xref: /openbmc/qemu/hw/i386/pc.c (revision eeceb084)
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 
57 /* debug PC/ISA interrupts */
58 //#define DEBUG_IRQ
59 
60 #ifdef DEBUG_IRQ
61 #define DPRINTF(fmt, ...)                                       \
62     do { printf("CPUIRQ: " fmt , ## __VA_ARGS__); } while (0)
63 #else
64 #define DPRINTF(fmt, ...)
65 #endif
66 
67 /* Leave a chunk of memory at the top of RAM for the BIOS ACPI tables.  */
68 #define ACPI_DATA_SIZE       0x10000
69 #define BIOS_CFG_IOPORT 0x510
70 #define FW_CFG_ACPI_TABLES (FW_CFG_ARCH_LOCAL + 0)
71 #define FW_CFG_SMBIOS_ENTRIES (FW_CFG_ARCH_LOCAL + 1)
72 #define FW_CFG_IRQ0_OVERRIDE (FW_CFG_ARCH_LOCAL + 2)
73 #define FW_CFG_E820_TABLE (FW_CFG_ARCH_LOCAL + 3)
74 #define FW_CFG_HPET (FW_CFG_ARCH_LOCAL + 4)
75 
76 #define IO_APIC_DEFAULT_ADDRESS 0xfec00000
77 
78 #define E820_NR_ENTRIES		16
79 
80 struct e820_entry {
81     uint64_t address;
82     uint64_t length;
83     uint32_t type;
84 } QEMU_PACKED __attribute((__aligned__(4)));
85 
86 struct e820_table {
87     uint32_t count;
88     struct e820_entry entry[E820_NR_ENTRIES];
89 } QEMU_PACKED __attribute((__aligned__(4)));
90 
91 static struct e820_table e820_table;
92 struct hpet_fw_config hpet_cfg = {.count = UINT8_MAX};
93 
94 void gsi_handler(void *opaque, int n, int level)
95 {
96     GSIState *s = opaque;
97 
98     DPRINTF("pc: %s GSI %d\n", level ? "raising" : "lowering", n);
99     if (n < ISA_NUM_IRQS) {
100         qemu_set_irq(s->i8259_irq[n], level);
101     }
102     qemu_set_irq(s->ioapic_irq[n], level);
103 }
104 
105 static void ioport80_write(void *opaque, hwaddr addr, uint64_t data,
106                            unsigned size)
107 {
108 }
109 
110 static uint64_t ioport80_read(void *opaque, hwaddr addr, unsigned size)
111 {
112     return 0xffffffffffffffffULL;
113 }
114 
115 /* MSDOS compatibility mode FPU exception support */
116 static qemu_irq ferr_irq;
117 
118 void pc_register_ferr_irq(qemu_irq irq)
119 {
120     ferr_irq = irq;
121 }
122 
123 /* XXX: add IGNNE support */
124 void cpu_set_ferr(CPUX86State *s)
125 {
126     qemu_irq_raise(ferr_irq);
127 }
128 
129 static void ioportF0_write(void *opaque, hwaddr addr, uint64_t data,
130                            unsigned size)
131 {
132     qemu_irq_lower(ferr_irq);
133 }
134 
135 static uint64_t ioportF0_read(void *opaque, hwaddr addr, unsigned size)
136 {
137     return 0xffffffffffffffffULL;
138 }
139 
140 /* TSC handling */
141 uint64_t cpu_get_tsc(CPUX86State *env)
142 {
143     return cpu_get_ticks();
144 }
145 
146 /* SMM support */
147 
148 static cpu_set_smm_t smm_set;
149 static void *smm_arg;
150 
151 void cpu_smm_register(cpu_set_smm_t callback, void *arg)
152 {
153     assert(smm_set == NULL);
154     assert(smm_arg == NULL);
155     smm_set = callback;
156     smm_arg = arg;
157 }
158 
159 void cpu_smm_update(CPUX86State *env)
160 {
161     if (smm_set && smm_arg && env == first_cpu)
162         smm_set(!!(env->hflags & HF_SMM_MASK), smm_arg);
163 }
164 
165 
166 /* IRQ handling */
167 int cpu_get_pic_interrupt(CPUX86State *env)
168 {
169     int intno;
170 
171     intno = apic_get_interrupt(env->apic_state);
172     if (intno >= 0) {
173         return intno;
174     }
175     /* read the irq from the PIC */
176     if (!apic_accept_pic_intr(env->apic_state)) {
177         return -1;
178     }
179 
180     intno = pic_read_irq(isa_pic);
181     return intno;
182 }
183 
184 static void pic_irq_request(void *opaque, int irq, int level)
185 {
186     CPUX86State *env = first_cpu;
187 
188     DPRINTF("pic_irqs: %s irq %d\n", level? "raise" : "lower", irq);
189     if (env->apic_state) {
190         while (env) {
191             if (apic_accept_pic_intr(env->apic_state)) {
192                 apic_deliver_pic_intr(env->apic_state, level);
193             }
194             env = env->next_cpu;
195         }
196     } else {
197         CPUState *cs = CPU(x86_env_get_cpu(env));
198         if (level) {
199             cpu_interrupt(cs, CPU_INTERRUPT_HARD);
200         } else {
201             cpu_reset_interrupt(cs, CPU_INTERRUPT_HARD);
202         }
203     }
204 }
205 
206 /* PC cmos mappings */
207 
208 #define REG_EQUIPMENT_BYTE          0x14
209 
210 static int cmos_get_fd_drive_type(FDriveType fd0)
211 {
212     int val;
213 
214     switch (fd0) {
215     case FDRIVE_DRV_144:
216         /* 1.44 Mb 3"5 drive */
217         val = 4;
218         break;
219     case FDRIVE_DRV_288:
220         /* 2.88 Mb 3"5 drive */
221         val = 5;
222         break;
223     case FDRIVE_DRV_120:
224         /* 1.2 Mb 5"5 drive */
225         val = 2;
226         break;
227     case FDRIVE_DRV_NONE:
228     default:
229         val = 0;
230         break;
231     }
232     return val;
233 }
234 
235 static void cmos_init_hd(ISADevice *s, int type_ofs, int info_ofs,
236                          int16_t cylinders, int8_t heads, int8_t sectors)
237 {
238     rtc_set_memory(s, type_ofs, 47);
239     rtc_set_memory(s, info_ofs, cylinders);
240     rtc_set_memory(s, info_ofs + 1, cylinders >> 8);
241     rtc_set_memory(s, info_ofs + 2, heads);
242     rtc_set_memory(s, info_ofs + 3, 0xff);
243     rtc_set_memory(s, info_ofs + 4, 0xff);
244     rtc_set_memory(s, info_ofs + 5, 0xc0 | ((heads > 8) << 3));
245     rtc_set_memory(s, info_ofs + 6, cylinders);
246     rtc_set_memory(s, info_ofs + 7, cylinders >> 8);
247     rtc_set_memory(s, info_ofs + 8, sectors);
248 }
249 
250 /* convert boot_device letter to something recognizable by the bios */
251 static int boot_device2nibble(char boot_device)
252 {
253     switch(boot_device) {
254     case 'a':
255     case 'b':
256         return 0x01; /* floppy boot */
257     case 'c':
258         return 0x02; /* hard drive boot */
259     case 'd':
260         return 0x03; /* CD-ROM boot */
261     case 'n':
262         return 0x04; /* Network boot */
263     }
264     return 0;
265 }
266 
267 static int set_boot_dev(ISADevice *s, const char *boot_device, int fd_bootchk)
268 {
269 #define PC_MAX_BOOT_DEVICES 3
270     int nbds, bds[3] = { 0, };
271     int i;
272 
273     nbds = strlen(boot_device);
274     if (nbds > PC_MAX_BOOT_DEVICES) {
275         error_report("Too many boot devices for PC");
276         return(1);
277     }
278     for (i = 0; i < nbds; i++) {
279         bds[i] = boot_device2nibble(boot_device[i]);
280         if (bds[i] == 0) {
281             error_report("Invalid boot device for PC: '%c'",
282                          boot_device[i]);
283             return(1);
284         }
285     }
286     rtc_set_memory(s, 0x3d, (bds[1] << 4) | bds[0]);
287     rtc_set_memory(s, 0x38, (bds[2] << 4) | (fd_bootchk ? 0x0 : 0x1));
288     return(0);
289 }
290 
291 static int pc_boot_set(void *opaque, const char *boot_device)
292 {
293     return set_boot_dev(opaque, boot_device, 0);
294 }
295 
296 typedef struct pc_cmos_init_late_arg {
297     ISADevice *rtc_state;
298     BusState *idebus[2];
299 } pc_cmos_init_late_arg;
300 
301 static void pc_cmos_init_late(void *opaque)
302 {
303     pc_cmos_init_late_arg *arg = opaque;
304     ISADevice *s = arg->rtc_state;
305     int16_t cylinders;
306     int8_t heads, sectors;
307     int val;
308     int i, trans;
309 
310     val = 0;
311     if (ide_get_geometry(arg->idebus[0], 0,
312                          &cylinders, &heads, &sectors) >= 0) {
313         cmos_init_hd(s, 0x19, 0x1b, cylinders, heads, sectors);
314         val |= 0xf0;
315     }
316     if (ide_get_geometry(arg->idebus[0], 1,
317                          &cylinders, &heads, &sectors) >= 0) {
318         cmos_init_hd(s, 0x1a, 0x24, cylinders, heads, sectors);
319         val |= 0x0f;
320     }
321     rtc_set_memory(s, 0x12, val);
322 
323     val = 0;
324     for (i = 0; i < 4; i++) {
325         /* NOTE: ide_get_geometry() returns the physical
326            geometry.  It is always such that: 1 <= sects <= 63, 1
327            <= heads <= 16, 1 <= cylinders <= 16383. The BIOS
328            geometry can be different if a translation is done. */
329         if (ide_get_geometry(arg->idebus[i / 2], i % 2,
330                              &cylinders, &heads, &sectors) >= 0) {
331             trans = ide_get_bios_chs_trans(arg->idebus[i / 2], i % 2) - 1;
332             assert((trans & ~3) == 0);
333             val |= trans << (i * 2);
334         }
335     }
336     rtc_set_memory(s, 0x39, val);
337 
338     qemu_unregister_reset(pc_cmos_init_late, opaque);
339 }
340 
341 void pc_cmos_init(ram_addr_t ram_size, ram_addr_t above_4g_mem_size,
342                   const char *boot_device,
343                   ISADevice *floppy, BusState *idebus0, BusState *idebus1,
344                   ISADevice *s)
345 {
346     int val, nb, i;
347     FDriveType fd_type[2] = { FDRIVE_DRV_NONE, FDRIVE_DRV_NONE };
348     static pc_cmos_init_late_arg arg;
349 
350     /* various important CMOS locations needed by PC/Bochs bios */
351 
352     /* memory size */
353     /* base memory (first MiB) */
354     val = MIN(ram_size / 1024, 640);
355     rtc_set_memory(s, 0x15, val);
356     rtc_set_memory(s, 0x16, val >> 8);
357     /* extended memory (next 64MiB) */
358     if (ram_size > 1024 * 1024) {
359         val = (ram_size - 1024 * 1024) / 1024;
360     } else {
361         val = 0;
362     }
363     if (val > 65535)
364         val = 65535;
365     rtc_set_memory(s, 0x17, val);
366     rtc_set_memory(s, 0x18, val >> 8);
367     rtc_set_memory(s, 0x30, val);
368     rtc_set_memory(s, 0x31, val >> 8);
369     /* memory between 16MiB and 4GiB */
370     if (ram_size > 16 * 1024 * 1024) {
371         val = (ram_size - 16 * 1024 * 1024) / 65536;
372     } else {
373         val = 0;
374     }
375     if (val > 65535)
376         val = 65535;
377     rtc_set_memory(s, 0x34, val);
378     rtc_set_memory(s, 0x35, val >> 8);
379     /* memory above 4GiB */
380     val = above_4g_mem_size / 65536;
381     rtc_set_memory(s, 0x5b, val);
382     rtc_set_memory(s, 0x5c, val >> 8);
383     rtc_set_memory(s, 0x5d, val >> 16);
384 
385     /* set the number of CPU */
386     rtc_set_memory(s, 0x5f, smp_cpus - 1);
387 
388     /* set boot devices, and disable floppy signature check if requested */
389     if (set_boot_dev(s, boot_device, fd_bootchk)) {
390         exit(1);
391     }
392 
393     /* floppy type */
394     if (floppy) {
395         for (i = 0; i < 2; i++) {
396             fd_type[i] = isa_fdc_get_drive_type(floppy, i);
397         }
398     }
399     val = (cmos_get_fd_drive_type(fd_type[0]) << 4) |
400         cmos_get_fd_drive_type(fd_type[1]);
401     rtc_set_memory(s, 0x10, val);
402 
403     val = 0;
404     nb = 0;
405     if (fd_type[0] < FDRIVE_DRV_NONE) {
406         nb++;
407     }
408     if (fd_type[1] < FDRIVE_DRV_NONE) {
409         nb++;
410     }
411     switch (nb) {
412     case 0:
413         break;
414     case 1:
415         val |= 0x01; /* 1 drive, ready for boot */
416         break;
417     case 2:
418         val |= 0x41; /* 2 drives, ready for boot */
419         break;
420     }
421     val |= 0x02; /* FPU is there */
422     val |= 0x04; /* PS/2 mouse installed */
423     rtc_set_memory(s, REG_EQUIPMENT_BYTE, val);
424 
425     /* hard drives */
426     arg.rtc_state = s;
427     arg.idebus[0] = idebus0;
428     arg.idebus[1] = idebus1;
429     qemu_register_reset(pc_cmos_init_late, &arg);
430 }
431 
432 #define TYPE_PORT92 "port92"
433 #define PORT92(obj) OBJECT_CHECK(Port92State, (obj), TYPE_PORT92)
434 
435 /* port 92 stuff: could be split off */
436 typedef struct Port92State {
437     ISADevice parent_obj;
438 
439     MemoryRegion io;
440     uint8_t outport;
441     qemu_irq *a20_out;
442 } Port92State;
443 
444 static void port92_write(void *opaque, hwaddr addr, uint64_t val,
445                          unsigned size)
446 {
447     Port92State *s = opaque;
448 
449     DPRINTF("port92: write 0x%02x\n", val);
450     s->outport = val;
451     qemu_set_irq(*s->a20_out, (val >> 1) & 1);
452     if (val & 1) {
453         qemu_system_reset_request();
454     }
455 }
456 
457 static uint64_t port92_read(void *opaque, hwaddr addr,
458                             unsigned size)
459 {
460     Port92State *s = opaque;
461     uint32_t ret;
462 
463     ret = s->outport;
464     DPRINTF("port92: read 0x%02x\n", ret);
465     return ret;
466 }
467 
468 static void port92_init(ISADevice *dev, qemu_irq *a20_out)
469 {
470     Port92State *s = PORT92(dev);
471 
472     s->a20_out = a20_out;
473 }
474 
475 static const VMStateDescription vmstate_port92_isa = {
476     .name = "port92",
477     .version_id = 1,
478     .minimum_version_id = 1,
479     .minimum_version_id_old = 1,
480     .fields      = (VMStateField []) {
481         VMSTATE_UINT8(outport, Port92State),
482         VMSTATE_END_OF_LIST()
483     }
484 };
485 
486 static void port92_reset(DeviceState *d)
487 {
488     Port92State *s = PORT92(d);
489 
490     s->outport &= ~1;
491 }
492 
493 static const MemoryRegionOps port92_ops = {
494     .read = port92_read,
495     .write = port92_write,
496     .impl = {
497         .min_access_size = 1,
498         .max_access_size = 1,
499     },
500     .endianness = DEVICE_LITTLE_ENDIAN,
501 };
502 
503 static int port92_initfn(ISADevice *dev)
504 {
505     Port92State *s = PORT92(dev);
506 
507     memory_region_init_io(&s->io, &port92_ops, s, "port92", 1);
508     isa_register_ioport(dev, &s->io, 0x92);
509 
510     s->outport = 0;
511     return 0;
512 }
513 
514 static void port92_class_initfn(ObjectClass *klass, void *data)
515 {
516     DeviceClass *dc = DEVICE_CLASS(klass);
517     ISADeviceClass *ic = ISA_DEVICE_CLASS(klass);
518     ic->init = port92_initfn;
519     dc->no_user = 1;
520     dc->reset = port92_reset;
521     dc->vmsd = &vmstate_port92_isa;
522 }
523 
524 static const TypeInfo port92_info = {
525     .name          = TYPE_PORT92,
526     .parent        = TYPE_ISA_DEVICE,
527     .instance_size = sizeof(Port92State),
528     .class_init    = port92_class_initfn,
529 };
530 
531 static void port92_register_types(void)
532 {
533     type_register_static(&port92_info);
534 }
535 
536 type_init(port92_register_types)
537 
538 static void handle_a20_line_change(void *opaque, int irq, int level)
539 {
540     X86CPU *cpu = opaque;
541 
542     /* XXX: send to all CPUs ? */
543     /* XXX: add logic to handle multiple A20 line sources */
544     x86_cpu_set_a20(cpu, level);
545 }
546 
547 int e820_add_entry(uint64_t address, uint64_t length, uint32_t type)
548 {
549     int index = le32_to_cpu(e820_table.count);
550     struct e820_entry *entry;
551 
552     if (index >= E820_NR_ENTRIES)
553         return -EBUSY;
554     entry = &e820_table.entry[index++];
555 
556     entry->address = cpu_to_le64(address);
557     entry->length = cpu_to_le64(length);
558     entry->type = cpu_to_le32(type);
559 
560     e820_table.count = cpu_to_le32(index);
561     return index;
562 }
563 
564 /* Calculates the limit to CPU APIC ID values
565  *
566  * This function returns the limit for the APIC ID value, so that all
567  * CPU APIC IDs are < pc_apic_id_limit().
568  *
569  * This is used for FW_CFG_MAX_CPUS. See comments on bochs_bios_init().
570  */
571 static unsigned int pc_apic_id_limit(unsigned int max_cpus)
572 {
573     return x86_cpu_apic_id_from_index(max_cpus - 1) + 1;
574 }
575 
576 static void *bochs_bios_init(void)
577 {
578     void *fw_cfg;
579     uint8_t *smbios_table;
580     size_t smbios_len;
581     uint64_t *numa_fw_cfg;
582     int i, j;
583     unsigned int apic_id_limit = pc_apic_id_limit(max_cpus);
584 
585     fw_cfg = fw_cfg_init(BIOS_CFG_IOPORT, BIOS_CFG_IOPORT + 1, 0, 0);
586     /* FW_CFG_MAX_CPUS is a bit confusing/problematic on x86:
587      *
588      * SeaBIOS needs FW_CFG_MAX_CPUS for CPU hotplug, but the CPU hotplug
589      * QEMU<->SeaBIOS interface is not based on the "CPU index", but on the APIC
590      * ID of hotplugged CPUs[1]. This means that FW_CFG_MAX_CPUS is not the
591      * "maximum number of CPUs", but the "limit to the APIC ID values SeaBIOS
592      * may see".
593      *
594      * So, this means we must not use max_cpus, here, but the maximum possible
595      * APIC ID value, plus one.
596      *
597      * [1] The only kind of "CPU identifier" used between SeaBIOS and QEMU is
598      *     the APIC ID, not the "CPU index"
599      */
600     fw_cfg_add_i16(fw_cfg, FW_CFG_MAX_CPUS, (uint16_t)apic_id_limit);
601     fw_cfg_add_i32(fw_cfg, FW_CFG_ID, 1);
602     fw_cfg_add_i64(fw_cfg, FW_CFG_RAM_SIZE, (uint64_t)ram_size);
603     fw_cfg_add_bytes(fw_cfg, FW_CFG_ACPI_TABLES,
604                      acpi_tables, acpi_tables_len);
605     fw_cfg_add_i32(fw_cfg, FW_CFG_IRQ0_OVERRIDE, kvm_allows_irq0_override());
606 
607     smbios_table = smbios_get_table(&smbios_len);
608     if (smbios_table)
609         fw_cfg_add_bytes(fw_cfg, FW_CFG_SMBIOS_ENTRIES,
610                          smbios_table, smbios_len);
611     fw_cfg_add_bytes(fw_cfg, FW_CFG_E820_TABLE,
612                      &e820_table, sizeof(e820_table));
613 
614     fw_cfg_add_bytes(fw_cfg, FW_CFG_HPET, &hpet_cfg, sizeof(hpet_cfg));
615     /* allocate memory for the NUMA channel: one (64bit) word for the number
616      * of nodes, one word for each VCPU->node and one word for each node to
617      * hold the amount of memory.
618      */
619     numa_fw_cfg = g_new0(uint64_t, 1 + apic_id_limit + nb_numa_nodes);
620     numa_fw_cfg[0] = cpu_to_le64(nb_numa_nodes);
621     for (i = 0; i < max_cpus; i++) {
622         unsigned int apic_id = x86_cpu_apic_id_from_index(i);
623         assert(apic_id < apic_id_limit);
624         for (j = 0; j < nb_numa_nodes; j++) {
625             if (test_bit(i, node_cpumask[j])) {
626                 numa_fw_cfg[apic_id + 1] = cpu_to_le64(j);
627                 break;
628             }
629         }
630     }
631     for (i = 0; i < nb_numa_nodes; i++) {
632         numa_fw_cfg[apic_id_limit + 1 + i] = cpu_to_le64(node_mem[i]);
633     }
634     fw_cfg_add_bytes(fw_cfg, FW_CFG_NUMA, numa_fw_cfg,
635                      (1 + apic_id_limit + nb_numa_nodes) *
636                      sizeof(*numa_fw_cfg));
637 
638     return fw_cfg;
639 }
640 
641 static long get_file_size(FILE *f)
642 {
643     long where, size;
644 
645     /* XXX: on Unix systems, using fstat() probably makes more sense */
646 
647     where = ftell(f);
648     fseek(f, 0, SEEK_END);
649     size = ftell(f);
650     fseek(f, where, SEEK_SET);
651 
652     return size;
653 }
654 
655 static void load_linux(void *fw_cfg,
656                        const char *kernel_filename,
657                        const char *initrd_filename,
658                        const char *kernel_cmdline,
659                        hwaddr max_ram_size)
660 {
661     uint16_t protocol;
662     int setup_size, kernel_size, initrd_size = 0, cmdline_size;
663     uint32_t initrd_max;
664     uint8_t header[8192], *setup, *kernel, *initrd_data;
665     hwaddr real_addr, prot_addr, cmdline_addr, initrd_addr = 0;
666     FILE *f;
667     char *vmode;
668 
669     /* Align to 16 bytes as a paranoia measure */
670     cmdline_size = (strlen(kernel_cmdline)+16) & ~15;
671 
672     /* load the kernel header */
673     f = fopen(kernel_filename, "rb");
674     if (!f || !(kernel_size = get_file_size(f)) ||
675         fread(header, 1, MIN(ARRAY_SIZE(header), kernel_size), f) !=
676         MIN(ARRAY_SIZE(header), kernel_size)) {
677         fprintf(stderr, "qemu: could not load kernel '%s': %s\n",
678                 kernel_filename, strerror(errno));
679         exit(1);
680     }
681 
682     /* kernel protocol version */
683 #if 0
684     fprintf(stderr, "header magic: %#x\n", ldl_p(header+0x202));
685 #endif
686     if (ldl_p(header+0x202) == 0x53726448) {
687         protocol = lduw_p(header+0x206);
688     } else {
689         /* This looks like a multiboot kernel. If it is, let's stop
690            treating it like a Linux kernel. */
691         if (load_multiboot(fw_cfg, f, kernel_filename, initrd_filename,
692                            kernel_cmdline, kernel_size, header)) {
693             return;
694         }
695         protocol = 0;
696     }
697 
698     if (protocol < 0x200 || !(header[0x211] & 0x01)) {
699         /* Low kernel */
700         real_addr    = 0x90000;
701         cmdline_addr = 0x9a000 - cmdline_size;
702         prot_addr    = 0x10000;
703     } else if (protocol < 0x202) {
704         /* High but ancient kernel */
705         real_addr    = 0x90000;
706         cmdline_addr = 0x9a000 - cmdline_size;
707         prot_addr    = 0x100000;
708     } else {
709         /* High and recent kernel */
710         real_addr    = 0x10000;
711         cmdline_addr = 0x20000;
712         prot_addr    = 0x100000;
713     }
714 
715 #if 0
716     fprintf(stderr,
717             "qemu: real_addr     = 0x" TARGET_FMT_plx "\n"
718             "qemu: cmdline_addr  = 0x" TARGET_FMT_plx "\n"
719             "qemu: prot_addr     = 0x" TARGET_FMT_plx "\n",
720             real_addr,
721             cmdline_addr,
722             prot_addr);
723 #endif
724 
725     /* highest address for loading the initrd */
726     if (protocol >= 0x203) {
727         initrd_max = ldl_p(header+0x22c);
728     } else {
729         initrd_max = 0x37ffffff;
730     }
731 
732     if (initrd_max >= max_ram_size-ACPI_DATA_SIZE)
733     	initrd_max = max_ram_size-ACPI_DATA_SIZE-1;
734 
735     fw_cfg_add_i32(fw_cfg, FW_CFG_CMDLINE_ADDR, cmdline_addr);
736     fw_cfg_add_i32(fw_cfg, FW_CFG_CMDLINE_SIZE, strlen(kernel_cmdline)+1);
737     fw_cfg_add_string(fw_cfg, FW_CFG_CMDLINE_DATA, kernel_cmdline);
738 
739     if (protocol >= 0x202) {
740         stl_p(header+0x228, cmdline_addr);
741     } else {
742         stw_p(header+0x20, 0xA33F);
743         stw_p(header+0x22, cmdline_addr-real_addr);
744     }
745 
746     /* handle vga= parameter */
747     vmode = strstr(kernel_cmdline, "vga=");
748     if (vmode) {
749         unsigned int video_mode;
750         /* skip "vga=" */
751         vmode += 4;
752         if (!strncmp(vmode, "normal", 6)) {
753             video_mode = 0xffff;
754         } else if (!strncmp(vmode, "ext", 3)) {
755             video_mode = 0xfffe;
756         } else if (!strncmp(vmode, "ask", 3)) {
757             video_mode = 0xfffd;
758         } else {
759             video_mode = strtol(vmode, NULL, 0);
760         }
761         stw_p(header+0x1fa, video_mode);
762     }
763 
764     /* loader type */
765     /* High nybble = B reserved for QEMU; low nybble is revision number.
766        If this code is substantially changed, you may want to consider
767        incrementing the revision. */
768     if (protocol >= 0x200) {
769         header[0x210] = 0xB0;
770     }
771     /* heap */
772     if (protocol >= 0x201) {
773         header[0x211] |= 0x80;	/* CAN_USE_HEAP */
774         stw_p(header+0x224, cmdline_addr-real_addr-0x200);
775     }
776 
777     /* load initrd */
778     if (initrd_filename) {
779         if (protocol < 0x200) {
780             fprintf(stderr, "qemu: linux kernel too old to load a ram disk\n");
781             exit(1);
782         }
783 
784         initrd_size = get_image_size(initrd_filename);
785         if (initrd_size < 0) {
786             fprintf(stderr, "qemu: error reading initrd %s\n",
787                     initrd_filename);
788             exit(1);
789         }
790 
791         initrd_addr = (initrd_max-initrd_size) & ~4095;
792 
793         initrd_data = g_malloc(initrd_size);
794         load_image(initrd_filename, initrd_data);
795 
796         fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_ADDR, initrd_addr);
797         fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_SIZE, initrd_size);
798         fw_cfg_add_bytes(fw_cfg, FW_CFG_INITRD_DATA, initrd_data, initrd_size);
799 
800         stl_p(header+0x218, initrd_addr);
801         stl_p(header+0x21c, initrd_size);
802     }
803 
804     /* load kernel and setup */
805     setup_size = header[0x1f1];
806     if (setup_size == 0) {
807         setup_size = 4;
808     }
809     setup_size = (setup_size+1)*512;
810     kernel_size -= setup_size;
811 
812     setup  = g_malloc(setup_size);
813     kernel = g_malloc(kernel_size);
814     fseek(f, 0, SEEK_SET);
815     if (fread(setup, 1, setup_size, f) != setup_size) {
816         fprintf(stderr, "fread() failed\n");
817         exit(1);
818     }
819     if (fread(kernel, 1, kernel_size, f) != kernel_size) {
820         fprintf(stderr, "fread() failed\n");
821         exit(1);
822     }
823     fclose(f);
824     memcpy(setup, header, MIN(sizeof(header), setup_size));
825 
826     fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_ADDR, prot_addr);
827     fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_SIZE, kernel_size);
828     fw_cfg_add_bytes(fw_cfg, FW_CFG_KERNEL_DATA, kernel, kernel_size);
829 
830     fw_cfg_add_i32(fw_cfg, FW_CFG_SETUP_ADDR, real_addr);
831     fw_cfg_add_i32(fw_cfg, FW_CFG_SETUP_SIZE, setup_size);
832     fw_cfg_add_bytes(fw_cfg, FW_CFG_SETUP_DATA, setup, setup_size);
833 
834     option_rom[nb_option_roms].name = "linuxboot.bin";
835     option_rom[nb_option_roms].bootindex = 0;
836     nb_option_roms++;
837 }
838 
839 #define NE2000_NB_MAX 6
840 
841 static const int ne2000_io[NE2000_NB_MAX] = { 0x300, 0x320, 0x340, 0x360,
842                                               0x280, 0x380 };
843 static const int ne2000_irq[NE2000_NB_MAX] = { 9, 10, 11, 3, 4, 5 };
844 
845 static const int parallel_io[MAX_PARALLEL_PORTS] = { 0x378, 0x278, 0x3bc };
846 static const int parallel_irq[MAX_PARALLEL_PORTS] = { 7, 7, 7 };
847 
848 void pc_init_ne2k_isa(ISABus *bus, NICInfo *nd)
849 {
850     static int nb_ne2k = 0;
851 
852     if (nb_ne2k == NE2000_NB_MAX)
853         return;
854     isa_ne2000_init(bus, ne2000_io[nb_ne2k],
855                     ne2000_irq[nb_ne2k], nd);
856     nb_ne2k++;
857 }
858 
859 DeviceState *cpu_get_current_apic(void)
860 {
861     if (cpu_single_env) {
862         return cpu_single_env->apic_state;
863     } else {
864         return NULL;
865     }
866 }
867 
868 void pc_acpi_smi_interrupt(void *opaque, int irq, int level)
869 {
870     X86CPU *cpu = opaque;
871 
872     if (level) {
873         cpu_interrupt(CPU(cpu), CPU_INTERRUPT_SMI);
874     }
875 }
876 
877 void pc_cpus_init(const char *cpu_model)
878 {
879     int i;
880 
881     /* init CPUs */
882     if (cpu_model == NULL) {
883 #ifdef TARGET_X86_64
884         cpu_model = "qemu64";
885 #else
886         cpu_model = "qemu32";
887 #endif
888     }
889 
890     for (i = 0; i < smp_cpus; i++) {
891         if (!cpu_x86_init(cpu_model)) {
892             exit(1);
893         }
894     }
895 }
896 
897 void pc_acpi_init(const char *default_dsdt)
898 {
899     char *filename;
900 
901     if (acpi_tables != NULL) {
902         /* manually set via -acpitable, leave it alone */
903         return;
904     }
905 
906     filename = qemu_find_file(QEMU_FILE_TYPE_BIOS, default_dsdt);
907     if (filename == NULL) {
908         fprintf(stderr, "WARNING: failed to find %s\n", default_dsdt);
909     } else {
910         char *arg;
911         QemuOpts *opts;
912         Error *err = NULL;
913 
914         arg = g_strdup_printf("file=%s", filename);
915 
916         /* creates a deep copy of "arg" */
917         opts = qemu_opts_parse(qemu_find_opts("acpi"), arg, 0);
918         g_assert(opts != NULL);
919 
920         acpi_table_add(opts, &err);
921         if (err) {
922             fprintf(stderr, "WARNING: failed to load %s: %s\n", filename,
923                     error_get_pretty(err));
924             error_free(err);
925         }
926         g_free(arg);
927         g_free(filename);
928     }
929 }
930 
931 void *pc_memory_init(MemoryRegion *system_memory,
932                     const char *kernel_filename,
933                     const char *kernel_cmdline,
934                     const char *initrd_filename,
935                     ram_addr_t below_4g_mem_size,
936                     ram_addr_t above_4g_mem_size,
937                     MemoryRegion *rom_memory,
938                     MemoryRegion **ram_memory)
939 {
940     int linux_boot, i;
941     MemoryRegion *ram, *option_rom_mr;
942     MemoryRegion *ram_below_4g, *ram_above_4g;
943     void *fw_cfg;
944 
945     linux_boot = (kernel_filename != NULL);
946 
947     /* Allocate RAM.  We allocate it as a single memory region and use
948      * aliases to address portions of it, mostly for backwards compatibility
949      * with older qemus that used qemu_ram_alloc().
950      */
951     ram = g_malloc(sizeof(*ram));
952     memory_region_init_ram(ram, "pc.ram",
953                            below_4g_mem_size + above_4g_mem_size);
954     vmstate_register_ram_global(ram);
955     *ram_memory = ram;
956     ram_below_4g = g_malloc(sizeof(*ram_below_4g));
957     memory_region_init_alias(ram_below_4g, "ram-below-4g", ram,
958                              0, below_4g_mem_size);
959     memory_region_add_subregion(system_memory, 0, ram_below_4g);
960     if (above_4g_mem_size > 0) {
961         ram_above_4g = g_malloc(sizeof(*ram_above_4g));
962         memory_region_init_alias(ram_above_4g, "ram-above-4g", ram,
963                                  below_4g_mem_size, above_4g_mem_size);
964         memory_region_add_subregion(system_memory, 0x100000000ULL,
965                                     ram_above_4g);
966     }
967 
968 
969     /* Initialize PC system firmware */
970     pc_system_firmware_init(rom_memory);
971 
972     option_rom_mr = g_malloc(sizeof(*option_rom_mr));
973     memory_region_init_ram(option_rom_mr, "pc.rom", PC_ROM_SIZE);
974     vmstate_register_ram_global(option_rom_mr);
975     memory_region_add_subregion_overlap(rom_memory,
976                                         PC_ROM_MIN_VGA,
977                                         option_rom_mr,
978                                         1);
979 
980     fw_cfg = bochs_bios_init();
981     rom_set_fw(fw_cfg);
982 
983     if (linux_boot) {
984         load_linux(fw_cfg, kernel_filename, initrd_filename, kernel_cmdline, below_4g_mem_size);
985     }
986 
987     for (i = 0; i < nb_option_roms; i++) {
988         rom_add_option(option_rom[i].name, option_rom[i].bootindex);
989     }
990     return fw_cfg;
991 }
992 
993 qemu_irq *pc_allocate_cpu_irq(void)
994 {
995     return qemu_allocate_irqs(pic_irq_request, NULL, 1);
996 }
997 
998 DeviceState *pc_vga_init(ISABus *isa_bus, PCIBus *pci_bus)
999 {
1000     DeviceState *dev = NULL;
1001 
1002     if (pci_bus) {
1003         PCIDevice *pcidev = pci_vga_init(pci_bus);
1004         dev = pcidev ? &pcidev->qdev : NULL;
1005     } else if (isa_bus) {
1006         ISADevice *isadev = isa_vga_init(isa_bus);
1007         dev = isadev ? &isadev->qdev : NULL;
1008     }
1009     return dev;
1010 }
1011 
1012 static void cpu_request_exit(void *opaque, int irq, int level)
1013 {
1014     CPUX86State *env = cpu_single_env;
1015 
1016     if (env && level) {
1017         cpu_exit(env);
1018     }
1019 }
1020 
1021 static const MemoryRegionOps ioport80_io_ops = {
1022     .write = ioport80_write,
1023     .read = ioport80_read,
1024     .endianness = DEVICE_NATIVE_ENDIAN,
1025     .impl = {
1026         .min_access_size = 1,
1027         .max_access_size = 1,
1028     },
1029 };
1030 
1031 static const MemoryRegionOps ioportF0_io_ops = {
1032     .write = ioportF0_write,
1033     .read = ioportF0_read,
1034     .endianness = DEVICE_NATIVE_ENDIAN,
1035     .impl = {
1036         .min_access_size = 1,
1037         .max_access_size = 1,
1038     },
1039 };
1040 
1041 void pc_basic_device_init(ISABus *isa_bus, qemu_irq *gsi,
1042                           ISADevice **rtc_state,
1043                           ISADevice **floppy,
1044                           bool no_vmport)
1045 {
1046     int i;
1047     DriveInfo *fd[MAX_FD];
1048     DeviceState *hpet = NULL;
1049     int pit_isa_irq = 0;
1050     qemu_irq pit_alt_irq = NULL;
1051     qemu_irq rtc_irq = NULL;
1052     qemu_irq *a20_line;
1053     ISADevice *i8042, *port92, *vmmouse, *pit = NULL;
1054     qemu_irq *cpu_exit_irq;
1055     MemoryRegion *ioport80_io = g_new(MemoryRegion, 1);
1056     MemoryRegion *ioportF0_io = g_new(MemoryRegion, 1);
1057 
1058     memory_region_init_io(ioport80_io, &ioport80_io_ops, NULL, "ioport80", 1);
1059     memory_region_add_subregion(isa_bus->address_space_io, 0x80, ioport80_io);
1060 
1061     memory_region_init_io(ioportF0_io, &ioportF0_io_ops, NULL, "ioportF0", 1);
1062     memory_region_add_subregion(isa_bus->address_space_io, 0xf0, ioportF0_io);
1063 
1064     /*
1065      * Check if an HPET shall be created.
1066      *
1067      * Without KVM_CAP_PIT_STATE2, we cannot switch off the in-kernel PIT
1068      * when the HPET wants to take over. Thus we have to disable the latter.
1069      */
1070     if (!no_hpet && (!kvm_irqchip_in_kernel() || kvm_has_pit_state2())) {
1071         hpet = sysbus_try_create_simple("hpet", HPET_BASE, NULL);
1072 
1073         if (hpet) {
1074             for (i = 0; i < GSI_NUM_PINS; i++) {
1075                 sysbus_connect_irq(SYS_BUS_DEVICE(hpet), i, gsi[i]);
1076             }
1077             pit_isa_irq = -1;
1078             pit_alt_irq = qdev_get_gpio_in(hpet, HPET_LEGACY_PIT_INT);
1079             rtc_irq = qdev_get_gpio_in(hpet, HPET_LEGACY_RTC_INT);
1080         }
1081     }
1082     *rtc_state = rtc_init(isa_bus, 2000, rtc_irq);
1083 
1084     qemu_register_boot_set(pc_boot_set, *rtc_state);
1085 
1086     if (!xen_enabled()) {
1087         if (kvm_irqchip_in_kernel()) {
1088             pit = kvm_pit_init(isa_bus, 0x40);
1089         } else {
1090             pit = pit_init(isa_bus, 0x40, pit_isa_irq, pit_alt_irq);
1091         }
1092         if (hpet) {
1093             /* connect PIT to output control line of the HPET */
1094             qdev_connect_gpio_out(hpet, 0, qdev_get_gpio_in(&pit->qdev, 0));
1095         }
1096         pcspk_init(isa_bus, pit);
1097     }
1098 
1099     for(i = 0; i < MAX_SERIAL_PORTS; i++) {
1100         if (serial_hds[i]) {
1101             serial_isa_init(isa_bus, i, serial_hds[i]);
1102         }
1103     }
1104 
1105     for(i = 0; i < MAX_PARALLEL_PORTS; i++) {
1106         if (parallel_hds[i]) {
1107             parallel_init(isa_bus, i, parallel_hds[i]);
1108         }
1109     }
1110 
1111     a20_line = qemu_allocate_irqs(handle_a20_line_change,
1112                                   x86_env_get_cpu(first_cpu), 2);
1113     i8042 = isa_create_simple(isa_bus, "i8042");
1114     i8042_setup_a20_line(i8042, &a20_line[0]);
1115     if (!no_vmport) {
1116         vmport_init(isa_bus);
1117         vmmouse = isa_try_create(isa_bus, "vmmouse");
1118     } else {
1119         vmmouse = NULL;
1120     }
1121     if (vmmouse) {
1122         qdev_prop_set_ptr(&vmmouse->qdev, "ps2_mouse", i8042);
1123         qdev_init_nofail(&vmmouse->qdev);
1124     }
1125     port92 = isa_create_simple(isa_bus, "port92");
1126     port92_init(port92, &a20_line[1]);
1127 
1128     cpu_exit_irq = qemu_allocate_irqs(cpu_request_exit, NULL, 1);
1129     DMA_init(0, cpu_exit_irq);
1130 
1131     for(i = 0; i < MAX_FD; i++) {
1132         fd[i] = drive_get(IF_FLOPPY, 0, i);
1133     }
1134     *floppy = fdctrl_init_isa(isa_bus, fd);
1135 }
1136 
1137 void pc_nic_init(ISABus *isa_bus, PCIBus *pci_bus)
1138 {
1139     int i;
1140 
1141     for (i = 0; i < nb_nics; i++) {
1142         NICInfo *nd = &nd_table[i];
1143 
1144         if (!pci_bus || (nd->model && strcmp(nd->model, "ne2k_isa") == 0)) {
1145             pc_init_ne2k_isa(isa_bus, nd);
1146         } else {
1147             pci_nic_init_nofail(nd, "e1000", NULL);
1148         }
1149     }
1150 }
1151 
1152 void pc_pci_device_init(PCIBus *pci_bus)
1153 {
1154     int max_bus;
1155     int bus;
1156 
1157     max_bus = drive_get_max_bus(IF_SCSI);
1158     for (bus = 0; bus <= max_bus; bus++) {
1159         pci_create_simple(pci_bus, -1, "lsi53c895a");
1160     }
1161 }
1162 
1163 void ioapic_init_gsi(GSIState *gsi_state, const char *parent_name)
1164 {
1165     DeviceState *dev;
1166     SysBusDevice *d;
1167     unsigned int i;
1168 
1169     if (kvm_irqchip_in_kernel()) {
1170         dev = qdev_create(NULL, "kvm-ioapic");
1171     } else {
1172         dev = qdev_create(NULL, "ioapic");
1173     }
1174     if (parent_name) {
1175         object_property_add_child(object_resolve_path(parent_name, NULL),
1176                                   "ioapic", OBJECT(dev), NULL);
1177     }
1178     qdev_init_nofail(dev);
1179     d = SYS_BUS_DEVICE(dev);
1180     sysbus_mmio_map(d, 0, IO_APIC_DEFAULT_ADDRESS);
1181 
1182     for (i = 0; i < IOAPIC_NUM_PINS; i++) {
1183         gsi_state->ioapic_irq[i] = qdev_get_gpio_in(dev, i);
1184     }
1185 }
1186