bsd-user/main.c

   1 /*
   2  *  qemu bsd user main
   3  *
   4  *  Copyright (c) 2003-2008 Fabrice Bellard
   5  *  Copyright (c) 2013-14 Stacey Son
   6  *
   7  *  This program is free software; you can redistribute it and/or modify
   8  *  it under the terms of the GNU General Public License as published by
   9  *  the Free Software Foundation; either version 2 of the License, or
  10  *  (at your option) any later version.
  11  *
  12  *  This program is distributed in the hope that it will be useful,
  13  *  but WITHOUT ANY WARRANTY; without even the implied warranty of
  14  *  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  15  *  GNU General Public License for more details.
  16  *
  17  *  You should have received a copy of the GNU General Public License
  18  *  along with this program; if not, see <http://www.gnu.org/licenses/>.
  19  */
  20
  21 #include "qemu/osdep.h"
  22 #include <sys/resource.h>
  23 #include <sys/sysctl.h>
  24
  25 #include "qemu/help-texts.h"
  26 #include "qemu/units.h"
  27 #include "qemu/accel.h"
  28 #include "qemu-version.h"
  29 #include <machine/trap.h>
  30
  31 #include "qapi/error.h"
  32 #include "qemu.h"
  33 #include "qemu/config-file.h"
  34 #include "qemu/error-report.h"
  35 #include "qemu/path.h"
  36 #include "qemu/help_option.h"
  37 #include "qemu/module.h"
  38 #include "exec/exec-all.h"
  39 #include "user/guest-base.h"
  40 #include "tcg/startup.h"
  41 #include "qemu/timer.h"
  42 #include "qemu/envlist.h"
  43 #include "qemu/cutils.h"
  44 #include "exec/log.h"
  45 #include "trace/control.h"
  46 #include "crypto/init.h"
  47 #include "qemu/guest-random.h"
  48 #include "gdbstub/user.h"
  49
  50 #include "host-os.h"
  51 #include "target_arch_cpu.h"
  52
  53
  54 /*
  55  * TODO: Remove these and rely only on qemu_real_host_page_size().
  56  */
  57 uintptr_t qemu_host_page_size;
  58 intptr_t qemu_host_page_mask;
  59
  60 static bool opt_one_insn_per_tb;
  61 uintptr_t guest_base;
  62 bool have_guest_base;
  63 /*
  64  * When running 32-on-64 we should make sure we can fit all of the possible
  65  * guest address space into a contiguous chunk of virtual host memory.
  66  *
  67  * This way we will never overlap with our own libraries or binaries or stack
  68  * or anything else that QEMU maps.
  69  *
  70  * Many cpus reserve the high bit (or more than one for some 64-bit cpus)
  71  * of the address for the kernel.  Some cpus rely on this and user space
  72  * uses the high bit(s) for pointer tagging and the like.  For them, we
  73  * must preserve the expected address space.
  74  */
  75 #ifndef MAX_RESERVED_VA
  76 # if HOST_LONG_BITS > TARGET_VIRT_ADDR_SPACE_BITS
  77 #  if TARGET_VIRT_ADDR_SPACE_BITS == 32 && \
  78       (TARGET_LONG_BITS == 32 || defined(TARGET_ABI32))
  79 #   define MAX_RESERVED_VA  0xfffffffful
  80 #  else
  81 #   define MAX_RESERVED_VA  ((1ul << TARGET_VIRT_ADDR_SPACE_BITS) - 1)
  82 #  endif
  83 # else
  84 #  define MAX_RESERVED_VA  0
  85 # endif
  86 #endif
  87
  88 /*
  89  * That said, reserving *too* much vm space via mmap can run into problems
  90  * with rlimits, oom due to page table creation, etc.  We will still try it,
  91  * if directed by the command-line option, but not by default.
  92  */
  93 #if HOST_LONG_BITS == 64 && TARGET_VIRT_ADDR_SPACE_BITS <= 32
  94 unsigned long reserved_va = MAX_RESERVED_VA;
  95 #else
  96 unsigned long reserved_va;
  97 #endif
  98
  99 const char *interp_prefix = CONFIG_QEMU_INTERP_PREFIX;
 100 const char *qemu_uname_release;
 101 char qemu_proc_pathname[PATH_MAX];  /* full path to exeutable */
 102
 103 unsigned long target_maxtsiz = TARGET_MAXTSIZ;   /* max text size */
 104 unsigned long target_dfldsiz = TARGET_DFLDSIZ;   /* initial data size limit */
 105 unsigned long target_maxdsiz = TARGET_MAXDSIZ;   /* max data size */
 106 unsigned long target_dflssiz = TARGET_DFLSSIZ;   /* initial data size limit */
 107 unsigned long target_maxssiz = TARGET_MAXSSIZ;   /* max stack size */
 108 unsigned long target_sgrowsiz = TARGET_SGROWSIZ; /* amount to grow stack */
 109
 110 /* Helper routines for implementing atomic operations. */
 111
 112 void fork_start(void)
 113 {
 114     start_exclusive();
 115     cpu_list_lock();
 116     mmap_fork_start();
 117     gdbserver_fork_start();
 118 }
 119
 120 void fork_end(pid_t pid)
 121 {
 122     bool child = pid == 0;
 123
 124     if (child) {
 125         CPUState *cpu, *next_cpu;
 126         /*
 127          * Child processes created by fork() only have a single thread.  Discard
 128          * information about the parent threads.
 129          */
 130         CPU_FOREACH_SAFE(cpu, next_cpu) {
 131             if (cpu != thread_cpu) {
 132                 QTAILQ_REMOVE_RCU(&cpus_queue, cpu, node);
 133             }
 134         }
 135         mmap_fork_end(child);
 136         /*
 137          * qemu_init_cpu_list() takes care of reinitializing the exclusive
 138          * state, so we don't need to end_exclusive() here.
 139          */
 140         qemu_init_cpu_list();
 141         get_task_state(thread_cpu)->ts_tid = qemu_get_thread_id();
 142         gdbserver_fork_end(thread_cpu, pid);
 143     } else {
 144         mmap_fork_end(child);
 145         cpu_list_unlock();
 146         gdbserver_fork_end(thread_cpu, pid);
 147         end_exclusive();
 148     }
 149 }
 150
 151 void cpu_loop(CPUArchState *env)
 152 {
 153     target_cpu_loop(env);
 154 }
 155
 156 static void usage(void)
 157 {
 158     printf("qemu-" TARGET_NAME " version " QEMU_FULL_VERSION
 159            "\n" QEMU_COPYRIGHT "\n"
 160            "usage: qemu-" TARGET_NAME " [options] program [arguments...]\n"
 161            "BSD CPU emulator (compiled for %s emulation)\n"
 162            "\n"
 163            "Standard options:\n"
 164            "-h                print this help\n"
 165            "-g port           wait gdb connection to port\n"
 166            "-L path           set the elf interpreter prefix (default=%s)\n"
 167            "-s size           set the stack size in bytes (default=%ld)\n"
 168            "-cpu model        select CPU (-cpu help for list)\n"
 169            "-drop-ld-preload  drop LD_PRELOAD for target process\n"
 170            "-E var=value      sets/modifies targets environment variable(s)\n"
 171            "-U var            unsets targets environment variable(s)\n"
 172            "-B address        set guest_base address to address\n"
 173            "\n"
 174            "Debug options:\n"
 175            "-d item1[,...]    enable logging of specified items\n"
 176            "                  (use '-d help' for a list of log items)\n"
 177            "-D logfile        write logs to 'logfile' (default stderr)\n"
 178            "-one-insn-per-tb  run with one guest instruction per emulated TB\n"
 179            "-strace           log system calls\n"
 180            "-trace            [[enable=]<pattern>][,events=<file>][,file=<file>]\n"
 181            "                  specify tracing options\n"
 182            "\n"
 183            "Environment variables:\n"
 184            "QEMU_STRACE       Print system calls and arguments similar to the\n"
 185            "                  'strace' program.  Enable by setting to any value.\n"
 186            "You can use -E and -U options to set/unset environment variables\n"
 187            "for target process.  It is possible to provide several variables\n"
 188            "by repeating the option.  For example:\n"
 189            "    -E var1=val2 -E var2=val2 -U LD_PRELOAD -U LD_DEBUG\n"
 190            "Note that if you provide several changes to single variable\n"
 191            "last change will stay in effect.\n"
 192            "\n"
 193            QEMU_HELP_BOTTOM "\n"
 194            ,
 195            TARGET_NAME,
 196            interp_prefix,
 197            target_dflssiz);
 198     exit(1);
 199 }
 200
 201 __thread CPUState *thread_cpu;
 202
 203 void stop_all_tasks(void)
 204 {
 205     /*
 206      * We trust when using NPTL (pthreads) start_exclusive() handles thread
 207      * stopping correctly.
 208      */
 209     start_exclusive();
 210 }
 211
 212 bool qemu_cpu_is_self(CPUState *cpu)
 213 {
 214     return thread_cpu == cpu;
 215 }
 216
 217 void qemu_cpu_kick(CPUState *cpu)
 218 {
 219     cpu_exit(cpu);
 220 }
 221
 222 /* Assumes contents are already zeroed.  */
 223 static void init_task_state(TaskState *ts)
 224 {
 225     ts->sigaltstack_used = (struct target_sigaltstack) {
 226         .ss_sp = 0,
 227         .ss_size = 0,
 228         .ss_flags = TARGET_SS_DISABLE,
 229     };
 230 }
 231
 232 void gemu_log(const char *fmt, ...)
 233 {
 234     va_list ap;
 235
 236     va_start(ap, fmt);
 237     vfprintf(stderr, fmt, ap);
 238     va_end(ap);
 239 }
 240
 241 static void
 242 adjust_ssize(void)
 243 {
 244     struct rlimit rl;
 245
 246     if (getrlimit(RLIMIT_STACK, &rl) != 0) {
 247         return;
 248     }
 249
 250     target_maxssiz = MIN(target_maxssiz, rl.rlim_max);
 251     target_dflssiz = MIN(MAX(target_dflssiz, rl.rlim_cur), target_maxssiz);
 252
 253     rl.rlim_max = target_maxssiz;
 254     rl.rlim_cur = target_dflssiz;
 255     setrlimit(RLIMIT_STACK, &rl);
 256 }
 257
 258 static void save_proc_pathname(char *argv0)
 259 {
 260     int mib[4];
 261     size_t len;
 262
 263     mib[0] = CTL_KERN;
 264     mib[1] = KERN_PROC;
 265     mib[2] = KERN_PROC_PATHNAME;
 266     mib[3] = -1;
 267
 268     len = sizeof(qemu_proc_pathname);
 269     if (sysctl(mib, 4, qemu_proc_pathname, &len, NULL, 0)) {
 270         perror("sysctl");
 271     }
 272 }
 273
 274 int main(int argc, char **argv)
 275 {
 276     const char *filename;
 277     const char *cpu_model;
 278     const char *cpu_type;
 279     const char *log_file = NULL;
 280     const char *log_mask = NULL;
 281     const char *seed_optarg = NULL;
 282     struct target_pt_regs regs1, *regs = &regs1;
 283     struct image_info info1, *info = &info1;
 284     struct bsd_binprm bprm;
 285     TaskState *ts;
 286     CPUArchState *env;
 287     CPUState *cpu;
 288     int optind, rv;
 289     const char *r;
 290     const char *gdbstub = NULL;
 291     char **target_environ, **wrk;
 292     envlist_t *envlist = NULL;
 293     char *argv0 = NULL;
 294
 295     adjust_ssize();
 296
 297     if (argc <= 1) {
 298         usage();
 299     }
 300
 301     save_proc_pathname(argv[0]);
 302
 303     error_init(argv[0]);
 304     module_call_init(MODULE_INIT_TRACE);
 305     qemu_init_cpu_list();
 306     module_call_init(MODULE_INIT_QOM);
 307
 308     envlist = envlist_create();
 309
 310     /*
 311      * add current environment into the list
 312      * envlist_setenv adds to the front of the list; to preserve environ
 313      * order add from back to front
 314      */
 315     for (wrk = environ; *wrk != NULL; wrk++) {
 316         continue;
 317     }
 318     while (wrk != environ) {
 319         wrk--;
 320         (void) envlist_setenv(envlist, *wrk);
 321     }
 322
 323     qemu_host_page_size = getpagesize();
 324     qemu_host_page_size = MAX(qemu_host_page_size, TARGET_PAGE_SIZE);
 325
 326     cpu_model = NULL;
 327
 328     qemu_add_opts(&qemu_trace_opts);
 329
 330     optind = 1;
 331     for (;;) {
 332         if (optind >= argc) {
 333             break;
 334         }
 335         r = argv[optind];
 336         if (r[0] != '-') {
 337             break;
 338         }
 339         optind++;
 340         r++;
 341         if (!strcmp(r, "-")) {
 342             break;
 343         } else if (!strcmp(r, "d")) {
 344             if (optind >= argc) {
 345                 break;
 346             }
 347             log_mask = argv[optind++];
 348         } else if (!strcmp(r, "D")) {
 349             if (optind >= argc) {
 350                 break;
 351             }
 352             log_file = argv[optind++];
 353         } else if (!strcmp(r, "E")) {
 354             r = argv[optind++];
 355             if (envlist_setenv(envlist, r) != 0) {
 356                 usage();
 357             }
 358         } else if (!strcmp(r, "ignore-environment")) {
 359             envlist_free(envlist);
 360             envlist = envlist_create();
 361         } else if (!strcmp(r, "U")) {
 362             r = argv[optind++];
 363             if (envlist_unsetenv(envlist, r) != 0) {
 364                 usage();
 365             }
 366         } else if (!strcmp(r, "s")) {
 367             r = argv[optind++];
 368             rv = qemu_strtoul(r, &r, 0, &target_dflssiz);
 369             if (rv < 0 || target_dflssiz <= 0) {
 370                 usage();
 371             }
 372             if (*r == 'M') {
 373                 target_dflssiz *= 1024 * 1024;
 374             } else if (*r == 'k' || *r == 'K') {
 375                 target_dflssiz *= 1024;
 376             }
 377             if (target_dflssiz > target_maxssiz) {
 378                 usage();
 379             }
 380         } else if (!strcmp(r, "L")) {
 381             interp_prefix = argv[optind++];
 382         } else if (!strcmp(r, "p")) {
 383             unsigned size, want = qemu_real_host_page_size();
 384
 385             r = argv[optind++];
 386             if (qemu_strtoui(r, NULL, 10, &size) || size != want) {
 387                 warn_report("Deprecated page size option cannot "
 388                             "change host page size (%u)", want);
 389             }
 390         } else if (!strcmp(r, "g")) {
 391             gdbstub = g_strdup(argv[optind++]);
 392         } else if (!strcmp(r, "r")) {
 393             qemu_uname_release = argv[optind++];
 394         } else if (!strcmp(r, "cpu")) {
 395             cpu_model = argv[optind++];
 396             if (is_help_option(cpu_model)) {
 397                 list_cpus();
 398                 exit(1);
 399             }
 400         } else if (!strcmp(r, "B")) {
 401             rv = qemu_strtoul(argv[optind++], NULL, 0, &guest_base);
 402             if (rv < 0) {
 403                 usage();
 404             }
 405             have_guest_base = true;
 406         } else if (!strcmp(r, "drop-ld-preload")) {
 407             (void) envlist_unsetenv(envlist, "LD_PRELOAD");
 408         } else if (!strcmp(r, "seed")) {
 409             seed_optarg = optarg;
 410         } else if (!strcmp(r, "one-insn-per-tb")) {
 411             opt_one_insn_per_tb = true;
 412         } else if (!strcmp(r, "strace")) {
 413             do_strace = 1;
 414         } else if (!strcmp(r, "trace")) {
 415             trace_opt_parse(optarg);
 416         } else if (!strcmp(r, "0")) {
 417             argv0 = argv[optind++];
 418         } else {
 419             usage();
 420         }
 421     }
 422
 423     qemu_host_page_mask = -qemu_host_page_size;
 424
 425     /* init debug */
 426     {
 427         int mask = 0;
 428         if (log_mask) {
 429             mask = qemu_str_to_log_mask(log_mask);
 430             if (!mask) {
 431                 qemu_print_log_usage(stdout);
 432                 exit(1);
 433             }
 434         }
 435         qemu_set_log_filename_flags(log_file, mask, &error_fatal);
 436     }
 437
 438     if (optind >= argc) {
 439         usage();
 440     }
 441     filename = argv[optind];
 442     if (argv0) {
 443         argv[optind] = argv0;
 444     }
 445
 446     if (!trace_init_backends()) {
 447         exit(1);
 448     }
 449     trace_init_file();
 450
 451     /* Zero out regs */
 452     memset(regs, 0, sizeof(struct target_pt_regs));
 453
 454     /* Zero bsd params */
 455     memset(&bprm, 0, sizeof(bprm));
 456
 457     /* Zero out image_info */
 458     memset(info, 0, sizeof(struct image_info));
 459
 460     /* Scan interp_prefix dir for replacement files. */
 461     init_paths(interp_prefix);
 462
 463     if (cpu_model == NULL) {
 464         cpu_model = TARGET_DEFAULT_CPU_MODEL;
 465     }
 466
 467     cpu_type = parse_cpu_option(cpu_model);
 468
 469     /* init tcg before creating CPUs and to get qemu_host_page_size */
 470     {
 471         AccelState *accel = current_accel();
 472         AccelClass *ac = ACCEL_GET_CLASS(accel);
 473
 474         accel_init_interfaces(ac);
 475         object_property_set_bool(OBJECT(accel), "one-insn-per-tb",
 476                                  opt_one_insn_per_tb, &error_abort);
 477         ac->init_machine(NULL);
 478     }
 479     cpu = cpu_create(cpu_type);
 480     env = cpu_env(cpu);
 481     cpu_reset(cpu);
 482     thread_cpu = cpu;
 483
 484     if (getenv("QEMU_STRACE")) {
 485         do_strace = 1;
 486     }
 487
 488     target_environ = envlist_to_environ(envlist, NULL);
 489     envlist_free(envlist);
 490
 491     {
 492         Error *err = NULL;
 493         if (seed_optarg != NULL) {
 494             qemu_guest_random_seed_main(seed_optarg, &err);
 495         } else {
 496             qcrypto_init(&err);
 497         }
 498         if (err) {
 499             error_reportf_err(err, "cannot initialize crypto: ");
 500             exit(1);
 501         }
 502     }
 503
 504     /*
 505      * Now that page sizes are configured we can do
 506      * proper page alignment for guest_base.
 507      */
 508     if (have_guest_base) {
 509         if (guest_base & ~qemu_host_page_mask) {
 510             error_report("Selected guest base not host page aligned");
 511             exit(1);
 512         }
 513     }
 514
 515     /*
 516      * If reserving host virtual address space, do so now.
 517      * Combined with '-B', ensure that the chosen range is free.
 518      */
 519     if (reserved_va) {
 520         void *p;
 521
 522         if (have_guest_base) {
 523             p = mmap((void *)guest_base, reserved_va + 1, PROT_NONE,
 524                      MAP_ANON | MAP_PRIVATE | MAP_FIXED | MAP_EXCL, -1, 0);
 525         } else {
 526             p = mmap(NULL, reserved_va + 1, PROT_NONE,
 527                      MAP_ANON | MAP_PRIVATE, -1, 0);
 528         }
 529         if (p == MAP_FAILED) {
 530             const char *err = strerror(errno);
 531             char *sz = size_to_str(reserved_va + 1);
 532
 533             if (have_guest_base) {
 534                 error_report("Cannot allocate %s bytes at -B %p for guest "
 535                              "address space: %s", sz, (void *)guest_base, err);
 536             } else {
 537                 error_report("Cannot allocate %s bytes for guest "
 538                              "address space: %s", sz, err);
 539             }
 540             exit(1);
 541         }
 542         guest_base = (uintptr_t)p;
 543         have_guest_base = true;
 544
 545         /* Ensure that mmap_next_start is within range. */
 546         if (reserved_va <= mmap_next_start) {
 547             mmap_next_start = (reserved_va / 4 * 3)
 548                               & TARGET_PAGE_MASK & qemu_host_page_mask;
 549         }
 550     }
 551
 552     if (loader_exec(filename, argv + optind, target_environ, regs, info,
 553                     &bprm) != 0) {
 554         printf("Error loading %s\n", filename);
 555         _exit(1);
 556     }
 557
 558     for (wrk = target_environ; *wrk; wrk++) {
 559         g_free(*wrk);
 560     }
 561
 562     g_free(target_environ);
 563
 564     if (qemu_loglevel_mask(CPU_LOG_PAGE)) {
 565         FILE *f = qemu_log_trylock();
 566         if (f) {
 567             fprintf(f, "guest_base  %p\n", (void *)guest_base);
 568             fprintf(f, "page layout changed following binary load\n");
 569             page_dump(f);
 570
 571             fprintf(f, "end_code    0x" TARGET_ABI_FMT_lx "\n",
 572                     info->end_code);
 573             fprintf(f, "start_code  0x" TARGET_ABI_FMT_lx "\n",
 574                     info->start_code);
 575             fprintf(f, "start_data  0x" TARGET_ABI_FMT_lx "\n",
 576                     info->start_data);
 577             fprintf(f, "end_data    0x" TARGET_ABI_FMT_lx "\n",
 578                     info->end_data);
 579             fprintf(f, "start_stack 0x" TARGET_ABI_FMT_lx "\n",
 580                     info->start_stack);
 581             fprintf(f, "brk         0x" TARGET_ABI_FMT_lx "\n", info->brk);
 582             fprintf(f, "entry       0x" TARGET_ABI_FMT_lx "\n", info->entry);
 583
 584             qemu_log_unlock(f);
 585         }
 586     }
 587
 588     /* build Task State */
 589     ts = g_new0(TaskState, 1);
 590     init_task_state(ts);
 591     ts->info = info;
 592     ts->bprm = &bprm;
 593     cpu->opaque = ts;
 594
 595     target_set_brk(info->brk);
 596     syscall_init();
 597     signal_init();
 598
 599     /*
 600      * Now that we've loaded the binary, GUEST_BASE is fixed.  Delay
 601      * generating the prologue until now so that the prologue can take
 602      * the real value of GUEST_BASE into account.
 603      */
 604     tcg_prologue_init();
 605
 606     target_cpu_init(env, regs);
 607
 608     if (gdbstub) {
 609         gdbserver_start(gdbstub);
 610         gdb_handlesig(cpu, 0, NULL, NULL, 0);
 611     }
 612     cpu_loop(env);
 613     /* never exits */
 614     return 0;
 615 }