hw/i386/x86.c

   1 /*
   2  * Copyright (c) 2003-2004 Fabrice Bellard
   3  * Copyright (c) 2019 Red Hat, Inc.
   4  *
   5  * Permission is hereby granted, free of charge, to any person obtaining a copy
   6  * of this software and associated documentation files (the "Software"), to deal
   7  * in the Software without restriction, including without limitation the rights
   8  * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
   9  * copies of the Software, and to permit persons to whom the Software is
  10  * furnished to do so, subject to the following conditions:
  11  *
  12  * The above copyright notice and this permission notice shall be included in
  13  * all copies or substantial portions of the Software.
  14  *
  15  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
  16  * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
  17  * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
  18  * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
  19  * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
  20  * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
  21  * THE SOFTWARE.
  22  */
  23 #include "qemu/osdep.h"
  24 #include "qemu/error-report.h"
  25 #include "qemu/option.h"
  26 #include "qemu/cutils.h"
  27 #include "qemu/units.h"
  28 #include "qemu-common.h"
  29 #include "qapi/error.h"
  30 #include "qapi/qmp/qerror.h"
  31 #include "qapi/qapi-visit-common.h"
  32 #include "qapi/visitor.h"
  33 #include "sysemu/qtest.h"
  34 #include "sysemu/numa.h"
  35 #include "sysemu/replay.h"
  36 #include "sysemu/sysemu.h"
  37
  38 #include "hw/i386/x86.h"
  39 #include "target/i386/cpu.h"
  40 #include "hw/i386/topology.h"
  41 #include "hw/i386/fw_cfg.h"
  42
  43 #include "hw/acpi/cpu_hotplug.h"
  44 #include "hw/nmi.h"
  45 #include "hw/loader.h"
  46 #include "multiboot.h"
  47 #include "elf.h"
  48 #include "standard-headers/asm-x86/bootparam.h"
  49
  50 #define BIOS_FILENAME "bios.bin"
  51
  52 /* Physical Address of PVH entry point read from kernel ELF NOTE */
  53 static size_t pvh_start_addr;
  54
  55 /*
  56  * Calculates initial APIC ID for a specific CPU index
  57  *
  58  * Currently we need to be able to calculate the APIC ID from the CPU index
  59  * alone (without requiring a CPU object), as the QEMU<->Seabios interfaces have
  60  * no concept of "CPU index", and the NUMA tables on fw_cfg need the APIC ID of
  61  * all CPUs up to max_cpus.
  62  */
  63 uint32_t x86_cpu_apic_id_from_index(X86MachineState *x86ms,
  64                                     unsigned int cpu_index)
  65 {
  66     MachineState *ms = MACHINE(x86ms);
  67     X86MachineClass *x86mc = X86_MACHINE_GET_CLASS(x86ms);
  68     uint32_t correct_id;
  69     static bool warned;
  70
  71     correct_id = x86_apicid_from_cpu_idx(x86ms->smp_dies, ms->smp.cores,
  72                                          ms->smp.threads, cpu_index);
  73     if (x86mc->compat_apic_id_mode) {
  74         if (cpu_index != correct_id && !warned && !qtest_enabled()) {
  75             error_report("APIC IDs set in compatibility mode, "
  76                          "CPU topology won't match the configuration");
  77             warned = true;
  78         }
  79         return cpu_index;
  80     } else {
  81         return correct_id;
  82     }
  83 }
  84
  85
  86 void x86_cpu_new(X86MachineState *x86ms, int64_t apic_id, Error **errp)
  87 {
  88     Object *cpu = NULL;
  89     Error *local_err = NULL;
  90     CPUX86State *env = NULL;
  91
  92     cpu = object_new(MACHINE(x86ms)->cpu_type);
  93
  94     env = &X86_CPU(cpu)->env;
  95     env->nr_dies = x86ms->smp_dies;
  96
  97     object_property_set_uint(cpu, apic_id, "apic-id", &local_err);
  98     object_property_set_bool(cpu, true, "realized", &local_err);
  99
 100     object_unref(cpu);
 101     error_propagate(errp, local_err);
 102 }
 103
 104 void x86_cpus_init(X86MachineState *x86ms, int default_cpu_version)
 105 {
 106     int i;
 107     const CPUArchIdList *possible_cpus;
 108     MachineState *ms = MACHINE(x86ms);
 109     MachineClass *mc = MACHINE_GET_CLASS(x86ms);
 110
 111     x86_cpu_set_default_version(default_cpu_version);
 112
 113     /*
 114      * Calculates the limit to CPU APIC ID values
 115      *
 116      * Limit for the APIC ID value, so that all
 117      * CPU APIC IDs are < x86ms->apic_id_limit.
 118      *
 119      * This is used for FW_CFG_MAX_CPUS. See comments on fw_cfg_arch_create().
 120      */
 121     x86ms->apic_id_limit = x86_cpu_apic_id_from_index(x86ms,
 122                                                       ms->smp.max_cpus - 1) + 1;
 123     possible_cpus = mc->possible_cpu_arch_ids(ms);
 124     for (i = 0; i < ms->smp.cpus; i++) {
 125         x86_cpu_new(x86ms, possible_cpus->cpus[i].arch_id, &error_fatal);
 126     }
 127 }
 128
 129 CpuInstanceProperties
 130 x86_cpu_index_to_props(MachineState *ms, unsigned cpu_index)
 131 {
 132     MachineClass *mc = MACHINE_GET_CLASS(ms);
 133     const CPUArchIdList *possible_cpus = mc->possible_cpu_arch_ids(ms);
 134
 135     assert(cpu_index < possible_cpus->len);
 136     return possible_cpus->cpus[cpu_index].props;
 137 }
 138
 139 int64_t x86_get_default_cpu_node_id(const MachineState *ms, int idx)
 140 {
 141    X86CPUTopoInfo topo;
 142    X86MachineState *x86ms = X86_MACHINE(ms);
 143
 144    assert(idx < ms->possible_cpus->len);
 145    x86_topo_ids_from_apicid(ms->possible_cpus->cpus[idx].arch_id,
 146                             x86ms->smp_dies, ms->smp.cores,
 147                             ms->smp.threads, &topo);
 148    return topo.pkg_id % ms->numa_state->num_nodes;
 149 }
 150
 151 const CPUArchIdList *x86_possible_cpu_arch_ids(MachineState *ms)
 152 {
 153     X86MachineState *x86ms = X86_MACHINE(ms);
 154     int i;
 155     unsigned int max_cpus = ms->smp.max_cpus;
 156
 157     if (ms->possible_cpus) {
 158         /*
 159          * make sure that max_cpus hasn't changed since the first use, i.e.
 160          * -smp hasn't been parsed after it
 161          */
 162         assert(ms->possible_cpus->len == max_cpus);
 163         return ms->possible_cpus;
 164     }
 165
 166     ms->possible_cpus = g_malloc0(sizeof(CPUArchIdList) +
 167                                   sizeof(CPUArchId) * max_cpus);
 168     ms->possible_cpus->len = max_cpus;
 169     for (i = 0; i < ms->possible_cpus->len; i++) {
 170         X86CPUTopoInfo topo;
 171
 172         ms->possible_cpus->cpus[i].type = ms->cpu_type;
 173         ms->possible_cpus->cpus[i].vcpus_count = 1;
 174         ms->possible_cpus->cpus[i].arch_id =
 175             x86_cpu_apic_id_from_index(x86ms, i);
 176         x86_topo_ids_from_apicid(ms->possible_cpus->cpus[i].arch_id,
 177                                  x86ms->smp_dies, ms->smp.cores,
 178                                  ms->smp.threads, &topo);
 179         ms->possible_cpus->cpus[i].props.has_socket_id = true;
 180         ms->possible_cpus->cpus[i].props.socket_id = topo.pkg_id;
 181         if (x86ms->smp_dies > 1) {
 182             ms->possible_cpus->cpus[i].props.has_die_id = true;
 183             ms->possible_cpus->cpus[i].props.die_id = topo.die_id;
 184         }
 185         ms->possible_cpus->cpus[i].props.has_core_id = true;
 186         ms->possible_cpus->cpus[i].props.core_id = topo.core_id;
 187         ms->possible_cpus->cpus[i].props.has_thread_id = true;
 188         ms->possible_cpus->cpus[i].props.thread_id = topo.smt_id;
 189     }
 190     return ms->possible_cpus;
 191 }
 192
 193 static void x86_nmi(NMIState *n, int cpu_index, Error **errp)
 194 {
 195     /* cpu index isn't used */
 196     CPUState *cs;
 197
 198     CPU_FOREACH(cs) {
 199         X86CPU *cpu = X86_CPU(cs);
 200
 201         if (!cpu->apic_state) {
 202             cpu_interrupt(cs, CPU_INTERRUPT_NMI);
 203         } else {
 204             apic_deliver_nmi(cpu->apic_state);
 205         }
 206     }
 207 }
 208
 209 static long get_file_size(FILE *f)
 210 {
 211     long where, size;
 212
 213     /* XXX: on Unix systems, using fstat() probably makes more sense */
 214
 215     where = ftell(f);
 216     fseek(f, 0, SEEK_END);
 217     size = ftell(f);
 218     fseek(f, where, SEEK_SET);
 219
 220     return size;
 221 }
 222
 223 struct setup_data {
 224     uint64_t next;
 225     uint32_t type;
 226     uint32_t len;
 227     uint8_t data[0];
 228 } __attribute__((packed));
 229
 230
 231 /*
 232  * The entry point into the kernel for PVH boot is different from
 233  * the native entry point.  The PVH entry is defined by the x86/HVM
 234  * direct boot ABI and is available in an ELFNOTE in the kernel binary.
 235  *
 236  * This function is passed to load_elf() when it is called from
 237  * load_elfboot() which then additionally checks for an ELF Note of
 238  * type XEN_ELFNOTE_PHYS32_ENTRY and passes it to this function to
 239  * parse the PVH entry address from the ELF Note.
 240  *
 241  * Due to trickery in elf_opts.h, load_elf() is actually available as
 242  * load_elf32() or load_elf64() and this routine needs to be able
 243  * to deal with being called as 32 or 64 bit.
 244  *
 245  * The address of the PVH entry point is saved to the 'pvh_start_addr'
 246  * global variable.  (although the entry point is 32-bit, the kernel
 247  * binary can be either 32-bit or 64-bit).
 248  */
 249 static uint64_t read_pvh_start_addr(void *arg1, void *arg2, bool is64)
 250 {
 251     size_t *elf_note_data_addr;
 252
 253     /* Check if ELF Note header passed in is valid */
 254     if (arg1 == NULL) {
 255         return 0;
 256     }
 257
 258     if (is64) {
 259         struct elf64_note *nhdr64 = (struct elf64_note *)arg1;
 260         uint64_t nhdr_size64 = sizeof(struct elf64_note);
 261         uint64_t phdr_align = *(uint64_t *)arg2;
 262         uint64_t nhdr_namesz = nhdr64->n_namesz;
 263
 264         elf_note_data_addr =
 265             ((void *)nhdr64) + nhdr_size64 +
 266             QEMU_ALIGN_UP(nhdr_namesz, phdr_align);
 267     } else {
 268         struct elf32_note *nhdr32 = (struct elf32_note *)arg1;
 269         uint32_t nhdr_size32 = sizeof(struct elf32_note);
 270         uint32_t phdr_align = *(uint32_t *)arg2;
 271         uint32_t nhdr_namesz = nhdr32->n_namesz;
 272
 273         elf_note_data_addr =
 274             ((void *)nhdr32) + nhdr_size32 +
 275             QEMU_ALIGN_UP(nhdr_namesz, phdr_align);
 276     }
 277
 278     pvh_start_addr = *elf_note_data_addr;
 279
 280     return pvh_start_addr;
 281 }
 282
 283 static bool load_elfboot(const char *kernel_filename,
 284                          int kernel_file_size,
 285                          uint8_t *header,
 286                          size_t pvh_xen_start_addr,
 287                          FWCfgState *fw_cfg)
 288 {
 289     uint32_t flags = 0;
 290     uint32_t mh_load_addr = 0;
 291     uint32_t elf_kernel_size = 0;
 292     uint64_t elf_entry;
 293     uint64_t elf_low, elf_high;
 294     int kernel_size;
 295
 296     if (ldl_p(header) != 0x464c457f) {
 297         return false; /* no elfboot */
 298     }
 299
 300     bool elf_is64 = header[EI_CLASS] == ELFCLASS64;
 301     flags = elf_is64 ?
 302         ((Elf64_Ehdr *)header)->e_flags : ((Elf32_Ehdr *)header)->e_flags;
 303
 304     if (flags & 0x00010004) { /* LOAD_ELF_HEADER_HAS_ADDR */
 305         error_report("elfboot unsupported flags = %x", flags);
 306         exit(1);
 307     }
 308
 309     uint64_t elf_note_type = XEN_ELFNOTE_PHYS32_ENTRY;
 310     kernel_size = load_elf(kernel_filename, read_pvh_start_addr,
 311                            NULL, &elf_note_type, &elf_entry,
 312                            &elf_low, &elf_high, 0, I386_ELF_MACHINE,
 313                            0, 0);
 314
 315     if (kernel_size < 0) {
 316         error_report("Error while loading elf kernel");
 317         exit(1);
 318     }
 319     mh_load_addr = elf_low;
 320     elf_kernel_size = elf_high - elf_low;
 321
 322     if (pvh_start_addr == 0) {
 323         error_report("Error loading uncompressed kernel without PVH ELF Note");
 324         exit(1);
 325     }
 326     fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_ENTRY, pvh_start_addr);
 327     fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_ADDR, mh_load_addr);
 328     fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_SIZE, elf_kernel_size);
 329
 330     return true;
 331 }
 332
 333 void x86_load_linux(X86MachineState *x86ms,
 334                     FWCfgState *fw_cfg,
 335                     int acpi_data_size,
 336                     bool pvh_enabled,
 337                     bool linuxboot_dma_enabled)
 338 {
 339     uint16_t protocol;
 340     int setup_size, kernel_size, cmdline_size;
 341     int dtb_size, setup_data_offset;
 342     uint32_t initrd_max;
 343     uint8_t header[8192], *setup, *kernel;
 344     hwaddr real_addr, prot_addr, cmdline_addr, initrd_addr = 0;
 345     FILE *f;
 346     char *vmode;
 347     MachineState *machine = MACHINE(x86ms);
 348     struct setup_data *setup_data;
 349     const char *kernel_filename = machine->kernel_filename;
 350     const char *initrd_filename = machine->initrd_filename;
 351     const char *dtb_filename = machine->dtb;
 352     const char *kernel_cmdline = machine->kernel_cmdline;
 353
 354     /* Align to 16 bytes as a paranoia measure */
 355     cmdline_size = (strlen(kernel_cmdline) + 16) & ~15;
 356
 357     /* load the kernel header */
 358     f = fopen(kernel_filename, "rb");
 359     if (!f) {
 360         fprintf(stderr, "qemu: could not open kernel file '%s': %s\n",
 361                 kernel_filename, strerror(errno));
 362         exit(1);
 363     }
 364
 365     kernel_size = get_file_size(f);
 366     if (!kernel_size ||
 367         fread(header, 1, MIN(ARRAY_SIZE(header), kernel_size), f) !=
 368         MIN(ARRAY_SIZE(header), kernel_size)) {
 369         fprintf(stderr, "qemu: could not load kernel '%s': %s\n",
 370                 kernel_filename, strerror(errno));
 371         exit(1);
 372     }
 373
 374     /* kernel protocol version */
 375     if (ldl_p(header + 0x202) == 0x53726448) {
 376         protocol = lduw_p(header + 0x206);
 377     } else {
 378         /*
 379          * This could be a multiboot kernel. If it is, let's stop treating it
 380          * like a Linux kernel.
 381          * Note: some multiboot images could be in the ELF format (the same of
 382          * PVH), so we try multiboot first since we check the multiboot magic
 383          * header before to load it.
 384          */
 385         if (load_multiboot(fw_cfg, f, kernel_filename, initrd_filename,
 386                            kernel_cmdline, kernel_size, header)) {
 387             return;
 388         }
 389         /*
 390          * Check if the file is an uncompressed kernel file (ELF) and load it,
 391          * saving the PVH entry point used by the x86/HVM direct boot ABI.
 392          * If load_elfboot() is successful, populate the fw_cfg info.
 393          */
 394         if (pvh_enabled &&
 395             load_elfboot(kernel_filename, kernel_size,
 396                          header, pvh_start_addr, fw_cfg)) {
 397             fclose(f);
 398
 399             fw_cfg_add_i32(fw_cfg, FW_CFG_CMDLINE_SIZE,
 400                 strlen(kernel_cmdline) + 1);
 401             fw_cfg_add_string(fw_cfg, FW_CFG_CMDLINE_DATA, kernel_cmdline);
 402
 403             fw_cfg_add_i32(fw_cfg, FW_CFG_SETUP_SIZE, sizeof(header));
 404             fw_cfg_add_bytes(fw_cfg, FW_CFG_SETUP_DATA,
 405                              header, sizeof(header));
 406
 407             /* load initrd */
 408             if (initrd_filename) {
 409                 GMappedFile *mapped_file;
 410                 gsize initrd_size;
 411                 gchar *initrd_data;
 412                 GError *gerr = NULL;
 413
 414                 mapped_file = g_mapped_file_new(initrd_filename, false, &gerr);
 415                 if (!mapped_file) {
 416                     fprintf(stderr, "qemu: error reading initrd %s: %s\n",
 417                             initrd_filename, gerr->message);
 418                     exit(1);
 419                 }
 420                 x86ms->initrd_mapped_file = mapped_file;
 421
 422                 initrd_data = g_mapped_file_get_contents(mapped_file);
 423                 initrd_size = g_mapped_file_get_length(mapped_file);
 424                 initrd_max = x86ms->below_4g_mem_size - acpi_data_size - 1;
 425                 if (initrd_size >= initrd_max) {
 426                     fprintf(stderr, "qemu: initrd is too large, cannot support."
 427                             "(max: %"PRIu32", need %"PRId64")\n",
 428                             initrd_max, (uint64_t)initrd_size);
 429                     exit(1);
 430                 }
 431
 432                 initrd_addr = (initrd_max - initrd_size) & ~4095;
 433
 434                 fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_ADDR, initrd_addr);
 435                 fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_SIZE, initrd_size);
 436                 fw_cfg_add_bytes(fw_cfg, FW_CFG_INITRD_DATA, initrd_data,
 437                                  initrd_size);
 438             }
 439
 440             option_rom[nb_option_roms].bootindex = 0;
 441             option_rom[nb_option_roms].name = "pvh.bin";
 442             nb_option_roms++;
 443
 444             return;
 445         }
 446         protocol = 0;
 447     }
 448
 449     if (protocol < 0x200 || !(header[0x211] & 0x01)) {
 450         /* Low kernel */
 451         real_addr    = 0x90000;
 452         cmdline_addr = 0x9a000 - cmdline_size;
 453         prot_addr    = 0x10000;
 454     } else if (protocol < 0x202) {
 455         /* High but ancient kernel */
 456         real_addr    = 0x90000;
 457         cmdline_addr = 0x9a000 - cmdline_size;
 458         prot_addr    = 0x100000;
 459     } else {
 460         /* High and recent kernel */
 461         real_addr    = 0x10000;
 462         cmdline_addr = 0x20000;
 463         prot_addr    = 0x100000;
 464     }
 465
 466     /* highest address for loading the initrd */
 467     if (protocol >= 0x20c &&
 468         lduw_p(header + 0x236) & XLF_CAN_BE_LOADED_ABOVE_4G) {
 469         /*
 470          * Linux has supported initrd up to 4 GB for a very long time (2007,
 471          * long before XLF_CAN_BE_LOADED_ABOVE_4G which was added in 2013),
 472          * though it only sets initrd_max to 2 GB to "work around bootloader
 473          * bugs". Luckily, QEMU firmware(which does something like bootloader)
 474          * has supported this.
 475          *
 476          * It's believed that if XLF_CAN_BE_LOADED_ABOVE_4G is set, initrd can
 477          * be loaded into any address.
 478          *
 479          * In addition, initrd_max is uint32_t simply because QEMU doesn't
 480          * support the 64-bit boot protocol (specifically the ext_ramdisk_image
 481          * field).
 482          *
 483          * Therefore here just limit initrd_max to UINT32_MAX simply as well.
 484          */
 485         initrd_max = UINT32_MAX;
 486     } else if (protocol >= 0x203) {
 487         initrd_max = ldl_p(header + 0x22c);
 488     } else {
 489         initrd_max = 0x37ffffff;
 490     }
 491
 492     if (initrd_max >= x86ms->below_4g_mem_size - acpi_data_size) {
 493         initrd_max = x86ms->below_4g_mem_size - acpi_data_size - 1;
 494     }
 495
 496     fw_cfg_add_i32(fw_cfg, FW_CFG_CMDLINE_ADDR, cmdline_addr);
 497     fw_cfg_add_i32(fw_cfg, FW_CFG_CMDLINE_SIZE, strlen(kernel_cmdline) + 1);
 498     fw_cfg_add_string(fw_cfg, FW_CFG_CMDLINE_DATA, kernel_cmdline);
 499
 500     if (protocol >= 0x202) {
 501         stl_p(header + 0x228, cmdline_addr);
 502     } else {
 503         stw_p(header + 0x20, 0xA33F);
 504         stw_p(header + 0x22, cmdline_addr - real_addr);
 505     }
 506
 507     /* handle vga= parameter */
 508     vmode = strstr(kernel_cmdline, "vga=");
 509     if (vmode) {
 510         unsigned int video_mode;
 511         int ret;
 512         /* skip "vga=" */
 513         vmode += 4;
 514         if (!strncmp(vmode, "normal", 6)) {
 515             video_mode = 0xffff;
 516         } else if (!strncmp(vmode, "ext", 3)) {
 517             video_mode = 0xfffe;
 518         } else if (!strncmp(vmode, "ask", 3)) {
 519             video_mode = 0xfffd;
 520         } else {
 521             ret = qemu_strtoui(vmode, NULL, 0, &video_mode);
 522             if (ret != 0) {
 523                 fprintf(stderr, "qemu: can't parse 'vga' parameter: %s\n",
 524                         strerror(-ret));
 525                 exit(1);
 526             }
 527         }
 528         stw_p(header + 0x1fa, video_mode);
 529     }
 530
 531     /* loader type */
 532     /*
 533      * High nybble = B reserved for QEMU; low nybble is revision number.
 534      * If this code is substantially changed, you may want to consider
 535      * incrementing the revision.
 536      */
 537     if (protocol >= 0x200) {
 538         header[0x210] = 0xB0;
 539     }
 540     /* heap */
 541     if (protocol >= 0x201) {
 542         header[0x211] |= 0x80; /* CAN_USE_HEAP */
 543         stw_p(header + 0x224, cmdline_addr - real_addr - 0x200);
 544     }
 545
 546     /* load initrd */
 547     if (initrd_filename) {
 548         GMappedFile *mapped_file;
 549         gsize initrd_size;
 550         gchar *initrd_data;
 551         GError *gerr = NULL;
 552
 553         if (protocol < 0x200) {
 554             fprintf(stderr, "qemu: linux kernel too old to load a ram disk\n");
 555             exit(1);
 556         }
 557
 558         mapped_file = g_mapped_file_new(initrd_filename, false, &gerr);
 559         if (!mapped_file) {
 560             fprintf(stderr, "qemu: error reading initrd %s: %s\n",
 561                     initrd_filename, gerr->message);
 562             exit(1);
 563         }
 564         x86ms->initrd_mapped_file = mapped_file;
 565
 566         initrd_data = g_mapped_file_get_contents(mapped_file);
 567         initrd_size = g_mapped_file_get_length(mapped_file);
 568         if (initrd_size >= initrd_max) {
 569             fprintf(stderr, "qemu: initrd is too large, cannot support."
 570                     "(max: %"PRIu32", need %"PRId64")\n",
 571                     initrd_max, (uint64_t)initrd_size);
 572             exit(1);
 573         }
 574
 575         initrd_addr = (initrd_max - initrd_size) & ~4095;
 576
 577         fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_ADDR, initrd_addr);
 578         fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_SIZE, initrd_size);
 579         fw_cfg_add_bytes(fw_cfg, FW_CFG_INITRD_DATA, initrd_data, initrd_size);
 580
 581         stl_p(header + 0x218, initrd_addr);
 582         stl_p(header + 0x21c, initrd_size);
 583     }
 584
 585     /* load kernel and setup */
 586     setup_size = header[0x1f1];
 587     if (setup_size == 0) {
 588         setup_size = 4;
 589     }
 590     setup_size = (setup_size + 1) * 512;
 591     if (setup_size > kernel_size) {
 592         fprintf(stderr, "qemu: invalid kernel header\n");
 593         exit(1);
 594     }
 595     kernel_size -= setup_size;
 596
 597     setup  = g_malloc(setup_size);
 598     kernel = g_malloc(kernel_size);
 599     fseek(f, 0, SEEK_SET);
 600     if (fread(setup, 1, setup_size, f) != setup_size) {
 601         fprintf(stderr, "fread() failed\n");
 602         exit(1);
 603     }
 604     if (fread(kernel, 1, kernel_size, f) != kernel_size) {
 605         fprintf(stderr, "fread() failed\n");
 606         exit(1);
 607     }
 608     fclose(f);
 609
 610     /* append dtb to kernel */
 611     if (dtb_filename) {
 612         if (protocol < 0x209) {
 613             fprintf(stderr, "qemu: Linux kernel too old to load a dtb\n");
 614             exit(1);
 615         }
 616
 617         dtb_size = get_image_size(dtb_filename);
 618         if (dtb_size <= 0) {
 619             fprintf(stderr, "qemu: error reading dtb %s: %s\n",
 620                     dtb_filename, strerror(errno));
 621             exit(1);
 622         }
 623
 624         setup_data_offset = QEMU_ALIGN_UP(kernel_size, 16);
 625         kernel_size = setup_data_offset + sizeof(struct setup_data) + dtb_size;
 626         kernel = g_realloc(kernel, kernel_size);
 627
 628         stq_p(header + 0x250, prot_addr + setup_data_offset);
 629
 630         setup_data = (struct setup_data *)(kernel + setup_data_offset);
 631         setup_data->next = 0;
 632         setup_data->type = cpu_to_le32(SETUP_DTB);
 633         setup_data->len = cpu_to_le32(dtb_size);
 634
 635         load_image_size(dtb_filename, setup_data->data, dtb_size);
 636     }
 637
 638     memcpy(setup, header, MIN(sizeof(header), setup_size));
 639
 640     fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_ADDR, prot_addr);
 641     fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_SIZE, kernel_size);
 642     fw_cfg_add_bytes(fw_cfg, FW_CFG_KERNEL_DATA, kernel, kernel_size);
 643
 644     fw_cfg_add_i32(fw_cfg, FW_CFG_SETUP_ADDR, real_addr);
 645     fw_cfg_add_i32(fw_cfg, FW_CFG_SETUP_SIZE, setup_size);
 646     fw_cfg_add_bytes(fw_cfg, FW_CFG_SETUP_DATA, setup, setup_size);
 647
 648     option_rom[nb_option_roms].bootindex = 0;
 649     option_rom[nb_option_roms].name = "linuxboot.bin";
 650     if (linuxboot_dma_enabled && fw_cfg_dma_enabled(fw_cfg)) {
 651         option_rom[nb_option_roms].name = "linuxboot_dma.bin";
 652     }
 653     nb_option_roms++;
 654 }
 655
 656 void x86_bios_rom_init(MemoryRegion *rom_memory, bool isapc_ram_fw)
 657 {
 658     char *filename;
 659     MemoryRegion *bios, *isa_bios;
 660     int bios_size, isa_bios_size;
 661     int ret;
 662
 663     /* BIOS load */
 664     if (bios_name == NULL) {
 665         bios_name = BIOS_FILENAME;
 666     }
 667     filename = qemu_find_file(QEMU_FILE_TYPE_BIOS, bios_name);
 668     if (filename) {
 669         bios_size = get_image_size(filename);
 670     } else {
 671         bios_size = -1;
 672     }
 673     if (bios_size <= 0 ||
 674         (bios_size % 65536) != 0) {
 675         goto bios_error;
 676     }
 677     bios = g_malloc(sizeof(*bios));
 678     memory_region_init_ram(bios, NULL, "pc.bios", bios_size, &error_fatal);
 679     if (!isapc_ram_fw) {
 680         memory_region_set_readonly(bios, true);
 681     }
 682     ret = rom_add_file_fixed(bios_name, (uint32_t)(-bios_size), -1);
 683     if (ret != 0) {
 684     bios_error:
 685         fprintf(stderr, "qemu: could not load PC BIOS '%s'\n", bios_name);
 686         exit(1);
 687     }
 688     g_free(filename);
 689
 690     /* map the last 128KB of the BIOS in ISA space */
 691     isa_bios_size = MIN(bios_size, 128 * KiB);
 692     isa_bios = g_malloc(sizeof(*isa_bios));
 693     memory_region_init_alias(isa_bios, NULL, "isa-bios", bios,
 694                              bios_size - isa_bios_size, isa_bios_size);
 695     memory_region_add_subregion_overlap(rom_memory,
 696                                         0x100000 - isa_bios_size,
 697                                         isa_bios,
 698                                         1);
 699     if (!isapc_ram_fw) {
 700         memory_region_set_readonly(isa_bios, true);
 701     }
 702
 703     /* map all the bios at the top of memory */
 704     memory_region_add_subregion(rom_memory,
 705                                 (uint32_t)(-bios_size),
 706                                 bios);
 707 }
 708
 709 static void x86_machine_get_max_ram_below_4g(Object *obj, Visitor *v,
 710                                              const char *name, void *opaque,
 711                                              Error **errp)
 712 {
 713     X86MachineState *x86ms = X86_MACHINE(obj);
 714     uint64_t value = x86ms->max_ram_below_4g;
 715
 716     visit_type_size(v, name, &value, errp);
 717 }
 718
 719 static void x86_machine_set_max_ram_below_4g(Object *obj, Visitor *v,
 720                                              const char *name, void *opaque,
 721                                              Error **errp)
 722 {
 723     X86MachineState *x86ms = X86_MACHINE(obj);
 724     Error *error = NULL;
 725     uint64_t value;
 726
 727     visit_type_size(v, name, &value, &error);
 728     if (error) {
 729         error_propagate(errp, error);
 730         return;
 731     }
 732     if (value > 4 * GiB) {
 733         error_setg(&error,
 734                    "Machine option 'max-ram-below-4g=%"PRIu64
 735                    "' expects size less than or equal to 4G", value);
 736         error_propagate(errp, error);
 737         return;
 738     }
 739
 740     if (value < 1 * MiB) {
 741         warn_report("Only %" PRIu64 " bytes of RAM below the 4GiB boundary,"
 742                     "BIOS may not work with less than 1MiB", value);
 743     }
 744
 745     x86ms->max_ram_below_4g = value;
 746 }
 747
 748 static void x86_machine_initfn(Object *obj)
 749 {
 750     X86MachineState *x86ms = X86_MACHINE(obj);
 751
 752     x86ms->max_ram_below_4g = 0; /* use default */
 753     x86ms->smp_dies = 1;
 754 }
 755
 756 static void x86_machine_class_init(ObjectClass *oc, void *data)
 757 {
 758     MachineClass *mc = MACHINE_CLASS(oc);
 759     X86MachineClass *x86mc = X86_MACHINE_CLASS(oc);
 760     NMIClass *nc = NMI_CLASS(oc);
 761
 762     mc->cpu_index_to_instance_props = x86_cpu_index_to_props;
 763     mc->get_default_cpu_node_id = x86_get_default_cpu_node_id;
 764     mc->possible_cpu_arch_ids = x86_possible_cpu_arch_ids;
 765     x86mc->compat_apic_id_mode = false;
 766     nc->nmi_monitor_handler = x86_nmi;
 767
 768     object_class_property_add(oc, X86_MACHINE_MAX_RAM_BELOW_4G, "size",
 769         x86_machine_get_max_ram_below_4g, x86_machine_set_max_ram_below_4g,
 770         NULL, NULL, &error_abort);
 771
 772     object_class_property_set_description(oc, X86_MACHINE_MAX_RAM_BELOW_4G,
 773         "Maximum ram below the 4G boundary (32bit boundary)", &error_abort);
 774 }
 775
 776 static const TypeInfo x86_machine_info = {
 777     .name = TYPE_X86_MACHINE,
 778     .parent = TYPE_MACHINE,
 779     .abstract = true,
 780     .instance_size = sizeof(X86MachineState),
 781     .instance_init = x86_machine_initfn,
 782     .class_size = sizeof(X86MachineClass),
 783     .class_init = x86_machine_class_init,
 784     .interfaces = (InterfaceInfo[]) {
 785          { TYPE_NMI },
 786          { }
 787     },
 788 };
 789
 790 static void x86_machine_register_types(void)
 791 {
 792     type_register_static(&x86_machine_info);
 793 }
 794
 795 type_init(x86_machine_register_types)