xref: /openbmc/qemu/hw/i386/pc.c (revision 10358b6a)
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 #define ACPI_DATA_SIZE       0x10000
77 #define BIOS_CFG_IOPORT 0x510
78 #define FW_CFG_ACPI_TABLES (FW_CFG_ARCH_LOCAL + 0)
79 #define FW_CFG_SMBIOS_ENTRIES (FW_CFG_ARCH_LOCAL + 1)
80 #define FW_CFG_IRQ0_OVERRIDE (FW_CFG_ARCH_LOCAL + 2)
81 #define FW_CFG_E820_TABLE (FW_CFG_ARCH_LOCAL + 3)
82 #define FW_CFG_HPET (FW_CFG_ARCH_LOCAL + 4)
83 
84 #define E820_NR_ENTRIES		16
85 
86 struct e820_entry {
87     uint64_t address;
88     uint64_t length;
89     uint32_t type;
90 } QEMU_PACKED __attribute((__aligned__(4)));
91 
92 struct e820_table {
93     uint32_t count;
94     struct e820_entry entry[E820_NR_ENTRIES];
95 } QEMU_PACKED __attribute((__aligned__(4)));
96 
97 static struct e820_table e820_reserve;
98 static struct e820_entry *e820_table;
99 static unsigned e820_entries;
100 struct hpet_fw_config hpet_cfg = {.count = UINT8_MAX};
101 
102 void gsi_handler(void *opaque, int n, int level)
103 {
104     GSIState *s = opaque;
105 
106     DPRINTF("pc: %s GSI %d\n", level ? "raising" : "lowering", n);
107     if (n < ISA_NUM_IRQS) {
108         qemu_set_irq(s->i8259_irq[n], level);
109     }
110     qemu_set_irq(s->ioapic_irq[n], level);
111 }
112 
113 static void ioport80_write(void *opaque, hwaddr addr, uint64_t data,
114                            unsigned size)
115 {
116 }
117 
118 static uint64_t ioport80_read(void *opaque, hwaddr addr, unsigned size)
119 {
120     return 0xffffffffffffffffULL;
121 }
122 
123 /* MSDOS compatibility mode FPU exception support */
124 static qemu_irq ferr_irq;
125 
126 void pc_register_ferr_irq(qemu_irq irq)
127 {
128     ferr_irq = irq;
129 }
130 
131 /* XXX: add IGNNE support */
132 void cpu_set_ferr(CPUX86State *s)
133 {
134     qemu_irq_raise(ferr_irq);
135 }
136 
137 static void ioportF0_write(void *opaque, hwaddr addr, uint64_t data,
138                            unsigned size)
139 {
140     qemu_irq_lower(ferr_irq);
141 }
142 
143 static uint64_t ioportF0_read(void *opaque, hwaddr addr, unsigned size)
144 {
145     return 0xffffffffffffffffULL;
146 }
147 
148 /* TSC handling */
149 uint64_t cpu_get_tsc(CPUX86State *env)
150 {
151     return cpu_get_ticks();
152 }
153 
154 /* SMM support */
155 
156 static cpu_set_smm_t smm_set;
157 static void *smm_arg;
158 
159 void cpu_smm_register(cpu_set_smm_t callback, void *arg)
160 {
161     assert(smm_set == NULL);
162     assert(smm_arg == NULL);
163     smm_set = callback;
164     smm_arg = arg;
165 }
166 
167 void cpu_smm_update(CPUX86State *env)
168 {
169     if (smm_set && smm_arg && CPU(x86_env_get_cpu(env)) == first_cpu) {
170         smm_set(!!(env->hflags & HF_SMM_MASK), smm_arg);
171     }
172 }
173 
174 
175 /* IRQ handling */
176 int cpu_get_pic_interrupt(CPUX86State *env)
177 {
178     X86CPU *cpu = x86_env_get_cpu(env);
179     int intno;
180 
181     intno = apic_get_interrupt(cpu->apic_state);
182     if (intno >= 0) {
183         return intno;
184     }
185     /* read the irq from the PIC */
186     if (!apic_accept_pic_intr(cpu->apic_state)) {
187         return -1;
188     }
189 
190     intno = pic_read_irq(isa_pic);
191     return intno;
192 }
193 
194 static void pic_irq_request(void *opaque, int irq, int level)
195 {
196     CPUState *cs = first_cpu;
197     X86CPU *cpu = X86_CPU(cs);
198 
199     DPRINTF("pic_irqs: %s irq %d\n", level? "raise" : "lower", irq);
200     if (cpu->apic_state) {
201         CPU_FOREACH(cs) {
202             cpu = X86_CPU(cs);
203             if (apic_accept_pic_intr(cpu->apic_state)) {
204                 apic_deliver_pic_intr(cpu->apic_state, level);
205             }
206         }
207     } else {
208         if (level) {
209             cpu_interrupt(cs, CPU_INTERRUPT_HARD);
210         } else {
211             cpu_reset_interrupt(cs, CPU_INTERRUPT_HARD);
212         }
213     }
214 }
215 
216 /* PC cmos mappings */
217 
218 #define REG_EQUIPMENT_BYTE          0x14
219 
220 static int cmos_get_fd_drive_type(FDriveType fd0)
221 {
222     int val;
223 
224     switch (fd0) {
225     case FDRIVE_DRV_144:
226         /* 1.44 Mb 3"5 drive */
227         val = 4;
228         break;
229     case FDRIVE_DRV_288:
230         /* 2.88 Mb 3"5 drive */
231         val = 5;
232         break;
233     case FDRIVE_DRV_120:
234         /* 1.2 Mb 5"5 drive */
235         val = 2;
236         break;
237     case FDRIVE_DRV_NONE:
238     default:
239         val = 0;
240         break;
241     }
242     return val;
243 }
244 
245 static void cmos_init_hd(ISADevice *s, int type_ofs, int info_ofs,
246                          int16_t cylinders, int8_t heads, int8_t sectors)
247 {
248     rtc_set_memory(s, type_ofs, 47);
249     rtc_set_memory(s, info_ofs, cylinders);
250     rtc_set_memory(s, info_ofs + 1, cylinders >> 8);
251     rtc_set_memory(s, info_ofs + 2, heads);
252     rtc_set_memory(s, info_ofs + 3, 0xff);
253     rtc_set_memory(s, info_ofs + 4, 0xff);
254     rtc_set_memory(s, info_ofs + 5, 0xc0 | ((heads > 8) << 3));
255     rtc_set_memory(s, info_ofs + 6, cylinders);
256     rtc_set_memory(s, info_ofs + 7, cylinders >> 8);
257     rtc_set_memory(s, info_ofs + 8, sectors);
258 }
259 
260 /* convert boot_device letter to something recognizable by the bios */
261 static int boot_device2nibble(char boot_device)
262 {
263     switch(boot_device) {
264     case 'a':
265     case 'b':
266         return 0x01; /* floppy boot */
267     case 'c':
268         return 0x02; /* hard drive boot */
269     case 'd':
270         return 0x03; /* CD-ROM boot */
271     case 'n':
272         return 0x04; /* Network boot */
273     }
274     return 0;
275 }
276 
277 static int set_boot_dev(ISADevice *s, const char *boot_device)
278 {
279 #define PC_MAX_BOOT_DEVICES 3
280     int nbds, bds[3] = { 0, };
281     int i;
282 
283     nbds = strlen(boot_device);
284     if (nbds > PC_MAX_BOOT_DEVICES) {
285         error_report("Too many boot devices for PC");
286         return(1);
287     }
288     for (i = 0; i < nbds; i++) {
289         bds[i] = boot_device2nibble(boot_device[i]);
290         if (bds[i] == 0) {
291             error_report("Invalid boot device for PC: '%c'",
292                          boot_device[i]);
293             return(1);
294         }
295     }
296     rtc_set_memory(s, 0x3d, (bds[1] << 4) | bds[0]);
297     rtc_set_memory(s, 0x38, (bds[2] << 4) | (fd_bootchk ? 0x0 : 0x1));
298     return(0);
299 }
300 
301 static int pc_boot_set(void *opaque, const char *boot_device)
302 {
303     return set_boot_dev(opaque, boot_device);
304 }
305 
306 typedef struct pc_cmos_init_late_arg {
307     ISADevice *rtc_state;
308     BusState *idebus[2];
309 } pc_cmos_init_late_arg;
310 
311 static void pc_cmos_init_late(void *opaque)
312 {
313     pc_cmos_init_late_arg *arg = opaque;
314     ISADevice *s = arg->rtc_state;
315     int16_t cylinders;
316     int8_t heads, sectors;
317     int val;
318     int i, trans;
319 
320     val = 0;
321     if (ide_get_geometry(arg->idebus[0], 0,
322                          &cylinders, &heads, &sectors) >= 0) {
323         cmos_init_hd(s, 0x19, 0x1b, cylinders, heads, sectors);
324         val |= 0xf0;
325     }
326     if (ide_get_geometry(arg->idebus[0], 1,
327                          &cylinders, &heads, &sectors) >= 0) {
328         cmos_init_hd(s, 0x1a, 0x24, cylinders, heads, sectors);
329         val |= 0x0f;
330     }
331     rtc_set_memory(s, 0x12, val);
332 
333     val = 0;
334     for (i = 0; i < 4; i++) {
335         /* NOTE: ide_get_geometry() returns the physical
336            geometry.  It is always such that: 1 <= sects <= 63, 1
337            <= heads <= 16, 1 <= cylinders <= 16383. The BIOS
338            geometry can be different if a translation is done. */
339         if (ide_get_geometry(arg->idebus[i / 2], i % 2,
340                              &cylinders, &heads, &sectors) >= 0) {
341             trans = ide_get_bios_chs_trans(arg->idebus[i / 2], i % 2) - 1;
342             assert((trans & ~3) == 0);
343             val |= trans << (i * 2);
344         }
345     }
346     rtc_set_memory(s, 0x39, val);
347 
348     qemu_unregister_reset(pc_cmos_init_late, opaque);
349 }
350 
351 typedef struct RTCCPUHotplugArg {
352     Notifier cpu_added_notifier;
353     ISADevice *rtc_state;
354 } RTCCPUHotplugArg;
355 
356 static void rtc_notify_cpu_added(Notifier *notifier, void *data)
357 {
358     RTCCPUHotplugArg *arg = container_of(notifier, RTCCPUHotplugArg,
359                                          cpu_added_notifier);
360     ISADevice *s = arg->rtc_state;
361 
362     /* increment the number of CPUs */
363     rtc_set_memory(s, 0x5f, rtc_get_memory(s, 0x5f) + 1);
364 }
365 
366 void pc_cmos_init(ram_addr_t ram_size, ram_addr_t above_4g_mem_size,
367                   const char *boot_device,
368                   ISADevice *floppy, BusState *idebus0, BusState *idebus1,
369                   ISADevice *s)
370 {
371     int val, nb, i;
372     FDriveType fd_type[2] = { FDRIVE_DRV_NONE, FDRIVE_DRV_NONE };
373     static pc_cmos_init_late_arg arg;
374     static RTCCPUHotplugArg cpu_hotplug_cb;
375 
376     /* various important CMOS locations needed by PC/Bochs bios */
377 
378     /* memory size */
379     /* base memory (first MiB) */
380     val = MIN(ram_size / 1024, 640);
381     rtc_set_memory(s, 0x15, val);
382     rtc_set_memory(s, 0x16, val >> 8);
383     /* extended memory (next 64MiB) */
384     if (ram_size > 1024 * 1024) {
385         val = (ram_size - 1024 * 1024) / 1024;
386     } else {
387         val = 0;
388     }
389     if (val > 65535)
390         val = 65535;
391     rtc_set_memory(s, 0x17, val);
392     rtc_set_memory(s, 0x18, val >> 8);
393     rtc_set_memory(s, 0x30, val);
394     rtc_set_memory(s, 0x31, val >> 8);
395     /* memory between 16MiB and 4GiB */
396     if (ram_size > 16 * 1024 * 1024) {
397         val = (ram_size - 16 * 1024 * 1024) / 65536;
398     } else {
399         val = 0;
400     }
401     if (val > 65535)
402         val = 65535;
403     rtc_set_memory(s, 0x34, val);
404     rtc_set_memory(s, 0x35, val >> 8);
405     /* memory above 4GiB */
406     val = above_4g_mem_size / 65536;
407     rtc_set_memory(s, 0x5b, val);
408     rtc_set_memory(s, 0x5c, val >> 8);
409     rtc_set_memory(s, 0x5d, val >> 16);
410 
411     /* set the number of CPU */
412     rtc_set_memory(s, 0x5f, smp_cpus - 1);
413     /* init CPU hotplug notifier */
414     cpu_hotplug_cb.rtc_state = s;
415     cpu_hotplug_cb.cpu_added_notifier.notify = rtc_notify_cpu_added;
416     qemu_register_cpu_added_notifier(&cpu_hotplug_cb.cpu_added_notifier);
417 
418     if (set_boot_dev(s, boot_device)) {
419         exit(1);
420     }
421 
422     /* floppy type */
423     if (floppy) {
424         for (i = 0; i < 2; i++) {
425             fd_type[i] = isa_fdc_get_drive_type(floppy, i);
426         }
427     }
428     val = (cmos_get_fd_drive_type(fd_type[0]) << 4) |
429         cmos_get_fd_drive_type(fd_type[1]);
430     rtc_set_memory(s, 0x10, val);
431 
432     val = 0;
433     nb = 0;
434     if (fd_type[0] < FDRIVE_DRV_NONE) {
435         nb++;
436     }
437     if (fd_type[1] < FDRIVE_DRV_NONE) {
438         nb++;
439     }
440     switch (nb) {
441     case 0:
442         break;
443     case 1:
444         val |= 0x01; /* 1 drive, ready for boot */
445         break;
446     case 2:
447         val |= 0x41; /* 2 drives, ready for boot */
448         break;
449     }
450     val |= 0x02; /* FPU is there */
451     val |= 0x04; /* PS/2 mouse installed */
452     rtc_set_memory(s, REG_EQUIPMENT_BYTE, val);
453 
454     /* hard drives */
455     arg.rtc_state = s;
456     arg.idebus[0] = idebus0;
457     arg.idebus[1] = idebus1;
458     qemu_register_reset(pc_cmos_init_late, &arg);
459 }
460 
461 #define TYPE_PORT92 "port92"
462 #define PORT92(obj) OBJECT_CHECK(Port92State, (obj), TYPE_PORT92)
463 
464 /* port 92 stuff: could be split off */
465 typedef struct Port92State {
466     ISADevice parent_obj;
467 
468     MemoryRegion io;
469     uint8_t outport;
470     qemu_irq *a20_out;
471 } Port92State;
472 
473 static void port92_write(void *opaque, hwaddr addr, uint64_t val,
474                          unsigned size)
475 {
476     Port92State *s = opaque;
477     int oldval = s->outport;
478 
479     DPRINTF("port92: write 0x%02x\n", val);
480     s->outport = val;
481     qemu_set_irq(*s->a20_out, (val >> 1) & 1);
482     if ((val & 1) && !(oldval & 1)) {
483         qemu_system_reset_request();
484     }
485 }
486 
487 static uint64_t port92_read(void *opaque, hwaddr addr,
488                             unsigned size)
489 {
490     Port92State *s = opaque;
491     uint32_t ret;
492 
493     ret = s->outport;
494     DPRINTF("port92: read 0x%02x\n", ret);
495     return ret;
496 }
497 
498 static void port92_init(ISADevice *dev, qemu_irq *a20_out)
499 {
500     Port92State *s = PORT92(dev);
501 
502     s->a20_out = a20_out;
503 }
504 
505 static const VMStateDescription vmstate_port92_isa = {
506     .name = "port92",
507     .version_id = 1,
508     .minimum_version_id = 1,
509     .fields = (VMStateField[]) {
510         VMSTATE_UINT8(outport, Port92State),
511         VMSTATE_END_OF_LIST()
512     }
513 };
514 
515 static void port92_reset(DeviceState *d)
516 {
517     Port92State *s = PORT92(d);
518 
519     s->outport &= ~1;
520 }
521 
522 static const MemoryRegionOps port92_ops = {
523     .read = port92_read,
524     .write = port92_write,
525     .impl = {
526         .min_access_size = 1,
527         .max_access_size = 1,
528     },
529     .endianness = DEVICE_LITTLE_ENDIAN,
530 };
531 
532 static void port92_initfn(Object *obj)
533 {
534     Port92State *s = PORT92(obj);
535 
536     memory_region_init_io(&s->io, OBJECT(s), &port92_ops, s, "port92", 1);
537 
538     s->outport = 0;
539 }
540 
541 static void port92_realizefn(DeviceState *dev, Error **errp)
542 {
543     ISADevice *isadev = ISA_DEVICE(dev);
544     Port92State *s = PORT92(dev);
545 
546     isa_register_ioport(isadev, &s->io, 0x92);
547 }
548 
549 static void port92_class_initfn(ObjectClass *klass, void *data)
550 {
551     DeviceClass *dc = DEVICE_CLASS(klass);
552 
553     dc->realize = port92_realizefn;
554     dc->reset = port92_reset;
555     dc->vmsd = &vmstate_port92_isa;
556     /*
557      * Reason: unlike ordinary ISA devices, this one needs additional
558      * wiring: its A20 output line needs to be wired up by
559      * port92_init().
560      */
561     dc->cannot_instantiate_with_device_add_yet = true;
562 }
563 
564 static const TypeInfo port92_info = {
565     .name          = TYPE_PORT92,
566     .parent        = TYPE_ISA_DEVICE,
567     .instance_size = sizeof(Port92State),
568     .instance_init = port92_initfn,
569     .class_init    = port92_class_initfn,
570 };
571 
572 static void port92_register_types(void)
573 {
574     type_register_static(&port92_info);
575 }
576 
577 type_init(port92_register_types)
578 
579 static void handle_a20_line_change(void *opaque, int irq, int level)
580 {
581     X86CPU *cpu = opaque;
582 
583     /* XXX: send to all CPUs ? */
584     /* XXX: add logic to handle multiple A20 line sources */
585     x86_cpu_set_a20(cpu, level);
586 }
587 
588 int e820_add_entry(uint64_t address, uint64_t length, uint32_t type)
589 {
590     int index = le32_to_cpu(e820_reserve.count);
591     struct e820_entry *entry;
592 
593     if (type != E820_RAM) {
594         /* old FW_CFG_E820_TABLE entry -- reservations only */
595         if (index >= E820_NR_ENTRIES) {
596             return -EBUSY;
597         }
598         entry = &e820_reserve.entry[index++];
599 
600         entry->address = cpu_to_le64(address);
601         entry->length = cpu_to_le64(length);
602         entry->type = cpu_to_le32(type);
603 
604         e820_reserve.count = cpu_to_le32(index);
605     }
606 
607     /* new "etc/e820" file -- include ram too */
608     e820_table = g_realloc(e820_table,
609                            sizeof(struct e820_entry) * (e820_entries+1));
610     e820_table[e820_entries].address = cpu_to_le64(address);
611     e820_table[e820_entries].length = cpu_to_le64(length);
612     e820_table[e820_entries].type = cpu_to_le32(type);
613     e820_entries++;
614 
615     return e820_entries;
616 }
617 
618 int e820_get_num_entries(void)
619 {
620     return e820_entries;
621 }
622 
623 bool e820_get_entry(int idx, uint32_t type, uint64_t *address, uint64_t *length)
624 {
625     if (idx < e820_entries && e820_table[idx].type == cpu_to_le32(type)) {
626         *address = le64_to_cpu(e820_table[idx].address);
627         *length = le64_to_cpu(e820_table[idx].length);
628         return true;
629     }
630     return false;
631 }
632 
633 /* Calculates the limit to CPU APIC ID values
634  *
635  * This function returns the limit for the APIC ID value, so that all
636  * CPU APIC IDs are < pc_apic_id_limit().
637  *
638  * This is used for FW_CFG_MAX_CPUS. See comments on bochs_bios_init().
639  */
640 static unsigned int pc_apic_id_limit(unsigned int max_cpus)
641 {
642     return x86_cpu_apic_id_from_index(max_cpus - 1) + 1;
643 }
644 
645 static FWCfgState *bochs_bios_init(void)
646 {
647     FWCfgState *fw_cfg;
648     uint8_t *smbios_tables, *smbios_anchor;
649     size_t smbios_tables_len, smbios_anchor_len;
650     uint64_t *numa_fw_cfg;
651     int i, j;
652     unsigned int apic_id_limit = pc_apic_id_limit(max_cpus);
653 
654     fw_cfg = fw_cfg_init(BIOS_CFG_IOPORT, BIOS_CFG_IOPORT + 1, 0, 0);
655     /* FW_CFG_MAX_CPUS is a bit confusing/problematic on x86:
656      *
657      * SeaBIOS needs FW_CFG_MAX_CPUS for CPU hotplug, but the CPU hotplug
658      * QEMU<->SeaBIOS interface is not based on the "CPU index", but on the APIC
659      * ID of hotplugged CPUs[1]. This means that FW_CFG_MAX_CPUS is not the
660      * "maximum number of CPUs", but the "limit to the APIC ID values SeaBIOS
661      * may see".
662      *
663      * So, this means we must not use max_cpus, here, but the maximum possible
664      * APIC ID value, plus one.
665      *
666      * [1] The only kind of "CPU identifier" used between SeaBIOS and QEMU is
667      *     the APIC ID, not the "CPU index"
668      */
669     fw_cfg_add_i16(fw_cfg, FW_CFG_MAX_CPUS, (uint16_t)apic_id_limit);
670     fw_cfg_add_i32(fw_cfg, FW_CFG_ID, 1);
671     fw_cfg_add_i64(fw_cfg, FW_CFG_RAM_SIZE, (uint64_t)ram_size);
672     fw_cfg_add_bytes(fw_cfg, FW_CFG_ACPI_TABLES,
673                      acpi_tables, acpi_tables_len);
674     fw_cfg_add_i32(fw_cfg, FW_CFG_IRQ0_OVERRIDE, kvm_allows_irq0_override());
675 
676     smbios_tables = smbios_get_table_legacy(&smbios_tables_len);
677     if (smbios_tables) {
678         fw_cfg_add_bytes(fw_cfg, FW_CFG_SMBIOS_ENTRIES,
679                          smbios_tables, smbios_tables_len);
680     }
681 
682     smbios_get_tables(&smbios_tables, &smbios_tables_len,
683                       &smbios_anchor, &smbios_anchor_len);
684     if (smbios_anchor) {
685         fw_cfg_add_file(fw_cfg, "etc/smbios/smbios-tables",
686                         smbios_tables, smbios_tables_len);
687         fw_cfg_add_file(fw_cfg, "etc/smbios/smbios-anchor",
688                         smbios_anchor, smbios_anchor_len);
689     }
690 
691     fw_cfg_add_bytes(fw_cfg, FW_CFG_E820_TABLE,
692                      &e820_reserve, sizeof(e820_reserve));
693     fw_cfg_add_file(fw_cfg, "etc/e820", e820_table,
694                     sizeof(struct e820_entry) * e820_entries);
695 
696     fw_cfg_add_bytes(fw_cfg, FW_CFG_HPET, &hpet_cfg, sizeof(hpet_cfg));
697     /* allocate memory for the NUMA channel: one (64bit) word for the number
698      * of nodes, one word for each VCPU->node and one word for each node to
699      * hold the amount of memory.
700      */
701     numa_fw_cfg = g_new0(uint64_t, 1 + apic_id_limit + nb_numa_nodes);
702     numa_fw_cfg[0] = cpu_to_le64(nb_numa_nodes);
703     for (i = 0; i < max_cpus; i++) {
704         unsigned int apic_id = x86_cpu_apic_id_from_index(i);
705         assert(apic_id < apic_id_limit);
706         for (j = 0; j < nb_numa_nodes; j++) {
707             if (test_bit(i, numa_info[j].node_cpu)) {
708                 numa_fw_cfg[apic_id + 1] = cpu_to_le64(j);
709                 break;
710             }
711         }
712     }
713     for (i = 0; i < nb_numa_nodes; i++) {
714         numa_fw_cfg[apic_id_limit + 1 + i] = cpu_to_le64(numa_info[i].node_mem);
715     }
716     fw_cfg_add_bytes(fw_cfg, FW_CFG_NUMA, numa_fw_cfg,
717                      (1 + apic_id_limit + nb_numa_nodes) *
718                      sizeof(*numa_fw_cfg));
719 
720     return fw_cfg;
721 }
722 
723 static long get_file_size(FILE *f)
724 {
725     long where, size;
726 
727     /* XXX: on Unix systems, using fstat() probably makes more sense */
728 
729     where = ftell(f);
730     fseek(f, 0, SEEK_END);
731     size = ftell(f);
732     fseek(f, where, SEEK_SET);
733 
734     return size;
735 }
736 
737 static void load_linux(FWCfgState *fw_cfg,
738                        const char *kernel_filename,
739                        const char *initrd_filename,
740                        const char *kernel_cmdline,
741                        hwaddr max_ram_size)
742 {
743     uint16_t protocol;
744     int setup_size, kernel_size, initrd_size = 0, cmdline_size;
745     uint32_t initrd_max;
746     uint8_t header[8192], *setup, *kernel, *initrd_data;
747     hwaddr real_addr, prot_addr, cmdline_addr, initrd_addr = 0;
748     FILE *f;
749     char *vmode;
750 
751     /* Align to 16 bytes as a paranoia measure */
752     cmdline_size = (strlen(kernel_cmdline)+16) & ~15;
753 
754     /* load the kernel header */
755     f = fopen(kernel_filename, "rb");
756     if (!f || !(kernel_size = get_file_size(f)) ||
757         fread(header, 1, MIN(ARRAY_SIZE(header), kernel_size), f) !=
758         MIN(ARRAY_SIZE(header), kernel_size)) {
759         fprintf(stderr, "qemu: could not load kernel '%s': %s\n",
760                 kernel_filename, strerror(errno));
761         exit(1);
762     }
763 
764     /* kernel protocol version */
765 #if 0
766     fprintf(stderr, "header magic: %#x\n", ldl_p(header+0x202));
767 #endif
768     if (ldl_p(header+0x202) == 0x53726448) {
769         protocol = lduw_p(header+0x206);
770     } else {
771         /* This looks like a multiboot kernel. If it is, let's stop
772            treating it like a Linux kernel. */
773         if (load_multiboot(fw_cfg, f, kernel_filename, initrd_filename,
774                            kernel_cmdline, kernel_size, header)) {
775             return;
776         }
777         protocol = 0;
778     }
779 
780     if (protocol < 0x200 || !(header[0x211] & 0x01)) {
781         /* Low kernel */
782         real_addr    = 0x90000;
783         cmdline_addr = 0x9a000 - cmdline_size;
784         prot_addr    = 0x10000;
785     } else if (protocol < 0x202) {
786         /* High but ancient kernel */
787         real_addr    = 0x90000;
788         cmdline_addr = 0x9a000 - cmdline_size;
789         prot_addr    = 0x100000;
790     } else {
791         /* High and recent kernel */
792         real_addr    = 0x10000;
793         cmdline_addr = 0x20000;
794         prot_addr    = 0x100000;
795     }
796 
797 #if 0
798     fprintf(stderr,
799             "qemu: real_addr     = 0x" TARGET_FMT_plx "\n"
800             "qemu: cmdline_addr  = 0x" TARGET_FMT_plx "\n"
801             "qemu: prot_addr     = 0x" TARGET_FMT_plx "\n",
802             real_addr,
803             cmdline_addr,
804             prot_addr);
805 #endif
806 
807     /* highest address for loading the initrd */
808     if (protocol >= 0x203) {
809         initrd_max = ldl_p(header+0x22c);
810     } else {
811         initrd_max = 0x37ffffff;
812     }
813 
814     if (initrd_max >= max_ram_size-ACPI_DATA_SIZE)
815     	initrd_max = max_ram_size-ACPI_DATA_SIZE-1;
816 
817     fw_cfg_add_i32(fw_cfg, FW_CFG_CMDLINE_ADDR, cmdline_addr);
818     fw_cfg_add_i32(fw_cfg, FW_CFG_CMDLINE_SIZE, strlen(kernel_cmdline)+1);
819     fw_cfg_add_string(fw_cfg, FW_CFG_CMDLINE_DATA, kernel_cmdline);
820 
821     if (protocol >= 0x202) {
822         stl_p(header+0x228, cmdline_addr);
823     } else {
824         stw_p(header+0x20, 0xA33F);
825         stw_p(header+0x22, cmdline_addr-real_addr);
826     }
827 
828     /* handle vga= parameter */
829     vmode = strstr(kernel_cmdline, "vga=");
830     if (vmode) {
831         unsigned int video_mode;
832         /* skip "vga=" */
833         vmode += 4;
834         if (!strncmp(vmode, "normal", 6)) {
835             video_mode = 0xffff;
836         } else if (!strncmp(vmode, "ext", 3)) {
837             video_mode = 0xfffe;
838         } else if (!strncmp(vmode, "ask", 3)) {
839             video_mode = 0xfffd;
840         } else {
841             video_mode = strtol(vmode, NULL, 0);
842         }
843         stw_p(header+0x1fa, video_mode);
844     }
845 
846     /* loader type */
847     /* High nybble = B reserved for QEMU; low nybble is revision number.
848        If this code is substantially changed, you may want to consider
849        incrementing the revision. */
850     if (protocol >= 0x200) {
851         header[0x210] = 0xB0;
852     }
853     /* heap */
854     if (protocol >= 0x201) {
855         header[0x211] |= 0x80;	/* CAN_USE_HEAP */
856         stw_p(header+0x224, cmdline_addr-real_addr-0x200);
857     }
858 
859     /* load initrd */
860     if (initrd_filename) {
861         if (protocol < 0x200) {
862             fprintf(stderr, "qemu: linux kernel too old to load a ram disk\n");
863             exit(1);
864         }
865 
866         initrd_size = get_image_size(initrd_filename);
867         if (initrd_size < 0) {
868             fprintf(stderr, "qemu: error reading initrd %s: %s\n",
869                     initrd_filename, strerror(errno));
870             exit(1);
871         }
872 
873         initrd_addr = (initrd_max-initrd_size) & ~4095;
874 
875         initrd_data = g_malloc(initrd_size);
876         load_image(initrd_filename, initrd_data);
877 
878         fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_ADDR, initrd_addr);
879         fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_SIZE, initrd_size);
880         fw_cfg_add_bytes(fw_cfg, FW_CFG_INITRD_DATA, initrd_data, initrd_size);
881 
882         stl_p(header+0x218, initrd_addr);
883         stl_p(header+0x21c, initrd_size);
884     }
885 
886     /* load kernel and setup */
887     setup_size = header[0x1f1];
888     if (setup_size == 0) {
889         setup_size = 4;
890     }
891     setup_size = (setup_size+1)*512;
892     kernel_size -= setup_size;
893 
894     setup  = g_malloc(setup_size);
895     kernel = g_malloc(kernel_size);
896     fseek(f, 0, SEEK_SET);
897     if (fread(setup, 1, setup_size, f) != setup_size) {
898         fprintf(stderr, "fread() failed\n");
899         exit(1);
900     }
901     if (fread(kernel, 1, kernel_size, f) != kernel_size) {
902         fprintf(stderr, "fread() failed\n");
903         exit(1);
904     }
905     fclose(f);
906     memcpy(setup, header, MIN(sizeof(header), setup_size));
907 
908     fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_ADDR, prot_addr);
909     fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_SIZE, kernel_size);
910     fw_cfg_add_bytes(fw_cfg, FW_CFG_KERNEL_DATA, kernel, kernel_size);
911 
912     fw_cfg_add_i32(fw_cfg, FW_CFG_SETUP_ADDR, real_addr);
913     fw_cfg_add_i32(fw_cfg, FW_CFG_SETUP_SIZE, setup_size);
914     fw_cfg_add_bytes(fw_cfg, FW_CFG_SETUP_DATA, setup, setup_size);
915 
916     option_rom[nb_option_roms].name = "linuxboot.bin";
917     option_rom[nb_option_roms].bootindex = 0;
918     nb_option_roms++;
919 }
920 
921 #define NE2000_NB_MAX 6
922 
923 static const int ne2000_io[NE2000_NB_MAX] = { 0x300, 0x320, 0x340, 0x360,
924                                               0x280, 0x380 };
925 static const int ne2000_irq[NE2000_NB_MAX] = { 9, 10, 11, 3, 4, 5 };
926 
927 void pc_init_ne2k_isa(ISABus *bus, NICInfo *nd)
928 {
929     static int nb_ne2k = 0;
930 
931     if (nb_ne2k == NE2000_NB_MAX)
932         return;
933     isa_ne2000_init(bus, ne2000_io[nb_ne2k],
934                     ne2000_irq[nb_ne2k], nd);
935     nb_ne2k++;
936 }
937 
938 DeviceState *cpu_get_current_apic(void)
939 {
940     if (current_cpu) {
941         X86CPU *cpu = X86_CPU(current_cpu);
942         return cpu->apic_state;
943     } else {
944         return NULL;
945     }
946 }
947 
948 void pc_acpi_smi_interrupt(void *opaque, int irq, int level)
949 {
950     X86CPU *cpu = opaque;
951 
952     if (level) {
953         cpu_interrupt(CPU(cpu), CPU_INTERRUPT_SMI);
954     }
955 }
956 
957 static X86CPU *pc_new_cpu(const char *cpu_model, int64_t apic_id,
958                           DeviceState *icc_bridge, Error **errp)
959 {
960     X86CPU *cpu;
961     Error *local_err = NULL;
962 
963     cpu = cpu_x86_create(cpu_model, icc_bridge, &local_err);
964     if (local_err != NULL) {
965         error_propagate(errp, local_err);
966         return NULL;
967     }
968 
969     object_property_set_int(OBJECT(cpu), apic_id, "apic-id", &local_err);
970     object_property_set_bool(OBJECT(cpu), true, "realized", &local_err);
971 
972     if (local_err) {
973         error_propagate(errp, local_err);
974         object_unref(OBJECT(cpu));
975         cpu = NULL;
976     }
977     return cpu;
978 }
979 
980 static const char *current_cpu_model;
981 
982 void pc_hot_add_cpu(const int64_t id, Error **errp)
983 {
984     DeviceState *icc_bridge;
985     int64_t apic_id = x86_cpu_apic_id_from_index(id);
986 
987     if (id < 0) {
988         error_setg(errp, "Invalid CPU id: %" PRIi64, id);
989         return;
990     }
991 
992     if (cpu_exists(apic_id)) {
993         error_setg(errp, "Unable to add CPU: %" PRIi64
994                    ", it already exists", id);
995         return;
996     }
997 
998     if (id >= max_cpus) {
999         error_setg(errp, "Unable to add CPU: %" PRIi64
1000                    ", max allowed: %d", id, max_cpus - 1);
1001         return;
1002     }
1003 
1004     if (apic_id >= ACPI_CPU_HOTPLUG_ID_LIMIT) {
1005         error_setg(errp, "Unable to add CPU: %" PRIi64
1006                    ", resulting APIC ID (%" PRIi64 ") is too large",
1007                    id, apic_id);
1008         return;
1009     }
1010 
1011     icc_bridge = DEVICE(object_resolve_path_type("icc-bridge",
1012                                                  TYPE_ICC_BRIDGE, NULL));
1013     pc_new_cpu(current_cpu_model, apic_id, icc_bridge, errp);
1014 }
1015 
1016 void pc_cpus_init(const char *cpu_model, DeviceState *icc_bridge)
1017 {
1018     int i;
1019     X86CPU *cpu = NULL;
1020     Error *error = NULL;
1021     unsigned long apic_id_limit;
1022 
1023     /* init CPUs */
1024     if (cpu_model == NULL) {
1025 #ifdef TARGET_X86_64
1026         cpu_model = "qemu64";
1027 #else
1028         cpu_model = "qemu32";
1029 #endif
1030     }
1031     current_cpu_model = cpu_model;
1032 
1033     apic_id_limit = pc_apic_id_limit(max_cpus);
1034     if (apic_id_limit > ACPI_CPU_HOTPLUG_ID_LIMIT) {
1035         error_report("max_cpus is too large. APIC ID of last CPU is %lu",
1036                      apic_id_limit - 1);
1037         exit(1);
1038     }
1039 
1040     for (i = 0; i < smp_cpus; i++) {
1041         cpu = pc_new_cpu(cpu_model, x86_cpu_apic_id_from_index(i),
1042                          icc_bridge, &error);
1043         if (error) {
1044             error_report("%s", error_get_pretty(error));
1045             error_free(error);
1046             exit(1);
1047         }
1048     }
1049 
1050     /* map APIC MMIO area if CPU has APIC */
1051     if (cpu && cpu->apic_state) {
1052         /* XXX: what if the base changes? */
1053         sysbus_mmio_map_overlap(SYS_BUS_DEVICE(icc_bridge), 0,
1054                                 APIC_DEFAULT_ADDRESS, 0x1000);
1055     }
1056 
1057     /* tell smbios about cpuid version and features */
1058     smbios_set_cpuid(cpu->env.cpuid_version, cpu->env.features[FEAT_1_EDX]);
1059 }
1060 
1061 /* pci-info ROM file. Little endian format */
1062 typedef struct PcRomPciInfo {
1063     uint64_t w32_min;
1064     uint64_t w32_max;
1065     uint64_t w64_min;
1066     uint64_t w64_max;
1067 } PcRomPciInfo;
1068 
1069 static void pc_fw_cfg_guest_info(PcGuestInfo *guest_info)
1070 {
1071     PcRomPciInfo *info;
1072     Object *pci_info;
1073     bool ambiguous = false;
1074 
1075     if (!guest_info->has_pci_info || !guest_info->fw_cfg) {
1076         return;
1077     }
1078     pci_info = object_resolve_path_type("", TYPE_PCI_HOST_BRIDGE, &ambiguous);
1079     g_assert(!ambiguous);
1080     if (!pci_info) {
1081         return;
1082     }
1083 
1084     info = g_malloc(sizeof *info);
1085     info->w32_min = cpu_to_le64(object_property_get_int(pci_info,
1086                                 PCI_HOST_PROP_PCI_HOLE_START, NULL));
1087     info->w32_max = cpu_to_le64(object_property_get_int(pci_info,
1088                                 PCI_HOST_PROP_PCI_HOLE_END, NULL));
1089     info->w64_min = cpu_to_le64(object_property_get_int(pci_info,
1090                                 PCI_HOST_PROP_PCI_HOLE64_START, NULL));
1091     info->w64_max = cpu_to_le64(object_property_get_int(pci_info,
1092                                 PCI_HOST_PROP_PCI_HOLE64_END, NULL));
1093     /* Pass PCI hole info to guest via a side channel.
1094      * Required so guest PCI enumeration does the right thing. */
1095     fw_cfg_add_file(guest_info->fw_cfg, "etc/pci-info", info, sizeof *info);
1096 }
1097 
1098 typedef struct PcGuestInfoState {
1099     PcGuestInfo info;
1100     Notifier machine_done;
1101 } PcGuestInfoState;
1102 
1103 static
1104 void pc_guest_info_machine_done(Notifier *notifier, void *data)
1105 {
1106     PcGuestInfoState *guest_info_state = container_of(notifier,
1107                                                       PcGuestInfoState,
1108                                                       machine_done);
1109     pc_fw_cfg_guest_info(&guest_info_state->info);
1110     acpi_setup(&guest_info_state->info);
1111 }
1112 
1113 PcGuestInfo *pc_guest_info_init(ram_addr_t below_4g_mem_size,
1114                                 ram_addr_t above_4g_mem_size)
1115 {
1116     PcGuestInfoState *guest_info_state = g_malloc0(sizeof *guest_info_state);
1117     PcGuestInfo *guest_info = &guest_info_state->info;
1118     int i, j;
1119 
1120     guest_info->ram_size_below_4g = below_4g_mem_size;
1121     guest_info->ram_size = below_4g_mem_size + above_4g_mem_size;
1122     guest_info->apic_id_limit = pc_apic_id_limit(max_cpus);
1123     guest_info->apic_xrupt_override = kvm_allows_irq0_override();
1124     guest_info->numa_nodes = nb_numa_nodes;
1125     guest_info->node_mem = g_malloc0(guest_info->numa_nodes *
1126                                     sizeof *guest_info->node_mem);
1127     for (i = 0; i < nb_numa_nodes; i++) {
1128         guest_info->node_mem[i] = numa_info[i].node_mem;
1129     }
1130 
1131     guest_info->node_cpu = g_malloc0(guest_info->apic_id_limit *
1132                                      sizeof *guest_info->node_cpu);
1133 
1134     for (i = 0; i < max_cpus; i++) {
1135         unsigned int apic_id = x86_cpu_apic_id_from_index(i);
1136         assert(apic_id < guest_info->apic_id_limit);
1137         for (j = 0; j < nb_numa_nodes; j++) {
1138             if (test_bit(i, numa_info[j].node_cpu)) {
1139                 guest_info->node_cpu[apic_id] = j;
1140                 break;
1141             }
1142         }
1143     }
1144 
1145     guest_info_state->machine_done.notify = pc_guest_info_machine_done;
1146     qemu_add_machine_init_done_notifier(&guest_info_state->machine_done);
1147     return guest_info;
1148 }
1149 
1150 /* setup pci memory address space mapping into system address space */
1151 void pc_pci_as_mapping_init(Object *owner, MemoryRegion *system_memory,
1152                             MemoryRegion *pci_address_space)
1153 {
1154     /* Set to lower priority than RAM */
1155     memory_region_add_subregion_overlap(system_memory, 0x0,
1156                                         pci_address_space, -1);
1157 }
1158 
1159 void pc_acpi_init(const char *default_dsdt)
1160 {
1161     char *filename;
1162 
1163     if (acpi_tables != NULL) {
1164         /* manually set via -acpitable, leave it alone */
1165         return;
1166     }
1167 
1168     filename = qemu_find_file(QEMU_FILE_TYPE_BIOS, default_dsdt);
1169     if (filename == NULL) {
1170         fprintf(stderr, "WARNING: failed to find %s\n", default_dsdt);
1171     } else {
1172         char *arg;
1173         QemuOpts *opts;
1174         Error *err = NULL;
1175 
1176         arg = g_strdup_printf("file=%s", filename);
1177 
1178         /* creates a deep copy of "arg" */
1179         opts = qemu_opts_parse(qemu_find_opts("acpi"), arg, 0);
1180         g_assert(opts != NULL);
1181 
1182         acpi_table_add_builtin(opts, &err);
1183         if (err) {
1184             error_report("WARNING: failed to load %s: %s", filename,
1185                          error_get_pretty(err));
1186             error_free(err);
1187         }
1188         g_free(arg);
1189         g_free(filename);
1190     }
1191 }
1192 
1193 FWCfgState *pc_memory_init(MachineState *machine,
1194                            MemoryRegion *system_memory,
1195                            ram_addr_t below_4g_mem_size,
1196                            ram_addr_t above_4g_mem_size,
1197                            MemoryRegion *rom_memory,
1198                            MemoryRegion **ram_memory,
1199                            PcGuestInfo *guest_info)
1200 {
1201     int linux_boot, i;
1202     MemoryRegion *ram, *option_rom_mr;
1203     MemoryRegion *ram_below_4g, *ram_above_4g;
1204     FWCfgState *fw_cfg;
1205     PCMachineState *pcms = PC_MACHINE(machine);
1206 
1207     assert(machine->ram_size == below_4g_mem_size + above_4g_mem_size);
1208 
1209     linux_boot = (machine->kernel_filename != NULL);
1210 
1211     /* Allocate RAM.  We allocate it as a single memory region and use
1212      * aliases to address portions of it, mostly for backwards compatibility
1213      * with older qemus that used qemu_ram_alloc().
1214      */
1215     ram = g_malloc(sizeof(*ram));
1216     memory_region_allocate_system_memory(ram, NULL, "pc.ram",
1217                                          machine->ram_size);
1218     *ram_memory = ram;
1219     ram_below_4g = g_malloc(sizeof(*ram_below_4g));
1220     memory_region_init_alias(ram_below_4g, NULL, "ram-below-4g", ram,
1221                              0, below_4g_mem_size);
1222     memory_region_add_subregion(system_memory, 0, ram_below_4g);
1223     e820_add_entry(0, below_4g_mem_size, E820_RAM);
1224     if (above_4g_mem_size > 0) {
1225         ram_above_4g = g_malloc(sizeof(*ram_above_4g));
1226         memory_region_init_alias(ram_above_4g, NULL, "ram-above-4g", ram,
1227                                  below_4g_mem_size, above_4g_mem_size);
1228         memory_region_add_subregion(system_memory, 0x100000000ULL,
1229                                     ram_above_4g);
1230         e820_add_entry(0x100000000ULL, above_4g_mem_size, E820_RAM);
1231     }
1232 
1233     if (!guest_info->has_reserved_memory &&
1234         (machine->ram_slots ||
1235          (machine->maxram_size > machine->ram_size))) {
1236         MachineClass *mc = MACHINE_GET_CLASS(machine);
1237 
1238         error_report("\"-memory 'slots|maxmem'\" is not supported by: %s",
1239                      mc->name);
1240         exit(EXIT_FAILURE);
1241     }
1242 
1243     /* initialize hotplug memory address space */
1244     if (guest_info->has_reserved_memory &&
1245         (machine->ram_size < machine->maxram_size)) {
1246         ram_addr_t hotplug_mem_size =
1247             machine->maxram_size - machine->ram_size;
1248 
1249         if (machine->ram_slots > ACPI_MAX_RAM_SLOTS) {
1250             error_report("unsupported amount of memory slots: %"PRIu64,
1251                          machine->ram_slots);
1252             exit(EXIT_FAILURE);
1253         }
1254 
1255         pcms->hotplug_memory_base =
1256             ROUND_UP(0x100000000ULL + above_4g_mem_size, 1ULL << 30);
1257 
1258         if ((pcms->hotplug_memory_base + hotplug_mem_size) <
1259             hotplug_mem_size) {
1260             error_report("unsupported amount of maximum memory: " RAM_ADDR_FMT,
1261                          machine->maxram_size);
1262             exit(EXIT_FAILURE);
1263         }
1264 
1265         memory_region_init(&pcms->hotplug_memory, OBJECT(pcms),
1266                            "hotplug-memory", hotplug_mem_size);
1267         memory_region_add_subregion(system_memory, pcms->hotplug_memory_base,
1268                                     &pcms->hotplug_memory);
1269     }
1270 
1271     /* Initialize PC system firmware */
1272     pc_system_firmware_init(rom_memory, guest_info->isapc_ram_fw);
1273 
1274     option_rom_mr = g_malloc(sizeof(*option_rom_mr));
1275     memory_region_init_ram(option_rom_mr, NULL, "pc.rom", PC_ROM_SIZE);
1276     vmstate_register_ram_global(option_rom_mr);
1277     memory_region_add_subregion_overlap(rom_memory,
1278                                         PC_ROM_MIN_VGA,
1279                                         option_rom_mr,
1280                                         1);
1281 
1282     fw_cfg = bochs_bios_init();
1283     rom_set_fw(fw_cfg);
1284 
1285     if (guest_info->has_reserved_memory && pcms->hotplug_memory_base) {
1286         uint64_t *val = g_malloc(sizeof(*val));
1287         *val = cpu_to_le64(ROUND_UP(pcms->hotplug_memory_base, 0x1ULL << 30));
1288         fw_cfg_add_file(fw_cfg, "etc/reserved-memory-end", val, sizeof(*val));
1289     }
1290 
1291     if (linux_boot) {
1292         load_linux(fw_cfg, machine->kernel_filename, machine->initrd_filename,
1293                    machine->kernel_cmdline, below_4g_mem_size);
1294     }
1295 
1296     for (i = 0; i < nb_option_roms; i++) {
1297         rom_add_option(option_rom[i].name, option_rom[i].bootindex);
1298     }
1299     guest_info->fw_cfg = fw_cfg;
1300     return fw_cfg;
1301 }
1302 
1303 qemu_irq *pc_allocate_cpu_irq(void)
1304 {
1305     return qemu_allocate_irqs(pic_irq_request, NULL, 1);
1306 }
1307 
1308 DeviceState *pc_vga_init(ISABus *isa_bus, PCIBus *pci_bus)
1309 {
1310     DeviceState *dev = NULL;
1311 
1312     if (pci_bus) {
1313         PCIDevice *pcidev = pci_vga_init(pci_bus);
1314         dev = pcidev ? &pcidev->qdev : NULL;
1315     } else if (isa_bus) {
1316         ISADevice *isadev = isa_vga_init(isa_bus);
1317         dev = isadev ? DEVICE(isadev) : NULL;
1318     }
1319     return dev;
1320 }
1321 
1322 static void cpu_request_exit(void *opaque, int irq, int level)
1323 {
1324     CPUState *cpu = current_cpu;
1325 
1326     if (cpu && level) {
1327         cpu_exit(cpu);
1328     }
1329 }
1330 
1331 static const MemoryRegionOps ioport80_io_ops = {
1332     .write = ioport80_write,
1333     .read = ioport80_read,
1334     .endianness = DEVICE_NATIVE_ENDIAN,
1335     .impl = {
1336         .min_access_size = 1,
1337         .max_access_size = 1,
1338     },
1339 };
1340 
1341 static const MemoryRegionOps ioportF0_io_ops = {
1342     .write = ioportF0_write,
1343     .read = ioportF0_read,
1344     .endianness = DEVICE_NATIVE_ENDIAN,
1345     .impl = {
1346         .min_access_size = 1,
1347         .max_access_size = 1,
1348     },
1349 };
1350 
1351 void pc_basic_device_init(ISABus *isa_bus, qemu_irq *gsi,
1352                           ISADevice **rtc_state,
1353                           ISADevice **floppy,
1354                           bool no_vmport,
1355                           uint32 hpet_irqs)
1356 {
1357     int i;
1358     DriveInfo *fd[MAX_FD];
1359     DeviceState *hpet = NULL;
1360     int pit_isa_irq = 0;
1361     qemu_irq pit_alt_irq = NULL;
1362     qemu_irq rtc_irq = NULL;
1363     qemu_irq *a20_line;
1364     ISADevice *i8042, *port92, *vmmouse, *pit = NULL;
1365     qemu_irq *cpu_exit_irq;
1366     MemoryRegion *ioport80_io = g_new(MemoryRegion, 1);
1367     MemoryRegion *ioportF0_io = g_new(MemoryRegion, 1);
1368 
1369     memory_region_init_io(ioport80_io, NULL, &ioport80_io_ops, NULL, "ioport80", 1);
1370     memory_region_add_subregion(isa_bus->address_space_io, 0x80, ioport80_io);
1371 
1372     memory_region_init_io(ioportF0_io, NULL, &ioportF0_io_ops, NULL, "ioportF0", 1);
1373     memory_region_add_subregion(isa_bus->address_space_io, 0xf0, ioportF0_io);
1374 
1375     /*
1376      * Check if an HPET shall be created.
1377      *
1378      * Without KVM_CAP_PIT_STATE2, we cannot switch off the in-kernel PIT
1379      * when the HPET wants to take over. Thus we have to disable the latter.
1380      */
1381     if (!no_hpet && (!kvm_irqchip_in_kernel() || kvm_has_pit_state2())) {
1382         /* In order to set property, here not using sysbus_try_create_simple */
1383         hpet = qdev_try_create(NULL, TYPE_HPET);
1384         if (hpet) {
1385             /* For pc-piix-*, hpet's intcap is always IRQ2. For pc-q35-1.7
1386              * and earlier, use IRQ2 for compat. Otherwise, use IRQ16~23,
1387              * IRQ8 and IRQ2.
1388              */
1389             uint8_t compat = object_property_get_int(OBJECT(hpet),
1390                     HPET_INTCAP, NULL);
1391             if (!compat) {
1392                 qdev_prop_set_uint32(hpet, HPET_INTCAP, hpet_irqs);
1393             }
1394             qdev_init_nofail(hpet);
1395             sysbus_mmio_map(SYS_BUS_DEVICE(hpet), 0, HPET_BASE);
1396 
1397             for (i = 0; i < GSI_NUM_PINS; i++) {
1398                 sysbus_connect_irq(SYS_BUS_DEVICE(hpet), i, gsi[i]);
1399             }
1400             pit_isa_irq = -1;
1401             pit_alt_irq = qdev_get_gpio_in(hpet, HPET_LEGACY_PIT_INT);
1402             rtc_irq = qdev_get_gpio_in(hpet, HPET_LEGACY_RTC_INT);
1403         }
1404     }
1405     *rtc_state = rtc_init(isa_bus, 2000, rtc_irq);
1406 
1407     qemu_register_boot_set(pc_boot_set, *rtc_state);
1408 
1409     if (!xen_enabled()) {
1410         if (kvm_irqchip_in_kernel()) {
1411             pit = kvm_pit_init(isa_bus, 0x40);
1412         } else {
1413             pit = pit_init(isa_bus, 0x40, pit_isa_irq, pit_alt_irq);
1414         }
1415         if (hpet) {
1416             /* connect PIT to output control line of the HPET */
1417             qdev_connect_gpio_out(hpet, 0, qdev_get_gpio_in(DEVICE(pit), 0));
1418         }
1419         pcspk_init(isa_bus, pit);
1420     }
1421 
1422     for(i = 0; i < MAX_SERIAL_PORTS; i++) {
1423         if (serial_hds[i]) {
1424             serial_isa_init(isa_bus, i, serial_hds[i]);
1425         }
1426     }
1427 
1428     for(i = 0; i < MAX_PARALLEL_PORTS; i++) {
1429         if (parallel_hds[i]) {
1430             parallel_init(isa_bus, i, parallel_hds[i]);
1431         }
1432     }
1433 
1434     a20_line = qemu_allocate_irqs(handle_a20_line_change, first_cpu, 2);
1435     i8042 = isa_create_simple(isa_bus, "i8042");
1436     i8042_setup_a20_line(i8042, &a20_line[0]);
1437     if (!no_vmport) {
1438         vmport_init(isa_bus);
1439         vmmouse = isa_try_create(isa_bus, "vmmouse");
1440     } else {
1441         vmmouse = NULL;
1442     }
1443     if (vmmouse) {
1444         DeviceState *dev = DEVICE(vmmouse);
1445         qdev_prop_set_ptr(dev, "ps2_mouse", i8042);
1446         qdev_init_nofail(dev);
1447     }
1448     port92 = isa_create_simple(isa_bus, "port92");
1449     port92_init(port92, &a20_line[1]);
1450 
1451     cpu_exit_irq = qemu_allocate_irqs(cpu_request_exit, NULL, 1);
1452     DMA_init(0, cpu_exit_irq);
1453 
1454     for(i = 0; i < MAX_FD; i++) {
1455         fd[i] = drive_get(IF_FLOPPY, 0, i);
1456     }
1457     *floppy = fdctrl_init_isa(isa_bus, fd);
1458 }
1459 
1460 void pc_nic_init(ISABus *isa_bus, PCIBus *pci_bus)
1461 {
1462     int i;
1463 
1464     for (i = 0; i < nb_nics; i++) {
1465         NICInfo *nd = &nd_table[i];
1466 
1467         if (!pci_bus || (nd->model && strcmp(nd->model, "ne2k_isa") == 0)) {
1468             pc_init_ne2k_isa(isa_bus, nd);
1469         } else {
1470             pci_nic_init_nofail(nd, pci_bus, "e1000", NULL);
1471         }
1472     }
1473 }
1474 
1475 void pc_pci_device_init(PCIBus *pci_bus)
1476 {
1477     int max_bus;
1478     int bus;
1479 
1480     max_bus = drive_get_max_bus(IF_SCSI);
1481     for (bus = 0; bus <= max_bus; bus++) {
1482         pci_create_simple(pci_bus, -1, "lsi53c895a");
1483     }
1484 }
1485 
1486 void ioapic_init_gsi(GSIState *gsi_state, const char *parent_name)
1487 {
1488     DeviceState *dev;
1489     SysBusDevice *d;
1490     unsigned int i;
1491 
1492     if (kvm_irqchip_in_kernel()) {
1493         dev = qdev_create(NULL, "kvm-ioapic");
1494     } else {
1495         dev = qdev_create(NULL, "ioapic");
1496     }
1497     if (parent_name) {
1498         object_property_add_child(object_resolve_path(parent_name, NULL),
1499                                   "ioapic", OBJECT(dev), NULL);
1500     }
1501     qdev_init_nofail(dev);
1502     d = SYS_BUS_DEVICE(dev);
1503     sysbus_mmio_map(d, 0, IO_APIC_DEFAULT_ADDRESS);
1504 
1505     for (i = 0; i < IOAPIC_NUM_PINS; i++) {
1506         gsi_state->ioapic_irq[i] = qdev_get_gpio_in(dev, i);
1507     }
1508 }
1509 
1510 static void pc_generic_machine_class_init(ObjectClass *oc, void *data)
1511 {
1512     MachineClass *mc = MACHINE_CLASS(oc);
1513     QEMUMachine *qm = data;
1514 
1515     mc->name = qm->name;
1516     mc->alias = qm->alias;
1517     mc->desc = qm->desc;
1518     mc->init = qm->init;
1519     mc->reset = qm->reset;
1520     mc->hot_add_cpu = qm->hot_add_cpu;
1521     mc->kvm_type = qm->kvm_type;
1522     mc->block_default_type = qm->block_default_type;
1523     mc->max_cpus = qm->max_cpus;
1524     mc->no_serial = qm->no_serial;
1525     mc->no_parallel = qm->no_parallel;
1526     mc->use_virtcon = qm->use_virtcon;
1527     mc->use_sclp = qm->use_sclp;
1528     mc->no_floppy = qm->no_floppy;
1529     mc->no_cdrom = qm->no_cdrom;
1530     mc->no_sdcard = qm->no_sdcard;
1531     mc->is_default = qm->is_default;
1532     mc->default_machine_opts = qm->default_machine_opts;
1533     mc->default_boot_order = qm->default_boot_order;
1534     mc->compat_props = qm->compat_props;
1535     mc->hw_version = qm->hw_version;
1536 }
1537 
1538 void qemu_register_pc_machine(QEMUMachine *m)
1539 {
1540     char *name = g_strconcat(m->name, TYPE_MACHINE_SUFFIX, NULL);
1541     TypeInfo ti = {
1542         .name       = name,
1543         .parent     = TYPE_PC_MACHINE,
1544         .class_init = pc_generic_machine_class_init,
1545         .class_data = (void *)m,
1546     };
1547 
1548     type_register(&ti);
1549     g_free(name);
1550 }
1551 
1552 static void pc_dimm_plug(HotplugHandler *hotplug_dev,
1553                          DeviceState *dev, Error **errp)
1554 {
1555     int slot;
1556     HotplugHandlerClass *hhc;
1557     Error *local_err = NULL;
1558     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1559     MachineState *machine = MACHINE(hotplug_dev);
1560     PCDIMMDevice *dimm = PC_DIMM(dev);
1561     PCDIMMDeviceClass *ddc = PC_DIMM_GET_CLASS(dimm);
1562     MemoryRegion *mr = ddc->get_memory_region(dimm);
1563     uint64_t addr = object_property_get_int(OBJECT(dimm), PC_DIMM_ADDR_PROP,
1564                                             &local_err);
1565     if (local_err) {
1566         goto out;
1567     }
1568 
1569     addr = pc_dimm_get_free_addr(pcms->hotplug_memory_base,
1570                                  memory_region_size(&pcms->hotplug_memory),
1571                                  !addr ? NULL : &addr,
1572                                  memory_region_size(mr), &local_err);
1573     if (local_err) {
1574         goto out;
1575     }
1576 
1577     object_property_set_int(OBJECT(dev), addr, PC_DIMM_ADDR_PROP, &local_err);
1578     if (local_err) {
1579         goto out;
1580     }
1581     trace_mhp_pc_dimm_assigned_address(addr);
1582 
1583     slot = object_property_get_int(OBJECT(dev), PC_DIMM_SLOT_PROP, &local_err);
1584     if (local_err) {
1585         goto out;
1586     }
1587 
1588     slot = pc_dimm_get_free_slot(slot == PC_DIMM_UNASSIGNED_SLOT ? NULL : &slot,
1589                                  machine->ram_slots, &local_err);
1590     if (local_err) {
1591         goto out;
1592     }
1593     object_property_set_int(OBJECT(dev), slot, PC_DIMM_SLOT_PROP, &local_err);
1594     if (local_err) {
1595         goto out;
1596     }
1597     trace_mhp_pc_dimm_assigned_slot(slot);
1598 
1599     if (!pcms->acpi_dev) {
1600         error_setg(&local_err,
1601                    "memory hotplug is not enabled: missing acpi device");
1602         goto out;
1603     }
1604 
1605     memory_region_add_subregion(&pcms->hotplug_memory,
1606                                 addr - pcms->hotplug_memory_base, mr);
1607     vmstate_register_ram(mr, dev);
1608 
1609     hhc = HOTPLUG_HANDLER_GET_CLASS(pcms->acpi_dev);
1610     hhc->plug(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &local_err);
1611 out:
1612     error_propagate(errp, local_err);
1613 }
1614 
1615 static void pc_machine_device_plug_cb(HotplugHandler *hotplug_dev,
1616                                       DeviceState *dev, Error **errp)
1617 {
1618     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
1619         pc_dimm_plug(hotplug_dev, dev, errp);
1620     }
1621 }
1622 
1623 static HotplugHandler *pc_get_hotpug_handler(MachineState *machine,
1624                                              DeviceState *dev)
1625 {
1626     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(machine);
1627 
1628     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
1629         return HOTPLUG_HANDLER(machine);
1630     }
1631 
1632     return pcmc->get_hotplug_handler ?
1633         pcmc->get_hotplug_handler(machine, dev) : NULL;
1634 }
1635 
1636 static void
1637 pc_machine_get_hotplug_memory_region_size(Object *obj, Visitor *v, void *opaque,
1638                                           const char *name, Error **errp)
1639 {
1640     PCMachineState *pcms = PC_MACHINE(obj);
1641     int64_t value = memory_region_size(&pcms->hotplug_memory);
1642 
1643     visit_type_int(v, &value, name, errp);
1644 }
1645 
1646 static void pc_machine_get_max_ram_below_4g(Object *obj, Visitor *v,
1647                                          void *opaque, const char *name,
1648                                          Error **errp)
1649 {
1650     PCMachineState *pcms = PC_MACHINE(obj);
1651     uint64_t value = pcms->max_ram_below_4g;
1652 
1653     visit_type_size(v, &value, name, errp);
1654 }
1655 
1656 static void pc_machine_set_max_ram_below_4g(Object *obj, Visitor *v,
1657                                          void *opaque, const char *name,
1658                                          Error **errp)
1659 {
1660     PCMachineState *pcms = PC_MACHINE(obj);
1661     Error *error = NULL;
1662     uint64_t value;
1663 
1664     visit_type_size(v, &value, name, &error);
1665     if (error) {
1666         error_propagate(errp, error);
1667         return;
1668     }
1669     if (value > (1ULL << 32)) {
1670         error_set(&error, ERROR_CLASS_GENERIC_ERROR,
1671                   "Machine option 'max-ram-below-4g=%"PRIu64
1672                   "' expects size less than or equal to 4G", value);
1673         error_propagate(errp, error);
1674         return;
1675     }
1676 
1677     if (value < (1ULL << 20)) {
1678         error_report("Warning: small max_ram_below_4g(%"PRIu64
1679                      ") less than 1M.  BIOS may not work..",
1680                      value);
1681     }
1682 
1683     pcms->max_ram_below_4g = value;
1684 }
1685 
1686 static void pc_machine_initfn(Object *obj)
1687 {
1688     PCMachineState *pcms = PC_MACHINE(obj);
1689 
1690     object_property_add(obj, PC_MACHINE_MEMHP_REGION_SIZE, "int",
1691                         pc_machine_get_hotplug_memory_region_size,
1692                         NULL, NULL, NULL, NULL);
1693     pcms->max_ram_below_4g = 1ULL << 32; /* 4G */
1694     object_property_add(obj, PC_MACHINE_MAX_RAM_BELOW_4G, "size",
1695                         pc_machine_get_max_ram_below_4g,
1696                         pc_machine_set_max_ram_below_4g,
1697                         NULL, NULL, NULL);
1698 }
1699 
1700 static void pc_machine_class_init(ObjectClass *oc, void *data)
1701 {
1702     MachineClass *mc = MACHINE_CLASS(oc);
1703     PCMachineClass *pcmc = PC_MACHINE_CLASS(oc);
1704     HotplugHandlerClass *hc = HOTPLUG_HANDLER_CLASS(oc);
1705 
1706     pcmc->get_hotplug_handler = mc->get_hotplug_handler;
1707     mc->get_hotplug_handler = pc_get_hotpug_handler;
1708     hc->plug = pc_machine_device_plug_cb;
1709 }
1710 
1711 static const TypeInfo pc_machine_info = {
1712     .name = TYPE_PC_MACHINE,
1713     .parent = TYPE_MACHINE,
1714     .abstract = true,
1715     .instance_size = sizeof(PCMachineState),
1716     .instance_init = pc_machine_initfn,
1717     .class_size = sizeof(PCMachineClass),
1718     .class_init = pc_machine_class_init,
1719     .interfaces = (InterfaceInfo[]) {
1720          { TYPE_HOTPLUG_HANDLER },
1721          { }
1722     },
1723 };
1724 
1725 static void pc_machine_register_types(void)
1726 {
1727     type_register_static(&pc_machine_info);
1728 }
1729 
1730 type_init(pc_machine_register_types)
1731