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