enabled = value;
2567 }
2568 
2569 static bool pc_machine_get_sata(Object *obj, Error **errp)
2570 {
2571     PCMachineState *pcms = PC_MACHINE(obj);
2572 
2573     return pcms->sata_enabled;
2574 }
2575 
2576 static void pc_machine_set_sata(Object *obj, bool value, Error **errp)
2577 {
2578     PCMachineState *pcms = PC_MACHINE(obj);
2579 
2580     pcms->sata_enabled = value;
2581 }
2582 
2583 static bool pc_machine_get_pit(Object *obj, Error **errp)
2584 {
2585     PCMachineState *pcms = PC_MACHINE(obj);
2586 
2587     return pcms->pit_enabled;
2588 }
2589 
2590 static void pc_machine_set_pit(Object *obj, bool value, Error **errp)
2591 {
2592     PCMachineState *pcms = PC_MACHINE(obj);
2593 
2594     pcms->pit_enabled = value;
2595 }
2596 
2597 static void pc_machine_initfn(Object *obj)
2598 {
2599     PCMachineState *pcms = PC_MACHINE(obj);
2600 
2601     pcms->max_ram_below_4g = 0; /* use default */
2602     pcms->smm = ON_OFF_AUTO_AUTO;
2603     pcms->vmport = ON_OFF_AUTO_AUTO;
2604     /* nvdimm is disabled on default. */
2605     pcms->acpi_nvdimm_state.is_enabled = false;
2606     /* acpi build is enabled by default if machine supports it */
2607     pcms->acpi_build_enabled = PC_MACHINE_GET_CLASS(pcms)->has_acpi_build;
2608     pcms->smbus_enabled = true;
2609     pcms->sata_enabled = true;
2610     pcms->pit_enabled = true;
2611 }
2612 
2613 static void pc_machine_reset(void)
2614 {
2615     CPUState *cs;
2616     X86CPU *cpu;
2617 
2618     qemu_devices_reset();
2619 
2620     /* Reset APIC after devices have been reset to cancel
2621      * any changes that qemu_devices_reset() might have done.
2622      */
2623     CPU_FOREACH(cs) {
2624         cpu = X86_CPU(cs);
2625 
2626         if (cpu->apic_state) {
2627             device_reset(cpu->apic_state);
2628         }
2629     }
2630 }
2631 
2632 static CpuInstanceProperties
2633 pc_cpu_index_to_props(MachineState *ms, unsigned cpu_index)
2634 {
2635     MachineClass *mc = MACHINE_GET_CLASS(ms);
2636     const CPUArchIdList *possible_cpus = mc->possible_cpu_arch_ids(ms);
2637 
2638     assert(cpu_index < possible_cpus->len);
2639     return possible_cpus->cpus[cpu_index].props;
2640 }
2641 
2642 static int64_t pc_get_default_cpu_node_id(const MachineState *ms, int idx)
2643 {
2644    X86CPUTopoInfo topo;
2645 
2646    assert(idx < ms->possible_cpus->len);
2647    x86_topo_ids_from_apicid(ms->possible_cpus->cpus[idx].arch_id,
2648                             smp_cores, smp_threads, &topo);
2649    return topo.pkg_id % nb_numa_nodes;
2650 }
2651 
2652 static const CPUArchIdList *pc_possible_cpu_arch_ids(MachineState *ms)
2653 {
2654     int i;
2655 
2656     if (ms->possible_cpus) {
2657         /*
2658          * make sure that max_cpus hasn't changed since the first use, i.e.
2659          * -smp hasn't been parsed after it
2660         */
2661         assert(ms->possible_cpus->len == max_cpus);
2662         return ms->possible_cpus;
2663     }
2664 
2665     ms->possible_cpus = g_malloc0(sizeof(CPUArchIdList) +
2666                                   sizeof(CPUArchId) * max_cpus);
2667     ms->possible_cpus->len = max_cpus;
2668     for (i = 0; i < ms->possible_cpus->len; i++) {
2669         X86CPUTopoInfo topo;
2670 
2671         ms->possible_cpus->cpus[i].type = ms->cpu_type;
2672         ms->possible_cpus->cpus[i].vcpus_count = 1;
2673         ms->possible_cpus->cpus[i].arch_id = x86_cpu_apic_id_from_index(i);
2674         x86_topo_ids_from_apicid(ms->possible_cpus->cpus[i].arch_id,
2675                                  smp_cores, smp_threads, &topo);
2676         ms->possible_cpus->cpus[i].props.has_socket_id = true;
2677         ms->possible_cpus->cpus[i].props.socket_id = topo.pkg_id;
2678         ms->possible_cpus->cpus[i].props.has_core_id = true;
2679         ms->possible_cpus->cpus[i].props.core_id = topo.core_id;
2680         ms->possible_cpus->cpus[i].props.has_thread_id = true;
2681         ms->possible_cpus->cpus[i].props.thread_id = topo.smt_id;
2682     }
2683     return ms->possible_cpus;
2684 }
2685 
2686 static void x86_nmi(NMIState *n, int cpu_index, Error **errp)
2687 {
2688     /* cpu index isn't used */
2689     CPUState *cs;
2690 
2691     CPU_FOREACH(cs) {
2692         X86CPU *cpu = X86_CPU(cs);
2693 
2694         if (!cpu->apic_state) {
2695             cpu_interrupt(cs, CPU_INTERRUPT_NMI);
2696         } else {
2697             apic_deliver_nmi(cpu->apic_state);
2698         }
2699     }
2700 }
2701 
2702 static void pc_machine_class_init(ObjectClass *oc, void *data)
2703 {
2704     MachineClass *mc = MACHINE_CLASS(oc);
2705     PCMachineClass *pcmc = PC_MACHINE_CLASS(oc);
2706     HotplugHandlerClass *hc = HOTPLUG_HANDLER_CLASS(oc);
2707     NMIClass *nc = NMI_CLASS(oc);
2708 
2709     pcmc->pci_enabled = true;
2710     pcmc->has_acpi_build = true;
2711     pcmc->rsdp_in_ram = true;
2712     pcmc->smbios_defaults = true;
2713     pcmc->smbios_uuid_encoded = true;
2714     pcmc->gigabyte_align = true;
2715     pcmc->has_reserved_memory = true;
2716     pcmc->kvmclock_enabled = true;
2717     pcmc->enforce_aligned_dimm = true;
2718     /* BIOS ACPI tables: 128K. Other BIOS datastructures: less than 4K reported
2719      * to be used at the moment, 32K should be enough for a while.  */
2720     pcmc->acpi_data_size = 0x20000 + 0x8000;
2721     pcmc->save_tsc_khz = true;
2722     pcmc->linuxboot_dma_enabled = true;
2723     assert(!mc->get_hotplug_handler);
2724     mc->get_hotplug_handler = pc_get_hotpug_handler;
2725     mc->cpu_index_to_instance_props = pc_cpu_index_to_props;
2726     mc->get_default_cpu_node_id = pc_get_default_cpu_node_id;
2727     mc->possible_cpu_arch_ids = pc_possible_cpu_arch_ids;
2728     mc->auto_enable_numa_with_memhp = true;
2729     mc->has_hotpluggable_cpus = true;
2730     mc->default_boot_order = "cad";
2731     mc->hot_add_cpu = pc_hot_add_cpu;
2732     mc->block_default_type = IF_IDE;
2733     mc->max_cpus = 255;
2734     mc->reset = pc_machine_reset;
2735     hc->pre_plug = pc_machine_device_pre_plug_cb;
2736     hc->plug = pc_machine_device_plug_cb;
2737     hc->unplug_request = pc_machine_device_unplug_request_cb;
2738     hc->unplug = pc_machine_device_unplug_cb;
2739     nc->nmi_monitor_handler = x86_nmi;
2740     mc->default_cpu_type = TARGET_DEFAULT_CPU_TYPE;
2741 
2742     object_class_property_add(oc, PC_MACHINE_DEVMEM_REGION_SIZE, "int",
2743         pc_machine_get_device_memory_region_size, NULL,
2744         NULL, NULL, &error_abort);
2745 
2746     object_class_property_add(oc, PC_MACHINE_MAX_RAM_BELOW_4G, "size",
2747         pc_machine_get_max_ram_below_4g, pc_machine_set_max_ram_below_4g,
2748         NULL, NULL, &error_abort);
2749 
2750     object_class_property_set_description(oc, PC_MACHINE_MAX_RAM_BELOW_4G,
2751         "Maximum ram below the 4G boundary (32bit boundary)", &error_abort);
2752 
2753     object_class_property_add(oc, PC_MACHINE_SMM, "OnOffAuto",
2754         pc_machine_get_smm, pc_machine_set_smm,
2755         NULL, NULL, &error_abort);
2756     object_class_property_set_description(oc, PC_MACHINE_SMM,
2757         "Enable SMM (pc & q35)", &error_abort);
2758 
2759     object_class_property_add(oc, PC_MACHINE_VMPORT, "OnOffAuto",
2760         pc_machine_get_vmport, pc_machine_set_vmport,
2761         NULL, NULL, &error_abort);
2762     object_class_property_set_description(oc, PC_MACHINE_VMPORT,
2763         "Enable vmport (pc & q35)", &error_abort);
2764 
2765     object_class_property_add_bool(oc, PC_MACHINE_NVDIMM,
2766         pc_machine_get_nvdimm, pc_machine_set_nvdimm, &error_abort);
2767 
2768     object_class_property_add_str(oc, PC_MACHINE_NVDIMM_PERSIST,
2769         pc_machine_get_nvdimm_persistence,
2770         pc_machine_set_nvdimm_persistence, &error_abort);
2771 
2772     object_class_property_add_bool(oc, PC_MACHINE_SMBUS,
2773         pc_machine_get_smbus, pc_machine_set_smbus, &error_abort);
2774 
2775     object_class_property_add_bool(oc, PC_MACHINE_SATA,
2776         pc_machine_get_sata, pc_machine_set_sata, &error_abort);
2777 
2778     object_class_property_add_bool(oc, PC_MACHINE_PIT,
2779         pc_machine_get_pit, pc_machine_set_pit, &error_abort);
2780 }
2781 
2782 static const TypeInfo pc_machine_info = {
2783     .name = TYPE_PC_MACHINE,
2784     .parent = TYPE_MACHINE,
2785     .abstract = true,
2786     .instance_size = sizeof(PCMachineState),
2787     .instance_init = pc_machine_initfn,
2788     .class_size = sizeof(PCMachineClass),
2789     .class_init = pc_machine_class_init,
2790     .interfaces = (InterfaceInfo[]) {
2791          { TYPE_HOTPLUG_HANDLER },
2792          { TYPE_NMI },
2793          { }
2794     },
2795 };
2796 
2797 static void pc_machine_register_types(void)
2798 {
2799     type_register_static(&pc_machine_info);
2800 }
2801 
2802 type_init(pc_machine_register_types)
2803