xref: /openbmc/qemu/hw/i386/pc.c (revision 461a2753)
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 #include "hw/hw.h"
25 #include "hw/i386/pc.h"
26 #include "hw/char/serial.h"
27 #include "hw/i386/apic.h"
28 #include "hw/block/fdc.h"
29 #include "hw/ide.h"
30 #include "hw/pci/pci.h"
31 #include "monitor/monitor.h"
32 #include "hw/nvram/fw_cfg.h"
33 #include "hw/timer/hpet.h"
34 #include "hw/i386/smbios.h"
35 #include "hw/loader.h"
36 #include "elf.h"
37 #include "multiboot.h"
38 #include "hw/timer/mc146818rtc.h"
39 #include "hw/timer/i8254.h"
40 #include "hw/audio/pcspk.h"
41 #include "hw/pci/msi.h"
42 #include "hw/sysbus.h"
43 #include "sysemu/sysemu.h"
44 #include "sysemu/kvm.h"
45 #include "kvm_i386.h"
46 #include "hw/xen/xen.h"
47 #include "sysemu/blockdev.h"
48 #include "hw/block/block.h"
49 #include "ui/qemu-spice.h"
50 #include "exec/memory.h"
51 #include "exec/address-spaces.h"
52 #include "sysemu/arch_init.h"
53 #include "qemu/bitmap.h"
54 #include "qemu/config-file.h"
55 #include "hw/acpi/acpi.h"
56 #include "hw/acpi/cpu_hotplug.h"
57 #include "hw/cpu/icc_bus.h"
58 #include "hw/boards.h"
59 #include "hw/pci/pci_host.h"
60 #include "acpi-build.h"
61 #include "hw/mem/pc-dimm.h"
62 #include "trace.h"
63 #include "qapi/visitor.h"
64 
65 /* debug PC/ISA interrupts */
66 //#define DEBUG_IRQ
67 
68 #ifdef DEBUG_IRQ
69 #define DPRINTF(fmt, ...)                                       \
70     do { printf("CPUIRQ: " fmt , ## __VA_ARGS__); } while (0)
71 #else
72 #define DPRINTF(fmt, ...)
73 #endif
74 
75 /* Leave a chunk of memory at the top of RAM for the BIOS ACPI tables.  */
76 unsigned acpi_data_size = 0x20000;
77 void pc_set_legacy_acpi_data_size(void)
78 {
79     acpi_data_size = 0x10000;
80 }
81 
82 #define BIOS_CFG_IOPORT 0x510
83 #define FW_CFG_ACPI_TABLES (FW_CFG_ARCH_LOCAL + 0)
84 #define FW_CFG_SMBIOS_ENTRIES (FW_CFG_ARCH_LOCAL + 1)
85 #define FW_CFG_IRQ0_OVERRIDE (FW_CFG_ARCH_LOCAL + 2)
86 #define FW_CFG_E820_TABLE (FW_CFG_ARCH_LOCAL + 3)
87 #define FW_CFG_HPET (FW_CFG_ARCH_LOCAL + 4)
88 
89 #define E820_NR_ENTRIES		16
90 
91 struct e820_entry {
92     uint64_t address;
93     uint64_t length;
94     uint32_t type;
95 } QEMU_PACKED __attribute((__aligned__(4)));
96 
97 struct e820_table {
98     uint32_t count;
99     struct e820_entry entry[E820_NR_ENTRIES];
100 } QEMU_PACKED __attribute((__aligned__(4)));
101 
102 static struct e820_table e820_reserve;
103 static struct e820_entry *e820_table;
104 static unsigned e820_entries;
105 struct hpet_fw_config hpet_cfg = {.count = UINT8_MAX};
106 
107 void gsi_handler(void *opaque, int n, int level)
108 {
109     GSIState *s = opaque;
110 
111     DPRINTF("pc: %s GSI %d\n", level ? "raising" : "lowering", n);
112     if (n < ISA_NUM_IRQS) {
113         qemu_set_irq(s->i8259_irq[n], level);
114     }
115     qemu_set_irq(s->ioapic_irq[n], level);
116 }
117 
118 static void ioport80_write(void *opaque, hwaddr addr, uint64_t data,
119                            unsigned size)
120 {
121 }
122 
123 static uint64_t ioport80_read(void *opaque, hwaddr addr, unsigned size)
124 {
125     return 0xffffffffffffffffULL;
126 }
127 
128 /* MSDOS compatibility mode FPU exception support */
129 static qemu_irq ferr_irq;
130 
131 void pc_register_ferr_irq(qemu_irq irq)
132 {
133     ferr_irq = irq;
134 }
135 
136 /* XXX: add IGNNE support */
137 void cpu_set_ferr(CPUX86State *s)
138 {
139     qemu_irq_raise(ferr_irq);
140 }
141 
142 static void ioportF0_write(void *opaque, hwaddr addr, uint64_t data,
143                            unsigned size)
144 {
145     qemu_irq_lower(ferr_irq);
146 }
147 
148 static uint64_t ioportF0_read(void *opaque, hwaddr addr, unsigned size)
149 {
150     return 0xffffffffffffffffULL;
151 }
152 
153 /* TSC handling */
154 uint64_t cpu_get_tsc(CPUX86State *env)
155 {
156     return cpu_get_ticks();
157 }
158 
159 /* SMM support */
160 
161 static cpu_set_smm_t smm_set;
162 static void *smm_arg;
163 
164 void cpu_smm_register(cpu_set_smm_t callback, void *arg)
165 {
166     assert(smm_set == NULL);
167     assert(smm_arg == NULL);
168     smm_set = callback;
169     smm_arg = arg;
170 }
171 
172 void cpu_smm_update(CPUX86State *env)
173 {
174     if (smm_set && smm_arg && CPU(x86_env_get_cpu(env)) == first_cpu) {
175         smm_set(!!(env->hflags & HF_SMM_MASK), smm_arg);
176     }
177 }
178 
179 
180 /* IRQ handling */
181 int cpu_get_pic_interrupt(CPUX86State *env)
182 {
183     X86CPU *cpu = x86_env_get_cpu(env);
184     int intno;
185 
186     intno = apic_get_interrupt(cpu->apic_state);
187     if (intno >= 0) {
188         return intno;
189     }
190     /* read the irq from the PIC */
191     if (!apic_accept_pic_intr(cpu->apic_state)) {
192         return -1;
193     }
194 
195     intno = pic_read_irq(isa_pic);
196     return intno;
197 }
198 
199 static void pic_irq_request(void *opaque, int irq, int level)
200 {
201     CPUState *cs = first_cpu;
202     X86CPU *cpu = X86_CPU(cs);
203 
204     DPRINTF("pic_irqs: %s irq %d\n", level? "raise" : "lower", irq);
205     if (cpu->apic_state) {
206         CPU_FOREACH(cs) {
207             cpu = X86_CPU(cs);
208             if (apic_accept_pic_intr(cpu->apic_state)) {
209                 apic_deliver_pic_intr(cpu->apic_state, level);
210             }
211         }
212     } else {
213         if (level) {
214             cpu_interrupt(cs, CPU_INTERRUPT_HARD);
215         } else {
216             cpu_reset_interrupt(cs, CPU_INTERRUPT_HARD);
217         }
218     }
219 }
220 
221 /* PC cmos mappings */
222 
223 #define REG_EQUIPMENT_BYTE          0x14
224 
225 static int cmos_get_fd_drive_type(FDriveType fd0)
226 {
227     int val;
228 
229     switch (fd0) {
230     case FDRIVE_DRV_144:
231         /* 1.44 Mb 3"5 drive */
232         val = 4;
233         break;
234     case FDRIVE_DRV_288:
235         /* 2.88 Mb 3"5 drive */
236         val = 5;
237         break;
238     case FDRIVE_DRV_120:
239         /* 1.2 Mb 5"5 drive */
240         val = 2;
241         break;
242     case FDRIVE_DRV_NONE:
243     default:
244         val = 0;
245         break;
246     }
247     return val;
248 }
249 
250 static void cmos_init_hd(ISADevice *s, int type_ofs, int info_ofs,
251                          int16_t cylinders, int8_t heads, int8_t sectors)
252 {
253     rtc_set_memory(s, type_ofs, 47);
254     rtc_set_memory(s, info_ofs, cylinders);
255     rtc_set_memory(s, info_ofs + 1, cylinders >> 8);
256     rtc_set_memory(s, info_ofs + 2, heads);
257     rtc_set_memory(s, info_ofs + 3, 0xff);
258     rtc_set_memory(s, info_ofs + 4, 0xff);
259     rtc_set_memory(s, info_ofs + 5, 0xc0 | ((heads > 8) << 3));
260     rtc_set_memory(s, info_ofs + 6, cylinders);
261     rtc_set_memory(s, info_ofs + 7, cylinders >> 8);
262     rtc_set_memory(s, info_ofs + 8, sectors);
263 }
264 
265 /* convert boot_device letter to something recognizable by the bios */
266 static int boot_device2nibble(char boot_device)
267 {
268     switch(boot_device) {
269     case 'a':
270     case 'b':
271         return 0x01; /* floppy boot */
272     case 'c':
273         return 0x02; /* hard drive boot */
274     case 'd':
275         return 0x03; /* CD-ROM boot */
276     case 'n':
277         return 0x04; /* Network boot */
278     }
279     return 0;
280 }
281 
282 static int set_boot_dev(ISADevice *s, const char *boot_device)
283 {
284 #define PC_MAX_BOOT_DEVICES 3
285     int nbds, bds[3] = { 0, };
286     int i;
287 
288     nbds = strlen(boot_device);
289     if (nbds > PC_MAX_BOOT_DEVICES) {
290         error_report("Too many boot devices for PC");
291         return(1);
292     }
293     for (i = 0; i < nbds; i++) {
294         bds[i] = boot_device2nibble(boot_device[i]);
295         if (bds[i] == 0) {
296             error_report("Invalid boot device for PC: '%c'",
297                          boot_device[i]);
298             return(1);
299         }
300     }
301     rtc_set_memory(s, 0x3d, (bds[1] << 4) | bds[0]);
302     rtc_set_memory(s, 0x38, (bds[2] << 4) | (fd_bootchk ? 0x0 : 0x1));
303     return(0);
304 }
305 
306 static int pc_boot_set(void *opaque, const char *boot_device)
307 {
308     return set_boot_dev(opaque, boot_device);
309 }
310 
311 typedef struct pc_cmos_init_late_arg {
312     ISADevice *rtc_state;
313     BusState *idebus[2];
314 } pc_cmos_init_late_arg;
315 
316 static void pc_cmos_init_late(void *opaque)
317 {
318     pc_cmos_init_late_arg *arg = opaque;
319     ISADevice *s = arg->rtc_state;
320     int16_t cylinders;
321     int8_t heads, sectors;
322     int val;
323     int i, trans;
324 
325     val = 0;
326     if (ide_get_geometry(arg->idebus[0], 0,
327                          &cylinders, &heads, &sectors) >= 0) {
328         cmos_init_hd(s, 0x19, 0x1b, cylinders, heads, sectors);
329         val |= 0xf0;
330     }
331     if (ide_get_geometry(arg->idebus[0], 1,
332                          &cylinders, &heads, &sectors) >= 0) {
333         cmos_init_hd(s, 0x1a, 0x24, cylinders, heads, sectors);
334         val |= 0x0f;
335     }
336     rtc_set_memory(s, 0x12, val);
337 
338     val = 0;
339     for (i = 0; i < 4; i++) {
340         /* NOTE: ide_get_geometry() returns the physical
341            geometry.  It is always such that: 1 <= sects <= 63, 1
342            <= heads <= 16, 1 <= cylinders <= 16383. The BIOS
343            geometry can be different if a translation is done. */
344         if (ide_get_geometry(arg->idebus[i / 2], i % 2,
345                              &cylinders, &heads, &sectors) >= 0) {
346             trans = ide_get_bios_chs_trans(arg->idebus[i / 2], i % 2) - 1;
347             assert((trans & ~3) == 0);
348             val |= trans << (i * 2);
349         }
350     }
351     rtc_set_memory(s, 0x39, val);
352 
353     qemu_unregister_reset(pc_cmos_init_late, opaque);
354 }
355 
356 typedef struct RTCCPUHotplugArg {
357     Notifier cpu_added_notifier;
358     ISADevice *rtc_state;
359 } RTCCPUHotplugArg;
360 
361 static void rtc_notify_cpu_added(Notifier *notifier, void *data)
362 {
363     RTCCPUHotplugArg *arg = container_of(notifier, RTCCPUHotplugArg,
364                                          cpu_added_notifier);
365     ISADevice *s = arg->rtc_state;
366 
367     /* increment the number of CPUs */
368     rtc_set_memory(s, 0x5f, rtc_get_memory(s, 0x5f) + 1);
369 }
370 
371 void pc_cmos_init(ram_addr_t ram_size, ram_addr_t above_4g_mem_size,
372                   const char *boot_device,
373                   ISADevice *floppy, BusState *idebus0, BusState *idebus1,
374                   ISADevice *s)
375 {
376     int val, nb, i;
377     FDriveType fd_type[2] = { FDRIVE_DRV_NONE, FDRIVE_DRV_NONE };
378     static pc_cmos_init_late_arg arg;
379     static RTCCPUHotplugArg cpu_hotplug_cb;
380 
381     /* various important CMOS locations needed by PC/Bochs bios */
382 
383     /* memory size */
384     /* base memory (first MiB) */
385     val = MIN(ram_size / 1024, 640);
386     rtc_set_memory(s, 0x15, val);
387     rtc_set_memory(s, 0x16, val >> 8);
388     /* extended memory (next 64MiB) */
389     if (ram_size > 1024 * 1024) {
390         val = (ram_size - 1024 * 1024) / 1024;
391     } else {
392         val = 0;
393     }
394     if (val > 65535)
395         val = 65535;
396     rtc_set_memory(s, 0x17, val);
397     rtc_set_memory(s, 0x18, val >> 8);
398     rtc_set_memory(s, 0x30, val);
399     rtc_set_memory(s, 0x31, val >> 8);
400     /* memory between 16MiB and 4GiB */
401     if (ram_size > 16 * 1024 * 1024) {
402         val = (ram_size - 16 * 1024 * 1024) / 65536;
403     } else {
404         val = 0;
405     }
406     if (val > 65535)
407         val = 65535;
408     rtc_set_memory(s, 0x34, val);
409     rtc_set_memory(s, 0x35, val >> 8);
410     /* memory above 4GiB */
411     val = above_4g_mem_size / 65536;
412     rtc_set_memory(s, 0x5b, val);
413     rtc_set_memory(s, 0x5c, val >> 8);
414     rtc_set_memory(s, 0x5d, val >> 16);
415 
416     /* set the number of CPU */
417     rtc_set_memory(s, 0x5f, smp_cpus - 1);
418     /* init CPU hotplug notifier */
419     cpu_hotplug_cb.rtc_state = s;
420     cpu_hotplug_cb.cpu_added_notifier.notify = rtc_notify_cpu_added;
421     qemu_register_cpu_added_notifier(&cpu_hotplug_cb.cpu_added_notifier);
422 
423     if (set_boot_dev(s, boot_device)) {
424         exit(1);
425     }
426 
427     /* floppy type */
428     if (floppy) {
429         for (i = 0; i < 2; i++) {
430             fd_type[i] = isa_fdc_get_drive_type(floppy, i);
431         }
432     }
433     val = (cmos_get_fd_drive_type(fd_type[0]) << 4) |
434         cmos_get_fd_drive_type(fd_type[1]);
435     rtc_set_memory(s, 0x10, val);
436 
437     val = 0;
438     nb = 0;
439     if (fd_type[0] < FDRIVE_DRV_NONE) {
440         nb++;
441     }
442     if (fd_type[1] < FDRIVE_DRV_NONE) {
443         nb++;
444     }
445     switch (nb) {
446     case 0:
447         break;
448     case 1:
449         val |= 0x01; /* 1 drive, ready for boot */
450         break;
451     case 2:
452         val |= 0x41; /* 2 drives, ready for boot */
453         break;
454     }
455     val |= 0x02; /* FPU is there */
456     val |= 0x04; /* PS/2 mouse installed */
457     rtc_set_memory(s, REG_EQUIPMENT_BYTE, val);
458 
459     /* hard drives */
460     arg.rtc_state = s;
461     arg.idebus[0] = idebus0;
462     arg.idebus[1] = idebus1;
463     qemu_register_reset(pc_cmos_init_late, &arg);
464 }
465 
466 #define TYPE_PORT92 "port92"
467 #define PORT92(obj) OBJECT_CHECK(Port92State, (obj), TYPE_PORT92)
468 
469 /* port 92 stuff: could be split off */
470 typedef struct Port92State {
471     ISADevice parent_obj;
472 
473     MemoryRegion io;
474     uint8_t outport;
475     qemu_irq *a20_out;
476 } Port92State;
477 
478 static void port92_write(void *opaque, hwaddr addr, uint64_t val,
479                          unsigned size)
480 {
481     Port92State *s = opaque;
482     int oldval = s->outport;
483 
484     DPRINTF("port92: write 0x%02" PRIx64 "\n", val);
485     s->outport = val;
486     qemu_set_irq(*s->a20_out, (val >> 1) & 1);
487     if ((val & 1) && !(oldval & 1)) {
488         qemu_system_reset_request();
489     }
490 }
491 
492 static uint64_t port92_read(void *opaque, hwaddr addr,
493                             unsigned size)
494 {
495     Port92State *s = opaque;
496     uint32_t ret;
497 
498     ret = s->outport;
499     DPRINTF("port92: read 0x%02x\n", ret);
500     return ret;
501 }
502 
503 static void port92_init(ISADevice *dev, qemu_irq *a20_out)
504 {
505     Port92State *s = PORT92(dev);
506 
507     s->a20_out = a20_out;
508 }
509 
510 static const VMStateDescription vmstate_port92_isa = {
511     .name = "port92",
512     .version_id = 1,
513     .minimum_version_id = 1,
514     .fields = (VMStateField[]) {
515         VMSTATE_UINT8(outport, Port92State),
516         VMSTATE_END_OF_LIST()
517     }
518 };
519 
520 static void port92_reset(DeviceState *d)
521 {
522     Port92State *s = PORT92(d);
523 
524     s->outport &= ~1;
525 }
526 
527 static const MemoryRegionOps port92_ops = {
528     .read = port92_read,
529     .write = port92_write,
530     .impl = {
531         .min_access_size = 1,
532         .max_access_size = 1,
533     },
534     .endianness = DEVICE_LITTLE_ENDIAN,
535 };
536 
537 static void port92_initfn(Object *obj)
538 {
539     Port92State *s = PORT92(obj);
540 
541     memory_region_init_io(&s->io, OBJECT(s), &port92_ops, s, "port92", 1);
542 
543     s->outport = 0;
544 }
545 
546 static void port92_realizefn(DeviceState *dev, Error **errp)
547 {
548     ISADevice *isadev = ISA_DEVICE(dev);
549     Port92State *s = PORT92(dev);
550 
551     isa_register_ioport(isadev, &s->io, 0x92);
552 }
553 
554 static void port92_class_initfn(ObjectClass *klass, void *data)
555 {
556     DeviceClass *dc = DEVICE_CLASS(klass);
557 
558     dc->realize = port92_realizefn;
559     dc->reset = port92_reset;
560     dc->vmsd = &vmstate_port92_isa;
561     /*
562      * Reason: unlike ordinary ISA devices, this one needs additional
563      * wiring: its A20 output line needs to be wired up by
564      * port92_init().
565      */
566     dc->cannot_instantiate_with_device_add_yet = true;
567 }
568 
569 static const TypeInfo port92_info = {
570     .name          = TYPE_PORT92,
571     .parent        = TYPE_ISA_DEVICE,
572     .instance_size = sizeof(Port92State),
573     .instance_init = port92_initfn,
574     .class_init    = port92_class_initfn,
575 };
576 
577 static void port92_register_types(void)
578 {
579     type_register_static(&port92_info);
580 }
581 
582 type_init(port92_register_types)
583 
584 static void handle_a20_line_change(void *opaque, int irq, int level)
585 {
586     X86CPU *cpu = opaque;
587 
588     /* XXX: send to all CPUs ? */
589     /* XXX: add logic to handle multiple A20 line sources */
590     x86_cpu_set_a20(cpu, level);
591 }
592 
593 int e820_add_entry(uint64_t address, uint64_t length, uint32_t type)
594 {
595     int index = le32_to_cpu(e820_reserve.count);
596     struct e820_entry *entry;
597 
598     if (type != E820_RAM) {
599         /* old FW_CFG_E820_TABLE entry -- reservations only */
600         if (index >= E820_NR_ENTRIES) {
601             return -EBUSY;
602         }
603         entry = &e820_reserve.entry[index++];
604 
605         entry->address = cpu_to_le64(address);
606         entry->length = cpu_to_le64(length);
607         entry->type = cpu_to_le32(type);
608 
609         e820_reserve.count = cpu_to_le32(index);
610     }
611 
612     /* new "etc/e820" file -- include ram too */
613     e820_table = g_realloc(e820_table,
614                            sizeof(struct e820_entry) * (e820_entries+1));
615     e820_table[e820_entries].address = cpu_to_le64(address);
616     e820_table[e820_entries].length = cpu_to_le64(length);
617     e820_table[e820_entries].type = cpu_to_le32(type);
618     e820_entries++;
619 
620     return e820_entries;
621 }
622 
623 int e820_get_num_entries(void)
624 {
625     return e820_entries;
626 }
627 
628 bool e820_get_entry(int idx, uint32_t type, uint64_t *address, uint64_t *length)
629 {
630     if (idx < e820_entries && e820_table[idx].type == cpu_to_le32(type)) {
631         *address = le64_to_cpu(e820_table[idx].address);
632         *length = le64_to_cpu(e820_table[idx].length);
633         return true;
634     }
635     return false;
636 }
637 
638 /* Calculates the limit to CPU APIC ID values
639  *
640  * This function returns the limit for the APIC ID value, so that all
641  * CPU APIC IDs are < pc_apic_id_limit().
642  *
643  * This is used for FW_CFG_MAX_CPUS. See comments on bochs_bios_init().
644  */
645 static unsigned int pc_apic_id_limit(unsigned int max_cpus)
646 {
647     return x86_cpu_apic_id_from_index(max_cpus - 1) + 1;
648 }
649 
650 static FWCfgState *bochs_bios_init(void)
651 {
652     FWCfgState *fw_cfg;
653     uint8_t *smbios_tables, *smbios_anchor;
654     size_t smbios_tables_len, smbios_anchor_len;
655     uint64_t *numa_fw_cfg;
656     int i, j;
657     unsigned int apic_id_limit = pc_apic_id_limit(max_cpus);
658 
659     fw_cfg = fw_cfg_init(BIOS_CFG_IOPORT, BIOS_CFG_IOPORT + 1, 0, 0);
660     /* FW_CFG_MAX_CPUS is a bit confusing/problematic on x86:
661      *
662      * SeaBIOS needs FW_CFG_MAX_CPUS for CPU hotplug, but the CPU hotplug
663      * QEMU<->SeaBIOS interface is not based on the "CPU index", but on the APIC
664      * ID of hotplugged CPUs[1]. This means that FW_CFG_MAX_CPUS is not the
665      * "maximum number of CPUs", but the "limit to the APIC ID values SeaBIOS
666      * may see".
667      *
668      * So, this means we must not use max_cpus, here, but the maximum possible
669      * APIC ID value, plus one.
670      *
671      * [1] The only kind of "CPU identifier" used between SeaBIOS and QEMU is
672      *     the APIC ID, not the "CPU index"
673      */
674     fw_cfg_add_i16(fw_cfg, FW_CFG_MAX_CPUS, (uint16_t)apic_id_limit);
675     fw_cfg_add_i32(fw_cfg, FW_CFG_ID, 1);
676     fw_cfg_add_i64(fw_cfg, FW_CFG_RAM_SIZE, (uint64_t)ram_size);
677     fw_cfg_add_bytes(fw_cfg, FW_CFG_ACPI_TABLES,
678                      acpi_tables, acpi_tables_len);
679     fw_cfg_add_i32(fw_cfg, FW_CFG_IRQ0_OVERRIDE, kvm_allows_irq0_override());
680 
681     smbios_tables = smbios_get_table_legacy(&smbios_tables_len);
682     if (smbios_tables) {
683         fw_cfg_add_bytes(fw_cfg, FW_CFG_SMBIOS_ENTRIES,
684                          smbios_tables, smbios_tables_len);
685     }
686 
687     smbios_get_tables(&smbios_tables, &smbios_tables_len,
688                       &smbios_anchor, &smbios_anchor_len);
689     if (smbios_anchor) {
690         fw_cfg_add_file(fw_cfg, "etc/smbios/smbios-tables",
691                         smbios_tables, smbios_tables_len);
692         fw_cfg_add_file(fw_cfg, "etc/smbios/smbios-anchor",
693                         smbios_anchor, smbios_anchor_len);
694     }
695 
696     fw_cfg_add_bytes(fw_cfg, FW_CFG_E820_TABLE,
697                      &e820_reserve, sizeof(e820_reserve));
698     fw_cfg_add_file(fw_cfg, "etc/e820", e820_table,
699                     sizeof(struct e820_entry) * e820_entries);
700 
701     fw_cfg_add_bytes(fw_cfg, FW_CFG_HPET, &hpet_cfg, sizeof(hpet_cfg));
702     /* allocate memory for the NUMA channel: one (64bit) word for the number
703      * of nodes, one word for each VCPU->node and one word for each node to
704      * hold the amount of memory.
705      */
706     numa_fw_cfg = g_new0(uint64_t, 1 + apic_id_limit + nb_numa_nodes);
707     numa_fw_cfg[0] = cpu_to_le64(nb_numa_nodes);
708     for (i = 0; i < max_cpus; i++) {
709         unsigned int apic_id = x86_cpu_apic_id_from_index(i);
710         assert(apic_id < apic_id_limit);
711         for (j = 0; j < nb_numa_nodes; j++) {
712             if (test_bit(i, numa_info[j].node_cpu)) {
713                 numa_fw_cfg[apic_id + 1] = cpu_to_le64(j);
714                 break;
715             }
716         }
717     }
718     for (i = 0; i < nb_numa_nodes; i++) {
719         numa_fw_cfg[apic_id_limit + 1 + i] = cpu_to_le64(numa_info[i].node_mem);
720     }
721     fw_cfg_add_bytes(fw_cfg, FW_CFG_NUMA, numa_fw_cfg,
722                      (1 + apic_id_limit + nb_numa_nodes) *
723                      sizeof(*numa_fw_cfg));
724 
725     return fw_cfg;
726 }
727 
728 static long get_file_size(FILE *f)
729 {
730     long where, size;
731 
732     /* XXX: on Unix systems, using fstat() probably makes more sense */
733 
734     where = ftell(f);
735     fseek(f, 0, SEEK_END);
736     size = ftell(f);
737     fseek(f, where, SEEK_SET);
738 
739     return size;
740 }
741 
742 static void load_linux(FWCfgState *fw_cfg,
743                        const char *kernel_filename,
744                        const char *initrd_filename,
745                        const char *kernel_cmdline,
746                        hwaddr max_ram_size)
747 {
748     uint16_t protocol;
749     int setup_size, kernel_size, initrd_size = 0, cmdline_size;
750     uint32_t initrd_max;
751     uint8_t header[8192], *setup, *kernel, *initrd_data;
752     hwaddr real_addr, prot_addr, cmdline_addr, initrd_addr = 0;
753     FILE *f;
754     char *vmode;
755 
756     /* Align to 16 bytes as a paranoia measure */
757     cmdline_size = (strlen(kernel_cmdline)+16) & ~15;
758 
759     /* load the kernel header */
760     f = fopen(kernel_filename, "rb");
761     if (!f || !(kernel_size = get_file_size(f)) ||
762         fread(header, 1, MIN(ARRAY_SIZE(header), kernel_size), f) !=
763         MIN(ARRAY_SIZE(header), kernel_size)) {
764         fprintf(stderr, "qemu: could not load kernel '%s': %s\n",
765                 kernel_filename, strerror(errno));
766         exit(1);
767     }
768 
769     /* kernel protocol version */
770 #if 0
771     fprintf(stderr, "header magic: %#x\n", ldl_p(header+0x202));
772 #endif
773     if (ldl_p(header+0x202) == 0x53726448) {
774         protocol = lduw_p(header+0x206);
775     } else {
776         /* This looks like a multiboot kernel. If it is, let's stop
777            treating it like a Linux kernel. */
778         if (load_multiboot(fw_cfg, f, kernel_filename, initrd_filename,
779                            kernel_cmdline, kernel_size, header)) {
780             return;
781         }
782         protocol = 0;
783     }
784 
785     if (protocol < 0x200 || !(header[0x211] & 0x01)) {
786         /* Low kernel */
787         real_addr    = 0x90000;
788         cmdline_addr = 0x9a000 - cmdline_size;
789         prot_addr    = 0x10000;
790     } else if (protocol < 0x202) {
791         /* High but ancient kernel */
792         real_addr    = 0x90000;
793         cmdline_addr = 0x9a000 - cmdline_size;
794         prot_addr    = 0x100000;
795     } else {
796         /* High and recent kernel */
797         real_addr    = 0x10000;
798         cmdline_addr = 0x20000;
799         prot_addr    = 0x100000;
800     }
801 
802 #if 0
803     fprintf(stderr,
804             "qemu: real_addr     = 0x" TARGET_FMT_plx "\n"
805             "qemu: cmdline_addr  = 0x" TARGET_FMT_plx "\n"
806             "qemu: prot_addr     = 0x" TARGET_FMT_plx "\n",
807             real_addr,
808             cmdline_addr,
809             prot_addr);
810 #endif
811 
812     /* highest address for loading the initrd */
813     if (protocol >= 0x203) {
814         initrd_max = ldl_p(header+0x22c);
815     } else {
816         initrd_max = 0x37ffffff;
817     }
818 
819     if (initrd_max >= max_ram_size - acpi_data_size) {
820         initrd_max = max_ram_size - acpi_data_size - 1;
821     }
822 
823     fw_cfg_add_i32(fw_cfg, FW_CFG_CMDLINE_ADDR, cmdline_addr);
824     fw_cfg_add_i32(fw_cfg, FW_CFG_CMDLINE_SIZE, strlen(kernel_cmdline)+1);
825     fw_cfg_add_string(fw_cfg, FW_CFG_CMDLINE_DATA, kernel_cmdline);
826 
827     if (protocol >= 0x202) {
828         stl_p(header+0x228, cmdline_addr);
829     } else {
830         stw_p(header+0x20, 0xA33F);
831         stw_p(header+0x22, cmdline_addr-real_addr);
832     }
833 
834     /* handle vga= parameter */
835     vmode = strstr(kernel_cmdline, "vga=");
836     if (vmode) {
837         unsigned int video_mode;
838         /* skip "vga=" */
839         vmode += 4;
840         if (!strncmp(vmode, "normal", 6)) {
841             video_mode = 0xffff;
842         } else if (!strncmp(vmode, "ext", 3)) {
843             video_mode = 0xfffe;
844         } else if (!strncmp(vmode, "ask", 3)) {
845             video_mode = 0xfffd;
846         } else {
847             video_mode = strtol(vmode, NULL, 0);
848         }
849         stw_p(header+0x1fa, video_mode);
850     }
851 
852     /* loader type */
853     /* High nybble = B reserved for QEMU; low nybble is revision number.
854        If this code is substantially changed, you may want to consider
855        incrementing the revision. */
856     if (protocol >= 0x200) {
857         header[0x210] = 0xB0;
858     }
859     /* heap */
860     if (protocol >= 0x201) {
861         header[0x211] |= 0x80;	/* CAN_USE_HEAP */
862         stw_p(header+0x224, cmdline_addr-real_addr-0x200);
863     }
864 
865     /* load initrd */
866     if (initrd_filename) {
867         if (protocol < 0x200) {
868             fprintf(stderr, "qemu: linux kernel too old to load a ram disk\n");
869             exit(1);
870         }
871 
872         initrd_size = get_image_size(initrd_filename);
873         if (initrd_size < 0) {
874             fprintf(stderr, "qemu: error reading initrd %s: %s\n",
875                     initrd_filename, strerror(errno));
876             exit(1);
877         }
878 
879         initrd_addr = (initrd_max-initrd_size) & ~4095;
880 
881         initrd_data = g_malloc(initrd_size);
882         load_image(initrd_filename, initrd_data);
883 
884         fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_ADDR, initrd_addr);
885         fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_SIZE, initrd_size);
886         fw_cfg_add_bytes(fw_cfg, FW_CFG_INITRD_DATA, initrd_data, initrd_size);
887 
888         stl_p(header+0x218, initrd_addr);
889         stl_p(header+0x21c, initrd_size);
890     }
891 
892     /* load kernel and setup */
893     setup_size = header[0x1f1];
894     if (setup_size == 0) {
895         setup_size = 4;
896     }
897     setup_size = (setup_size+1)*512;
898     kernel_size -= setup_size;
899 
900     setup  = g_malloc(setup_size);
901     kernel = g_malloc(kernel_size);
902     fseek(f, 0, SEEK_SET);
903     if (fread(setup, 1, setup_size, f) != setup_size) {
904         fprintf(stderr, "fread() failed\n");
905         exit(1);
906     }
907     if (fread(kernel, 1, kernel_size, f) != kernel_size) {
908         fprintf(stderr, "fread() failed\n");
909         exit(1);
910     }
911     fclose(f);
912     memcpy(setup, header, MIN(sizeof(header), setup_size));
913 
914     fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_ADDR, prot_addr);
915     fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_SIZE, kernel_size);
916     fw_cfg_add_bytes(fw_cfg, FW_CFG_KERNEL_DATA, kernel, kernel_size);
917 
918     fw_cfg_add_i32(fw_cfg, FW_CFG_SETUP_ADDR, real_addr);
919     fw_cfg_add_i32(fw_cfg, FW_CFG_SETUP_SIZE, setup_size);
920     fw_cfg_add_bytes(fw_cfg, FW_CFG_SETUP_DATA, setup, setup_size);
921 
922     option_rom[nb_option_roms].name = "linuxboot.bin";
923     option_rom[nb_option_roms].bootindex = 0;
924     nb_option_roms++;
925 }
926 
927 #define NE2000_NB_MAX 6
928 
929 static const int ne2000_io[NE2000_NB_MAX] = { 0x300, 0x320, 0x340, 0x360,
930                                               0x280, 0x380 };
931 static const int ne2000_irq[NE2000_NB_MAX] = { 9, 10, 11, 3, 4, 5 };
932 
933 void pc_init_ne2k_isa(ISABus *bus, NICInfo *nd)
934 {
935     static int nb_ne2k = 0;
936 
937     if (nb_ne2k == NE2000_NB_MAX)
938         return;
939     isa_ne2000_init(bus, ne2000_io[nb_ne2k],
940                     ne2000_irq[nb_ne2k], nd);
941     nb_ne2k++;
942 }
943 
944 DeviceState *cpu_get_current_apic(void)
945 {
946     if (current_cpu) {
947         X86CPU *cpu = X86_CPU(current_cpu);
948         return cpu->apic_state;
949     } else {
950         return NULL;
951     }
952 }
953 
954 void pc_acpi_smi_interrupt(void *opaque, int irq, int level)
955 {
956     X86CPU *cpu = opaque;
957 
958     if (level) {
959         cpu_interrupt(CPU(cpu), CPU_INTERRUPT_SMI);
960     }
961 }
962 
963 static X86CPU *pc_new_cpu(const char *cpu_model, int64_t apic_id,
964                           DeviceState *icc_bridge, Error **errp)
965 {
966     X86CPU *cpu;
967     Error *local_err = NULL;
968 
969     cpu = cpu_x86_create(cpu_model, icc_bridge, &local_err);
970     if (local_err != NULL) {
971         error_propagate(errp, local_err);
972         return NULL;
973     }
974 
975     object_property_set_int(OBJECT(cpu), apic_id, "apic-id", &local_err);
976     object_property_set_bool(OBJECT(cpu), true, "realized", &local_err);
977 
978     if (local_err) {
979         error_propagate(errp, local_err);
980         object_unref(OBJECT(cpu));
981         cpu = NULL;
982     }
983     return cpu;
984 }
985 
986 static const char *current_cpu_model;
987 
988 void pc_hot_add_cpu(const int64_t id, Error **errp)
989 {
990     DeviceState *icc_bridge;
991     int64_t apic_id = x86_cpu_apic_id_from_index(id);
992 
993     if (id < 0) {
994         error_setg(errp, "Invalid CPU id: %" PRIi64, id);
995         return;
996     }
997 
998     if (cpu_exists(apic_id)) {
999         error_setg(errp, "Unable to add CPU: %" PRIi64
1000                    ", it already exists", id);
1001         return;
1002     }
1003 
1004     if (id >= max_cpus) {
1005         error_setg(errp, "Unable to add CPU: %" PRIi64
1006                    ", max allowed: %d", id, max_cpus - 1);
1007         return;
1008     }
1009 
1010     if (apic_id >= ACPI_CPU_HOTPLUG_ID_LIMIT) {
1011         error_setg(errp, "Unable to add CPU: %" PRIi64
1012                    ", resulting APIC ID (%" PRIi64 ") is too large",
1013                    id, apic_id);
1014         return;
1015     }
1016 
1017     icc_bridge = DEVICE(object_resolve_path_type("icc-bridge",
1018                                                  TYPE_ICC_BRIDGE, NULL));
1019     pc_new_cpu(current_cpu_model, apic_id, icc_bridge, errp);
1020 }
1021 
1022 void pc_cpus_init(const char *cpu_model, DeviceState *icc_bridge)
1023 {
1024     int i;
1025     X86CPU *cpu = NULL;
1026     Error *error = NULL;
1027     unsigned long apic_id_limit;
1028 
1029     /* init CPUs */
1030     if (cpu_model == NULL) {
1031 #ifdef TARGET_X86_64
1032         cpu_model = "qemu64";
1033 #else
1034         cpu_model = "qemu32";
1035 #endif
1036     }
1037     current_cpu_model = cpu_model;
1038 
1039     apic_id_limit = pc_apic_id_limit(max_cpus);
1040     if (apic_id_limit > ACPI_CPU_HOTPLUG_ID_LIMIT) {
1041         error_report("max_cpus is too large. APIC ID of last CPU is %lu",
1042                      apic_id_limit - 1);
1043         exit(1);
1044     }
1045 
1046     for (i = 0; i < smp_cpus; i++) {
1047         cpu = pc_new_cpu(cpu_model, x86_cpu_apic_id_from_index(i),
1048                          icc_bridge, &error);
1049         if (error) {
1050             error_report("%s", error_get_pretty(error));
1051             error_free(error);
1052             exit(1);
1053         }
1054     }
1055 
1056     /* map APIC MMIO area if CPU has APIC */
1057     if (cpu && cpu->apic_state) {
1058         /* XXX: what if the base changes? */
1059         sysbus_mmio_map_overlap(SYS_BUS_DEVICE(icc_bridge), 0,
1060                                 APIC_DEFAULT_ADDRESS, 0x1000);
1061     }
1062 
1063     /* tell smbios about cpuid version and features */
1064     smbios_set_cpuid(cpu->env.cpuid_version, cpu->env.features[FEAT_1_EDX]);
1065 }
1066 
1067 /* pci-info ROM file. Little endian format */
1068 typedef struct PcRomPciInfo {
1069     uint64_t w32_min;
1070     uint64_t w32_max;
1071     uint64_t w64_min;
1072     uint64_t w64_max;
1073 } PcRomPciInfo;
1074 
1075 typedef struct PcGuestInfoState {
1076     PcGuestInfo info;
1077     Notifier machine_done;
1078 } PcGuestInfoState;
1079 
1080 static
1081 void pc_guest_info_machine_done(Notifier *notifier, void *data)
1082 {
1083     PcGuestInfoState *guest_info_state = container_of(notifier,
1084                                                       PcGuestInfoState,
1085                                                       machine_done);
1086     acpi_setup(&guest_info_state->info);
1087 }
1088 
1089 PcGuestInfo *pc_guest_info_init(ram_addr_t below_4g_mem_size,
1090                                 ram_addr_t above_4g_mem_size)
1091 {
1092     PcGuestInfoState *guest_info_state = g_malloc0(sizeof *guest_info_state);
1093     PcGuestInfo *guest_info = &guest_info_state->info;
1094     int i, j;
1095 
1096     guest_info->ram_size_below_4g = below_4g_mem_size;
1097     guest_info->ram_size = below_4g_mem_size + above_4g_mem_size;
1098     guest_info->apic_id_limit = pc_apic_id_limit(max_cpus);
1099     guest_info->apic_xrupt_override = kvm_allows_irq0_override();
1100     guest_info->numa_nodes = nb_numa_nodes;
1101     guest_info->node_mem = g_malloc0(guest_info->numa_nodes *
1102                                     sizeof *guest_info->node_mem);
1103     for (i = 0; i < nb_numa_nodes; i++) {
1104         guest_info->node_mem[i] = numa_info[i].node_mem;
1105     }
1106 
1107     guest_info->node_cpu = g_malloc0(guest_info->apic_id_limit *
1108                                      sizeof *guest_info->node_cpu);
1109 
1110     for (i = 0; i < max_cpus; i++) {
1111         unsigned int apic_id = x86_cpu_apic_id_from_index(i);
1112         assert(apic_id < guest_info->apic_id_limit);
1113         for (j = 0; j < nb_numa_nodes; j++) {
1114             if (test_bit(i, numa_info[j].node_cpu)) {
1115                 guest_info->node_cpu[apic_id] = j;
1116                 break;
1117             }
1118         }
1119     }
1120 
1121     guest_info_state->machine_done.notify = pc_guest_info_machine_done;
1122     qemu_add_machine_init_done_notifier(&guest_info_state->machine_done);
1123     return guest_info;
1124 }
1125 
1126 /* setup pci memory address space mapping into system address space */
1127 void pc_pci_as_mapping_init(Object *owner, MemoryRegion *system_memory,
1128                             MemoryRegion *pci_address_space)
1129 {
1130     /* Set to lower priority than RAM */
1131     memory_region_add_subregion_overlap(system_memory, 0x0,
1132                                         pci_address_space, -1);
1133 }
1134 
1135 void pc_acpi_init(const char *default_dsdt)
1136 {
1137     char *filename;
1138 
1139     if (acpi_tables != NULL) {
1140         /* manually set via -acpitable, leave it alone */
1141         return;
1142     }
1143 
1144     filename = qemu_find_file(QEMU_FILE_TYPE_BIOS, default_dsdt);
1145     if (filename == NULL) {
1146         fprintf(stderr, "WARNING: failed to find %s\n", default_dsdt);
1147     } else {
1148         char *arg;
1149         QemuOpts *opts;
1150         Error *err = NULL;
1151 
1152         arg = g_strdup_printf("file=%s", filename);
1153 
1154         /* creates a deep copy of "arg" */
1155         opts = qemu_opts_parse(qemu_find_opts("acpi"), arg, 0);
1156         g_assert(opts != NULL);
1157 
1158         acpi_table_add_builtin(opts, &err);
1159         if (err) {
1160             error_report("WARNING: failed to load %s: %s", filename,
1161                          error_get_pretty(err));
1162             error_free(err);
1163         }
1164         g_free(arg);
1165         g_free(filename);
1166     }
1167 }
1168 
1169 FWCfgState *xen_load_linux(const char *kernel_filename,
1170                            const char *kernel_cmdline,
1171                            const char *initrd_filename,
1172                            ram_addr_t below_4g_mem_size,
1173                            PcGuestInfo *guest_info)
1174 {
1175     int i;
1176     FWCfgState *fw_cfg;
1177 
1178     assert(kernel_filename != NULL);
1179 
1180     fw_cfg = fw_cfg_init(BIOS_CFG_IOPORT, BIOS_CFG_IOPORT + 1, 0, 0);
1181     rom_set_fw(fw_cfg);
1182 
1183     load_linux(fw_cfg, kernel_filename, initrd_filename,
1184                kernel_cmdline, below_4g_mem_size);
1185     for (i = 0; i < nb_option_roms; i++) {
1186         assert(!strcmp(option_rom[i].name, "linuxboot.bin") ||
1187                !strcmp(option_rom[i].name, "multiboot.bin"));
1188         rom_add_option(option_rom[i].name, option_rom[i].bootindex);
1189     }
1190     guest_info->fw_cfg = fw_cfg;
1191     return fw_cfg;
1192 }
1193 
1194 FWCfgState *pc_memory_init(MachineState *machine,
1195                            MemoryRegion *system_memory,
1196                            ram_addr_t below_4g_mem_size,
1197                            ram_addr_t above_4g_mem_size,
1198                            MemoryRegion *rom_memory,
1199                            MemoryRegion **ram_memory,
1200                            PcGuestInfo *guest_info)
1201 {
1202     int linux_boot, i;
1203     MemoryRegion *ram, *option_rom_mr;
1204     MemoryRegion *ram_below_4g, *ram_above_4g;
1205     FWCfgState *fw_cfg;
1206     PCMachineState *pcms = PC_MACHINE(machine);
1207 
1208     assert(machine->ram_size == below_4g_mem_size + above_4g_mem_size);
1209 
1210     linux_boot = (machine->kernel_filename != NULL);
1211 
1212     /* Allocate RAM.  We allocate it as a single memory region and use
1213      * aliases to address portions of it, mostly for backwards compatibility
1214      * with older qemus that used qemu_ram_alloc().
1215      */
1216     ram = g_malloc(sizeof(*ram));
1217     memory_region_allocate_system_memory(ram, NULL, "pc.ram",
1218                                          machine->ram_size);
1219     *ram_memory = ram;
1220     ram_below_4g = g_malloc(sizeof(*ram_below_4g));
1221     memory_region_init_alias(ram_below_4g, NULL, "ram-below-4g", ram,
1222                              0, below_4g_mem_size);
1223     memory_region_add_subregion(system_memory, 0, ram_below_4g);
1224     e820_add_entry(0, below_4g_mem_size, E820_RAM);
1225     if (above_4g_mem_size > 0) {
1226         ram_above_4g = g_malloc(sizeof(*ram_above_4g));
1227         memory_region_init_alias(ram_above_4g, NULL, "ram-above-4g", ram,
1228                                  below_4g_mem_size, above_4g_mem_size);
1229         memory_region_add_subregion(system_memory, 0x100000000ULL,
1230                                     ram_above_4g);
1231         e820_add_entry(0x100000000ULL, above_4g_mem_size, E820_RAM);
1232     }
1233 
1234     if (!guest_info->has_reserved_memory &&
1235         (machine->ram_slots ||
1236          (machine->maxram_size > machine->ram_size))) {
1237         MachineClass *mc = MACHINE_GET_CLASS(machine);
1238 
1239         error_report("\"-memory 'slots|maxmem'\" is not supported by: %s",
1240                      mc->name);
1241         exit(EXIT_FAILURE);
1242     }
1243 
1244     /* initialize hotplug memory address space */
1245     if (guest_info->has_reserved_memory &&
1246         (machine->ram_size < machine->maxram_size)) {
1247         ram_addr_t hotplug_mem_size =
1248             machine->maxram_size - machine->ram_size;
1249 
1250         if (machine->ram_slots > ACPI_MAX_RAM_SLOTS) {
1251             error_report("unsupported amount of memory slots: %"PRIu64,
1252                          machine->ram_slots);
1253             exit(EXIT_FAILURE);
1254         }
1255 
1256         pcms->hotplug_memory_base =
1257             ROUND_UP(0x100000000ULL + above_4g_mem_size, 1ULL << 30);
1258 
1259         if ((pcms->hotplug_memory_base + hotplug_mem_size) <
1260             hotplug_mem_size) {
1261             error_report("unsupported amount of maximum memory: " RAM_ADDR_FMT,
1262                          machine->maxram_size);
1263             exit(EXIT_FAILURE);
1264         }
1265 
1266         memory_region_init(&pcms->hotplug_memory, OBJECT(pcms),
1267                            "hotplug-memory", hotplug_mem_size);
1268         memory_region_add_subregion(system_memory, pcms->hotplug_memory_base,
1269                                     &pcms->hotplug_memory);
1270     }
1271 
1272     /* Initialize PC system firmware */
1273     pc_system_firmware_init(rom_memory, guest_info->isapc_ram_fw);
1274 
1275     option_rom_mr = g_malloc(sizeof(*option_rom_mr));
1276     memory_region_init_ram(option_rom_mr, NULL, "pc.rom", PC_ROM_SIZE,
1277                            &error_abort);
1278     vmstate_register_ram_global(option_rom_mr);
1279     memory_region_add_subregion_overlap(rom_memory,
1280                                         PC_ROM_MIN_VGA,
1281                                         option_rom_mr,
1282                                         1);
1283 
1284     fw_cfg = bochs_bios_init();
1285     rom_set_fw(fw_cfg);
1286 
1287     if (guest_info->has_reserved_memory && pcms->hotplug_memory_base) {
1288         uint64_t *val = g_malloc(sizeof(*val));
1289         *val = cpu_to_le64(ROUND_UP(pcms->hotplug_memory_base, 0x1ULL << 30));
1290         fw_cfg_add_file(fw_cfg, "etc/reserved-memory-end", val, sizeof(*val));
1291     }
1292 
1293     if (linux_boot) {
1294         load_linux(fw_cfg, machine->kernel_filename, machine->initrd_filename,
1295                    machine->kernel_cmdline, below_4g_mem_size);
1296     }
1297 
1298     for (i = 0; i < nb_option_roms; i++) {
1299         rom_add_option(option_rom[i].name, option_rom[i].bootindex);
1300     }
1301     guest_info->fw_cfg = fw_cfg;
1302     return fw_cfg;
1303 }
1304 
1305 qemu_irq *pc_allocate_cpu_irq(void)
1306 {
1307     return qemu_allocate_irqs(pic_irq_request, NULL, 1);
1308 }
1309 
1310 DeviceState *pc_vga_init(ISABus *isa_bus, PCIBus *pci_bus)
1311 {
1312     DeviceState *dev = NULL;
1313 
1314     if (pci_bus) {
1315         PCIDevice *pcidev = pci_vga_init(pci_bus);
1316         dev = pcidev ? &pcidev->qdev : NULL;
1317     } else if (isa_bus) {
1318         ISADevice *isadev = isa_vga_init(isa_bus);
1319         dev = isadev ? DEVICE(isadev) : NULL;
1320     }
1321     return dev;
1322 }
1323 
1324 static void cpu_request_exit(void *opaque, int irq, int level)
1325 {
1326     CPUState *cpu = current_cpu;
1327 
1328     if (cpu && level) {
1329         cpu_exit(cpu);
1330     }
1331 }
1332 
1333 static const MemoryRegionOps ioport80_io_ops = {
1334     .write = ioport80_write,
1335     .read = ioport80_read,
1336     .endianness = DEVICE_NATIVE_ENDIAN,
1337     .impl = {
1338         .min_access_size = 1,
1339         .max_access_size = 1,
1340     },
1341 };
1342 
1343 static const MemoryRegionOps ioportF0_io_ops = {
1344     .write = ioportF0_write,
1345     .read = ioportF0_read,
1346     .endianness = DEVICE_NATIVE_ENDIAN,
1347     .impl = {
1348         .min_access_size = 1,
1349         .max_access_size = 1,
1350     },
1351 };
1352 
1353 void pc_basic_device_init(ISABus *isa_bus, qemu_irq *gsi,
1354                           ISADevice **rtc_state,
1355                           ISADevice **floppy,
1356                           bool no_vmport,
1357                           uint32 hpet_irqs)
1358 {
1359     int i;
1360     DriveInfo *fd[MAX_FD];
1361     DeviceState *hpet = NULL;
1362     int pit_isa_irq = 0;
1363     qemu_irq pit_alt_irq = NULL;
1364     qemu_irq rtc_irq = NULL;
1365     qemu_irq *a20_line;
1366     ISADevice *i8042, *port92, *vmmouse, *pit = NULL;
1367     qemu_irq *cpu_exit_irq;
1368     MemoryRegion *ioport80_io = g_new(MemoryRegion, 1);
1369     MemoryRegion *ioportF0_io = g_new(MemoryRegion, 1);
1370 
1371     memory_region_init_io(ioport80_io, NULL, &ioport80_io_ops, NULL, "ioport80", 1);
1372     memory_region_add_subregion(isa_bus->address_space_io, 0x80, ioport80_io);
1373 
1374     memory_region_init_io(ioportF0_io, NULL, &ioportF0_io_ops, NULL, "ioportF0", 1);
1375     memory_region_add_subregion(isa_bus->address_space_io, 0xf0, ioportF0_io);
1376 
1377     /*
1378      * Check if an HPET shall be created.
1379      *
1380      * Without KVM_CAP_PIT_STATE2, we cannot switch off the in-kernel PIT
1381      * when the HPET wants to take over. Thus we have to disable the latter.
1382      */
1383     if (!no_hpet && (!kvm_irqchip_in_kernel() || kvm_has_pit_state2())) {
1384         /* In order to set property, here not using sysbus_try_create_simple */
1385         hpet = qdev_try_create(NULL, TYPE_HPET);
1386         if (hpet) {
1387             /* For pc-piix-*, hpet's intcap is always IRQ2. For pc-q35-1.7
1388              * and earlier, use IRQ2 for compat. Otherwise, use IRQ16~23,
1389              * IRQ8 and IRQ2.
1390              */
1391             uint8_t compat = object_property_get_int(OBJECT(hpet),
1392                     HPET_INTCAP, NULL);
1393             if (!compat) {
1394                 qdev_prop_set_uint32(hpet, HPET_INTCAP, hpet_irqs);
1395             }
1396             qdev_init_nofail(hpet);
1397             sysbus_mmio_map(SYS_BUS_DEVICE(hpet), 0, HPET_BASE);
1398 
1399             for (i = 0; i < GSI_NUM_PINS; i++) {
1400                 sysbus_connect_irq(SYS_BUS_DEVICE(hpet), i, gsi[i]);
1401             }
1402             pit_isa_irq = -1;
1403             pit_alt_irq = qdev_get_gpio_in(hpet, HPET_LEGACY_PIT_INT);
1404             rtc_irq = qdev_get_gpio_in(hpet, HPET_LEGACY_RTC_INT);
1405         }
1406     }
1407     *rtc_state = rtc_init(isa_bus, 2000, rtc_irq);
1408 
1409     qemu_register_boot_set(pc_boot_set, *rtc_state);
1410 
1411     if (!xen_enabled()) {
1412         if (kvm_irqchip_in_kernel()) {
1413             pit = kvm_pit_init(isa_bus, 0x40);
1414         } else {
1415             pit = pit_init(isa_bus, 0x40, pit_isa_irq, pit_alt_irq);
1416         }
1417         if (hpet) {
1418             /* connect PIT to output control line of the HPET */
1419             qdev_connect_gpio_out(hpet, 0, qdev_get_gpio_in(DEVICE(pit), 0));
1420         }
1421         pcspk_init(isa_bus, pit);
1422     }
1423 
1424     for(i = 0; i < MAX_SERIAL_PORTS; i++) {
1425         if (serial_hds[i]) {
1426             serial_isa_init(isa_bus, i, serial_hds[i]);
1427         }
1428     }
1429 
1430     for(i = 0; i < MAX_PARALLEL_PORTS; i++) {
1431         if (parallel_hds[i]) {
1432             parallel_init(isa_bus, i, parallel_hds[i]);
1433         }
1434     }
1435 
1436     a20_line = qemu_allocate_irqs(handle_a20_line_change, first_cpu, 2);
1437     i8042 = isa_create_simple(isa_bus, "i8042");
1438     i8042_setup_a20_line(i8042, &a20_line[0]);
1439     if (!no_vmport) {
1440         vmport_init(isa_bus);
1441         vmmouse = isa_try_create(isa_bus, "vmmouse");
1442     } else {
1443         vmmouse = NULL;
1444     }
1445     if (vmmouse) {
1446         DeviceState *dev = DEVICE(vmmouse);
1447         qdev_prop_set_ptr(dev, "ps2_mouse", i8042);
1448         qdev_init_nofail(dev);
1449     }
1450     port92 = isa_create_simple(isa_bus, "port92");
1451     port92_init(port92, &a20_line[1]);
1452 
1453     cpu_exit_irq = qemu_allocate_irqs(cpu_request_exit, NULL, 1);
1454     DMA_init(0, cpu_exit_irq);
1455 
1456     for(i = 0; i < MAX_FD; i++) {
1457         fd[i] = drive_get(IF_FLOPPY, 0, i);
1458     }
1459     *floppy = fdctrl_init_isa(isa_bus, fd);
1460 }
1461 
1462 void pc_nic_init(ISABus *isa_bus, PCIBus *pci_bus)
1463 {
1464     int i;
1465 
1466     for (i = 0; i < nb_nics; i++) {
1467         NICInfo *nd = &nd_table[i];
1468 
1469         if (!pci_bus || (nd->model && strcmp(nd->model, "ne2k_isa") == 0)) {
1470             pc_init_ne2k_isa(isa_bus, nd);
1471         } else {
1472             pci_nic_init_nofail(nd, pci_bus, "e1000", NULL);
1473         }
1474     }
1475 }
1476 
1477 void pc_pci_device_init(PCIBus *pci_bus)
1478 {
1479     int max_bus;
1480     int bus;
1481 
1482     max_bus = drive_get_max_bus(IF_SCSI);
1483     for (bus = 0; bus <= max_bus; bus++) {
1484         pci_create_simple(pci_bus, -1, "lsi53c895a");
1485     }
1486 }
1487 
1488 void ioapic_init_gsi(GSIState *gsi_state, const char *parent_name)
1489 {
1490     DeviceState *dev;
1491     SysBusDevice *d;
1492     unsigned int i;
1493 
1494     if (kvm_irqchip_in_kernel()) {
1495         dev = qdev_create(NULL, "kvm-ioapic");
1496     } else {
1497         dev = qdev_create(NULL, "ioapic");
1498     }
1499     if (parent_name) {
1500         object_property_add_child(object_resolve_path(parent_name, NULL),
1501                                   "ioapic", OBJECT(dev), NULL);
1502     }
1503     qdev_init_nofail(dev);
1504     d = SYS_BUS_DEVICE(dev);
1505     sysbus_mmio_map(d, 0, IO_APIC_DEFAULT_ADDRESS);
1506 
1507     for (i = 0; i < IOAPIC_NUM_PINS; i++) {
1508         gsi_state->ioapic_irq[i] = qdev_get_gpio_in(dev, i);
1509     }
1510 }
1511 
1512 static void pc_generic_machine_class_init(ObjectClass *oc, void *data)
1513 {
1514     MachineClass *mc = MACHINE_CLASS(oc);
1515     QEMUMachine *qm = data;
1516 
1517     mc->name = qm->name;
1518     mc->alias = qm->alias;
1519     mc->desc = qm->desc;
1520     mc->init = qm->init;
1521     mc->reset = qm->reset;
1522     mc->hot_add_cpu = qm->hot_add_cpu;
1523     mc->kvm_type = qm->kvm_type;
1524     mc->block_default_type = qm->block_default_type;
1525     mc->max_cpus = qm->max_cpus;
1526     mc->no_serial = qm->no_serial;
1527     mc->no_parallel = qm->no_parallel;
1528     mc->use_virtcon = qm->use_virtcon;
1529     mc->use_sclp = qm->use_sclp;
1530     mc->no_floppy = qm->no_floppy;
1531     mc->no_cdrom = qm->no_cdrom;
1532     mc->no_sdcard = qm->no_sdcard;
1533     mc->is_default = qm->is_default;
1534     mc->default_machine_opts = qm->default_machine_opts;
1535     mc->default_boot_order = qm->default_boot_order;
1536     mc->compat_props = qm->compat_props;
1537     mc->hw_version = qm->hw_version;
1538 }
1539 
1540 void qemu_register_pc_machine(QEMUMachine *m)
1541 {
1542     char *name = g_strconcat(m->name, TYPE_MACHINE_SUFFIX, NULL);
1543     TypeInfo ti = {
1544         .name       = name,
1545         .parent     = TYPE_PC_MACHINE,
1546         .class_init = pc_generic_machine_class_init,
1547         .class_data = (void *)m,
1548     };
1549 
1550     type_register(&ti);
1551     g_free(name);
1552 }
1553 
1554 static void pc_dimm_plug(HotplugHandler *hotplug_dev,
1555                          DeviceState *dev, Error **errp)
1556 {
1557     int slot;
1558     HotplugHandlerClass *hhc;
1559     Error *local_err = NULL;
1560     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1561     MachineState *machine = MACHINE(hotplug_dev);
1562     PCDIMMDevice *dimm = PC_DIMM(dev);
1563     PCDIMMDeviceClass *ddc = PC_DIMM_GET_CLASS(dimm);
1564     MemoryRegion *mr = ddc->get_memory_region(dimm);
1565     uint64_t addr = object_property_get_int(OBJECT(dimm), PC_DIMM_ADDR_PROP,
1566                                             &local_err);
1567     if (local_err) {
1568         goto out;
1569     }
1570 
1571     addr = pc_dimm_get_free_addr(pcms->hotplug_memory_base,
1572                                  memory_region_size(&pcms->hotplug_memory),
1573                                  !addr ? NULL : &addr,
1574                                  memory_region_size(mr), &local_err);
1575     if (local_err) {
1576         goto out;
1577     }
1578 
1579     object_property_set_int(OBJECT(dev), addr, PC_DIMM_ADDR_PROP, &local_err);
1580     if (local_err) {
1581         goto out;
1582     }
1583     trace_mhp_pc_dimm_assigned_address(addr);
1584 
1585     slot = object_property_get_int(OBJECT(dev), PC_DIMM_SLOT_PROP, &local_err);
1586     if (local_err) {
1587         goto out;
1588     }
1589 
1590     slot = pc_dimm_get_free_slot(slot == PC_DIMM_UNASSIGNED_SLOT ? NULL : &slot,
1591                                  machine->ram_slots, &local_err);
1592     if (local_err) {
1593         goto out;
1594     }
1595     object_property_set_int(OBJECT(dev), slot, PC_DIMM_SLOT_PROP, &local_err);
1596     if (local_err) {
1597         goto out;
1598     }
1599     trace_mhp_pc_dimm_assigned_slot(slot);
1600 
1601     if (!pcms->acpi_dev) {
1602         error_setg(&local_err,
1603                    "memory hotplug is not enabled: missing acpi device");
1604         goto out;
1605     }
1606 
1607     memory_region_add_subregion(&pcms->hotplug_memory,
1608                                 addr - pcms->hotplug_memory_base, mr);
1609     vmstate_register_ram(mr, dev);
1610 
1611     hhc = HOTPLUG_HANDLER_GET_CLASS(pcms->acpi_dev);
1612     hhc->plug(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &local_err);
1613 out:
1614     error_propagate(errp, local_err);
1615 }
1616 
1617 static void pc_machine_device_plug_cb(HotplugHandler *hotplug_dev,
1618                                       DeviceState *dev, Error **errp)
1619 {
1620     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
1621         pc_dimm_plug(hotplug_dev, dev, errp);
1622     }
1623 }
1624 
1625 static HotplugHandler *pc_get_hotpug_handler(MachineState *machine,
1626                                              DeviceState *dev)
1627 {
1628     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(machine);
1629 
1630     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
1631         return HOTPLUG_HANDLER(machine);
1632     }
1633 
1634     return pcmc->get_hotplug_handler ?
1635         pcmc->get_hotplug_handler(machine, dev) : NULL;
1636 }
1637 
1638 static void
1639 pc_machine_get_hotplug_memory_region_size(Object *obj, Visitor *v, void *opaque,
1640                                           const char *name, Error **errp)
1641 {
1642     PCMachineState *pcms = PC_MACHINE(obj);
1643     int64_t value = memory_region_size(&pcms->hotplug_memory);
1644 
1645     visit_type_int(v, &value, name, errp);
1646 }
1647 
1648 static void pc_machine_get_max_ram_below_4g(Object *obj, Visitor *v,
1649                                          void *opaque, const char *name,
1650                                          Error **errp)
1651 {
1652     PCMachineState *pcms = PC_MACHINE(obj);
1653     uint64_t value = pcms->max_ram_below_4g;
1654 
1655     visit_type_size(v, &value, name, errp);
1656 }
1657 
1658 static void pc_machine_set_max_ram_below_4g(Object *obj, Visitor *v,
1659                                          void *opaque, const char *name,
1660                                          Error **errp)
1661 {
1662     PCMachineState *pcms = PC_MACHINE(obj);
1663     Error *error = NULL;
1664     uint64_t value;
1665 
1666     visit_type_size(v, &value, name, &error);
1667     if (error) {
1668         error_propagate(errp, error);
1669         return;
1670     }
1671     if (value > (1ULL << 32)) {
1672         error_set(&error, ERROR_CLASS_GENERIC_ERROR,
1673                   "Machine option 'max-ram-below-4g=%"PRIu64
1674                   "' expects size less than or equal to 4G", value);
1675         error_propagate(errp, error);
1676         return;
1677     }
1678 
1679     if (value < (1ULL << 20)) {
1680         error_report("Warning: small max_ram_below_4g(%"PRIu64
1681                      ") less than 1M.  BIOS may not work..",
1682                      value);
1683     }
1684 
1685     pcms->max_ram_below_4g = value;
1686 }
1687 
1688 static void pc_machine_initfn(Object *obj)
1689 {
1690     PCMachineState *pcms = PC_MACHINE(obj);
1691 
1692     object_property_add(obj, PC_MACHINE_MEMHP_REGION_SIZE, "int",
1693                         pc_machine_get_hotplug_memory_region_size,
1694                         NULL, NULL, NULL, NULL);
1695     pcms->max_ram_below_4g = 1ULL << 32; /* 4G */
1696     object_property_add(obj, PC_MACHINE_MAX_RAM_BELOW_4G, "size",
1697                         pc_machine_get_max_ram_below_4g,
1698                         pc_machine_set_max_ram_below_4g,
1699                         NULL, NULL, NULL);
1700 }
1701 
1702 static void pc_machine_class_init(ObjectClass *oc, void *data)
1703 {
1704     MachineClass *mc = MACHINE_CLASS(oc);
1705     PCMachineClass *pcmc = PC_MACHINE_CLASS(oc);
1706     HotplugHandlerClass *hc = HOTPLUG_HANDLER_CLASS(oc);
1707 
1708     pcmc->get_hotplug_handler = mc->get_hotplug_handler;
1709     mc->get_hotplug_handler = pc_get_hotpug_handler;
1710     hc->plug = pc_machine_device_plug_cb;
1711 }
1712 
1713 static const TypeInfo pc_machine_info = {
1714     .name = TYPE_PC_MACHINE,
1715     .parent = TYPE_MACHINE,
1716     .abstract = true,
1717     .instance_size = sizeof(PCMachineState),
1718     .instance_init = pc_machine_initfn,
1719     .class_size = sizeof(PCMachineClass),
1720     .class_init = pc_machine_class_init,
1721     .interfaces = (InterfaceInfo[]) {
1722          { TYPE_HOTPLUG_HANDLER },
1723          { }
1724     },
1725 };
1726 
1727 static void pc_machine_register_types(void)
1728 {
1729     type_register_static(&pc_machine_info);
1730 }
1731 
1732 type_init(pc_machine_register_types)
1733