qemu/hw/pc.c
<<
>>
Prefs
   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.h"
  25#include "pc.h"
  26#include "apic.h"
  27#include "fdc.h"
  28#include "ide.h"
  29#include "pci.h"
  30#include "vmware_vga.h"
  31#include "monitor.h"
  32#include "fw_cfg.h"
  33#include "hpet_emul.h"
  34#include "smbios.h"
  35#include "loader.h"
  36#include "elf.h"
  37#include "multiboot.h"
  38#include "mc146818rtc.h"
  39#include "i8254.h"
  40#include "pcspk.h"
  41#include "msi.h"
  42#include "sysbus.h"
  43#include "sysemu.h"
  44#include "kvm.h"
  45#include "xen.h"
  46#include "blockdev.h"
  47#include "ui/qemu-spice.h"
  48#include "memory.h"
  49#include "exec-memory.h"
  50#include "arch_init.h"
  51
  52/* output Bochs bios info messages */
  53//#define DEBUG_BIOS
  54
  55/* debug PC/ISA interrupts */
  56//#define DEBUG_IRQ
  57
  58#ifdef DEBUG_IRQ
  59#define DPRINTF(fmt, ...)                                       \
  60    do { printf("CPUIRQ: " fmt , ## __VA_ARGS__); } while (0)
  61#else
  62#define DPRINTF(fmt, ...)
  63#endif
  64
  65/* Leave a chunk of memory at the top of RAM for the BIOS ACPI tables.  */
  66#define ACPI_DATA_SIZE       0x10000
  67#define BIOS_CFG_IOPORT 0x510
  68#define FW_CFG_ACPI_TABLES (FW_CFG_ARCH_LOCAL + 0)
  69#define FW_CFG_SMBIOS_ENTRIES (FW_CFG_ARCH_LOCAL + 1)
  70#define FW_CFG_IRQ0_OVERRIDE (FW_CFG_ARCH_LOCAL + 2)
  71#define FW_CFG_E820_TABLE (FW_CFG_ARCH_LOCAL + 3)
  72#define FW_CFG_HPET (FW_CFG_ARCH_LOCAL + 4)
  73
  74#define MSI_ADDR_BASE 0xfee00000
  75
  76#define E820_NR_ENTRIES         16
  77
  78struct e820_entry {
  79    uint64_t address;
  80    uint64_t length;
  81    uint32_t type;
  82} QEMU_PACKED __attribute((__aligned__(4)));
  83
  84struct e820_table {
  85    uint32_t count;
  86    struct e820_entry entry[E820_NR_ENTRIES];
  87} QEMU_PACKED __attribute((__aligned__(4)));
  88
  89static struct e820_table e820_table;
  90struct hpet_fw_config hpet_cfg = {.count = UINT8_MAX};
  91
  92void gsi_handler(void *opaque, int n, int level)
  93{
  94    GSIState *s = opaque;
  95
  96    DPRINTF("pc: %s GSI %d\n", level ? "raising" : "lowering", n);
  97    if (n < ISA_NUM_IRQS) {
  98        qemu_set_irq(s->i8259_irq[n], level);
  99    }
 100    qemu_set_irq(s->ioapic_irq[n], level);
 101}
 102
 103static void ioport80_write(void *opaque, uint32_t addr, uint32_t data)
 104{
 105}
 106
 107/* MSDOS compatibility mode FPU exception support */
 108static qemu_irq ferr_irq;
 109
 110void pc_register_ferr_irq(qemu_irq irq)
 111{
 112    ferr_irq = irq;
 113}
 114
 115/* XXX: add IGNNE support */
 116void cpu_set_ferr(CPUX86State *s)
 117{
 118    qemu_irq_raise(ferr_irq);
 119}
 120
 121static void ioportF0_write(void *opaque, uint32_t addr, uint32_t data)
 122{
 123    qemu_irq_lower(ferr_irq);
 124}
 125
 126/* TSC handling */
 127uint64_t cpu_get_tsc(CPUX86State *env)
 128{
 129    return cpu_get_ticks();
 130}
 131
 132/* SMM support */
 133
 134static cpu_set_smm_t smm_set;
 135static void *smm_arg;
 136
 137void cpu_smm_register(cpu_set_smm_t callback, void *arg)
 138{
 139    assert(smm_set == NULL);
 140    assert(smm_arg == NULL);
 141    smm_set = callback;
 142    smm_arg = arg;
 143}
 144
 145void cpu_smm_update(CPUX86State *env)
 146{
 147    if (smm_set && smm_arg && env == first_cpu)
 148        smm_set(!!(env->hflags & HF_SMM_MASK), smm_arg);
 149}
 150
 151
 152/* IRQ handling */
 153int cpu_get_pic_interrupt(CPUX86State *env)
 154{
 155    int intno;
 156
 157    intno = apic_get_interrupt(env->apic_state);
 158    if (intno >= 0) {
 159        return intno;
 160    }
 161    /* read the irq from the PIC */
 162    if (!apic_accept_pic_intr(env->apic_state)) {
 163        return -1;
 164    }
 165
 166    intno = pic_read_irq(isa_pic);
 167    return intno;
 168}
 169
 170static void pic_irq_request(void *opaque, int irq, int level)
 171{
 172    CPUX86State *env = first_cpu;
 173
 174    DPRINTF("pic_irqs: %s irq %d\n", level? "raise" : "lower", irq);
 175    if (env->apic_state) {
 176        while (env) {
 177            if (apic_accept_pic_intr(env->apic_state)) {
 178                apic_deliver_pic_intr(env->apic_state, level);
 179            }
 180            env = env->next_cpu;
 181        }
 182    } else {
 183        if (level)
 184            cpu_interrupt(env, CPU_INTERRUPT_HARD);
 185        else
 186            cpu_reset_interrupt(env, CPU_INTERRUPT_HARD);
 187    }
 188}
 189
 190/* PC cmos mappings */
 191
 192#define REG_EQUIPMENT_BYTE          0x14
 193
 194static int cmos_get_fd_drive_type(FDriveType fd0)
 195{
 196    int val;
 197
 198    switch (fd0) {
 199    case FDRIVE_DRV_144:
 200        /* 1.44 Mb 3"5 drive */
 201        val = 4;
 202        break;
 203    case FDRIVE_DRV_288:
 204        /* 2.88 Mb 3"5 drive */
 205        val = 5;
 206        break;
 207    case FDRIVE_DRV_120:
 208        /* 1.2 Mb 5"5 drive */
 209        val = 2;
 210        break;
 211    case FDRIVE_DRV_NONE:
 212    default:
 213        val = 0;
 214        break;
 215    }
 216    return val;
 217}
 218
 219static void cmos_init_hd(int type_ofs, int info_ofs, BlockDriverState *hd,
 220                         ISADevice *s)
 221{
 222    int cylinders, heads, sectors;
 223    bdrv_get_geometry_hint(hd, &cylinders, &heads, &sectors);
 224    rtc_set_memory(s, type_ofs, 47);
 225    rtc_set_memory(s, info_ofs, cylinders);
 226    rtc_set_memory(s, info_ofs + 1, cylinders >> 8);
 227    rtc_set_memory(s, info_ofs + 2, heads);
 228    rtc_set_memory(s, info_ofs + 3, 0xff);
 229    rtc_set_memory(s, info_ofs + 4, 0xff);
 230    rtc_set_memory(s, info_ofs + 5, 0xc0 | ((heads > 8) << 3));
 231    rtc_set_memory(s, info_ofs + 6, cylinders);
 232    rtc_set_memory(s, info_ofs + 7, cylinders >> 8);
 233    rtc_set_memory(s, info_ofs + 8, sectors);
 234}
 235
 236/* convert boot_device letter to something recognizable by the bios */
 237static int boot_device2nibble(char boot_device)
 238{
 239    switch(boot_device) {
 240    case 'a':
 241    case 'b':
 242        return 0x01; /* floppy boot */
 243    case 'c':
 244        return 0x02; /* hard drive boot */
 245    case 'd':
 246        return 0x03; /* CD-ROM boot */
 247    case 'n':
 248        return 0x04; /* Network boot */
 249    }
 250    return 0;
 251}
 252
 253static int set_boot_dev(ISADevice *s, const char *boot_device, int fd_bootchk)
 254{
 255#define PC_MAX_BOOT_DEVICES 3
 256    int nbds, bds[3] = { 0, };
 257    int i;
 258
 259    nbds = strlen(boot_device);
 260    if (nbds > PC_MAX_BOOT_DEVICES) {
 261        error_report("Too many boot devices for PC");
 262        return(1);
 263    }
 264    for (i = 0; i < nbds; i++) {
 265        bds[i] = boot_device2nibble(boot_device[i]);
 266        if (bds[i] == 0) {
 267            error_report("Invalid boot device for PC: '%c'",
 268                         boot_device[i]);
 269            return(1);
 270        }
 271    }
 272    rtc_set_memory(s, 0x3d, (bds[1] << 4) | bds[0]);
 273    rtc_set_memory(s, 0x38, (bds[2] << 4) | (fd_bootchk ? 0x0 : 0x1));
 274    return(0);
 275}
 276
 277static int pc_boot_set(void *opaque, const char *boot_device)
 278{
 279    return set_boot_dev(opaque, boot_device, 0);
 280}
 281
 282typedef struct pc_cmos_init_late_arg {
 283    ISADevice *rtc_state;
 284    BusState *idebus0, *idebus1;
 285} pc_cmos_init_late_arg;
 286
 287static void pc_cmos_init_late(void *opaque)
 288{
 289    pc_cmos_init_late_arg *arg = opaque;
 290    ISADevice *s = arg->rtc_state;
 291    int val;
 292    BlockDriverState *hd_table[4];
 293    int i;
 294
 295    ide_get_bs(hd_table, arg->idebus0);
 296    ide_get_bs(hd_table + 2, arg->idebus1);
 297
 298    rtc_set_memory(s, 0x12, (hd_table[0] ? 0xf0 : 0) | (hd_table[1] ? 0x0f : 0));
 299    if (hd_table[0])
 300        cmos_init_hd(0x19, 0x1b, hd_table[0], s);
 301    if (hd_table[1])
 302        cmos_init_hd(0x1a, 0x24, hd_table[1], s);
 303
 304    val = 0;
 305    for (i = 0; i < 4; i++) {
 306        if (hd_table[i]) {
 307            int cylinders, heads, sectors, translation;
 308            /* NOTE: bdrv_get_geometry_hint() returns the physical
 309                geometry.  It is always such that: 1 <= sects <= 63, 1
 310                <= heads <= 16, 1 <= cylinders <= 16383. The BIOS
 311                geometry can be different if a translation is done. */
 312            translation = bdrv_get_translation_hint(hd_table[i]);
 313            if (translation == BIOS_ATA_TRANSLATION_AUTO) {
 314                bdrv_get_geometry_hint(hd_table[i], &cylinders, &heads, &sectors);
 315                if (cylinders <= 1024 && heads <= 16 && sectors <= 63) {
 316                    /* No translation. */
 317                    translation = 0;
 318                } else {
 319                    /* LBA translation. */
 320                    translation = 1;
 321                }
 322            } else {
 323                translation--;
 324            }
 325            val |= translation << (i * 2);
 326        }
 327    }
 328    rtc_set_memory(s, 0x39, val);
 329
 330    qemu_unregister_reset(pc_cmos_init_late, opaque);
 331}
 332
 333void pc_cmos_init(ram_addr_t ram_size, ram_addr_t above_4g_mem_size,
 334                  const char *boot_device,
 335                  ISADevice *floppy, BusState *idebus0, BusState *idebus1,
 336                  ISADevice *s)
 337{
 338    int val, nb, nb_heads, max_track, last_sect, i;
 339    FDriveType fd_type[2] = { FDRIVE_DRV_NONE, FDRIVE_DRV_NONE };
 340    FDriveRate rate;
 341    BlockDriverState *fd[MAX_FD];
 342    static pc_cmos_init_late_arg arg;
 343
 344    /* various important CMOS locations needed by PC/Bochs bios */
 345
 346    /* memory size */
 347    val = 640; /* base memory in K */
 348    rtc_set_memory(s, 0x15, val);
 349    rtc_set_memory(s, 0x16, val >> 8);
 350
 351    val = (ram_size / 1024) - 1024;
 352    if (val > 65535)
 353        val = 65535;
 354    rtc_set_memory(s, 0x17, val);
 355    rtc_set_memory(s, 0x18, val >> 8);
 356    rtc_set_memory(s, 0x30, val);
 357    rtc_set_memory(s, 0x31, val >> 8);
 358
 359    if (above_4g_mem_size) {
 360        rtc_set_memory(s, 0x5b, (unsigned int)above_4g_mem_size >> 16);
 361        rtc_set_memory(s, 0x5c, (unsigned int)above_4g_mem_size >> 24);
 362        rtc_set_memory(s, 0x5d, (uint64_t)above_4g_mem_size >> 32);
 363    }
 364
 365    if (ram_size > (16 * 1024 * 1024))
 366        val = (ram_size / 65536) - ((16 * 1024 * 1024) / 65536);
 367    else
 368        val = 0;
 369    if (val > 65535)
 370        val = 65535;
 371    rtc_set_memory(s, 0x34, val);
 372    rtc_set_memory(s, 0x35, val >> 8);
 373
 374    /* set the number of CPU */
 375    rtc_set_memory(s, 0x5f, smp_cpus - 1);
 376
 377    /* set boot devices, and disable floppy signature check if requested */
 378    if (set_boot_dev(s, boot_device, fd_bootchk)) {
 379        exit(1);
 380    }
 381
 382    /* floppy type */
 383    if (floppy) {
 384        fdc_get_bs(fd, floppy);
 385        for (i = 0; i < 2; i++) {
 386            if (fd[i]) {
 387                bdrv_get_floppy_geometry_hint(fd[i], &nb_heads, &max_track,
 388                                              &last_sect, FDRIVE_DRV_NONE,
 389                                              &fd_type[i], &rate);
 390            }
 391        }
 392    }
 393    val = (cmos_get_fd_drive_type(fd_type[0]) << 4) |
 394        cmos_get_fd_drive_type(fd_type[1]);
 395    rtc_set_memory(s, 0x10, val);
 396
 397    val = 0;
 398    nb = 0;
 399    if (fd_type[0] < FDRIVE_DRV_NONE) {
 400        nb++;
 401    }
 402    if (fd_type[1] < FDRIVE_DRV_NONE) {
 403        nb++;
 404    }
 405    switch (nb) {
 406    case 0:
 407        break;
 408    case 1:
 409        val |= 0x01; /* 1 drive, ready for boot */
 410        break;
 411    case 2:
 412        val |= 0x41; /* 2 drives, ready for boot */
 413        break;
 414    }
 415    val |= 0x02; /* FPU is there */
 416    val |= 0x04; /* PS/2 mouse installed */
 417    rtc_set_memory(s, REG_EQUIPMENT_BYTE, val);
 418
 419    /* hard drives */
 420    arg.rtc_state = s;
 421    arg.idebus0 = idebus0;
 422    arg.idebus1 = idebus1;
 423    qemu_register_reset(pc_cmos_init_late, &arg);
 424}
 425
 426/* port 92 stuff: could be split off */
 427typedef struct Port92State {
 428    ISADevice dev;
 429    MemoryRegion io;
 430    uint8_t outport;
 431    qemu_irq *a20_out;
 432} Port92State;
 433
 434static void port92_write(void *opaque, uint32_t addr, uint32_t val)
 435{
 436    Port92State *s = opaque;
 437
 438    DPRINTF("port92: write 0x%02x\n", val);
 439    s->outport = val;
 440    qemu_set_irq(*s->a20_out, (val >> 1) & 1);
 441    if (val & 1) {
 442        qemu_system_reset_request();
 443    }
 444}
 445
 446static uint32_t port92_read(void *opaque, uint32_t addr)
 447{
 448    Port92State *s = opaque;
 449    uint32_t ret;
 450
 451    ret = s->outport;
 452    DPRINTF("port92: read 0x%02x\n", ret);
 453    return ret;
 454}
 455
 456static void port92_init(ISADevice *dev, qemu_irq *a20_out)
 457{
 458    Port92State *s = DO_UPCAST(Port92State, dev, dev);
 459
 460    s->a20_out = a20_out;
 461}
 462
 463static const VMStateDescription vmstate_port92_isa = {
 464    .name = "port92",
 465    .version_id = 1,
 466    .minimum_version_id = 1,
 467    .minimum_version_id_old = 1,
 468    .fields      = (VMStateField []) {
 469        VMSTATE_UINT8(outport, Port92State),
 470        VMSTATE_END_OF_LIST()
 471    }
 472};
 473
 474static void port92_reset(DeviceState *d)
 475{
 476    Port92State *s = container_of(d, Port92State, dev.qdev);
 477
 478    s->outport &= ~1;
 479}
 480
 481static const MemoryRegionPortio port92_portio[] = {
 482    { 0, 1, 1, .read = port92_read, .write = port92_write },
 483    PORTIO_END_OF_LIST(),
 484};
 485
 486static const MemoryRegionOps port92_ops = {
 487    .old_portio = port92_portio
 488};
 489
 490static int port92_initfn(ISADevice *dev)
 491{
 492    Port92State *s = DO_UPCAST(Port92State, dev, dev);
 493
 494    memory_region_init_io(&s->io, &port92_ops, s, "port92", 1);
 495    isa_register_ioport(dev, &s->io, 0x92);
 496
 497    s->outport = 0;
 498    return 0;
 499}
 500
 501static void port92_class_initfn(ObjectClass *klass, void *data)
 502{
 503    DeviceClass *dc = DEVICE_CLASS(klass);
 504    ISADeviceClass *ic = ISA_DEVICE_CLASS(klass);
 505    ic->init = port92_initfn;
 506    dc->no_user = 1;
 507    dc->reset = port92_reset;
 508    dc->vmsd = &vmstate_port92_isa;
 509}
 510
 511static TypeInfo port92_info = {
 512    .name          = "port92",
 513    .parent        = TYPE_ISA_DEVICE,
 514    .instance_size = sizeof(Port92State),
 515    .class_init    = port92_class_initfn,
 516};
 517
 518static void port92_register_types(void)
 519{
 520    type_register_static(&port92_info);
 521}
 522
 523type_init(port92_register_types)
 524
 525static void handle_a20_line_change(void *opaque, int irq, int level)
 526{
 527    CPUX86State *cpu = opaque;
 528
 529    /* XXX: send to all CPUs ? */
 530    /* XXX: add logic to handle multiple A20 line sources */
 531    cpu_x86_set_a20(cpu, level);
 532}
 533
 534/***********************************************************/
 535/* Bochs BIOS debug ports */
 536
 537static void bochs_bios_write(void *opaque, uint32_t addr, uint32_t val)
 538{
 539    static const char shutdown_str[8] = "Shutdown";
 540    static int shutdown_index = 0;
 541
 542    switch(addr) {
 543        /* Bochs BIOS messages */
 544    case 0x400:
 545    case 0x401:
 546        /* used to be panic, now unused */
 547        break;
 548    case 0x402:
 549    case 0x403:
 550#ifdef DEBUG_BIOS
 551        fprintf(stderr, "%c", val);
 552#endif
 553        break;
 554    case 0x8900:
 555        /* same as Bochs power off */
 556        if (val == shutdown_str[shutdown_index]) {
 557            shutdown_index++;
 558            if (shutdown_index == 8) {
 559                shutdown_index = 0;
 560                qemu_system_shutdown_request();
 561            }
 562        } else {
 563            shutdown_index = 0;
 564        }
 565        break;
 566
 567        /* LGPL'ed VGA BIOS messages */
 568    case 0x501:
 569    case 0x502:
 570        exit((val << 1) | 1);
 571    case 0x500:
 572    case 0x503:
 573#ifdef DEBUG_BIOS
 574        fprintf(stderr, "%c", val);
 575#endif
 576        break;
 577    }
 578}
 579
 580int e820_add_entry(uint64_t address, uint64_t length, uint32_t type)
 581{
 582    int index = le32_to_cpu(e820_table.count);
 583    struct e820_entry *entry;
 584
 585    if (index >= E820_NR_ENTRIES)
 586        return -EBUSY;
 587    entry = &e820_table.entry[index++];
 588
 589    entry->address = cpu_to_le64(address);
 590    entry->length = cpu_to_le64(length);
 591    entry->type = cpu_to_le32(type);
 592
 593    e820_table.count = cpu_to_le32(index);
 594    return index;
 595}
 596
 597static void *bochs_bios_init(void)
 598{
 599    void *fw_cfg;
 600    uint8_t *smbios_table;
 601    size_t smbios_len;
 602    uint64_t *numa_fw_cfg;
 603    int i, j;
 604
 605    register_ioport_write(0x400, 1, 2, bochs_bios_write, NULL);
 606    register_ioport_write(0x401, 1, 2, bochs_bios_write, NULL);
 607    register_ioport_write(0x402, 1, 1, bochs_bios_write, NULL);
 608    register_ioport_write(0x403, 1, 1, bochs_bios_write, NULL);
 609    register_ioport_write(0x8900, 1, 1, bochs_bios_write, NULL);
 610
 611    register_ioport_write(0x501, 1, 1, bochs_bios_write, NULL);
 612    register_ioport_write(0x501, 1, 2, bochs_bios_write, NULL);
 613    register_ioport_write(0x502, 1, 2, bochs_bios_write, NULL);
 614    register_ioport_write(0x500, 1, 1, bochs_bios_write, NULL);
 615    register_ioport_write(0x503, 1, 1, bochs_bios_write, NULL);
 616
 617    fw_cfg = fw_cfg_init(BIOS_CFG_IOPORT, BIOS_CFG_IOPORT + 1, 0, 0);
 618
 619    fw_cfg_add_i32(fw_cfg, FW_CFG_ID, 1);
 620    fw_cfg_add_i64(fw_cfg, FW_CFG_RAM_SIZE, (uint64_t)ram_size);
 621    fw_cfg_add_bytes(fw_cfg, FW_CFG_ACPI_TABLES, (uint8_t *)acpi_tables,
 622                     acpi_tables_len);
 623    fw_cfg_add_i32(fw_cfg, FW_CFG_IRQ0_OVERRIDE, kvm_allows_irq0_override());
 624
 625    smbios_table = smbios_get_table(&smbios_len);
 626    if (smbios_table)
 627        fw_cfg_add_bytes(fw_cfg, FW_CFG_SMBIOS_ENTRIES,
 628                         smbios_table, smbios_len);
 629    fw_cfg_add_bytes(fw_cfg, FW_CFG_E820_TABLE, (uint8_t *)&e820_table,
 630                     sizeof(struct e820_table));
 631
 632    fw_cfg_add_bytes(fw_cfg, FW_CFG_HPET, (uint8_t *)&hpet_cfg,
 633                     sizeof(struct hpet_fw_config));
 634    /* allocate memory for the NUMA channel: one (64bit) word for the number
 635     * of nodes, one word for each VCPU->node and one word for each node to
 636     * hold the amount of memory.
 637     */
 638    numa_fw_cfg = g_malloc0((1 + max_cpus + nb_numa_nodes) * 8);
 639    numa_fw_cfg[0] = cpu_to_le64(nb_numa_nodes);
 640    for (i = 0; i < max_cpus; i++) {
 641        for (j = 0; j < nb_numa_nodes; j++) {
 642            if (node_cpumask[j] & (1 << i)) {
 643                numa_fw_cfg[i + 1] = cpu_to_le64(j);
 644                break;
 645            }
 646        }
 647    }
 648    for (i = 0; i < nb_numa_nodes; i++) {
 649        numa_fw_cfg[max_cpus + 1 + i] = cpu_to_le64(node_mem[i]);
 650    }
 651    fw_cfg_add_bytes(fw_cfg, FW_CFG_NUMA, (uint8_t *)numa_fw_cfg,
 652                     (1 + max_cpus + nb_numa_nodes) * 8);
 653
 654    return fw_cfg;
 655}
 656
 657static long get_file_size(FILE *f)
 658{
 659    long where, size;
 660
 661    /* XXX: on Unix systems, using fstat() probably makes more sense */
 662
 663    where = ftell(f);
 664    fseek(f, 0, SEEK_END);
 665    size = ftell(f);
 666    fseek(f, where, SEEK_SET);
 667
 668    return size;
 669}
 670
 671static void load_linux(void *fw_cfg,
 672                       const char *kernel_filename,
 673                       const char *initrd_filename,
 674                       const char *kernel_cmdline,
 675                       target_phys_addr_t max_ram_size)
 676{
 677    uint16_t protocol;
 678    int setup_size, kernel_size, initrd_size = 0, cmdline_size;
 679    uint32_t initrd_max;
 680    uint8_t header[8192], *setup, *kernel, *initrd_data;
 681    target_phys_addr_t real_addr, prot_addr, cmdline_addr, initrd_addr = 0;
 682    FILE *f;
 683    char *vmode;
 684
 685    /* Align to 16 bytes as a paranoia measure */
 686    cmdline_size = (strlen(kernel_cmdline)+16) & ~15;
 687
 688    /* load the kernel header */
 689    f = fopen(kernel_filename, "rb");
 690    if (!f || !(kernel_size = get_file_size(f)) ||
 691        fread(header, 1, MIN(ARRAY_SIZE(header), kernel_size), f) !=
 692        MIN(ARRAY_SIZE(header), kernel_size)) {
 693        fprintf(stderr, "qemu: could not load kernel '%s': %s\n",
 694                kernel_filename, strerror(errno));
 695        exit(1);
 696    }
 697
 698    /* kernel protocol version */
 699#if 0
 700    fprintf(stderr, "header magic: %#x\n", ldl_p(header+0x202));
 701#endif
 702    if (ldl_p(header+0x202) == 0x53726448)
 703        protocol = lduw_p(header+0x206);
 704    else {
 705        /* This looks like a multiboot kernel. If it is, let's stop
 706           treating it like a Linux kernel. */
 707        if (load_multiboot(fw_cfg, f, kernel_filename, initrd_filename,
 708                           kernel_cmdline, kernel_size, header))
 709            return;
 710        protocol = 0;
 711    }
 712
 713    if (protocol < 0x200 || !(header[0x211] & 0x01)) {
 714        /* Low kernel */
 715        real_addr    = 0x90000;
 716        cmdline_addr = 0x9a000 - cmdline_size;
 717        prot_addr    = 0x10000;
 718    } else if (protocol < 0x202) {
 719        /* High but ancient kernel */
 720        real_addr    = 0x90000;
 721        cmdline_addr = 0x9a000 - cmdline_size;
 722        prot_addr    = 0x100000;
 723    } else {
 724        /* High and recent kernel */
 725        real_addr    = 0x10000;
 726        cmdline_addr = 0x20000;
 727        prot_addr    = 0x100000;
 728    }
 729
 730#if 0
 731    fprintf(stderr,
 732            "qemu: real_addr     = 0x" TARGET_FMT_plx "\n"
 733            "qemu: cmdline_addr  = 0x" TARGET_FMT_plx "\n"
 734            "qemu: prot_addr     = 0x" TARGET_FMT_plx "\n",
 735            real_addr,
 736            cmdline_addr,
 737            prot_addr);
 738#endif
 739
 740    /* highest address for loading the initrd */
 741    if (protocol >= 0x203)
 742        initrd_max = ldl_p(header+0x22c);
 743    else
 744        initrd_max = 0x37ffffff;
 745
 746    if (initrd_max >= max_ram_size-ACPI_DATA_SIZE)
 747        initrd_max = max_ram_size-ACPI_DATA_SIZE-1;
 748
 749    fw_cfg_add_i32(fw_cfg, FW_CFG_CMDLINE_ADDR, cmdline_addr);
 750    fw_cfg_add_i32(fw_cfg, FW_CFG_CMDLINE_SIZE, strlen(kernel_cmdline)+1);
 751    fw_cfg_add_bytes(fw_cfg, FW_CFG_CMDLINE_DATA,
 752                     (uint8_t*)strdup(kernel_cmdline),
 753                     strlen(kernel_cmdline)+1);
 754
 755    if (protocol >= 0x202) {
 756        stl_p(header+0x228, cmdline_addr);
 757    } else {
 758        stw_p(header+0x20, 0xA33F);
 759        stw_p(header+0x22, cmdline_addr-real_addr);
 760    }
 761
 762    /* handle vga= parameter */
 763    vmode = strstr(kernel_cmdline, "vga=");
 764    if (vmode) {
 765        unsigned int video_mode;
 766        /* skip "vga=" */
 767        vmode += 4;
 768        if (!strncmp(vmode, "normal", 6)) {
 769            video_mode = 0xffff;
 770        } else if (!strncmp(vmode, "ext", 3)) {
 771            video_mode = 0xfffe;
 772        } else if (!strncmp(vmode, "ask", 3)) {
 773            video_mode = 0xfffd;
 774        } else {
 775            video_mode = strtol(vmode, NULL, 0);
 776        }
 777        stw_p(header+0x1fa, video_mode);
 778    }
 779
 780    /* loader type */
 781    /* High nybble = B reserved for QEMU; low nybble is revision number.
 782       If this code is substantially changed, you may want to consider
 783       incrementing the revision. */
 784    if (protocol >= 0x200)
 785        header[0x210] = 0xB0;
 786
 787    /* heap */
 788    if (protocol >= 0x201) {
 789        header[0x211] |= 0x80;  /* CAN_USE_HEAP */
 790        stw_p(header+0x224, cmdline_addr-real_addr-0x200);
 791    }
 792
 793    /* load initrd */
 794    if (initrd_filename) {
 795        if (protocol < 0x200) {
 796            fprintf(stderr, "qemu: linux kernel too old to load a ram disk\n");
 797            exit(1);
 798        }
 799
 800        initrd_size = get_image_size(initrd_filename);
 801        if (initrd_size < 0) {
 802            fprintf(stderr, "qemu: error reading initrd %s\n",
 803                    initrd_filename);
 804            exit(1);
 805        }
 806
 807        initrd_addr = (initrd_max-initrd_size) & ~4095;
 808
 809        initrd_data = g_malloc(initrd_size);
 810        load_image(initrd_filename, initrd_data);
 811
 812        fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_ADDR, initrd_addr);
 813        fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_SIZE, initrd_size);
 814        fw_cfg_add_bytes(fw_cfg, FW_CFG_INITRD_DATA, initrd_data, initrd_size);
 815
 816        stl_p(header+0x218, initrd_addr);
 817        stl_p(header+0x21c, initrd_size);
 818    }
 819
 820    /* load kernel and setup */
 821    setup_size = header[0x1f1];
 822    if (setup_size == 0)
 823        setup_size = 4;
 824    setup_size = (setup_size+1)*512;
 825    kernel_size -= setup_size;
 826
 827    setup  = g_malloc(setup_size);
 828    kernel = g_malloc(kernel_size);
 829    fseek(f, 0, SEEK_SET);
 830    if (fread(setup, 1, setup_size, f) != setup_size) {
 831        fprintf(stderr, "fread() failed\n");
 832        exit(1);
 833    }
 834    if (fread(kernel, 1, kernel_size, f) != kernel_size) {
 835        fprintf(stderr, "fread() failed\n");
 836        exit(1);
 837    }
 838    fclose(f);
 839    memcpy(setup, header, MIN(sizeof(header), setup_size));
 840
 841    fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_ADDR, prot_addr);
 842    fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_SIZE, kernel_size);
 843    fw_cfg_add_bytes(fw_cfg, FW_CFG_KERNEL_DATA, kernel, kernel_size);
 844
 845    fw_cfg_add_i32(fw_cfg, FW_CFG_SETUP_ADDR, real_addr);
 846    fw_cfg_add_i32(fw_cfg, FW_CFG_SETUP_SIZE, setup_size);
 847    fw_cfg_add_bytes(fw_cfg, FW_CFG_SETUP_DATA, setup, setup_size);
 848
 849    option_rom[nb_option_roms].name = "linuxboot.bin";
 850    option_rom[nb_option_roms].bootindex = 0;
 851    nb_option_roms++;
 852}
 853
 854#define NE2000_NB_MAX 6
 855
 856static const int ne2000_io[NE2000_NB_MAX] = { 0x300, 0x320, 0x340, 0x360,
 857                                              0x280, 0x380 };
 858static const int ne2000_irq[NE2000_NB_MAX] = { 9, 10, 11, 3, 4, 5 };
 859
 860static const int parallel_io[MAX_PARALLEL_PORTS] = { 0x378, 0x278, 0x3bc };
 861static const int parallel_irq[MAX_PARALLEL_PORTS] = { 7, 7, 7 };
 862
 863void pc_init_ne2k_isa(ISABus *bus, NICInfo *nd)
 864{
 865    static int nb_ne2k = 0;
 866
 867    if (nb_ne2k == NE2000_NB_MAX)
 868        return;
 869    isa_ne2000_init(bus, ne2000_io[nb_ne2k],
 870                    ne2000_irq[nb_ne2k], nd);
 871    nb_ne2k++;
 872}
 873
 874int cpu_is_bsp(CPUX86State *env)
 875{
 876    /* We hard-wire the BSP to the first CPU. */
 877    return env->cpu_index == 0;
 878}
 879
 880DeviceState *cpu_get_current_apic(void)
 881{
 882    if (cpu_single_env) {
 883        return cpu_single_env->apic_state;
 884    } else {
 885        return NULL;
 886    }
 887}
 888
 889static DeviceState *apic_init(void *env, uint8_t apic_id)
 890{
 891    DeviceState *dev;
 892    static int apic_mapped;
 893
 894    if (kvm_irqchip_in_kernel()) {
 895        dev = qdev_create(NULL, "kvm-apic");
 896    } else if (xen_enabled()) {
 897        dev = qdev_create(NULL, "xen-apic");
 898    } else {
 899        dev = qdev_create(NULL, "apic");
 900    }
 901
 902    qdev_prop_set_uint8(dev, "id", apic_id);
 903    qdev_prop_set_ptr(dev, "cpu_env", env);
 904    qdev_init_nofail(dev);
 905
 906    /* XXX: mapping more APICs at the same memory location */
 907    if (apic_mapped == 0) {
 908        /* NOTE: the APIC is directly connected to the CPU - it is not
 909           on the global memory bus. */
 910        /* XXX: what if the base changes? */
 911        sysbus_mmio_map(sysbus_from_qdev(dev), 0, MSI_ADDR_BASE);
 912        apic_mapped = 1;
 913    }
 914
 915    /* KVM does not support MSI yet. */
 916    if (!kvm_irqchip_in_kernel()) {
 917        msi_supported = true;
 918    }
 919
 920    if (xen_msi_support()) {
 921        msi_supported = true;
 922    }
 923
 924    return dev;
 925}
 926
 927void pc_acpi_smi_interrupt(void *opaque, int irq, int level)
 928{
 929    CPUX86State *s = opaque;
 930
 931    if (level) {
 932        cpu_interrupt(s, CPU_INTERRUPT_SMI);
 933    }
 934}
 935
 936static void pc_cpu_reset(void *opaque)
 937{
 938    CPUX86State *env = opaque;
 939
 940    cpu_state_reset(env);
 941    env->halted = !cpu_is_bsp(env);
 942}
 943
 944static CPUX86State *pc_new_cpu(const char *cpu_model)
 945{
 946    CPUX86State *env;
 947
 948    env = cpu_init(cpu_model);
 949    if (!env) {
 950        fprintf(stderr, "Unable to find x86 CPU definition\n");
 951        exit(1);
 952    }
 953    if ((env->cpuid_features & CPUID_APIC) || smp_cpus > 1) {
 954        env->apic_state = apic_init(env, env->cpuid_apic_id);
 955    }
 956    qemu_register_reset(pc_cpu_reset, env);
 957    pc_cpu_reset(env);
 958    return env;
 959}
 960
 961void pc_cpus_init(const char *cpu_model)
 962{
 963    int i;
 964
 965    /* init CPUs */
 966    if (cpu_model == NULL) {
 967#ifdef TARGET_X86_64
 968        cpu_model = "qemu64";
 969#else
 970        cpu_model = "qemu32";
 971#endif
 972    }
 973
 974    for(i = 0; i < smp_cpus; i++) {
 975        pc_new_cpu(cpu_model);
 976    }
 977}
 978
 979void pc_memory_init(MemoryRegion *system_memory,
 980                    const char *kernel_filename,
 981                    const char *kernel_cmdline,
 982                    const char *initrd_filename,
 983                    ram_addr_t below_4g_mem_size,
 984                    ram_addr_t above_4g_mem_size,
 985                    MemoryRegion *rom_memory,
 986                    MemoryRegion **ram_memory)
 987{
 988    int linux_boot, i;
 989    MemoryRegion *ram, *option_rom_mr;
 990    MemoryRegion *ram_below_4g, *ram_above_4g;
 991    void *fw_cfg;
 992
 993    linux_boot = (kernel_filename != NULL);
 994
 995    /* Allocate RAM.  We allocate it as a single memory region and use
 996     * aliases to address portions of it, mostly for backwards compatibility
 997     * with older qemus that used qemu_ram_alloc().
 998     */
 999    ram = g_malloc(sizeof(*ram));
1000    memory_region_init_ram(ram, "pc.ram",
1001                           below_4g_mem_size + above_4g_mem_size);
1002    vmstate_register_ram_global(ram);
1003    *ram_memory = ram;
1004    ram_below_4g = g_malloc(sizeof(*ram_below_4g));
1005    memory_region_init_alias(ram_below_4g, "ram-below-4g", ram,
1006                             0, below_4g_mem_size);
1007    memory_region_add_subregion(system_memory, 0, ram_below_4g);
1008    if (above_4g_mem_size > 0) {
1009        ram_above_4g = g_malloc(sizeof(*ram_above_4g));
1010        memory_region_init_alias(ram_above_4g, "ram-above-4g", ram,
1011                                 below_4g_mem_size, above_4g_mem_size);
1012        memory_region_add_subregion(system_memory, 0x100000000ULL,
1013                                    ram_above_4g);
1014    }
1015
1016
1017    /* Initialize PC system firmware */
1018    pc_system_firmware_init(rom_memory);
1019
1020    option_rom_mr = g_malloc(sizeof(*option_rom_mr));
1021    memory_region_init_ram(option_rom_mr, "pc.rom", PC_ROM_SIZE);
1022    vmstate_register_ram_global(option_rom_mr);
1023    memory_region_add_subregion_overlap(rom_memory,
1024                                        PC_ROM_MIN_VGA,
1025                                        option_rom_mr,
1026                                        1);
1027
1028    fw_cfg = bochs_bios_init();
1029    rom_set_fw(fw_cfg);
1030
1031    if (linux_boot) {
1032        load_linux(fw_cfg, kernel_filename, initrd_filename, kernel_cmdline, below_4g_mem_size);
1033    }
1034
1035    for (i = 0; i < nb_option_roms; i++) {
1036        rom_add_option(option_rom[i].name, option_rom[i].bootindex);
1037    }
1038}
1039
1040qemu_irq *pc_allocate_cpu_irq(void)
1041{
1042    return qemu_allocate_irqs(pic_irq_request, NULL, 1);
1043}
1044
1045DeviceState *pc_vga_init(ISABus *isa_bus, PCIBus *pci_bus)
1046{
1047    DeviceState *dev = NULL;
1048
1049    if (cirrus_vga_enabled) {
1050        if (pci_bus) {
1051            dev = pci_cirrus_vga_init(pci_bus);
1052        } else {
1053            dev = &isa_create_simple(isa_bus, "isa-cirrus-vga")->qdev;
1054        }
1055    } else if (vmsvga_enabled) {
1056        if (pci_bus) {
1057            dev = pci_vmsvga_init(pci_bus);
1058        } else {
1059            fprintf(stderr, "%s: vmware_vga: no PCI bus\n", __FUNCTION__);
1060        }
1061#ifdef CONFIG_SPICE
1062    } else if (qxl_enabled) {
1063        if (pci_bus) {
1064            dev = &pci_create_simple(pci_bus, -1, "qxl-vga")->qdev;
1065        } else {
1066            fprintf(stderr, "%s: qxl: no PCI bus\n", __FUNCTION__);
1067        }
1068#endif
1069    } else if (std_vga_enabled) {
1070        if (pci_bus) {
1071            dev = pci_vga_init(pci_bus);
1072        } else {
1073            dev = isa_vga_init(isa_bus);
1074        }
1075    }
1076
1077    return dev;
1078}
1079
1080static void cpu_request_exit(void *opaque, int irq, int level)
1081{
1082    CPUX86State *env = cpu_single_env;
1083
1084    if (env && level) {
1085        cpu_exit(env);
1086    }
1087}
1088
1089void pc_basic_device_init(ISABus *isa_bus, qemu_irq *gsi,
1090                          ISADevice **rtc_state,
1091                          ISADevice **floppy,
1092                          bool no_vmport)
1093{
1094    int i;
1095    DriveInfo *fd[MAX_FD];
1096    DeviceState *hpet = NULL;
1097    int pit_isa_irq = 0;
1098    qemu_irq pit_alt_irq = NULL;
1099    qemu_irq rtc_irq = NULL;
1100    qemu_irq *a20_line;
1101    ISADevice *i8042, *port92, *vmmouse, *pit = NULL;
1102    qemu_irq *cpu_exit_irq;
1103
1104    register_ioport_write(0x80, 1, 1, ioport80_write, NULL);
1105
1106    register_ioport_write(0xf0, 1, 1, ioportF0_write, NULL);
1107
1108    /*
1109     * Check if an HPET shall be created.
1110     *
1111     * Without KVM_CAP_PIT_STATE2, we cannot switch off the in-kernel PIT
1112     * when the HPET wants to take over. Thus we have to disable the latter.
1113     */
1114    if (!no_hpet && (!kvm_irqchip_in_kernel() || kvm_has_pit_state2())) {
1115        hpet = sysbus_try_create_simple("hpet", HPET_BASE, NULL);
1116
1117        if (hpet) {
1118            for (i = 0; i < GSI_NUM_PINS; i++) {
1119                sysbus_connect_irq(sysbus_from_qdev(hpet), i, gsi[i]);
1120            }
1121            pit_isa_irq = -1;
1122            pit_alt_irq = qdev_get_gpio_in(hpet, HPET_LEGACY_PIT_INT);
1123            rtc_irq = qdev_get_gpio_in(hpet, HPET_LEGACY_RTC_INT);
1124        }
1125    }
1126    *rtc_state = rtc_init(isa_bus, 2000, rtc_irq);
1127
1128    qemu_register_boot_set(pc_boot_set, *rtc_state);
1129
1130    if (!xen_enabled()) {
1131        if (kvm_irqchip_in_kernel()) {
1132            pit = kvm_pit_init(isa_bus, 0x40);
1133        } else {
1134            pit = pit_init(isa_bus, 0x40, pit_isa_irq, pit_alt_irq);
1135        }
1136        if (hpet) {
1137            /* connect PIT to output control line of the HPET */
1138            qdev_connect_gpio_out(hpet, 0, qdev_get_gpio_in(&pit->qdev, 0));
1139        }
1140        pcspk_init(isa_bus, pit);
1141    }
1142
1143    for(i = 0; i < MAX_SERIAL_PORTS; i++) {
1144        if (serial_hds[i]) {
1145            serial_isa_init(isa_bus, i, serial_hds[i]);
1146        }
1147    }
1148
1149    for(i = 0; i < MAX_PARALLEL_PORTS; i++) {
1150        if (parallel_hds[i]) {
1151            parallel_init(isa_bus, i, parallel_hds[i]);
1152        }
1153    }
1154
1155    a20_line = qemu_allocate_irqs(handle_a20_line_change, first_cpu, 2);
1156    i8042 = isa_create_simple(isa_bus, "i8042");
1157    i8042_setup_a20_line(i8042, &a20_line[0]);
1158    if (!no_vmport) {
1159        vmport_init(isa_bus);
1160        vmmouse = isa_try_create(isa_bus, "vmmouse");
1161    } else {
1162        vmmouse = NULL;
1163    }
1164    if (vmmouse) {
1165        qdev_prop_set_ptr(&vmmouse->qdev, "ps2_mouse", i8042);
1166        qdev_init_nofail(&vmmouse->qdev);
1167    }
1168    port92 = isa_create_simple(isa_bus, "port92");
1169    port92_init(port92, &a20_line[1]);
1170
1171    cpu_exit_irq = qemu_allocate_irqs(cpu_request_exit, NULL, 1);
1172    DMA_init(0, cpu_exit_irq);
1173
1174    for(i = 0; i < MAX_FD; i++) {
1175        fd[i] = drive_get(IF_FLOPPY, 0, i);
1176    }
1177    *floppy = fdctrl_init_isa(isa_bus, fd);
1178}
1179
1180void pc_pci_device_init(PCIBus *pci_bus)
1181{
1182    int max_bus;
1183    int bus;
1184
1185    max_bus = drive_get_max_bus(IF_SCSI);
1186    for (bus = 0; bus <= max_bus; bus++) {
1187        pci_create_simple(pci_bus, -1, "lsi53c895a");
1188    }
1189}
1190