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 #include "exec/page-vary.h"
  50
  51 #include "host-os.h"
  52 #include "target_arch_cpu.h"
  53
  54
  55 /*
  56  * TODO: Remove these and rely only on qemu_real_host_page_size().
  57  */
  58 uintptr_t qemu_host_page_size;
  59 intptr_t qemu_host_page_mask;
  60
  61 static bool opt_one_insn_per_tb;
  62 uintptr_t guest_base;
  63 bool have_guest_base;
  64 /*
  65  * When running 32-on-64 we should make sure we can fit all of the possible
  66  * guest address space into a contiguous chunk of virtual host memory.
  67  *
  68  * This way we will never overlap with our own libraries or binaries or stack
  69  * or anything else that QEMU maps.
  70  *
  71  * Many cpus reserve the high bit (or more than one for some 64-bit cpus)
  72  * of the address for the kernel.  Some cpus rely on this and user space
  73  * uses the high bit(s) for pointer tagging and the like.  For them, we
  74  * must preserve the expected address space.
  75  */
  76 #ifndef MAX_RESERVED_VA
  77 # if HOST_LONG_BITS > TARGET_VIRT_ADDR_SPACE_BITS
  78 #  if TARGET_VIRT_ADDR_SPACE_BITS == 32 && \
  79       (TARGET_LONG_BITS == 32 || defined(TARGET_ABI32))
  80 #   define MAX_RESERVED_VA(CPU)  0xfffffffful
  81 #  else
  82 #   define MAX_RESERVED_VA(CPU)  ((1ul << TARGET_VIRT_ADDR_SPACE_BITS) - 1)
  83 #  endif
  84 # else
  85 #  define MAX_RESERVED_VA(CPU)  0
  86 # endif
  87 #endif
  88
  89 unsigned long reserved_va;
  90
  91 const char *interp_prefix = CONFIG_QEMU_INTERP_PREFIX;
  92 const char *qemu_uname_release;
  93 char qemu_proc_pathname[PATH_MAX];  /* full path to exeutable */
  94
  95 unsigned long target_maxtsiz = TARGET_MAXTSIZ;   /* max text size */
  96 unsigned long target_dfldsiz = TARGET_DFLDSIZ;   /* initial data size limit */
  97 unsigned long target_maxdsiz = TARGET_MAXDSIZ;   /* max data size */
  98 unsigned long target_dflssiz = TARGET_DFLSSIZ;   /* initial data size limit */
  99 unsigned long target_maxssiz = TARGET_MAXSSIZ;   /* max stack size */
 100 unsigned long target_sgrowsiz = TARGET_SGROWSIZ; /* amount to grow stack */
 101
 102 /* Helper routines for implementing atomic operations. */
 103
 104 void fork_start(void)
 105 {
 106     start_exclusive();
 107     cpu_list_lock();
 108     mmap_fork_start();
 109     gdbserver_fork_start();
 110 }
 111
 112 void fork_end(pid_t pid)
 113 {
 114     bool child = pid == 0;
 115
 116     if (child) {
 117         CPUState *cpu, *next_cpu;
 118         /*
 119          * Child processes created by fork() only have a single thread.  Discard
 120          * information about the parent threads.
 121          */
 122         CPU_FOREACH_SAFE(cpu, next_cpu) {
 123             if (cpu != thread_cpu) {
 124                 QTAILQ_REMOVE_RCU(&cpus_queue, cpu, node);
 125             }
 126         }
 127         mmap_fork_end(child);
 128         /*
 129          * qemu_init_cpu_list() takes care of reinitializing the exclusive
 130          * state, so we don't need to end_exclusive() here.
 131          */
 132         qemu_init_cpu_list();
 133         get_task_state(thread_cpu)->ts_tid = qemu_get_thread_id();
 134         gdbserver_fork_end(thread_cpu, pid);
 135     } else {
 136         mmap_fork_end(child);
 137         cpu_list_unlock();
 138         gdbserver_fork_end(thread_cpu, pid);
 139         end_exclusive();
 140     }
 141 }
 142
 143 void cpu_loop(CPUArchState *env)
 144 {
 145     target_cpu_loop(env);
 146 }
 147
 148 static void usage(void)
 149 {
 150     printf("qemu-" TARGET_NAME " version " QEMU_FULL_VERSION
 151            "\n" QEMU_COPYRIGHT "\n"
 152            "usage: qemu-" TARGET_NAME " [options] program [arguments...]\n"
 153            "BSD CPU emulator (compiled for %s emulation)\n"
 154            "\n"
 155            "Standard options:\n"
 156            "-h                print this help\n"
 157            "-g port           wait gdb connection to port\n"
 158            "-L path           set the elf interpreter prefix (default=%s)\n"
 159            "-s size           set the stack size in bytes (default=%ld)\n"
 160            "-cpu model        select CPU (-cpu help for list)\n"
 161            "-drop-ld-preload  drop LD_PRELOAD for target process\n"
 162            "-E var=value      sets/modifies targets environment variable(s)\n"
 163            "-U var            unsets targets environment variable(s)\n"
 164            "-B address        set guest_base address to address\n"
 165            "\n"
 166            "Debug options:\n"
 167            "-d item1[,...]    enable logging of specified items\n"
 168            "                  (use '-d help' for a list of log items)\n"
 169            "-D logfile        write logs to 'logfile' (default stderr)\n"
 170            "-one-insn-per-tb  run with one guest instruction per emulated TB\n"
 171            "-strace           log system calls\n"
 172            "-trace            [[enable=]<pattern>][,events=<file>][,file=<file>]\n"
 173            "                  specify tracing options\n"
 174            "\n"
 175            "Environment variables:\n"
 176            "QEMU_STRACE       Print system calls and arguments similar to the\n"
 177            "                  'strace' program.  Enable by setting to any value.\n"
 178            "You can use -E and -U options to set/unset environment variables\n"
 179            "for target process.  It is possible to provide several variables\n"
 180            "by repeating the option.  For example:\n"
 181            "    -E var1=val2 -E var2=val2 -U LD_PRELOAD -U LD_DEBUG\n"
 182            "Note that if you provide several changes to single variable\n"
 183            "last change will stay in effect.\n"
 184            "\n"
 185            QEMU_HELP_BOTTOM "\n"
 186            ,
 187            TARGET_NAME,
 188            interp_prefix,
 189            target_dflssiz);
 190     exit(1);
 191 }
 192
 193 __thread CPUState *thread_cpu;
 194
 195 void stop_all_tasks(void)
 196 {
 197     /*
 198      * We trust when using NPTL (pthreads) start_exclusive() handles thread
 199      * stopping correctly.
 200      */
 201     start_exclusive();
 202 }
 203
 204 bool qemu_cpu_is_self(CPUState *cpu)
 205 {
 206     return thread_cpu == cpu;
 207 }
 208
 209 void qemu_cpu_kick(CPUState *cpu)
 210 {
 211     cpu_exit(cpu);
 212 }
 213
 214 /* Assumes contents are already zeroed.  */
 215 static void init_task_state(TaskState *ts)
 216 {
 217     ts->sigaltstack_used = (struct target_sigaltstack) {
 218         .ss_sp = 0,
 219         .ss_size = 0,
 220         .ss_flags = TARGET_SS_DISABLE,
 221     };
 222 }
 223
 224 void gemu_log(const char *fmt, ...)
 225 {
 226     va_list ap;
 227
 228     va_start(ap, fmt);
 229     vfprintf(stderr, fmt, ap);
 230     va_end(ap);
 231 }
 232
 233 static void
 234 adjust_ssize(void)
 235 {
 236     struct rlimit rl;
 237
 238     if (getrlimit(RLIMIT_STACK, &rl) != 0) {
 239         return;
 240     }
 241
 242     target_maxssiz = MIN(target_maxssiz, rl.rlim_max);
 243     target_dflssiz = MIN(MAX(target_dflssiz, rl.rlim_cur), target_maxssiz);
 244
 245     rl.rlim_max = target_maxssiz;
 246     rl.rlim_cur = target_dflssiz;
 247     setrlimit(RLIMIT_STACK, &rl);
 248 }
 249
 250 static void save_proc_pathname(char *argv0)
 251 {
 252     int mib[4];
 253     size_t len;
 254
 255     mib[0] = CTL_KERN;
 256     mib[1] = KERN_PROC;
 257     mib[2] = KERN_PROC_PATHNAME;
 258     mib[3] = -1;
 259
 260     len = sizeof(qemu_proc_pathname);
 261     if (sysctl(mib, 4, qemu_proc_pathname, &len, NULL, 0)) {
 262         perror("sysctl");
 263     }
 264 }
 265
 266 int main(int argc, char **argv)
 267 {
 268     const char *filename;
 269     const char *cpu_model;
 270     const char *cpu_type;
 271     const char *log_file = NULL;
 272     const char *log_mask = NULL;
 273     const char *seed_optarg = NULL;
 274     struct target_pt_regs regs1, *regs = &regs1;
 275     struct image_info info1, *info = &info1;
 276     struct bsd_binprm bprm;
 277     TaskState *ts;
 278     CPUArchState *env;
 279     CPUState *cpu;
 280     int optind, rv;
 281     const char *r;
 282     const char *gdbstub = NULL;
 283     char **target_environ, **wrk;
 284     envlist_t *envlist = NULL;
 285     char *argv0 = NULL;
 286     int host_page_size;
 287     unsigned long max_reserved_va;
 288
 289     adjust_ssize();
 290
 291     if (argc <= 1) {
 292         usage();
 293     }
 294
 295     save_proc_pathname(argv[0]);
 296
 297     error_init(argv[0]);
 298     module_call_init(MODULE_INIT_TRACE);
 299     qemu_init_cpu_list();
 300     module_call_init(MODULE_INIT_QOM);
 301
 302     envlist = envlist_create();
 303
 304     /*
 305      * add current environment into the list
 306      * envlist_setenv adds to the front of the list; to preserve environ
 307      * order add from back to front
 308      */
 309     for (wrk = environ; *wrk != NULL; wrk++) {
 310         continue;
 311     }
 312     while (wrk != environ) {
 313         wrk--;
 314         (void) envlist_setenv(envlist, *wrk);
 315     }
 316
 317     qemu_host_page_size = getpagesize();
 318     qemu_host_page_size = MAX(qemu_host_page_size, TARGET_PAGE_SIZE);
 319
 320     cpu_model = NULL;
 321
 322     qemu_add_opts(&qemu_trace_opts);
 323
 324     optind = 1;
 325     for (;;) {
 326         if (optind >= argc) {
 327             break;
 328         }
 329         r = argv[optind];
 330         if (r[0] != '-') {
 331             break;
 332         }
 333         optind++;
 334         r++;
 335         if (!strcmp(r, "-")) {
 336             break;
 337         } else if (!strcmp(r, "d")) {
 338             if (optind >= argc) {
 339                 break;
 340             }
 341             log_mask = argv[optind++];
 342         } else if (!strcmp(r, "D")) {
 343             if (optind >= argc) {
 344                 break;
 345             }
 346             log_file = argv[optind++];
 347         } else if (!strcmp(r, "E")) {
 348             r = argv[optind++];
 349             if (envlist_setenv(envlist, r) != 0) {
 350                 usage();
 351             }
 352         } else if (!strcmp(r, "ignore-environment")) {
 353             envlist_free(envlist);
 354             envlist = envlist_create();
 355         } else if (!strcmp(r, "U")) {
 356             r = argv[optind++];
 357             if (envlist_unsetenv(envlist, r) != 0) {
 358                 usage();
 359             }
 360         } else if (!strcmp(r, "s")) {
 361             r = argv[optind++];
 362             rv = qemu_strtoul(r, &r, 0, &target_dflssiz);
 363             if (rv < 0 || target_dflssiz <= 0) {
 364                 usage();
 365             }
 366             if (*r == 'M') {
 367                 target_dflssiz *= 1024 * 1024;
 368             } else if (*r == 'k' || *r == 'K') {
 369                 target_dflssiz *= 1024;
 370             }
 371             if (target_dflssiz > target_maxssiz) {
 372                 usage();
 373             }
 374         } else if (!strcmp(r, "L")) {
 375             interp_prefix = argv[optind++];
 376         } else if (!strcmp(r, "p")) {
 377             unsigned size, want = qemu_real_host_page_size();
 378
 379             r = argv[optind++];
 380             if (qemu_strtoui(r, NULL, 10, &size) || size != want) {
 381                 warn_report("Deprecated page size option cannot "
 382                             "change host page size (%u)", want);
 383             }
 384         } else if (!strcmp(r, "g")) {
 385             gdbstub = g_strdup(argv[optind++]);
 386         } else if (!strcmp(r, "r")) {
 387             qemu_uname_release = argv[optind++];
 388         } else if (!strcmp(r, "cpu")) {
 389             cpu_model = argv[optind++];
 390             if (is_help_option(cpu_model)) {
 391                 list_cpus();
 392                 exit(1);
 393             }
 394         } else if (!strcmp(r, "B")) {
 395             rv = qemu_strtoul(argv[optind++], NULL, 0, &guest_base);
 396             if (rv < 0) {
 397                 usage();
 398             }
 399             have_guest_base = true;
 400         } else if (!strcmp(r, "drop-ld-preload")) {
 401             (void) envlist_unsetenv(envlist, "LD_PRELOAD");
 402         } else if (!strcmp(r, "seed")) {
 403             seed_optarg = optarg;
 404         } else if (!strcmp(r, "one-insn-per-tb")) {
 405             opt_one_insn_per_tb = true;
 406         } else if (!strcmp(r, "strace")) {
 407             do_strace = 1;
 408         } else if (!strcmp(r, "trace")) {
 409             trace_opt_parse(optarg);
 410         } else if (!strcmp(r, "0")) {
 411             argv0 = argv[optind++];
 412         } else {
 413             usage();
 414         }
 415     }
 416
 417     qemu_host_page_mask = -qemu_host_page_size;
 418
 419     /* init debug */
 420     {
 421         int mask = 0;
 422         if (log_mask) {
 423             mask = qemu_str_to_log_mask(log_mask);
 424             if (!mask) {
 425                 qemu_print_log_usage(stdout);
 426                 exit(1);
 427             }
 428         }
 429         qemu_set_log_filename_flags(log_file, mask, &error_fatal);
 430     }
 431
 432     if (optind >= argc) {
 433         usage();
 434     }
 435     filename = argv[optind];
 436     if (argv0) {
 437         argv[optind] = argv0;
 438     }
 439
 440     if (!trace_init_backends()) {
 441         exit(1);
 442     }
 443     trace_init_file();
 444
 445     /* Zero out regs */
 446     memset(regs, 0, sizeof(struct target_pt_regs));
 447
 448     /* Zero bsd params */
 449     memset(&bprm, 0, sizeof(bprm));
 450
 451     /* Zero out image_info */
 452     memset(info, 0, sizeof(struct image_info));
 453
 454     /* Scan interp_prefix dir for replacement files. */
 455     init_paths(interp_prefix);
 456
 457     if (cpu_model == NULL) {
 458         cpu_model = TARGET_DEFAULT_CPU_MODEL;
 459     }
 460
 461     cpu_type = parse_cpu_option(cpu_model);
 462
 463     /* init tcg before creating CPUs and to get qemu_host_page_size */
 464     {
 465         AccelState *accel = current_accel();
 466         AccelClass *ac = ACCEL_GET_CLASS(accel);
 467
 468         accel_init_interfaces(ac);
 469         object_property_set_bool(OBJECT(accel), "one-insn-per-tb",
 470                                  opt_one_insn_per_tb, &error_abort);
 471         ac->init_machine(NULL);
 472     }
 473
 474     /*
 475      * Finalize page size before creating CPUs.
 476      * This will do nothing if !TARGET_PAGE_BITS_VARY.
 477      * The most efficient setting is to match the host.
 478      */
 479     host_page_size = qemu_real_host_page_size();
 480     set_preferred_target_page_bits(ctz32(host_page_size));
 481     finalize_target_page_bits();
 482
 483     cpu = cpu_create(cpu_type);
 484     env = cpu_env(cpu);
 485     cpu_reset(cpu);
 486     thread_cpu = cpu;
 487
 488     /*
 489      * Reserving too much vm space via mmap can run into problems with rlimits,
 490      * oom due to page table creation, etc.  We will still try it, if directed
 491      * by the command-line option, but not by default. Unless we're running a
 492      * target address space of 32 or fewer bits on a host with 64 bits.
 493      */
 494     max_reserved_va = MAX_RESERVED_VA(cpu);
 495     if (reserved_va != 0) {
 496         if ((reserved_va + 1) % host_page_size) {
 497             char *s = size_to_str(host_page_size);
 498             fprintf(stderr, "Reserved virtual address not aligned mod %s\n", s);
 499             g_free(s);
 500             exit(EXIT_FAILURE);
 501         }
 502         if (max_reserved_va && reserved_va > max_reserved_va) {
 503             fprintf(stderr, "Reserved virtual address too big\n");
 504             exit(EXIT_FAILURE);
 505         }
 506     } else if (HOST_LONG_BITS == 64 && TARGET_VIRT_ADDR_SPACE_BITS <= 32) {
 507         /* MAX_RESERVED_VA + 1 is a large power of 2, so is aligned. */
 508         reserved_va = max_reserved_va;
 509     }
 510
 511     if (getenv("QEMU_STRACE")) {
 512         do_strace = 1;
 513     }
 514
 515     target_environ = envlist_to_environ(envlist, NULL);
 516     envlist_free(envlist);
 517
 518     {
 519         Error *err = NULL;
 520         if (seed_optarg != NULL) {
 521             qemu_guest_random_seed_main(seed_optarg, &err);
 522         } else {
 523             qcrypto_init(&err);
 524         }
 525         if (err) {
 526             error_reportf_err(err, "cannot initialize crypto: ");
 527             exit(1);
 528         }
 529     }
 530
 531     /*
 532      * Now that page sizes are configured we can do
 533      * proper page alignment for guest_base.
 534      */
 535     if (have_guest_base) {
 536         if (guest_base & ~qemu_host_page_mask) {
 537             error_report("Selected guest base not host page aligned");
 538             exit(1);
 539         }
 540     }
 541
 542     /*
 543      * If reserving host virtual address space, do so now.
 544      * Combined with '-B', ensure that the chosen range is free.
 545      */
 546     if (reserved_va) {
 547         void *p;
 548
 549         if (have_guest_base) {
 550             p = mmap((void *)guest_base, reserved_va + 1, PROT_NONE,
 551                      MAP_ANON | MAP_PRIVATE | MAP_FIXED | MAP_EXCL, -1, 0);
 552         } else {
 553             p = mmap(NULL, reserved_va + 1, PROT_NONE,
 554                      MAP_ANON | MAP_PRIVATE, -1, 0);
 555         }
 556         if (p == MAP_FAILED) {
 557             const char *err = strerror(errno);
 558             char *sz = size_to_str(reserved_va + 1);
 559
 560             if (have_guest_base) {
 561                 error_report("Cannot allocate %s bytes at -B %p for guest "
 562                              "address space: %s", sz, (void *)guest_base, err);
 563             } else {
 564                 error_report("Cannot allocate %s bytes for guest "
 565                              "address space: %s", sz, err);
 566             }
 567             exit(1);
 568         }
 569         guest_base = (uintptr_t)p;
 570         have_guest_base = true;
 571
 572         /* Ensure that mmap_next_start is within range. */
 573         if (reserved_va <= mmap_next_start) {
 574             mmap_next_start = (reserved_va / 4 * 3)
 575                               & TARGET_PAGE_MASK & qemu_host_page_mask;
 576         }
 577     }
 578
 579     if (loader_exec(filename, argv + optind, target_environ, regs, info,
 580                     &bprm) != 0) {
 581         printf("Error loading %s\n", filename);
 582         _exit(1);
 583     }
 584
 585     for (wrk = target_environ; *wrk; wrk++) {
 586         g_free(*wrk);
 587     }
 588
 589     g_free(target_environ);
 590
 591     if (qemu_loglevel_mask(CPU_LOG_PAGE)) {
 592         FILE *f = qemu_log_trylock();
 593         if (f) {
 594             fprintf(f, "guest_base  %p\n", (void *)guest_base);
 595             fprintf(f, "page layout changed following binary load\n");
 596             page_dump(f);
 597
 598             fprintf(f, "end_code    0x" TARGET_ABI_FMT_lx "\n",
 599                     info->end_code);
 600             fprintf(f, "start_code  0x" TARGET_ABI_FMT_lx "\n",
 601                     info->start_code);
 602             fprintf(f, "start_data  0x" TARGET_ABI_FMT_lx "\n",
 603                     info->start_data);
 604             fprintf(f, "end_data    0x" TARGET_ABI_FMT_lx "\n",
 605                     info->end_data);
 606             fprintf(f, "start_stack 0x" TARGET_ABI_FMT_lx "\n",
 607                     info->start_stack);
 608             fprintf(f, "brk         0x" TARGET_ABI_FMT_lx "\n", info->brk);
 609             fprintf(f, "entry       0x" TARGET_ABI_FMT_lx "\n", info->entry);
 610
 611             qemu_log_unlock(f);
 612         }
 613     }
 614
 615     /* build Task State */
 616     ts = g_new0(TaskState, 1);
 617     init_task_state(ts);
 618     ts->info = info;
 619     ts->bprm = &bprm;
 620     cpu->opaque = ts;
 621
 622     target_set_brk(info->brk);
 623     syscall_init();
 624     signal_init();
 625
 626     /*
 627      * Now that we've loaded the binary, GUEST_BASE is fixed.  Delay
 628      * generating the prologue until now so that the prologue can take
 629      * the real value of GUEST_BASE into account.
 630      */
 631     tcg_prologue_init();
 632
 633     target_cpu_init(env, regs);
 634
 635     if (gdbstub) {
 636         gdbserver_start(gdbstub);
 637         gdb_handlesig(cpu, 0, NULL, NULL, 0);
 638     }
 639     cpu_loop(env);
 640     /* never exits */
 641     return 0;
 642 }