qemu/linux-user/main.c
<<
>>
Prefs
   1/*
   2 *  qemu user main
   3 *
   4 *  Copyright (c) 2003-2008 Fabrice Bellard
   5 *
   6 *  This program is free software; you can redistribute it and/or modify
   7 *  it under the terms of the GNU General Public License as published by
   8 *  the Free Software Foundation; either version 2 of the License, or
   9 *  (at your option) any later version.
  10 *
  11 *  This program is distributed in the hope that it will be useful,
  12 *  but WITHOUT ANY WARRANTY; without even the implied warranty of
  13 *  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  14 *  GNU General Public License for more details.
  15 *
  16 *  You should have received a copy of the GNU General Public License
  17 *  along with this program; if not, see <http://www.gnu.org/licenses/>.
  18 */
  19
  20#include "qemu/osdep.h"
  21#include "qemu/help-texts.h"
  22#include "qemu/units.h"
  23#include "qemu/accel.h"
  24#include "qemu-version.h"
  25#include <sys/syscall.h>
  26#include <sys/resource.h>
  27#include <sys/shm.h>
  28#include <linux/binfmts.h>
  29
  30#include "qapi/error.h"
  31#include "qemu.h"
  32#include "user-internals.h"
  33#include "qemu/path.h"
  34#include "qemu/queue.h"
  35#include "qemu/config-file.h"
  36#include "qemu/cutils.h"
  37#include "qemu/error-report.h"
  38#include "qemu/help_option.h"
  39#include "qemu/module.h"
  40#include "qemu/plugin.h"
  41#include "exec/exec-all.h"
  42#include "exec/gdbstub.h"
  43#include "gdbstub/user.h"
  44#include "tcg/tcg.h"
  45#include "qemu/timer.h"
  46#include "qemu/envlist.h"
  47#include "qemu/guest-random.h"
  48#include "elf.h"
  49#include "trace/control.h"
  50#include "target_elf.h"
  51#include "cpu_loop-common.h"
  52#include "crypto/init.h"
  53#include "fd-trans.h"
  54#include "signal-common.h"
  55#include "loader.h"
  56#include "user-mmap.h"
  57#include "accel/tcg/perf.h"
  58
  59#ifdef CONFIG_SEMIHOSTING
  60#include "semihosting/semihost.h"
  61#endif
  62
  63#ifndef AT_FLAGS_PRESERVE_ARGV0
  64#define AT_FLAGS_PRESERVE_ARGV0_BIT 0
  65#define AT_FLAGS_PRESERVE_ARGV0 (1 << AT_FLAGS_PRESERVE_ARGV0_BIT)
  66#endif
  67
  68char *exec_path;
  69char real_exec_path[PATH_MAX];
  70
  71int singlestep;
  72static const char *argv0;
  73static const char *gdbstub;
  74static envlist_t *envlist;
  75static const char *cpu_model;
  76static const char *cpu_type;
  77static const char *seed_optarg;
  78unsigned long mmap_min_addr;
  79uintptr_t guest_base;
  80bool have_guest_base;
  81
  82/*
  83 * Used to implement backwards-compatibility for the `-strace`, and
  84 * QEMU_STRACE options. Without this, the QEMU_LOG can be overwritten by
  85 * -strace, or vice versa.
  86 */
  87static bool enable_strace;
  88
  89/*
  90 * The last log mask given by the user in an environment variable or argument.
  91 * Used to support command line arguments overriding environment variables.
  92 */
  93static int last_log_mask;
  94static const char *last_log_filename;
  95
  96/*
  97 * When running 32-on-64 we should make sure we can fit all of the possible
  98 * guest address space into a contiguous chunk of virtual host memory.
  99 *
 100 * This way we will never overlap with our own libraries or binaries or stack
 101 * or anything else that QEMU maps.
 102 *
 103 * Many cpus reserve the high bit (or more than one for some 64-bit cpus)
 104 * of the address for the kernel.  Some cpus rely on this and user space
 105 * uses the high bit(s) for pointer tagging and the like.  For them, we
 106 * must preserve the expected address space.
 107 */
 108#ifndef MAX_RESERVED_VA
 109# if HOST_LONG_BITS > TARGET_VIRT_ADDR_SPACE_BITS
 110#  if TARGET_VIRT_ADDR_SPACE_BITS == 32 && \
 111      (TARGET_LONG_BITS == 32 || defined(TARGET_ABI32))
 112#   define MAX_RESERVED_VA(CPU)  0xfffffffful
 113#  else
 114#   define MAX_RESERVED_VA(CPU)  ((1ul << TARGET_VIRT_ADDR_SPACE_BITS) - 1)
 115#  endif
 116# else
 117#  define MAX_RESERVED_VA(CPU)  0
 118# endif
 119#endif
 120
 121unsigned long reserved_va;
 122
 123static void usage(int exitcode);
 124
 125static const char *interp_prefix = CONFIG_QEMU_INTERP_PREFIX;
 126const char *qemu_uname_release;
 127
 128#if !defined(TARGET_DEFAULT_STACK_SIZE)
 129/* XXX: on x86 MAP_GROWSDOWN only works if ESP <= address + 32, so
 130   we allocate a bigger stack. Need a better solution, for example
 131   by remapping the process stack directly at the right place */
 132#define TARGET_DEFAULT_STACK_SIZE       8 * 1024 * 1024UL
 133#endif
 134
 135unsigned long guest_stack_size = TARGET_DEFAULT_STACK_SIZE;
 136
 137/***********************************************************/
 138/* Helper routines for implementing atomic operations.  */
 139
 140/* Make sure everything is in a consistent state for calling fork().  */
 141void fork_start(void)
 142{
 143    start_exclusive();
 144    mmap_fork_start();
 145    cpu_list_lock();
 146    qemu_plugin_user_prefork_lock();
 147}
 148
 149void fork_end(int child)
 150{
 151    qemu_plugin_user_postfork(child);
 152    mmap_fork_end(child);
 153    if (child) {
 154        CPUState *cpu, *next_cpu;
 155        /* Child processes created by fork() only have a single thread.
 156           Discard information about the parent threads.  */
 157        CPU_FOREACH_SAFE(cpu, next_cpu) {
 158            if (cpu != thread_cpu) {
 159                QTAILQ_REMOVE_RCU(&cpus, cpu, node);
 160            }
 161        }
 162        qemu_init_cpu_list();
 163        gdbserver_fork(thread_cpu);
 164    } else {
 165        cpu_list_unlock();
 166    }
 167    /*
 168     * qemu_init_cpu_list() reinitialized the child exclusive state, but we
 169     * also need to keep current_cpu consistent, so call end_exclusive() for
 170     * both child and parent.
 171     */
 172    end_exclusive();
 173}
 174
 175__thread CPUState *thread_cpu;
 176
 177bool qemu_cpu_is_self(CPUState *cpu)
 178{
 179    return thread_cpu == cpu;
 180}
 181
 182void qemu_cpu_kick(CPUState *cpu)
 183{
 184    cpu_exit(cpu);
 185}
 186
 187void task_settid(TaskState *ts)
 188{
 189    if (ts->ts_tid == 0) {
 190        ts->ts_tid = (pid_t)syscall(SYS_gettid);
 191    }
 192}
 193
 194void stop_all_tasks(void)
 195{
 196    /*
 197     * We trust that when using NPTL, start_exclusive()
 198     * handles thread stopping correctly.
 199     */
 200    start_exclusive();
 201}
 202
 203/* Assumes contents are already zeroed.  */
 204void init_task_state(TaskState *ts)
 205{
 206    long ticks_per_sec;
 207    struct timespec bt;
 208
 209    ts->used = 1;
 210    ts->sigaltstack_used = (struct target_sigaltstack) {
 211        .ss_sp = 0,
 212        .ss_size = 0,
 213        .ss_flags = TARGET_SS_DISABLE,
 214    };
 215
 216    /* Capture task start time relative to system boot */
 217
 218    ticks_per_sec = sysconf(_SC_CLK_TCK);
 219
 220    if ((ticks_per_sec > 0) && !clock_gettime(CLOCK_BOOTTIME, &bt)) {
 221        /* start_boottime is expressed in clock ticks */
 222        ts->start_boottime = bt.tv_sec * (uint64_t) ticks_per_sec;
 223        ts->start_boottime += bt.tv_nsec * (uint64_t) ticks_per_sec /
 224                              NANOSECONDS_PER_SECOND;
 225    }
 226}
 227
 228CPUArchState *cpu_copy(CPUArchState *env)
 229{
 230    CPUState *cpu = env_cpu(env);
 231    CPUState *new_cpu = cpu_create(cpu_type);
 232    CPUArchState *new_env = new_cpu->env_ptr;
 233    CPUBreakpoint *bp;
 234
 235    /* Reset non arch specific state */
 236    cpu_reset(new_cpu);
 237
 238    new_cpu->tcg_cflags = cpu->tcg_cflags;
 239    memcpy(new_env, env, sizeof(CPUArchState));
 240#if defined(TARGET_I386) || defined(TARGET_X86_64)
 241    new_env->gdt.base = target_mmap(0, sizeof(uint64_t) * TARGET_GDT_ENTRIES,
 242                                    PROT_READ | PROT_WRITE,
 243                                    MAP_ANONYMOUS | MAP_PRIVATE, -1, 0);
 244    memcpy(g2h_untagged(new_env->gdt.base), g2h_untagged(env->gdt.base),
 245           sizeof(uint64_t) * TARGET_GDT_ENTRIES);
 246    OBJECT(new_cpu)->free = OBJECT(cpu)->free;
 247#endif
 248
 249    /* Clone all break/watchpoints.
 250       Note: Once we support ptrace with hw-debug register access, make sure
 251       BP_CPU break/watchpoints are handled correctly on clone. */
 252    QTAILQ_INIT(&new_cpu->breakpoints);
 253    QTAILQ_FOREACH(bp, &cpu->breakpoints, entry) {
 254        cpu_breakpoint_insert(new_cpu, bp->pc, bp->flags, NULL);
 255    }
 256
 257    return new_env;
 258}
 259
 260static void handle_arg_help(const char *arg)
 261{
 262    usage(EXIT_SUCCESS);
 263}
 264
 265static void handle_arg_log(const char *arg)
 266{
 267    last_log_mask = qemu_str_to_log_mask(arg);
 268    if (!last_log_mask) {
 269        qemu_print_log_usage(stdout);
 270        exit(EXIT_FAILURE);
 271    }
 272}
 273
 274static void handle_arg_dfilter(const char *arg)
 275{
 276    qemu_set_dfilter_ranges(arg, &error_fatal);
 277}
 278
 279static void handle_arg_log_filename(const char *arg)
 280{
 281    last_log_filename = arg;
 282}
 283
 284static void handle_arg_set_env(const char *arg)
 285{
 286    char *r, *p, *token;
 287    r = p = strdup(arg);
 288    while ((token = strsep(&p, ",")) != NULL) {
 289        if (envlist_setenv(envlist, token) != 0) {
 290            usage(EXIT_FAILURE);
 291        }
 292    }
 293    free(r);
 294}
 295
 296static void handle_arg_unset_env(const char *arg)
 297{
 298    char *r, *p, *token;
 299    r = p = strdup(arg);
 300    while ((token = strsep(&p, ",")) != NULL) {
 301        if (envlist_unsetenv(envlist, token) != 0) {
 302            usage(EXIT_FAILURE);
 303        }
 304    }
 305    free(r);
 306}
 307
 308static void handle_arg_argv0(const char *arg)
 309{
 310    argv0 = strdup(arg);
 311}
 312
 313static void handle_arg_stack_size(const char *arg)
 314{
 315    char *p;
 316    guest_stack_size = strtoul(arg, &p, 0);
 317    if (guest_stack_size == 0) {
 318        usage(EXIT_FAILURE);
 319    }
 320
 321    if (*p == 'M') {
 322        guest_stack_size *= MiB;
 323    } else if (*p == 'k' || *p == 'K') {
 324        guest_stack_size *= KiB;
 325    }
 326}
 327
 328static void handle_arg_ld_prefix(const char *arg)
 329{
 330    interp_prefix = strdup(arg);
 331}
 332
 333static void handle_arg_pagesize(const char *arg)
 334{
 335    qemu_host_page_size = atoi(arg);
 336    if (qemu_host_page_size == 0 ||
 337        (qemu_host_page_size & (qemu_host_page_size - 1)) != 0) {
 338        fprintf(stderr, "page size must be a power of two\n");
 339        exit(EXIT_FAILURE);
 340    }
 341}
 342
 343static void handle_arg_seed(const char *arg)
 344{
 345    seed_optarg = arg;
 346}
 347
 348static void handle_arg_gdb(const char *arg)
 349{
 350    gdbstub = g_strdup(arg);
 351}
 352
 353static void handle_arg_uname(const char *arg)
 354{
 355    qemu_uname_release = strdup(arg);
 356}
 357
 358static void handle_arg_cpu(const char *arg)
 359{
 360    cpu_model = strdup(arg);
 361    if (cpu_model == NULL || is_help_option(cpu_model)) {
 362        /* XXX: implement xxx_cpu_list for targets that still miss it */
 363#if defined(cpu_list)
 364        cpu_list();
 365#endif
 366        exit(EXIT_FAILURE);
 367    }
 368}
 369
 370static void handle_arg_guest_base(const char *arg)
 371{
 372    guest_base = strtol(arg, NULL, 0);
 373    have_guest_base = true;
 374}
 375
 376static void handle_arg_reserved_va(const char *arg)
 377{
 378    char *p;
 379    int shift = 0;
 380    unsigned long val;
 381
 382    val = strtoul(arg, &p, 0);
 383    switch (*p) {
 384    case 'k':
 385    case 'K':
 386        shift = 10;
 387        break;
 388    case 'M':
 389        shift = 20;
 390        break;
 391    case 'G':
 392        shift = 30;
 393        break;
 394    }
 395    if (shift) {
 396        unsigned long unshifted = val;
 397        p++;
 398        val <<= shift;
 399        if (val >> shift != unshifted) {
 400            fprintf(stderr, "Reserved virtual address too big\n");
 401            exit(EXIT_FAILURE);
 402        }
 403    }
 404    if (*p) {
 405        fprintf(stderr, "Unrecognised -R size suffix '%s'\n", p);
 406        exit(EXIT_FAILURE);
 407    }
 408    /* The representation is size - 1, with 0 remaining "default". */
 409    reserved_va = val ? val - 1 : 0;
 410}
 411
 412static void handle_arg_singlestep(const char *arg)
 413{
 414    singlestep = 1;
 415}
 416
 417static void handle_arg_strace(const char *arg)
 418{
 419    enable_strace = true;
 420}
 421
 422static void handle_arg_version(const char *arg)
 423{
 424    printf("qemu-" TARGET_NAME " version " QEMU_FULL_VERSION
 425           "\n" QEMU_COPYRIGHT "\n");
 426    exit(EXIT_SUCCESS);
 427}
 428
 429static void handle_arg_trace(const char *arg)
 430{
 431    trace_opt_parse(arg);
 432}
 433
 434#if defined(TARGET_XTENSA)
 435static void handle_arg_abi_call0(const char *arg)
 436{
 437    xtensa_set_abi_call0();
 438}
 439#endif
 440
 441static void handle_arg_perfmap(const char *arg)
 442{
 443    perf_enable_perfmap();
 444}
 445
 446static void handle_arg_jitdump(const char *arg)
 447{
 448    perf_enable_jitdump();
 449}
 450
 451static QemuPluginList plugins = QTAILQ_HEAD_INITIALIZER(plugins);
 452
 453#ifdef CONFIG_PLUGIN
 454static void handle_arg_plugin(const char *arg)
 455{
 456    qemu_plugin_opt_parse(arg, &plugins);
 457}
 458#endif
 459
 460struct qemu_argument {
 461    const char *argv;
 462    const char *env;
 463    bool has_arg;
 464    void (*handle_opt)(const char *arg);
 465    const char *example;
 466    const char *help;
 467};
 468
 469static const struct qemu_argument arg_table[] = {
 470    {"h",          "",                 false, handle_arg_help,
 471     "",           "print this help"},
 472    {"help",       "",                 false, handle_arg_help,
 473     "",           ""},
 474    {"g",          "QEMU_GDB",         true,  handle_arg_gdb,
 475     "port",       "wait gdb connection to 'port'"},
 476    {"L",          "QEMU_LD_PREFIX",   true,  handle_arg_ld_prefix,
 477     "path",       "set the elf interpreter prefix to 'path'"},
 478    {"s",          "QEMU_STACK_SIZE",  true,  handle_arg_stack_size,
 479     "size",       "set the stack size to 'size' bytes"},
 480    {"cpu",        "QEMU_CPU",         true,  handle_arg_cpu,
 481     "model",      "select CPU (-cpu help for list)"},
 482    {"E",          "QEMU_SET_ENV",     true,  handle_arg_set_env,
 483     "var=value",  "sets targets environment variable (see below)"},
 484    {"U",          "QEMU_UNSET_ENV",   true,  handle_arg_unset_env,
 485     "var",        "unsets targets environment variable (see below)"},
 486    {"0",          "QEMU_ARGV0",       true,  handle_arg_argv0,
 487     "argv0",      "forces target process argv[0] to be 'argv0'"},
 488    {"r",          "QEMU_UNAME",       true,  handle_arg_uname,
 489     "uname",      "set qemu uname release string to 'uname'"},
 490    {"B",          "QEMU_GUEST_BASE",  true,  handle_arg_guest_base,
 491     "address",    "set guest_base address to 'address'"},
 492    {"R",          "QEMU_RESERVED_VA", true,  handle_arg_reserved_va,
 493     "size",       "reserve 'size' bytes for guest virtual address space"},
 494    {"d",          "QEMU_LOG",         true,  handle_arg_log,
 495     "item[,...]", "enable logging of specified items "
 496     "(use '-d help' for a list of items)"},
 497    {"dfilter",    "QEMU_DFILTER",     true,  handle_arg_dfilter,
 498     "range[,...]","filter logging based on address range"},
 499    {"D",          "QEMU_LOG_FILENAME", true, handle_arg_log_filename,
 500     "logfile",     "write logs to 'logfile' (default stderr)"},
 501    {"p",          "QEMU_PAGESIZE",    true,  handle_arg_pagesize,
 502     "pagesize",   "set the host page size to 'pagesize'"},
 503    {"singlestep", "QEMU_SINGLESTEP",  false, handle_arg_singlestep,
 504     "",           "run in singlestep mode"},
 505    {"strace",     "QEMU_STRACE",      false, handle_arg_strace,
 506     "",           "log system calls"},
 507    {"seed",       "QEMU_RAND_SEED",   true,  handle_arg_seed,
 508     "",           "Seed for pseudo-random number generator"},
 509    {"trace",      "QEMU_TRACE",       true,  handle_arg_trace,
 510     "",           "[[enable=]<pattern>][,events=<file>][,file=<file>]"},
 511#ifdef CONFIG_PLUGIN
 512    {"plugin",     "QEMU_PLUGIN",      true,  handle_arg_plugin,
 513     "",           "[file=]<file>[,<argname>=<argvalue>]"},
 514#endif
 515    {"version",    "QEMU_VERSION",     false, handle_arg_version,
 516     "",           "display version information and exit"},
 517#if defined(TARGET_XTENSA)
 518    {"xtensa-abi-call0", "QEMU_XTENSA_ABI_CALL0", false, handle_arg_abi_call0,
 519     "",           "assume CALL0 Xtensa ABI"},
 520#endif
 521    {"perfmap",    "QEMU_PERFMAP",     false, handle_arg_perfmap,
 522     "",           "Generate a /tmp/perf-${pid}.map file for perf"},
 523    {"jitdump",    "QEMU_JITDUMP",     false, handle_arg_jitdump,
 524     "",           "Generate a jit-${pid}.dump file for perf"},
 525    {NULL, NULL, false, NULL, NULL, NULL}
 526};
 527
 528static void usage(int exitcode)
 529{
 530    const struct qemu_argument *arginfo;
 531    int maxarglen;
 532    int maxenvlen;
 533
 534    printf("usage: qemu-" TARGET_NAME " [options] program [arguments...]\n"
 535           "Linux CPU emulator (compiled for " TARGET_NAME " emulation)\n"
 536           "\n"
 537           "Options and associated environment variables:\n"
 538           "\n");
 539
 540    /* Calculate column widths. We must always have at least enough space
 541     * for the column header.
 542     */
 543    maxarglen = strlen("Argument");
 544    maxenvlen = strlen("Env-variable");
 545
 546    for (arginfo = arg_table; arginfo->handle_opt != NULL; arginfo++) {
 547        int arglen = strlen(arginfo->argv);
 548        if (arginfo->has_arg) {
 549            arglen += strlen(arginfo->example) + 1;
 550        }
 551        if (strlen(arginfo->env) > maxenvlen) {
 552            maxenvlen = strlen(arginfo->env);
 553        }
 554        if (arglen > maxarglen) {
 555            maxarglen = arglen;
 556        }
 557    }
 558
 559    printf("%-*s %-*s Description\n", maxarglen+1, "Argument",
 560            maxenvlen, "Env-variable");
 561
 562    for (arginfo = arg_table; arginfo->handle_opt != NULL; arginfo++) {
 563        if (arginfo->has_arg) {
 564            printf("-%s %-*s %-*s %s\n", arginfo->argv,
 565                   (int)(maxarglen - strlen(arginfo->argv) - 1),
 566                   arginfo->example, maxenvlen, arginfo->env, arginfo->help);
 567        } else {
 568            printf("-%-*s %-*s %s\n", maxarglen, arginfo->argv,
 569                    maxenvlen, arginfo->env,
 570                    arginfo->help);
 571        }
 572    }
 573
 574    printf("\n"
 575           "Defaults:\n"
 576           "QEMU_LD_PREFIX  = %s\n"
 577           "QEMU_STACK_SIZE = %ld byte\n",
 578           interp_prefix,
 579           guest_stack_size);
 580
 581    printf("\n"
 582           "You can use -E and -U options or the QEMU_SET_ENV and\n"
 583           "QEMU_UNSET_ENV environment variables to set and unset\n"
 584           "environment variables for the target process.\n"
 585           "It is possible to provide several variables by separating them\n"
 586           "by commas in getsubopt(3) style. Additionally it is possible to\n"
 587           "provide the -E and -U options multiple times.\n"
 588           "The following lines are equivalent:\n"
 589           "    -E var1=val2 -E var2=val2 -U LD_PRELOAD -U LD_DEBUG\n"
 590           "    -E var1=val2,var2=val2 -U LD_PRELOAD,LD_DEBUG\n"
 591           "    QEMU_SET_ENV=var1=val2,var2=val2 QEMU_UNSET_ENV=LD_PRELOAD,LD_DEBUG\n"
 592           "Note that if you provide several changes to a single variable\n"
 593           "the last change will stay in effect.\n"
 594           "\n"
 595           QEMU_HELP_BOTTOM "\n");
 596
 597    exit(exitcode);
 598}
 599
 600static int parse_args(int argc, char **argv)
 601{
 602    const char *r;
 603    int optind;
 604    const struct qemu_argument *arginfo;
 605
 606    for (arginfo = arg_table; arginfo->handle_opt != NULL; arginfo++) {
 607        if (arginfo->env == NULL) {
 608            continue;
 609        }
 610
 611        r = getenv(arginfo->env);
 612        if (r != NULL) {
 613            arginfo->handle_opt(r);
 614        }
 615    }
 616
 617    optind = 1;
 618    for (;;) {
 619        if (optind >= argc) {
 620            break;
 621        }
 622        r = argv[optind];
 623        if (r[0] != '-') {
 624            break;
 625        }
 626        optind++;
 627        r++;
 628        if (!strcmp(r, "-")) {
 629            break;
 630        }
 631        /* Treat --foo the same as -foo.  */
 632        if (r[0] == '-') {
 633            r++;
 634        }
 635
 636        for (arginfo = arg_table; arginfo->handle_opt != NULL; arginfo++) {
 637            if (!strcmp(r, arginfo->argv)) {
 638                if (arginfo->has_arg) {
 639                    if (optind >= argc) {
 640                        (void) fprintf(stderr,
 641                            "qemu: missing argument for option '%s'\n", r);
 642                        exit(EXIT_FAILURE);
 643                    }
 644                    arginfo->handle_opt(argv[optind]);
 645                    optind++;
 646                } else {
 647                    arginfo->handle_opt(NULL);
 648                }
 649                break;
 650            }
 651        }
 652
 653        /* no option matched the current argv */
 654        if (arginfo->handle_opt == NULL) {
 655            (void) fprintf(stderr, "qemu: unknown option '%s'\n", r);
 656            exit(EXIT_FAILURE);
 657        }
 658    }
 659
 660    if (optind >= argc) {
 661        (void) fprintf(stderr, "qemu: no user program specified\n");
 662        exit(EXIT_FAILURE);
 663    }
 664
 665    exec_path = argv[optind];
 666
 667    return optind;
 668}
 669
 670int main(int argc, char **argv, char **envp)
 671{
 672    struct target_pt_regs regs1, *regs = &regs1;
 673    struct image_info info1, *info = &info1;
 674    struct linux_binprm bprm;
 675    TaskState *ts;
 676    CPUArchState *env;
 677    CPUState *cpu;
 678    int optind;
 679    char **target_environ, **wrk;
 680    char **target_argv;
 681    int target_argc;
 682    int i;
 683    int ret;
 684    int execfd;
 685    unsigned long max_reserved_va;
 686    bool preserve_argv0;
 687
 688    error_init(argv[0]);
 689    module_call_init(MODULE_INIT_TRACE);
 690    qemu_init_cpu_list();
 691    module_call_init(MODULE_INIT_QOM);
 692
 693    envlist = envlist_create();
 694
 695    /* add current environment into the list */
 696    for (wrk = environ; *wrk != NULL; wrk++) {
 697        (void) envlist_setenv(envlist, *wrk);
 698    }
 699
 700    /* Read the stack limit from the kernel.  If it's "unlimited",
 701       then we can do little else besides use the default.  */
 702    {
 703        struct rlimit lim;
 704        if (getrlimit(RLIMIT_STACK, &lim) == 0
 705            && lim.rlim_cur != RLIM_INFINITY
 706            && lim.rlim_cur == (target_long)lim.rlim_cur
 707            && lim.rlim_cur > guest_stack_size) {
 708            guest_stack_size = lim.rlim_cur;
 709        }
 710    }
 711
 712    cpu_model = NULL;
 713
 714    qemu_add_opts(&qemu_trace_opts);
 715    qemu_plugin_add_opts();
 716
 717    optind = parse_args(argc, argv);
 718
 719    qemu_set_log_filename_flags(last_log_filename,
 720                                last_log_mask | (enable_strace * LOG_STRACE),
 721                                &error_fatal);
 722
 723    if (!trace_init_backends()) {
 724        exit(1);
 725    }
 726    trace_init_file();
 727    qemu_plugin_load_list(&plugins, &error_fatal);
 728
 729    /* Zero out regs */
 730    memset(regs, 0, sizeof(struct target_pt_regs));
 731
 732    /* Zero out image_info */
 733    memset(info, 0, sizeof(struct image_info));
 734
 735    memset(&bprm, 0, sizeof (bprm));
 736
 737    /* Scan interp_prefix dir for replacement files. */
 738    init_paths(interp_prefix);
 739
 740    init_qemu_uname_release();
 741
 742    /*
 743     * Manage binfmt-misc open-binary flag
 744     */
 745    execfd = qemu_getauxval(AT_EXECFD);
 746    if (execfd == 0) {
 747        execfd = open(exec_path, O_RDONLY);
 748        if (execfd < 0) {
 749            printf("Error while loading %s: %s\n", exec_path, strerror(errno));
 750            _exit(EXIT_FAILURE);
 751        }
 752    }
 753
 754    /* Resolve executable file name to full path name */
 755    if (realpath(exec_path, real_exec_path)) {
 756        exec_path = real_exec_path;
 757    }
 758
 759    /*
 760     * get binfmt_misc flags
 761     */
 762    preserve_argv0 = !!(qemu_getauxval(AT_FLAGS) & AT_FLAGS_PRESERVE_ARGV0);
 763
 764    /*
 765     * Manage binfmt-misc preserve-arg[0] flag
 766     *    argv[optind]     full path to the binary
 767     *    argv[optind + 1] original argv[0]
 768     */
 769    if (optind + 1 < argc && preserve_argv0) {
 770        optind++;
 771    }
 772
 773    if (cpu_model == NULL) {
 774        cpu_model = cpu_get_model(get_elf_eflags(execfd));
 775    }
 776    cpu_type = parse_cpu_option(cpu_model);
 777
 778    /* init tcg before creating CPUs and to get qemu_host_page_size */
 779    {
 780        AccelClass *ac = ACCEL_GET_CLASS(current_accel());
 781
 782        accel_init_interfaces(ac);
 783        ac->init_machine(NULL);
 784    }
 785    cpu = cpu_create(cpu_type);
 786    env = cpu->env_ptr;
 787    cpu_reset(cpu);
 788    thread_cpu = cpu;
 789
 790    /*
 791     * Reserving too much vm space via mmap can run into problems
 792     * with rlimits, oom due to page table creation, etc.  We will
 793     * still try it, if directed by the command-line option, but
 794     * not by default.
 795     */
 796    max_reserved_va = MAX_RESERVED_VA(cpu);
 797    if (reserved_va != 0) {
 798        if ((reserved_va + 1) % qemu_host_page_size) {
 799            char *s = size_to_str(qemu_host_page_size);
 800            fprintf(stderr, "Reserved virtual address not aligned mod %s\n", s);
 801            g_free(s);
 802            exit(EXIT_FAILURE);
 803        }
 804        if (max_reserved_va && reserved_va > max_reserved_va) {
 805            fprintf(stderr, "Reserved virtual address too big\n");
 806            exit(EXIT_FAILURE);
 807        }
 808    } else if (HOST_LONG_BITS == 64 && TARGET_VIRT_ADDR_SPACE_BITS <= 32) {
 809        /* MAX_RESERVED_VA + 1 is a large power of 2, so is aligned. */
 810        reserved_va = max_reserved_va;
 811    }
 812
 813    {
 814        Error *err = NULL;
 815        if (seed_optarg != NULL) {
 816            qemu_guest_random_seed_main(seed_optarg, &err);
 817        } else {
 818            qcrypto_init(&err);
 819        }
 820        if (err) {
 821            error_reportf_err(err, "cannot initialize crypto: ");
 822            exit(1);
 823        }
 824    }
 825
 826    target_environ = envlist_to_environ(envlist, NULL);
 827    envlist_free(envlist);
 828
 829    /*
 830     * Read in mmap_min_addr kernel parameter.  This value is used
 831     * When loading the ELF image to determine whether guest_base
 832     * is needed.  It is also used in mmap_find_vma.
 833     */
 834    {
 835        FILE *fp;
 836
 837        if ((fp = fopen("/proc/sys/vm/mmap_min_addr", "r")) != NULL) {
 838            unsigned long tmp;
 839            if (fscanf(fp, "%lu", &tmp) == 1 && tmp != 0) {
 840                mmap_min_addr = tmp;
 841                qemu_log_mask(CPU_LOG_PAGE, "host mmap_min_addr=0x%lx\n",
 842                              mmap_min_addr);
 843            }
 844            fclose(fp);
 845        }
 846    }
 847
 848    /*
 849     * We prefer to not make NULL pointers accessible to QEMU.
 850     * If we're in a chroot with no /proc, fall back to 1 page.
 851     */
 852    if (mmap_min_addr == 0) {
 853        mmap_min_addr = qemu_host_page_size;
 854        qemu_log_mask(CPU_LOG_PAGE,
 855                      "host mmap_min_addr=0x%lx (fallback)\n",
 856                      mmap_min_addr);
 857    }
 858
 859    /*
 860     * Prepare copy of argv vector for target.
 861     */
 862    target_argc = argc - optind;
 863    target_argv = calloc(target_argc + 1, sizeof (char *));
 864    if (target_argv == NULL) {
 865        (void) fprintf(stderr, "Unable to allocate memory for target_argv\n");
 866        exit(EXIT_FAILURE);
 867    }
 868
 869    /*
 870     * If argv0 is specified (using '-0' switch) we replace
 871     * argv[0] pointer with the given one.
 872     */
 873    i = 0;
 874    if (argv0 != NULL) {
 875        target_argv[i++] = strdup(argv0);
 876    }
 877    for (; i < target_argc; i++) {
 878        target_argv[i] = strdup(argv[optind + i]);
 879    }
 880    target_argv[target_argc] = NULL;
 881
 882    ts = g_new0(TaskState, 1);
 883    init_task_state(ts);
 884    /* build Task State */
 885    ts->info = info;
 886    ts->bprm = &bprm;
 887    cpu->opaque = ts;
 888    task_settid(ts);
 889
 890    fd_trans_init();
 891
 892    ret = loader_exec(execfd, exec_path, target_argv, target_environ, regs,
 893        info, &bprm);
 894    if (ret != 0) {
 895        printf("Error while loading %s: %s\n", exec_path, strerror(-ret));
 896        _exit(EXIT_FAILURE);
 897    }
 898
 899    for (wrk = target_environ; *wrk; wrk++) {
 900        g_free(*wrk);
 901    }
 902
 903    g_free(target_environ);
 904
 905    if (qemu_loglevel_mask(CPU_LOG_PAGE)) {
 906        FILE *f = qemu_log_trylock();
 907        if (f) {
 908            fprintf(f, "guest_base  %p\n", (void *)guest_base);
 909            fprintf(f, "page layout changed following binary load\n");
 910            page_dump(f);
 911
 912            fprintf(f, "start_brk   0x" TARGET_ABI_FMT_lx "\n",
 913                    info->start_brk);
 914            fprintf(f, "end_code    0x" TARGET_ABI_FMT_lx "\n",
 915                    info->end_code);
 916            fprintf(f, "start_code  0x" TARGET_ABI_FMT_lx "\n",
 917                    info->start_code);
 918            fprintf(f, "start_data  0x" TARGET_ABI_FMT_lx "\n",
 919                    info->start_data);
 920            fprintf(f, "end_data    0x" TARGET_ABI_FMT_lx "\n",
 921                    info->end_data);
 922            fprintf(f, "start_stack 0x" TARGET_ABI_FMT_lx "\n",
 923                    info->start_stack);
 924            fprintf(f, "brk         0x" TARGET_ABI_FMT_lx "\n",
 925                    info->brk);
 926            fprintf(f, "entry       0x" TARGET_ABI_FMT_lx "\n",
 927                    info->entry);
 928            fprintf(f, "argv_start  0x" TARGET_ABI_FMT_lx "\n",
 929                    info->argv);
 930            fprintf(f, "env_start   0x" TARGET_ABI_FMT_lx "\n",
 931                    info->envp);
 932            fprintf(f, "auxv_start  0x" TARGET_ABI_FMT_lx "\n",
 933                    info->saved_auxv);
 934            qemu_log_unlock(f);
 935        }
 936    }
 937
 938    target_set_brk(info->brk);
 939    syscall_init();
 940    signal_init();
 941
 942    /* Now that we've loaded the binary, GUEST_BASE is fixed.  Delay
 943       generating the prologue until now so that the prologue can take
 944       the real value of GUEST_BASE into account.  */
 945    tcg_prologue_init(tcg_ctx);
 946
 947    target_cpu_copy_regs(env, regs);
 948
 949    if (gdbstub) {
 950        if (gdbserver_start(gdbstub) < 0) {
 951            fprintf(stderr, "qemu: could not open gdbserver on %s\n",
 952                    gdbstub);
 953            exit(EXIT_FAILURE);
 954        }
 955        gdb_handlesig(cpu, 0);
 956    }
 957
 958#ifdef CONFIG_SEMIHOSTING
 959    qemu_semihosting_guestfd_init();
 960#endif
 961
 962    cpu_loop(env);
 963    /* never exits */
 964    return 0;
 965}
 966